add NEON GPU rasterizer

8 files changed, 12931 insertions, 0 deletions

diff --git a/plugins/gpu_neon/psx_gpu/common.h b/plugins/gpu_neon/psx_gpu/common.h
new file mode 100644
index 0000000..f299f79
--- /dev/null
+++ b/plugins/gpu_neon/psx_gpu/common.h
@@ -0,0 +1,22 @@
+#ifndef COMMON_H
+#define COMMON_H
+
+typedef signed char s8;
+typedef unsigned char u8;
+typedef signed short s16;
+typedef unsigned short u16;
+typedef signed int s32;
+typedef unsigned int u32;
+typedef signed long long int s64;
+typedef unsigned long long int u64;
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <sys/time.h>
+
+#include "vector_ops.h"
+#include "psx_gpu.h"
+
+#endif
+
diff --git a/plugins/gpu_neon/psx_gpu/psx_gpu.c b/plugins/gpu_neon/psx_gpu/psx_gpu.c
new file mode 100644
index 0000000..8d4b8b8
--- /dev/null
+++ b/plugins/gpu_neon/psx_gpu/psx_gpu.c
@@ -0,0 +1,4440 @@
+/*
+ * Copyright (C) 2011 Gilead Kutnick "Exophase" <exophase@gmail.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; either version 2 of
+ * the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+#include "common.h"
+
+u32 span_pixels = 0;
+u32 span_pixel_blocks = 0;
+u32 span_pixel_blocks_unaligned = 0;
+u32 spans = 0;
+u32 triangles = 0;
+u32 sprites = 0;
+u32 sprites_4bpp = 0;
+u32 sprites_8bpp = 0;
+u32 sprites_16bpp = 0;
+u32 sprite_blocks = 0;
+u32 sprites_untextured = 0;
+u32 lines = 0;
+u32 trivial_rejects = 0;
+u32 texels_4bpp = 0;
+u32 texels_8bpp = 0;
+u32 texels_16bpp = 0;
+u32 texel_blocks_4bpp = 0;
+u32 texel_blocks_8bpp = 0;
+u32 texel_blocks_16bpp = 0;
+u32 texel_blocks_untextured = 0;
+u32 blend_blocks = 0;
+u32 untextured_pixels = 0;
+u32 blend_pixels = 0;
+u32 transparent_pixels = 0;
+u32 render_buffer_flushes = 0;
+u32 state_changes = 0;
+u32 left_split_triangles = 0;
+u32 flat_triangles = 0;
+u32 clipped_triangles = 0;
+u32 zero_block_spans = 0;
+u32 texture_cache_loads = 0;
+u32 false_modulated_triangles = 0;
+u32 false_modulated_sprites = 0;
+
+u32 reciprocal_table[512];
+
+
+typedef s32 fixed_type;
+
+#define EDGE_STEP_BITS 32
+#define FIXED_BITS     12
+
+#define fixed_center(value)                                                    \
+  ((((fixed_type)(value)) << FIXED_BITS) + (1 << (FIXED_BITS - 1)))            \
+
+#define int_to_fixed(value)                                                    \
+  (((fixed_type)(value)) << FIXED_BITS)                                        \
+
+#define fixed_to_int(value)                                                    \
+  ((value) >> FIXED_BITS)                                                      \
+
+#define fixed_to_double(value)                                                 \
+  ((value) / (double)(1 << FIXED_BITS))                                        \
+
+#define double_to_fixed(value)                                                 \
+  (fixed_type)(((value) * (double)(1 << FIXED_BITS)))                          \
+
+typedef void (setup_blocks_function_type)(psx_gpu_struct *psx_gpu);
+typedef void (texture_blocks_function_type)(psx_gpu_struct *psx_gpu);
+typedef void (shade_blocks_function_type)(psx_gpu_struct *psx_gpu);
+typedef void (blend_blocks_function_type)(psx_gpu_struct *psx_gpu);
+
+typedef void (setup_sprite_function_type)(psx_gpu_struct *psx_gpu, s32 x,
+ s32 y, s32 u, s32 v, s32 width, s32 height, u32 color);
+
+struct render_block_handler_struct
+{
+  void *setup_blocks;
+  texture_blocks_function_type *texture_blocks;
+  shade_blocks_function_type *shade_blocks;
+  blend_blocks_function_type *blend_blocks; 
+};
+
+#ifndef PANDORA_BUILD
+
+u32 fixed_reciprocal(u32 denominator, u32 *_shift)
+{
+  u32 shift = __builtin_clz(denominator);
+  u32 denominator_normalized = denominator << shift;
+
+  double numerator = (1ULL << 62) + denominator_normalized;
+  double numerator_b;
+
+  double denominator_normalized_dp_b;
+  u64 denominator_normalized_dp_u64;
+
+  u32 reciprocal;
+  double reciprocal_dp;
+
+  u64 numerator_u64 = (denominator_normalized >> 10) |
+   ((u64)(62 + 1023) << 52);
+  *((u64 *)(&numerator_b)) = numerator_u64;
+
+  denominator_normalized_dp_u64 =
+   (u64)(denominator_normalized << 21) |
+   ((u64)((denominator_normalized >> 11) + ((1022 + 31) << 20)) << 32);
+  *((u64 *)(&denominator_normalized_dp_b)) = denominator_normalized_dp_u64;
+
+  // Implement with a DP divide
+  reciprocal_dp = numerator / denominator_normalized_dp_b;
+  reciprocal = reciprocal_dp;
+
+  if(reciprocal == 0x80000001)
+    reciprocal = 0x80000000;
+
+  *_shift = 62 - shift;
+  return reciprocal;
+}
+
+double reciprocal_estimate(double a)
+{
+  int q, s;
+  double r;
+
+  q = (int)(a * 512.0);
+  /* a in units of 1/512 rounded down */
+  r = 1.0 / (((double)q + 0.5) / 512.0); /* reciprocal r */
+  s = (int)(256.0 * r + 0.5);
+
+  /* r in units of 1/256 rounded to nearest */
+  
+  return (double)s / 256.0;
+}
+
+u32 reciprocal_estimate_u32(u32 value)
+{
+  u64 dp_value_u64;
+  volatile double dp_value;
+  volatile u64 *dp_value_ptr = (volatile u64 *)&dp_value;
+
+  if((value >> 31) == 0)
+    return 0xFFFFFFFF;
+
+  dp_value_u64 = (0x3FEULL << (31 + 21)) | ((u64)(value & 0x7FFFFFFF) << 21);
+
+  *dp_value_ptr = dp_value_u64;
+
+  dp_value = reciprocal_estimate(dp_value);
+  dp_value_u64 = *dp_value_ptr;
+
+  return (0x80000000 | ((dp_value_u64 >> 21) & 0x7FFFFFFF));
+}
+
+u32 fixed_reciprocal_nr(u32 value, u32 *_shift)
+{
+  u32 shift = __builtin_clz(value);
+  u32 value_normalized = value << shift;
+
+  *_shift = 62 - shift;
+
+  value_normalized -= 2;
+
+  u32 reciprocal_normalized = reciprocal_estimate_u32(value_normalized) >> 1;
+
+  u32 temp = -(((u64)value_normalized * (u32)reciprocal_normalized) >> 31);
+  reciprocal_normalized = (((u64)reciprocal_normalized * temp) >> 31);
+  temp = -(((u64)value_normalized * (u32)reciprocal_normalized) >> 31);
+  reciprocal_normalized = (((u64)reciprocal_normalized * temp) >> 31);
+  temp = -(((u64)value_normalized * (u32)reciprocal_normalized) >> 31);
+  reciprocal_normalized = (((u64)reciprocal_normalized * temp) >> 31);
+
+  return reciprocal_normalized;
+}
+
+#endif
+
+
+s32 triangle_signed_area_x2(s32 x0, s32 y0, s32 x1, s32 y1, s32 x2, s32 y2)
+{
+	return ((x1 - x0) * (y2 - y1)) - ((x2 - x1) * (y1 - y0));
+}
+
+u32 texture_region_mask(s32 x1, s32 y1, s32 x2, s32 y2)
+{
+  s32 coverage_x, coverage_y;
+
+  u32 mask_up_left;
+  u32 mask_down_right;
+
+  coverage_x = x2 >> 6;
+  coverage_y = y2 >> 8;
+
+  if(coverage_x < 0)
+    coverage_x = 0;
+
+  if(coverage_x > 31)
+    coverage_x = 31;
+
+  mask_down_right = ~(0xFFFFFFFF << (coverage_x + 1)) & 0xFFFF;
+
+  if(coverage_y >= 1)
+    mask_down_right |= mask_down_right << 16;
+
+  coverage_x = x1 >> 6;
+
+  mask_up_left = 0xFFFF0000 << coverage_x;
+  if(coverage_x < 0)
+    mask_up_left = 0xFFFF0000;
+
+  coverage_y = y1 >> 8;
+  if(coverage_y <= 0)
+    mask_up_left |= mask_up_left >> 16;
+
+  return mask_up_left & mask_down_right;
+}
+
+u32 invalidate_texture_cache_region(psx_gpu_struct *psx_gpu, u32 x1, u32 y1,
+ u32 x2, u32 y2)
+{
+  u32 mask = texture_region_mask(x1, y1, x2, y2);
+
+  psx_gpu->dirty_textures_4bpp_mask |= mask;
+  psx_gpu->dirty_textures_8bpp_mask |= mask;
+  psx_gpu->dirty_textures_8bpp_alternate_mask |= mask;
+
+  return mask;
+}
+
+u32 invalidate_texture_cache_region_viewport(psx_gpu_struct *psx_gpu, u32 x1,
+ u32 y1, u32 x2, u32 y2)
+{
+  u32 mask = texture_region_mask(x1, y1, x2, y2) &
+   psx_gpu->viewport_mask;
+  psx_gpu->dirty_textures_4bpp_mask |= mask;
+  psx_gpu->dirty_textures_8bpp_mask |= mask;
+  psx_gpu->dirty_textures_8bpp_alternate_mask |= mask;
+
+  return mask;
+}
+
+
+void update_texture_8bpp_cache_slice(psx_gpu_struct *psx_gpu,
+ u32 texture_page);
+
+#ifndef PANDORA_BUILD
+
+void update_texture_4bpp_cache(psx_gpu_struct *psx_gpu)
+{
+  u32 current_texture_page = psx_gpu->current_texture_page;
+  u8 *texture_page_ptr = psx_gpu->texture_page_ptr;
+  u16 *vram_ptr = psx_gpu->vram_ptr;
+
+  u32 texel_block;
+  u32 tile_x, tile_y;
+  u32 sub_x, sub_y;
+
+  vram_ptr += (current_texture_page >> 4) * 256 * 1024;
+  vram_ptr += (current_texture_page & 0xF) * 64;
+
+  texture_cache_loads++;
+
+  tile_y = 16;
+  tile_x = 16;
+  sub_x = 4;
+  sub_y = 16;
+
+  psx_gpu->dirty_textures_4bpp_mask &= ~(psx_gpu->current_texture_mask);
+
+  while(tile_y)
+  {
+    while(tile_x)
+    {
+      while(sub_y)
+      {
+        while(sub_x)
+        {
+          texel_block = *vram_ptr;
+          texture_page_ptr[0] = texel_block & 0xF;
+          texture_page_ptr[1] = (texel_block >> 4) & 0xF;
+          texture_page_ptr[2] = (texel_block >> 8) & 0xF;
+          texture_page_ptr[3] = texel_block >> 12;
+          
+          vram_ptr++;
+          texture_page_ptr += 4;
+
+          sub_x--;          
+        }
+
+        vram_ptr -= 4;
+        sub_x = 4;
+
+        sub_y--;
+        vram_ptr += 1024;
+      }
+
+      sub_y = 16;
+
+      vram_ptr -= (1024 * 16) - 4;
+      tile_x--;
+    }
+
+    tile_x = 16;
+
+    vram_ptr += (16 * 1024) - (4 * 16);
+    tile_y--;
+  }
+}
+
+void update_texture_8bpp_cache_slice(psx_gpu_struct *psx_gpu,
+ u32 texture_page)
+{
+  u16 *texture_page_ptr = psx_gpu->texture_page_ptr;
+  u16 *vram_ptr = psx_gpu->vram_ptr;
+
+  u32 tile_x, tile_y;
+  u32 sub_y;
+
+  vec_8x16u texels;
+
+  texture_cache_loads++;
+
+  vram_ptr += (texture_page >> 4) * 256 * 1024;
+  vram_ptr += (texture_page & 0xF) * 64;
+
+  if((texture_page ^ psx_gpu->current_texture_page) & 0x1)
+    texture_page_ptr += (8 * 16) * 8;
+
+  tile_x = 8;
+  tile_y = 16;
+
+  sub_y = 16;
+
+  while(tile_y)
+  {
+    while(tile_x)
+    {
+      while(sub_y)
+      {
+        load_128b(texels, vram_ptr);
+        store_128b(texels, texture_page_ptr);
+
+        texture_page_ptr += 8;
+        vram_ptr += 1024;
+
+        sub_y--;
+      }
+
+      sub_y = 16;
+
+      vram_ptr -= (1024 * 16);
+      vram_ptr += 8;
+
+      tile_x--;
+    }
+
+    tile_x = 8;
+
+    vram_ptr -= (8 * 8);
+    vram_ptr += (16 * 1024);
+
+    texture_page_ptr += (8 * 16) * 8;
+    tile_y--;
+  }
+}
+
+#endif
+
+
+void update_texture_8bpp_cache(psx_gpu_struct *psx_gpu)
+{
+  u32 current_texture_page = psx_gpu->current_texture_page;
+  u32 update_textures =
+   psx_gpu->dirty_textures_8bpp_mask & psx_gpu->current_texture_mask;
+
+  psx_gpu->dirty_textures_8bpp_mask &= ~update_textures;
+
+  if(update_textures & (1 << current_texture_page))
+  {
+    update_texture_8bpp_cache_slice(psx_gpu, current_texture_page);
+    update_textures &= ~(1 << current_texture_page);
+  }
+
+  if(update_textures)
+  {
+    u32 adjacent_texture_page = ((current_texture_page + 1) & 0xF) |
+     (current_texture_page & 0x10);
+
+    update_texture_8bpp_cache_slice(psx_gpu, adjacent_texture_page);
+  }
+}
+
+void setup_blocks_shaded_untextured_undithered_unswizzled_indirect(
+ psx_gpu_struct *psx_gpu);
+
+void flush_render_block_buffer(psx_gpu_struct *psx_gpu)
+{
+  if(psx_gpu->num_blocks)
+  {
+    render_block_handler_struct *render_block_handler =
+     psx_gpu->render_block_handler;
+
+    render_block_handler->texture_blocks(psx_gpu);
+    render_block_handler->shade_blocks(psx_gpu);
+    render_block_handler->blend_blocks(psx_gpu);
+
+    span_pixel_blocks += psx_gpu->num_blocks;
+    render_buffer_flushes++;
+
+    psx_gpu->num_blocks = 0;
+  }
+}
+
+
+void compute_all_gradients(psx_gpu_struct *psx_gpu, vertex_struct *a,
+ vertex_struct *b, vertex_struct *c);
+
+#ifndef PANDORA_BUILD
+
+#define setup_gradient_calculation_input(set, vertex)                          \
+  /* First type is:  uvrg bxxx xxxx                                          */\
+  /* Second type is: yyyy ybyy uvrg                                          */\
+  /* Since x_a and y_c are the same the same variable is used for both.      */\
+  x##set##_a_y##set##_c.e[0] = vertex->u;                                      \
+  x##set##_a_y##set##_c.e[1] = vertex->v;                                      \
+  x##set##_a_y##set##_c.e[2] = vertex->r;                                      \
+  x##set##_a_y##set##_c.e[3] = vertex->g;                                      \
+  dup_4x16b(x##set##_b, vertex->x);                                            \
+  dup_4x16b(x##set##_c, vertex->x);                                            \
+  dup_4x16b(y##set##_a, vertex->y);                                            \
+  dup_4x16b(y##set##_b, vertex->y);                                            \
+  x##set##_b.e[0] = vertex->b;                                                 \
+  y##set##_b.e[1] = vertex->b                                                  \
+  
+
+void compute_all_gradients(psx_gpu_struct *psx_gpu, vertex_struct *a,
+ vertex_struct *b, vertex_struct *c)
+{
+  u32 triangle_area = psx_gpu->triangle_area;
+  u32 winding_mask_scalar;
+
+  u32 triangle_area_shift;
+  u64 triangle_area_reciprocal =
+   fixed_reciprocal(triangle_area, &triangle_area_shift);
+  triangle_area_shift = -(triangle_area_shift - FIXED_BITS);
+
+  // ((x1 - x0) * (y2 - y1)) - ((x2 - x1) * (y1 - y0)) =
+  // ( d0       *  d1      ) - ( d2       *  d3      ) =
+  // ( m0                  ) - ( m1                  ) = gradient
+
+  // This is split to do 12 elements at a time over three sets: a, b, and c.
+  // Technically we only need to do 10 elements (uvrgb_x and uvrgb_y), so
+  // two of the slots are unused.
+
+  // Inputs are all 16-bit signed. The m0/m1 results are 32-bit signed, as
+  // is g.
+
+  vec_4x16s x0_a_y0_c, x0_b, x0_c;
+  vec_4x16s y0_a, y0_b;
+  vec_4x16s x1_a_y1_c, x1_b, x1_c;
+  vec_4x16s y1_a, y1_b;
+  vec_4x16s x2_a_y2_c, x2_b, x2_c;
+  vec_4x16s y2_a, y2_b;
+
+  vec_4x32u uvrg_base;
+  vec_4x32u b_base;
+  vec_4x32u const_0x8000;
+
+  vec_4x16s d0_a_d3_c, d0_b, d0_c;
+  vec_4x16s d1_a, d1_b, d1_c_d2_a;
+  vec_4x16s d2_b, d2_c;
+  vec_4x16s d3_a, d3_b;
+
+  vec_4x32s m0_a, m0_b, m0_c;
+  vec_4x32s m1_a, m1_b, m1_c;
+
+  vec_4x32u gradient_area_a, gradient_area_c;
+  vec_2x32u gradient_area_b;  
+
+  vec_4x32u gradient_area_sign_a, gradient_area_sign_c;
+  vec_2x32u gradient_area_sign_b;
+  vec_4x32u winding_mask;
+
+  vec_2x64u gradient_wide_a0, gradient_wide_a1;
+  vec_2x64u gradient_wide_c0, gradient_wide_c1;
+  vec_2x64u gradient_wide_b;
+
+  vec_4x32u gradient_a, gradient_c;
+  vec_2x32u gradient_b;
+  vec_16x8s gradient_shift;
+
+  setup_gradient_calculation_input(0, a);
+  setup_gradient_calculation_input(1, b);
+  setup_gradient_calculation_input(2, c);
+
+  dup_4x32b(const_0x8000, 0x8000);
+  shl_long_4x16b(uvrg_base, x0_a_y0_c, 16);
+  shl_long_4x16b(b_base, x0_b, 16);
+
+  add_4x32b(uvrg_base, uvrg_base, const_0x8000);
+  add_4x32b(b_base, b_base, const_0x8000);
+
+  // Can probably pair these, but it'll require careful register allocation
+  sub_4x16b(d0_a_d3_c, x1_a_y1_c, x0_a_y0_c);
+  sub_4x16b(d1_c_d2_a, x2_a_y2_c, x1_a_y1_c);
+
+  sub_4x16b(d0_b, x1_b, x0_b);
+  sub_4x16b(d0_c, x1_c, x0_c);
+
+  sub_4x16b(d1_a, y2_a, y1_a);
+  sub_4x16b(d1_b, y2_b, y1_b);
+
+  sub_4x16b(d2_b, x2_b, x1_b);
+  sub_4x16b(d2_c, x2_c, x1_c);
+
+  sub_4x16b(d3_a, y1_a, y0_a);
+  sub_4x16b(d3_b, y1_b, y0_b);
+
+  mul_long_4x16b(m0_a, d0_a_d3_c, d1_a);
+  mul_long_4x16b(m0_b, d0_b, d1_b);
+  mul_long_4x16b(m0_c, d0_c, d1_c_d2_a);
+
+  mul_long_4x16b(m1_a, d1_c_d2_a, d3_a);
+  mul_long_4x16b(m1_b, d2_b, d3_b);
+  mul_long_4x16b(m1_c, d2_c, d0_a_d3_c);
+
+  sub_4x32b(gradient_area_a, m0_a, m1_a);
+  sub_2x32b(gradient_area_b, m0_b.low, m1_b.low);
+  sub_4x32b(gradient_area_c, m0_c, m1_c);
+
+  cmpltz_4x32b(gradient_area_sign_a, gradient_area_a);
+  cmpltz_2x32b(gradient_area_sign_b, gradient_area_b);
+  cmpltz_4x32b(gradient_area_sign_c, gradient_area_c);
+
+  abs_4x32b(gradient_area_a, gradient_area_a);
+  abs_2x32b(gradient_area_b, gradient_area_b);
+  abs_4x32b(gradient_area_c, gradient_area_c);
+
+  winding_mask_scalar = -psx_gpu->triangle_winding;
+
+  dup_4x32b(winding_mask, winding_mask_scalar);
+  eor_4x32b(gradient_area_sign_a, gradient_area_sign_a, winding_mask);
+  eor_2x32b(gradient_area_sign_b, gradient_area_sign_b, winding_mask);
+  eor_4x32b(gradient_area_sign_c, gradient_area_sign_c, winding_mask);
+
+  mul_scalar_long_2x32b(gradient_wide_a0, 
+   vector_cast(vec_2x32s, gradient_area_a.low), 
+   (s64)triangle_area_reciprocal);
+  mul_scalar_long_2x32b(gradient_wide_a1,
+   vector_cast(vec_2x32s, gradient_area_a.high),
+   (s64)triangle_area_reciprocal);
+  mul_scalar_long_2x32b(gradient_wide_b, 
+   vector_cast(vec_2x32s, gradient_area_b),
+   (s64)triangle_area_reciprocal);
+  mul_scalar_long_2x32b(gradient_wide_c0, 
+   vector_cast(vec_2x32s, gradient_area_c.low),
+   (s64)triangle_area_reciprocal);
+  mul_scalar_long_2x32b(gradient_wide_c1, 
+   vector_cast(vec_2x32s, gradient_area_c.high),
+   (s64)triangle_area_reciprocal);
+
+  dup_16x8b(gradient_shift, triangle_area_shift);
+  shl_reg_2x64b(gradient_wide_a0, gradient_wide_a0,
+   vector_cast(vec_2x64u, gradient_shift));
+  shl_reg_2x64b(gradient_wide_a1, gradient_wide_a1,
+   vector_cast(vec_2x64u, gradient_shift));
+  shl_reg_2x64b(gradient_wide_b, gradient_wide_b,
+   vector_cast(vec_2x64u, gradient_shift));
+  shl_reg_2x64b(gradient_wide_c0, gradient_wide_c0,
+   vector_cast(vec_2x64u, gradient_shift));
+  shl_reg_2x64b(gradient_wide_c1, gradient_wide_c1,
+   vector_cast(vec_2x64u, gradient_shift));
+
+  mov_narrow_2x64b(gradient_a.low, gradient_wide_a0);
+  mov_narrow_2x64b(gradient_a.high, gradient_wide_a1);
+  mov_narrow_2x64b(gradient_b, gradient_wide_b);
+  mov_narrow_2x64b(gradient_c.low, gradient_wide_c0);
+  mov_narrow_2x64b(gradient_c.high, gradient_wide_c1);
+
+  shl_4x32b(gradient_a, gradient_a, 4);
+  shl_2x32b(gradient_b, gradient_b, 4);
+  shl_4x32b(gradient_c, gradient_c, 4);
+
+  eor_4x32b(gradient_a, gradient_a, gradient_area_sign_a);
+  eor_2x32b(gradient_b, gradient_b, gradient_area_sign_b);
+  eor_4x32b(gradient_c, gradient_c, gradient_area_sign_c);
+
+  sub_4x32b(gradient_a, gradient_a, gradient_area_sign_a);
+  sub_2x32b(gradient_b, gradient_b, gradient_area_sign_b);
+  sub_4x32b(gradient_c, gradient_c, gradient_area_sign_c);
+
+  u32 left_adjust = a->x;
+  mls_scalar_4x32b(uvrg_base, gradient_a, left_adjust);
+  mls_scalar_2x32b(b_base.low, gradient_b, left_adjust);
+
+  vec_4x32u uvrg_dx2;
+  vec_2x32u b_dx2;
+
+  vec_4x32u uvrg_dx3;
+  vec_2x32u b_dx3;
+
+  vec_4x32u zero;
+
+  eor_4x32b(zero, zero, zero);
+  add_4x32b(uvrg_dx2, gradient_a, gradient_a);
+  add_2x32b(b_dx2, gradient_b, gradient_b);
+  add_4x32b(uvrg_dx3, gradient_a, uvrg_dx2);
+  add_2x32b(b_dx3, gradient_b, b_dx2);
+
+  // Can be done with vst4, assuming that the zero, dx, dx2, and dx3 are
+  // lined up properly
+  psx_gpu->u_block_span.e[0] = zero.e[0];
+  psx_gpu->u_block_span.e[1] = gradient_a.e[0];
+  psx_gpu->u_block_span.e[2] = uvrg_dx2.e[0];
+  psx_gpu->u_block_span.e[3] = uvrg_dx3.e[0];
+
+  psx_gpu->v_block_span.e[0] = zero.e[1];
+  psx_gpu->v_block_span.e[1] = gradient_a.e[1];
+  psx_gpu->v_block_span.e[2] = uvrg_dx2.e[1];
+  psx_gpu->v_block_span.e[3] = uvrg_dx3.e[1];
+
+  psx_gpu->r_block_span.e[0] = zero.e[2];
+  psx_gpu->r_block_span.e[1] = gradient_a.e[2];
+  psx_gpu->r_block_span.e[2] = uvrg_dx2.e[2];
+  psx_gpu->r_block_span.e[3] = uvrg_dx3.e[2];
+
+  psx_gpu->g_block_span.e[0] = zero.e[3];
+  psx_gpu->g_block_span.e[1] = gradient_a.e[3];
+  psx_gpu->g_block_span.e[2] = uvrg_dx2.e[3];
+  psx_gpu->g_block_span.e[3] = uvrg_dx3.e[3];
+
+  psx_gpu->b_block_span.e[0] = zero.e[0];
+  psx_gpu->b_block_span.e[1] = gradient_b.e[0];
+  psx_gpu->b_block_span.e[2] = b_dx2.e[0];
+  psx_gpu->b_block_span.e[3] = b_dx3.e[0];
+
+  psx_gpu->uvrg = uvrg_base;
+  psx_gpu->b = b_base.e[0];
+
+  psx_gpu->uvrg_dx = gradient_a;
+  psx_gpu->uvrg_dy = gradient_c;
+  psx_gpu->b_dy = gradient_b.e[1];
+}
+#endif
+
+#define vector_check(_a, _b)                                                   \
+  if(memcmp(&_a, &_b, sizeof(_b)))                                             \
+  {                                                                            \
+    if(sizeof(_b) == 8)                                                        \
+    {                                                                          \
+      printf("mismatch on %s vs %s: (%x %x) vs (%x %x)\n",                     \
+       #_a, #_b, _a.e[0], _a.e[1], _b.e[0], _b.e[1]);                          \
+    }                                                                          \
+    else                                                                       \
+    {                                                                          \
+      printf("mismatch on %s vs %s: (%x %x %x %x) vs (%x %x %x %x)\n",         \
+       #_a, #_b, _a.e[0], _a.e[1], _a.e[2], _a.e[3], _b.e[0], _b.e[1],         \
+       _b.e[2], _b.e[3]);                                                      \
+    }                                                                          \
+  }                                                                            \
+
+#define scalar_check(_a, _b)                                                   \
+  if(_a != _b)                                                                 \
+    printf("mismatch on %s %s: %x vs %x\n", #_a, #_b, _a, _b)                  \
+
+
+#define setup_spans_prologue_alternate_yes()                                   \
+  vec_2x64s alternate_x;                                                       \
+  vec_2x64s alternate_dx_dy;                                                   \
+  vec_4x32s alternate_x_32;                                                    \
+  vec_2x32s alternate_x_16;                                                    \
+                                                                               \
+  vec_4x16u alternate_select;                                                  \
+  vec_4x16s y_mid_point;                                                       \
+                                                                               \
+  s32 y_b = v_b->y;                                                            \
+  s64 edge_alt;                                                                \
+  s32 edge_dx_dy_alt;                                                          \
+  u32 edge_shift_alt                                                           \
+
+#define setup_spans_prologue_alternate_no()                                    \
+
+#define setup_spans_prologue(alternate_active)                                 \
+  edge_data_struct *span_edge_data;                                            \
+  vec_4x32u *span_uvrg_offset;                                                 \
+  u32 *span_b_offset;                                                          \
+                                                                               \
+  s32 clip;                                                                    \
+                                                                               \
+  vec_2x64s edges_xy;                                                          \
+  vec_2x32s edges_dx_dy;                                                       \
+  vec_2x32u edge_shifts;                                                       \
+                                                                               \
+  vec_2x64s left_x, right_x;                                                   \
+  vec_2x64s left_dx_dy, right_dx_dy;                                           \
+  vec_4x32s left_x_32, right_x_32;                                             \
+  vec_8x16s left_right_x_16;                                                   \
+  vec_4x16s y_x4;                                                              \
+  vec_8x16s left_edge;                                                         \
+  vec_8x16s right_edge;                                                        \
+  vec_4x16u span_shift;                                                        \
+                                                                               \
+  vec_2x32u c_0x01;                                                            \
+  vec_4x16u c_0x04;                                                            \
+  vec_4x16u c_0xFFFE;                                                          \
+  vec_4x16u c_0x07;                                                            \
+                                                                               \
+  vec_2x32s x_starts;                                                          \
+  vec_2x32s x_ends;                                                            \
+                                                                               \
+  s32 x_a = v_a->x;                                                            \
+  s32 x_b = v_b->x;                                                            \
+  s32 x_c = v_c->x;                                                            \
+  s32 y_a = v_a->y;                                                            \
+  s32 y_c = v_c->y;                                                            \
+                                                                               \
+  vec_4x32u uvrg = psx_gpu->uvrg;                                              \
+  vec_4x32u uvrg_dy = psx_gpu->uvrg_dy;                                        \
+  u32 b = psx_gpu->b;                                                          \
+  u32 b_dy = psx_gpu->b_dy;                                                    \
+                                                                               \
+  dup_2x32b(c_0x01, 0x01);                                                     \
+  setup_spans_prologue_alternate_##alternate_active()                          \
+
+#define setup_spans_prologue_b()                                               \
+  span_edge_data = psx_gpu->span_edge_data;                                    \
+  span_uvrg_offset = psx_gpu->span_uvrg_offset;                                \
+  span_b_offset = psx_gpu->span_b_offset;                                      \
+                                                                               \
+  vec_8x16u c_0x0001;                                                          \
+                                                                               \
+  dup_8x16b(c_0x0001, 0x0001);                                                 \
+  dup_8x16b(left_edge, psx_gpu->viewport_start_x);                             \
+  dup_8x16b(right_edge, psx_gpu->viewport_end_x);                              \
+  add_8x16b(right_edge, right_edge, c_0x0001);                                 \
+  dup_4x16b(c_0x04, 0x04);                                                     \
+  dup_4x16b(c_0x07, 0x07);                                                     \
+  dup_4x16b(c_0xFFFE, 0xFFFE);                                                 \
+
+
+#define compute_edge_delta_x2()                                                \
+{                                                                              \
+  vec_2x32s heights;                                                           \
+  vec_2x32s height_reciprocals;                                                \
+  vec_2x32s heights_b;                                                         \
+  vec_4x32u widths;                                                            \
+                                                                               \
+  u32 edge_shift = reciprocal_table[height];                                   \
+                                                                               \
+  dup_2x32b(heights, height);                                                  \
+  sub_2x32b(widths, x_ends, x_starts);                                         \
+                                                                               \
+  dup_2x32b(edge_shifts, edge_shift);                                          \
+  sub_2x32b(heights_b, heights, c_0x01);                                       \
+  shr_2x32b(height_reciprocals, edge_shifts, 12);                              \
+                                                                               \
+  mla_2x32b(heights_b, x_starts, heights);                                     \
+  bic_immediate_4x16b(vector_cast(vec_4x16u, edge_shifts), 0xE0);              \
+  mul_2x32b(edges_dx_dy, widths, height_reciprocals);                          \
+  mul_long_2x32b(edges_xy, heights_b, height_reciprocals);                     \
+}                                                                              \
+
+#define compute_edge_delta_x3(start_c, height_a, height_b)                     \
+{                                                                              \
+  vec_2x32s heights;                                                           \
+  vec_2x32s height_reciprocals;                                                \
+  vec_2x32s heights_b;                                                         \
+  vec_2x32u widths;                                                            \
+                                                                               \
+  u32 width_alt;                                                               \
+  s32 height_b_alt;                                                            \
+  u32 height_reciprocal_alt;                                                   \
+                                                                               \
+  heights.e[0] = height_a;                                                     \
+  heights.e[1] = height_b;                                                     \
+                                                                               \
+  edge_shifts.e[0] = reciprocal_table[height_a];                               \
+  edge_shifts.e[1] = reciprocal_table[height_b];                               \
+  edge_shift_alt = reciprocal_table[height_minor_b];                           \
+                                                                               \
+  sub_2x32b(widths, x_ends, x_starts);                                         \
+  width_alt = x_c - start_c;                                                   \
+                                                                               \
+  shr_2x32b(height_reciprocals, edge_shifts, 12);                              \
+  height_reciprocal_alt = edge_shift_alt >> 12;                                \
+                                                                               \
+  bic_immediate_4x16b(vector_cast(vec_4x16u, edge_shifts), 0xE0);              \
+  edge_shift_alt &= 0x1F;                                                      \
+                                                                               \
+  sub_2x32b(heights_b, heights, c_0x01);                                       \
+  height_b_alt = height_minor_b - 1;                                           \
+                                                                               \
+  mla_2x32b(heights_b, x_starts, heights);                                     \
+  height_b_alt += height_minor_b * start_c;                                    \
+                                                                               \
+  mul_long_2x32b(edges_xy, heights_b, height_reciprocals);                     \
+  edge_alt = (s64)height_b_alt * height_reciprocal_alt;                        \
+                                                                               \
+  mul_2x32b(edges_dx_dy, widths, height_reciprocals);                          \
+  edge_dx_dy_alt = width_alt * height_reciprocal_alt;                          \
+}                                                                              \
+
+
+#define setup_spans_adjust_y_up()                                              \
+  sub_4x32b(y_x4, y_x4, c_0x04)                                                \
+
+#define setup_spans_adjust_y_down()                                            \
+  add_4x32b(y_x4, y_x4, c_0x04)                                                \
+
+#define setup_spans_adjust_interpolants_up()                                   \
+  sub_4x32b(uvrg, uvrg, uvrg_dy);                                              \
+  b -= b_dy                                                                    \
+
+#define setup_spans_adjust_interpolants_down()                                 \
+  add_4x32b(uvrg, uvrg, uvrg_dy);                                              \
+  b += b_dy                                                                    \
+
+
+#define setup_spans_clip_interpolants_increment()                              \
+  mla_scalar_4x32b(uvrg, uvrg_dy, clip);                                       \
+  b += b_dy * clip                                                             \
+
+#define setup_spans_clip_interpolants_decrement()                              \
+  mls_scalar_4x32b(uvrg, uvrg_dy, clip);                                       \
+  b -= b_dy * clip                                                             \
+
+#define setup_spans_clip_alternate_yes()                                       \
+  edge_alt += edge_dx_dy_alt * (s64)(clip)                                     \
+
+#define setup_spans_clip_alternate_no()                                        \
+
+#define setup_spans_clip(direction, alternate_active)                          \
+{                                                                              \
+  clipped_triangles++;                                                         \
+  mla_scalar_long_2x32b(edges_xy, edges_dx_dy, (s64)clip);                     \
+  setup_spans_clip_alternate_##alternate_active();                             \
+  setup_spans_clip_interpolants_##direction();                                 \
+}                                                                              \
+
+
+#define setup_spans_adjust_edges_alternate_no(left_index, right_index)         \
+{                                                                              \
+  vec_2x64u edge_shifts_64;                                                    \
+  vec_2x64s edges_dx_dy_64;                                                    \
+                                                                               \
+  mov_wide_2x32b(edge_shifts_64, edge_shifts);                                 \
+  shl_variable_2x64b(edges_xy, edges_xy, edge_shifts_64);                      \
+                                                                               \
+  mov_wide_2x32b(edges_dx_dy_64, edges_dx_dy);                                 \
+  shl_variable_2x64b(edges_dx_dy_64, edges_dx_dy_64, edge_shifts_64);          \
+                                                                               \
+  left_x.e[0] = edges_xy.e[left_index];                                        \
+  right_x.e[0] = edges_xy.e[right_index];                                      \
+                                                                               \
+  left_dx_dy.e[0] = edges_dx_dy_64.e[left_index];                              \
+  left_dx_dy.e[1] = edges_dx_dy_64.e[left_index];                              \
+  right_dx_dy.e[0] = edges_dx_dy_64.e[right_index];                            \
+  right_dx_dy.e[1] = edges_dx_dy_64.e[right_index];                            \
+                                                                               \
+  add_1x64b(left_x.high, left_x.low, left_dx_dy.low);                          \
+  add_1x64b(right_x.high, right_x.low, right_dx_dy.low);                       \
+                                                                               \
+  add_2x64b(left_dx_dy, left_dx_dy, left_dx_dy);                               \
+  add_2x64b(right_dx_dy, right_dx_dy, right_dx_dy);                            \
+}                                                                              \
+
+#define setup_spans_adjust_edges_alternate_yes(left_index, right_index)        \
+{                                                                              \
+  setup_spans_adjust_edges_alternate_no(left_index, right_index);              \
+  s64 edge_dx_dy_alt_64;                                                       \
+                                                                               \
+  dup_4x16b(y_mid_point, y_b);                                                 \
+                                                                               \
+  edge_alt <<= edge_shift_alt;                                                 \
+  edge_dx_dy_alt_64 = (s64)edge_dx_dy_alt << edge_shift_alt;                   \
+                                                                               \
+  alternate_x.e[0] = edge_alt;                                                 \
+  alternate_dx_dy.e[0] = edge_dx_dy_alt_64;                                    \
+  alternate_dx_dy.e[1] = edge_dx_dy_alt_64;                                    \
+                                                                               \
+  add_1x64b(alternate_x.high, alternate_x.low, alternate_dx_dy.low);           \
+  add_2x64b(alternate_dx_dy, alternate_dx_dy, alternate_dx_dy);                \
+}                                                                              \
+
+
+#define setup_spans_y_select_up()                                              \
+  cmplt_4x16b(alternate_select, y_x4, y_mid_point)                             \
+
+#define setup_spans_y_select_down()                                            \
+  cmpgt_4x16b(alternate_select, y_x4, y_mid_point)                             \
+
+#define setup_spans_y_select_alternate_yes(direction)                          \
+  setup_spans_y_select_##direction()                                           \
+
+#define setup_spans_y_select_alternate_no(direction)                           \
+
+#define setup_spans_alternate_select_left()                                    \
+  bit_4x16b(left_right_x_16.low, alternate_x_16, alternate_select)             \
+
+#define setup_spans_alternate_select_right()                                   \
+  bit_4x16b(left_right_x_16.high, alternate_x_16, alternate_select)            \
+
+#define setup_spans_alternate_select_none()                                    \
+
+#define setup_spans_increment_alternate_yes()                                  \
+  shr_narrow_2x64b(alternate_x_32.low, alternate_x, 32);                       \
+  add_2x64b(alternate_x, alternate_x, alternate_dx_dy);                        \
+  shr_narrow_2x64b(alternate_x_32.high, alternate_x, 32);                      \
+  add_2x64b(alternate_x, alternate_x, alternate_dx_dy);                        \
+  mov_narrow_4x32b(alternate_x_16, alternate_x_32)                             \
+
+#define setup_spans_increment_alternate_no()                                   \
+
+#define setup_spans_set_x4(alternate, direction, alternate_active)             \
+{                                                                              \
+  span_uvrg_offset[0] = uvrg;                                                  \
+  span_b_offset[0] = b;                                                        \
+  setup_spans_adjust_interpolants_##direction();                               \
+                                                                               \
+  span_uvrg_offset[1] = uvrg;                                                  \
+  span_b_offset[1] = b;                                                        \
+  setup_spans_adjust_interpolants_##direction();                               \
+                                                                               \
+  span_uvrg_offset[2] = uvrg;                                                  \
+  span_b_offset[2] = b;                                                        \
+  setup_spans_adjust_interpolants_##direction();                               \
+                                                                               \
+  span_uvrg_offset[3] = uvrg;                                                  \
+  span_b_offset[3] = b;                                                        \
+  setup_spans_adjust_interpolants_##direction();                               \
+                                                                               \
+  span_uvrg_offset += 4;                                                       \
+  span_b_offset += 4;                                                          \
+                                                                               \
+  shr_narrow_2x64b(left_x_32.low, left_x, 32);                                 \
+  shr_narrow_2x64b(right_x_32.low, right_x, 32);                               \
+                                                                               \
+  add_2x64b(left_x, left_x, left_dx_dy);                                       \
+  add_2x64b(right_x, right_x, right_dx_dy);                                    \
+                                                                               \
+  shr_narrow_2x64b(left_x_32.high, left_x, 32);                                \
+  shr_narrow_2x64b(right_x_32.high, right_x, 32);                              \
+                                                                               \
+  add_2x64b(left_x, left_x, left_dx_dy);                                       \
+  add_2x64b(right_x, right_x, right_dx_dy);                                    \
+                                                                               \
+  mov_narrow_4x32b(left_right_x_16.low, left_x_32);                            \
+  mov_narrow_4x32b(left_right_x_16.high, right_x_32);                          \
+                                                                               \
+  setup_spans_increment_alternate_##alternate_active();                        \
+  setup_spans_y_select_alternate_##alternate_active(direction);                \
+  setup_spans_alternate_select_##alternate();                                  \
+                                                                               \
+  max_8x16b(left_right_x_16, left_right_x_16, left_edge);                      \
+  min_8x16b(left_right_x_16, left_right_x_16, right_edge);                     \
+                                                                               \
+  sub_4x16b(left_right_x_16.high, left_right_x_16.high, left_right_x_16.low);  \
+  add_4x16b(left_right_x_16.high, left_right_x_16.high, c_0x07);               \
+  and_4x16b(span_shift, left_right_x_16.high, c_0x07);                         \
+  shl_variable_4x16b(span_shift, c_0xFFFE, span_shift);                        \
+  shr_4x16b(left_right_x_16.high, left_right_x_16.high, 3);                    \
+                                                                               \
+  u32 i;                                                                       \
+  for(i = 0; i < 4; i++)                                                       \
+  {                                                                            \
+    span_edge_data[i].left_x = left_right_x_16.low.e[i];                       \
+    span_edge_data[i].num_blocks = left_right_x_16.high.e[i];                  \
+    span_edge_data[i].right_mask = span_shift.e[i];                            \
+    span_edge_data[i].y = y_x4.e[i];                                           \
+  }                                                                            \
+                                                                               \
+  span_edge_data += 4;                                                         \
+                                                                               \
+  setup_spans_adjust_y_##direction();                                          \
+}                                                                              \
+
+
+#define setup_spans_alternate_adjust_yes()                                     \
+  edge_alt -= edge_dx_dy_alt * (s64)height_minor_a                             \
+
+#define setup_spans_alternate_adjust_no()                                      \
+
+
+#define setup_spans_down(left_index, right_index, alternate, alternate_active) \
+  setup_spans_alternate_adjust_##alternate_active();                           \
+  if(y_c > psx_gpu->viewport_end_y)                                            \
+    height -= y_c - psx_gpu->viewport_end_y - 1;                               \
+                                                                               \
+  clip = psx_gpu->viewport_start_y - y_a;                                      \
+  if(clip > 0)                                                                 \
+  {                                                                            \
+    height -= clip;                                                            \
+    y_a += clip;                                                               \
+    setup_spans_clip(increment, alternate_active);                             \
+  }                                                                            \
+                                                                               \
+  setup_spans_prologue_b();                                                    \
+                                                                               \
+  if(height > 0)                                                               \
+  {                                                                            \
+    y_x4.e[0] = y_a;                                                           \
+    y_x4.e[1] = y_a + 1;                                                       \
+    y_x4.e[2] = y_a + 2;                                                       \
+    y_x4.e[3] = y_a + 3;                                                       \
+    setup_spans_adjust_edges_alternate_##alternate_active(left_index,          \
+     right_index);                                                             \
+                                                                               \
+    psx_gpu->num_spans = height;                                               \
+    do                                                                         \
+    {                                                                          \
+      setup_spans_set_x4(alternate, down, alternate_active);                   \
+      height -= 4;                                                             \
+    } while(height > 0);                                                       \
+  }                                                                            \
+
+
+#define setup_spans_alternate_pre_increment_yes()                              \
+  edge_alt += edge_dx_dy_alt                                                   \
+
+#define setup_spans_alternate_pre_increment_no()                               \
+
+#define setup_spans_up_decrement_height_yes()                                  \
+  height--                                                                     \
+
+#define setup_spans_up_decrement_height_no()                                   \
+  {}                                                                           \
+
+#define setup_spans_up(left_index, right_index, alternate, alternate_active)   \
+  setup_spans_alternate_adjust_##alternate_active();                           \
+  y_a--;                                                                       \
+                                                                               \
+  if(y_c < psx_gpu->viewport_start_y)                                          \
+    height -= psx_gpu->viewport_start_y - y_c;                                 \
+  else                                                                         \
+    setup_spans_up_decrement_height_##alternate_active();                      \
+                                                                               \
+  clip = y_a - psx_gpu->viewport_end_y;                                        \
+  if(clip > 0)                                                                 \
+  {                                                                            \
+    height -= clip;                                                            \
+    y_a -= clip;                                                               \
+    setup_spans_clip(decrement, alternate_active);                             \
+  }                                                                            \
+                                                                               \
+  setup_spans_prologue_b();                                                    \
+                                                                               \
+  if(height > 0)                                                               \
+  {                                                                            \
+    y_x4.e[0] = y_a;                                                           \
+    y_x4.e[1] = y_a - 1;                                                       \
+    y_x4.e[2] = y_a - 2;                                                       \
+    y_x4.e[3] = y_a - 3;                                                       \
+    add_wide_2x32b(edges_xy, edges_xy, edges_dx_dy);                           \
+    setup_spans_alternate_pre_increment_##alternate_active();                  \
+    setup_spans_adjust_edges_alternate_##alternate_active(left_index,          \
+     right_index);                                                             \
+    setup_spans_adjust_interpolants_up();                                      \
+                                                                               \
+    psx_gpu->num_spans = height;                                               \
+    while(height > 0)                                                          \
+    {                                                                          \
+      setup_spans_set_x4(alternate, up, alternate_active);                     \
+      height -= 4;                                                             \
+    }                                                                          \
+  }                                                                            \
+
+#define index_left  0
+#define index_right 1
+
+#define setup_spans_up_up(minor, major)                                        \
+  setup_spans_prologue(yes);                                                   \
+  s32 height_minor_a = y_a - y_b;                                              \
+  s32 height_minor_b = y_b - y_c;                                              \
+  s32 height = y_a - y_c;                                                      \
+                                                                               \
+  dup_2x32b(x_starts, x_a);                                                    \
+  x_ends.e[0] = x_c;                                                           \
+  x_ends.e[1] = x_b;                                                           \
+                                                                               \
+  compute_edge_delta_x3(x_b, height, height_minor_a);                          \
+  setup_spans_up(index_##major, index_##minor, minor, yes)                     \
+
+
+void setup_spans_up_left(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
+ vertex_struct *v_b, vertex_struct *v_c);
+void setup_spans_up_right(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
+ vertex_struct *v_b, vertex_struct *v_c);
+void setup_spans_down_left(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
+ vertex_struct *v_b, vertex_struct *v_c);
+void setup_spans_down_right(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
+ vertex_struct *v_b, vertex_struct *v_c);
+void setup_spans_up_a(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
+ vertex_struct *v_b, vertex_struct *v_c);
+void setup_spans_up_b(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
+ vertex_struct *v_b, vertex_struct *v_c);
+void setup_spans_down_a(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
+ vertex_struct *v_b, vertex_struct *v_c);
+void setup_spans_down_b(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
+ vertex_struct *v_b, vertex_struct *v_c);
+void setup_spans_up_down(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
+ vertex_struct *v_b, vertex_struct *v_c);
+
+
+#ifndef PANDORA_BUILD
+
+void setup_spans_up_left(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
+ vertex_struct *v_b, vertex_struct *v_c)
+{
+  setup_spans_up_up(left, right);
+}
+
+void setup_spans_up_right(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
+ vertex_struct *v_b, vertex_struct *v_c)
+{
+  setup_spans_up_up(right, left);
+}
+
+#define setup_spans_down_down(minor, major)                                    \
+  setup_spans_prologue(yes);                                                   \
+  s32 height_minor_a = y_b - y_a;                                              \
+  s32 height_minor_b = y_c - y_b;                                              \
+  s32 height = y_c - y_a;                                                      \
+                                                                               \
+  dup_2x32b(x_starts, x_a);                                                    \
+  x_ends.e[0] = x_c;                                                           \
+  x_ends.e[1] = x_b;                                                           \
+                                                                               \
+  compute_edge_delta_x3(x_b, height, height_minor_a);                          \
+  setup_spans_down(index_##major, index_##minor, minor, yes)                   \
+
+void setup_spans_down_left(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
+ vertex_struct *v_b, vertex_struct *v_c)
+{
+  setup_spans_down_down(left, right);
+}
+
+void setup_spans_down_right(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
+ vertex_struct *v_b, vertex_struct *v_c)
+{
+  setup_spans_down_down(right, left);
+}
+
+#define setup_spans_up_flat()                                                  \
+  s32 height = y_a - y_c;                                                      \
+                                                                               \
+  flat_triangles++;                                                            \
+  compute_edge_delta_x2();                                                     \
+  setup_spans_up(index_left, index_right, none, no)                            \
+
+void setup_spans_up_a(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
+ vertex_struct *v_b, vertex_struct *v_c)
+{
+  setup_spans_prologue(no);
+  x_starts.e[0] = x_a;
+  x_starts.e[1] = x_b;
+  dup_2x32b(x_ends, x_c);
+
+  setup_spans_up_flat();
+}
+
+void setup_spans_up_b(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
+ vertex_struct *v_b, vertex_struct *v_c)
+{
+  setup_spans_prologue(no);
+  dup_2x32b(x_starts, x_a);
+  x_ends.e[0] = x_b;
+  x_ends.e[1] = x_c;
+
+  setup_spans_up_flat();
+}
+
+#define setup_spans_down_flat()                                                \
+  s32 height = y_c - y_a;                                                      \
+                                                                               \
+  flat_triangles++;                                                            \
+  compute_edge_delta_x2();                                                     \
+  setup_spans_down(index_left, index_right, none, no)                          \
+
+void setup_spans_down_a(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
+ vertex_struct *v_b, vertex_struct *v_c)
+{
+  setup_spans_prologue(no);
+  x_starts.e[0] = x_a;
+  x_starts.e[1] = x_b;
+  dup_2x32b(x_ends, x_c);
+
+  setup_spans_down_flat();
+}
+
+void setup_spans_down_b(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
+ vertex_struct *v_b, vertex_struct *v_c)
+{
+  setup_spans_prologue(no);
+  dup_2x32b(x_starts, x_a);
+  x_ends.e[0] = x_b;
+  x_ends.e[1] = x_c;
+
+  setup_spans_down_flat();
+}
+
+void setup_spans_up_down(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
+ vertex_struct *v_b, vertex_struct *v_c)
+{
+  setup_spans_prologue(no);
+
+  s32 y_b = v_b->y;
+  s64 edge_alt;
+  s32 edge_dx_dy_alt;
+  u32 edge_shift_alt;
+
+  s32 middle_y = y_a;
+  s32 height_minor_a = y_a - y_b;
+  s32 height_minor_b = y_c - y_a;
+  s32 height_major = y_c - y_b;
+
+  vec_2x64s edges_xy_b;
+  vec_2x32s edges_dx_dy_b;
+  vec_2x32u edge_shifts_b;
+
+  vec_2x32s height_increment;
+
+  x_starts.e[0] = x_a;
+  x_starts.e[1] = x_c;
+  dup_2x32b(x_ends, x_b);
+
+  compute_edge_delta_x3(x_a, height_minor_a, height_major);
+
+  height_increment.e[0] = 0;
+  height_increment.e[1] = height_minor_b;
+
+  mla_long_2x32b(edges_xy, edges_dx_dy, height_increment);
+
+  edges_xy_b.e[0] = edge_alt;
+  edges_xy_b.e[1] = edges_xy.e[1];
+
+  edge_shifts_b = edge_shifts;
+  edge_shifts_b.e[0] = edge_shift_alt;
+
+  neg_2x32b(edges_dx_dy_b, edges_dx_dy);
+  edges_dx_dy_b.e[0] = edge_dx_dy_alt;
+  
+  y_a--;
+
+  if(y_b < psx_gpu->viewport_start_y)
+    height_minor_a -= psx_gpu->viewport_start_y - y_b;
+
+  clip = y_a - psx_gpu->viewport_end_y;
+  if(clip > 0)
+  {
+    height_minor_a -= clip;
+    y_a -= clip;
+    setup_spans_clip(decrement, no);
+  }
+
+  setup_spans_prologue_b();
+
+  if(height_minor_a > 0)
+  {
+    y_x4.e[0] = y_a;
+    y_x4.e[1] = y_a - 1;
+    y_x4.e[2] = y_a - 2;
+    y_x4.e[3] = y_a - 3;
+    add_wide_2x32b(edges_xy, edges_xy, edges_dx_dy);
+    setup_spans_adjust_edges_alternate_no(index_left, index_right);
+    setup_spans_adjust_interpolants_up();
+
+    psx_gpu->num_spans = height_minor_a;
+    while(height_minor_a > 0)
+    {
+      setup_spans_set_x4(none, up, no);
+      height_minor_a -= 4;
+    }
+
+    span_edge_data += height_minor_a;
+    span_uvrg_offset += height_minor_a;
+    span_b_offset += height_minor_a;
+  }
+  
+  edges_xy = edges_xy_b;
+  edges_dx_dy = edges_dx_dy_b;
+  edge_shifts = edge_shifts_b;
+
+  uvrg = psx_gpu->uvrg;
+  b = psx_gpu->b;
+
+  y_a = middle_y;
+
+  if(y_c > psx_gpu->viewport_end_y)
+    height_minor_b -= y_c - psx_gpu->viewport_end_y - 1;
+
+  clip = psx_gpu->viewport_start_y - y_a;
+  if(clip > 0)
+  {
+    height_minor_b -= clip;
+    y_a += clip;
+    setup_spans_clip(increment, no);
+  }
+
+  if(height_minor_b > 0)
+  {
+    y_x4.e[0] = y_a;
+    y_x4.e[1] = y_a + 1;
+    y_x4.e[2] = y_a + 2;
+    y_x4.e[3] = y_a + 3;
+    setup_spans_adjust_edges_alternate_no(index_left, index_right);
+
+    psx_gpu->num_spans += height_minor_b;
+    do
+    {
+      setup_spans_set_x4(none, down, no);
+      height_minor_b -= 4;
+    } while(height_minor_b > 0);
+  }
+
+  left_split_triangles++;
+}
+
+#endif
+
+
+#define dither_table_entry_normal(value)                                       \
+  (value)                                                                      \
+
+
+#define setup_blocks_load_msb_mask_indirect()                                  \
+
+#define setup_blocks_load_msb_mask_direct()                                    \
+  vec_8x16u msb_mask;                                                          \
+  dup_8x16b(msb_mask, psx_gpu->mask_msb);                                      \
+
+
+#define setup_blocks_variables_shaded_textured(target)                         \
+  vec_4x32u u_block;                                                           \
+  vec_4x32u v_block;                                                           \
+  vec_4x32u r_block;                                                           \
+  vec_4x32u g_block;                                                           \
+  vec_4x32u b_block;                                                           \
+  vec_4x32u uvrg_dx = psx_gpu->uvrg_dx;                                        \
+  vec_4x32u uvrg_dx4;                                                          \
+  vec_4x32u uvrg_dx8;                                                          \
+  vec_4x32u uvrg;                                                              \
+  u32 b_dx = psx_gpu->b_block_span.e[1];                                       \
+  u32 b_dx4 = b_dx << 2;                                                       \
+  u32 b_dx8 = b_dx << 3;                                                       \
+  u32 b;                                                                       \
+                                                                               \
+  vec_16x8u texture_mask;                                                      \
+  shl_4x32b(uvrg_dx4, uvrg_dx, 2);                                             \
+  shl_4x32b(uvrg_dx8, uvrg_dx, 3);                                             \
+  dup_8x8b(texture_mask.low, psx_gpu->texture_mask_width);                     \
+  dup_8x8b(texture_mask.high, psx_gpu->texture_mask_height)                    \
+
+#define setup_blocks_variables_shaded_untextured(target)                       \
+  vec_4x32u r_block;                                                           \
+  vec_4x32u g_block;                                                           \
+  vec_4x32u b_block;                                                           \
+  vec_4x32u rgb_dx;                                                            \
+  vec_4x32u rgb_dx4;                                                           \
+  vec_4x32u rgb_dx8;                                                           \
+  vec_4x32u rgb;                                                               \
+                                                                               \
+  vec_8x8u d64_0x07;                                                           \
+  vec_8x8u d64_1;                                                              \
+  vec_8x8u d64_4;                                                              \
+  vec_8x8u d64_128;                                                            \
+                                                                               \
+  dup_8x8b(d64_0x07, 0x07);                                                    \
+  dup_8x8b(d64_1, 1);                                                          \
+  dup_8x8b(d64_4, 4);                                                          \
+  dup_8x8b(d64_128, 128);                                                      \
+                                                                               \
+  rgb_dx.low = psx_gpu->uvrg_dx.high;                                          \
+  rgb_dx.e[2] = psx_gpu->b_block_span.e[1];                                    \
+  shl_4x32b(rgb_dx4, rgb_dx, 2);                                               \
+  shl_4x32b(rgb_dx8, rgb_dx, 3)                                                \
+
+#define setup_blocks_variables_unshaded_textured(target)                       \
+  vec_4x32u u_block;                                                           \
+  vec_4x32u v_block;                                                           \
+  vec_2x32u uv_dx = psx_gpu->uvrg_dx.low;                                      \
+  vec_2x32u uv_dx4;                                                            \
+  vec_2x32u uv_dx8;                                                            \
+  vec_2x32u uv = psx_gpu->uvrg.low;                                            \
+                                                                               \
+  vec_16x8u texture_mask;                                                      \
+  shl_2x32b(uv_dx4, uv_dx, 2);                                                 \
+  shl_2x32b(uv_dx8, uv_dx, 3);                                                 \
+  dup_8x8b(texture_mask.low, psx_gpu->texture_mask_width);                     \
+  dup_8x8b(texture_mask.high, psx_gpu->texture_mask_height)                    \
+
+
+#define setup_blocks_variables_unshaded_untextured_direct()                    \
+  or_8x16b(colors, colors, msb_mask)                                           \
+
+#define setup_blocks_variables_unshaded_untextured_indirect()                  \
+
+#define setup_blocks_variables_unshaded_untextured(target)                     \
+  u32 color = psx_gpu->triangle_color;                                         \
+  vec_8x16u colors;                                                            \
+                                                                               \
+  u32 color_r = color & 0xFF;                                                  \
+  u32 color_g = (color >> 8) & 0xFF;                                           \
+  u32 color_b = (color >> 16) & 0xFF;                                          \
+                                                                               \
+  color = (color_r >> 3) | ((color_g >> 3) << 5) |                             \
+   ((color_b >> 3) << 10);                                                     \
+  dup_8x16b(colors, color);                                                    \
+  setup_blocks_variables_unshaded_untextured_##target()                        \
+
+#define setup_blocks_span_initialize_dithered_textured()                       \
+  vec_8x16u dither_offsets;                                                    \
+  shl_long_8x8b(dither_offsets, dither_offsets_short, 4)                       \
+
+#define setup_blocks_span_initialize_dithered_untextured()                     \
+  vec_8x8u dither_offsets;                                                     \
+  add_8x8b(dither_offsets, dither_offsets_short, d64_4)                        \
+
+#define setup_blocks_span_initialize_dithered(texturing)                       \
+  u32 dither_row = psx_gpu->dither_table[y & 0x3];                             \
+  u32 dither_shift = (span_edge_data->left_x & 0x3) * 8;                       \
+  vec_8x8s dither_offsets_short;                                               \
+                                                                               \
+  dither_row =                                                                 \
+   (dither_row >> dither_shift) | (dither_row << (32 - dither_shift));         \
+  dup_2x32b(vector_cast(vec_2x32u, dither_offsets_short), dither_row);         \
+  setup_blocks_span_initialize_dithered_##texturing()                          \
+
+#define setup_blocks_span_initialize_undithered(texturing)                     \
+
+
+#define setup_blocks_span_initialize_shaded_textured()                         \
+{                                                                              \
+  vec_4x32u block_span;                                                        \
+  u32 offset = span_edge_data->left_x;                                         \
+                                                                               \
+  uvrg = *span_uvrg_offset;                                                    \
+  mla_scalar_4x32b(uvrg, uvrg_dx, offset);                                     \
+  b = *span_b_offset;                                                          \
+  b += b_dx * offset;                                                          \
+                                                                               \
+  dup_4x32b(u_block, uvrg.e[0]);                                               \
+  dup_4x32b(v_block, uvrg.e[1]);                                               \
+  dup_4x32b(r_block, uvrg.e[2]);                                               \
+  dup_4x32b(g_block, uvrg.e[3]);                                               \
+  dup_4x32b(b_block, b);                                                       \
+                                                                               \
+  block_span = psx_gpu->u_block_span;                                          \
+  add_4x32b(u_block, u_block, block_span);                                     \
+  block_span = psx_gpu->v_block_span;                                          \
+  add_4x32b(v_block, v_block, block_span);                                     \
+  block_span = psx_gpu->r_block_span;                                          \
+  add_4x32b(r_block, r_block, block_span);                                     \
+  block_span = psx_gpu->g_block_span;                                          \
+  add_4x32b(g_block, g_block, block_span);                                     \
+  block_span = psx_gpu->b_block_span;                                          \
+  add_4x32b(b_block, b_block, block_span);                                     \
+}
+  
+#define setup_blocks_span_initialize_shaded_untextured()                       \
+{                                                                              \
+  vec_4x32u block_span;                                                        \
+  u32 offset = span_edge_data->left_x;                                         \
+                                                                               \
+  rgb.low = span_uvrg_offset->high;                                            \
+  rgb.high.e[0] = *span_b_offset;                                              \
+  mla_scalar_4x32b(rgb, rgb_dx, offset);                                       \
+                                                                               \
+  dup_4x32b(r_block, rgb.e[0]);                                                \
+  dup_4x32b(g_block, rgb.e[1]);                                                \
+  dup_4x32b(b_block, rgb.e[2]);                                                \
+                                                                               \
+  block_span = psx_gpu->r_block_span;                                          \
+  add_4x32b(r_block, r_block, block_span);                                     \
+  block_span = psx_gpu->g_block_span;                                          \
+  add_4x32b(g_block, g_block, block_span);                                     \
+  block_span = psx_gpu->b_block_span;                                          \
+  add_4x32b(b_block, b_block, block_span);                                     \
+}                                                                              \
+  
+#define setup_blocks_span_initialize_unshaded_textured()                       \
+{                                                                              \
+  vec_4x32u block_span;                                                        \
+  u32 offset = span_edge_data->left_x;                                         \
+                                                                               \
+  uv = span_uvrg_offset->low;                                                  \
+  mla_scalar_2x32b(uv, uv_dx, offset);                                         \
+                                                                               \
+  dup_4x32b(u_block, uv.e[0]);                                                 \
+  dup_4x32b(v_block, uv.e[1]);                                                 \
+                                                                               \
+  block_span = psx_gpu->u_block_span;                                          \
+  add_4x32b(u_block, u_block, block_span);                                     \
+  block_span = psx_gpu->v_block_span;                                          \
+  add_4x32b(v_block, v_block, block_span);                                     \
+}                                                                              \
+
+#define setup_blocks_span_initialize_unshaded_untextured()                     \
+
+
+#define setup_blocks_texture_swizzled()                                        \
+{                                                                              \
+  vec_8x8u u_saved = u;                                                        \
+  sli_8x8b(u, v, 4);                                                           \
+  sri_8x8b(v, u_saved, 4);                                                     \
+}                                                                              \
+
+#define setup_blocks_texture_unswizzled()                                      \
+
+#define setup_blocks_store_shaded_textured(swizzling, dithering, target,       \
+ edge_type)                                                                    \
+{                                                                              \
+  vec_8x16u u_whole;                                                           \
+  vec_8x16u v_whole;                                                           \
+  vec_8x16u r_whole;                                                           \
+  vec_8x16u g_whole;                                                           \
+  vec_8x16u b_whole;                                                           \
+                                                                               \
+  vec_8x8u u;                                                                  \
+  vec_8x8u v;                                                                  \
+  vec_8x8u r;                                                                  \
+  vec_8x8u g;                                                                  \
+  vec_8x8u b;                                                                  \
+  vec_8x16u uv;                                                                \
+                                                                               \
+  vec_4x32u dx4;                                                               \
+  vec_4x32u dx8;                                                               \
+                                                                               \
+  shr_narrow_4x32b(u_whole.low, u_block, 16);                                  \
+  shr_narrow_4x32b(v_whole.low, v_block, 16);                                  \
+  shr_narrow_4x32b(r_whole.low, r_block, 16);                                  \
+  shr_narrow_4x32b(g_whole.low, g_block, 16);                                  \
+  shr_narrow_4x32b(b_whole.low, b_block, 16);                                  \
+                                                                               \
+  dup_4x32b(dx4, uvrg_dx4.e[0]);                                               \
+  add_high_narrow_4x32b(u_whole.high, u_block, dx4);                           \
+  dup_4x32b(dx4, uvrg_dx4.e[1]);                                               \
+  add_high_narrow_4x32b(v_whole.high, v_block, dx4);                           \
+  dup_4x32b(dx4, uvrg_dx4.e[2]);                                               \
+  add_high_narrow_4x32b(r_whole.high, r_block, dx4);                           \
+  dup_4x32b(dx4, uvrg_dx4.e[3]);                                               \
+  add_high_narrow_4x32b(g_whole.high, g_block, dx4);                           \
+  dup_4x32b(dx4, b_dx4);                                                       \
+  add_high_narrow_4x32b(b_whole.high, b_block, dx4);                           \
+                                                                               \
+  mov_narrow_8x16b(u, u_whole);                                                \
+  mov_narrow_8x16b(v, v_whole);                                                \
+  mov_narrow_8x16b(r, r_whole);                                                \
+  mov_narrow_8x16b(g, g_whole);                                                \
+  mov_narrow_8x16b(b, b_whole);                                                \
+                                                                               \
+  dup_4x32b(dx8, uvrg_dx8.e[0]);                                               \
+  add_4x32b(u_block, u_block, dx8);                                            \
+  dup_4x32b(dx8, uvrg_dx8.e[1]);                                               \
+  add_4x32b(v_block, v_block, dx8);                                            \
+  dup_4x32b(dx8, uvrg_dx8.e[2]);                                               \
+  add_4x32b(r_block, r_block, dx8);                                            \
+  dup_4x32b(dx8, uvrg_dx8.e[3]);                                               \
+  add_4x32b(g_block, g_block, dx8);                                            \
+  dup_4x32b(dx8, b_dx8);                                                       \
+  add_4x32b(b_block, b_block, dx8);                                            \
+                                                                               \
+  and_8x8b(u, u, texture_mask.low);                                            \
+  and_8x8b(v, v, texture_mask.high);                                           \
+  setup_blocks_texture_##swizzling();                                          \
+                                                                               \
+  zip_8x16b(uv, u, v);                                                         \
+  block->uv = uv;                                                              \
+  block->r = r;                                                                \
+  block->g = g;                                                                \
+  block->b = b;                                                                \
+  block->dither_offsets = vector_cast(vec_8x16u, dither_offsets);              \
+  block->fb_ptr = fb_ptr;                                                      \
+}                                                                              \
+
+#define setup_blocks_store_unshaded_textured(swizzling, dithering, target,     \
+ edge_type)                                                                    \
+{                                                                              \
+  vec_8x16u u_whole;                                                           \
+  vec_8x16u v_whole;                                                           \
+                                                                               \
+  vec_8x8u u;                                                                  \
+  vec_8x8u v;                                                                  \
+  vec_8x16u uv;                                                                \
+                                                                               \
+  vec_4x32u dx4;                                                               \
+  vec_4x32u dx8;                                                               \
+                                                                               \
+  shr_narrow_4x32b(u_whole.low, u_block, 16);                                  \
+  shr_narrow_4x32b(v_whole.low, v_block, 16);                                  \
+                                                                               \
+  dup_4x32b(dx4, uv_dx4.e[0]);                                                 \
+  add_high_narrow_4x32b(u_whole.high, u_block, dx4);                           \
+  dup_4x32b(dx4, uv_dx4.e[1]);                                                 \
+  add_high_narrow_4x32b(v_whole.high, v_block, dx4);                           \
+                                                                               \
+  mov_narrow_8x16b(u, u_whole);                                                \
+  mov_narrow_8x16b(v, v_whole);                                                \
+                                                                               \
+  dup_4x32b(dx8, uv_dx8.e[0]);                                                 \
+  add_4x32b(u_block, u_block, dx8);                                            \
+  dup_4x32b(dx8, uv_dx8.e[1]);                                                 \
+  add_4x32b(v_block, v_block, dx8);                                            \
+                                                                               \
+  and_8x8b(u, u, texture_mask.low);                                            \
+  and_8x8b(v, v, texture_mask.high);                                           \
+  setup_blocks_texture_##swizzling();                                          \
+                                                                               \
+  zip_8x16b(uv, u, v);                                                         \
+  block->uv = uv;                                                              \
+  block->dither_offsets = vector_cast(vec_8x16u, dither_offsets);              \
+  block->fb_ptr = fb_ptr;                                                      \
+}                                                                              \
+
+#define setup_blocks_store_shaded_untextured_dithered()                        \
+  addq_8x8b(r, r, dither_offsets);                                             \
+  addq_8x8b(g, g, dither_offsets);                                             \
+  addq_8x8b(b, b, dither_offsets);                                             \
+                                                                               \
+  subq_8x8b(r, r, d64_4);                                                      \
+  subq_8x8b(g, g, d64_4);                                                      \
+  subq_8x8b(b, b, d64_4)                                                       \
+
+#define setup_blocks_store_shaded_untextured_undithered()                      \
+  
+
+#define setup_blocks_store_untextured_pixels_indirect_full(_pixels)            \
+  block->pixels = _pixels;                                                     \
+  block->fb_ptr = fb_ptr                                                       \
+
+#define setup_blocks_store_untextured_pixels_indirect_edge(_pixels)            \
+  block->pixels = _pixels;                                                     \
+  block->fb_ptr = fb_ptr                                                       \
+
+#define setup_blocks_store_shaded_untextured_seed_pixels_indirect()            \
+  mul_long_8x8b(pixels, r, d64_1)                                              \
+
+
+#define setup_blocks_store_untextured_pixels_direct_full(_pixels)              \
+  store_8x16b(_pixels, fb_ptr)                                                 \
+
+#define setup_blocks_store_untextured_pixels_direct_edge(_pixels)              \
+{                                                                              \
+  vec_8x16u fb_pixels;                                                         \
+  vec_8x16u draw_mask;                                                         \
+  vec_8x16u test_mask = psx_gpu->test_mask;                                    \
+                                                                               \
+  load_8x16b(fb_pixels, fb_ptr);                                               \
+  dup_8x16b(draw_mask, span_edge_data->right_mask);                            \
+  tst_8x16b(draw_mask, draw_mask, test_mask);                                  \
+  bif_8x16b(fb_pixels, _pixels, draw_mask);                                    \
+  store_8x16b(fb_pixels, fb_ptr);                                              \
+}                                                                              \
+
+#define setup_blocks_store_shaded_untextured_seed_pixels_direct()              \
+  pixels = msb_mask;                                                           \
+  mla_long_8x8b(pixels, r, d64_1)                                              \
+
+
+#define setup_blocks_store_shaded_untextured(swizzling, dithering, target,     \
+ edge_type)                                                                    \
+{                                                                              \
+  vec_8x16u r_whole;                                                           \
+  vec_8x16u g_whole;                                                           \
+  vec_8x16u b_whole;                                                           \
+                                                                               \
+  vec_8x8u r;                                                                  \
+  vec_8x8u g;                                                                  \
+  vec_8x8u b;                                                                  \
+                                                                               \
+  vec_4x32u dx4;                                                               \
+  vec_4x32u dx8;                                                               \
+                                                                               \
+  vec_8x16u pixels;                                                            \
+                                                                               \
+  shr_narrow_4x32b(r_whole.low, r_block, 16);                                  \
+  shr_narrow_4x32b(g_whole.low, g_block, 16);                                  \
+  shr_narrow_4x32b(b_whole.low, b_block, 16);                                  \
+                                                                               \
+  dup_4x32b(dx4, rgb_dx4.e[0]);                                                \
+  add_high_narrow_4x32b(r_whole.high, r_block, dx4);                           \
+  dup_4x32b(dx4, rgb_dx4.e[1]);                                                \
+  add_high_narrow_4x32b(g_whole.high, g_block, dx4);                           \
+  dup_4x32b(dx4, rgb_dx4.e[2]);                                                \
+  add_high_narrow_4x32b(b_whole.high, b_block, dx4);                           \
+                                                                               \
+  mov_narrow_8x16b(r, r_whole);                                                \
+  mov_narrow_8x16b(g, g_whole);                                                \
+  mov_narrow_8x16b(b, b_whole);                                                \
+                                                                               \
+  dup_4x32b(dx8, rgb_dx8.e[0]);                                                \
+  add_4x32b(r_block, r_block, dx8);                                            \
+  dup_4x32b(dx8, rgb_dx8.e[1]);                                                \
+  add_4x32b(g_block, g_block, dx8);                                            \
+  dup_4x32b(dx8, rgb_dx8.e[2]);                                                \
+  add_4x32b(b_block, b_block, dx8);                                            \
+                                                                               \
+  setup_blocks_store_shaded_untextured_##dithering();                          \
+                                                                               \
+  shr_8x8b(r, r, 3);                                                           \
+  bic_8x8b(g, g, d64_0x07);                                                    \
+  bic_8x8b(b, b, d64_0x07);                                                    \
+                                                                               \
+  setup_blocks_store_shaded_untextured_seed_pixels_##target();                 \
+  mla_long_8x8b(pixels, g, d64_4);                                             \
+  mla_long_8x8b(pixels, b, d64_128)                                            \
+                                                                               \
+  setup_blocks_store_untextured_pixels_##target##_##edge_type(pixels);         \
+}                                                                              \
+
+#define setup_blocks_store_unshaded_untextured(swizzling, dithering, target,   \
+ edge_type)                                                                    \
+  setup_blocks_store_untextured_pixels_##target##_##edge_type(colors)          \
+
+
+#define setup_blocks_store_draw_mask_textured_indirect(_block, bits)           \
+  (_block)->draw_mask_bits = bits                                              \
+
+#define setup_blocks_store_draw_mask_untextured_indirect(_block, bits)         \
+{                                                                              \
+  vec_8x16u bits_mask;                                                         \
+  vec_8x16u test_mask = psx_gpu->test_mask;                                    \
+  dup_8x16b(bits_mask, bits);                                                  \
+  tst_8x16b(bits_mask, bits_mask, test_mask);                                  \
+  (_block)->draw_mask = bits_mask;                                             \
+}                                                                              \
+
+#define setup_blocks_store_draw_mask_untextured_direct(_block, bits)           \
+
+
+#define setup_blocks_add_blocks_indirect()                                     \
+  num_blocks += span_num_blocks;                                               \
+                                                                               \
+  if(num_blocks > MAX_BLOCKS)                                                  \
+  {                                                                            \
+    psx_gpu->num_blocks = num_blocks - span_num_blocks;                        \
+    flush_render_block_buffer(psx_gpu);                                        \
+    num_blocks = span_num_blocks;                                              \
+    block = psx_gpu->blocks;                                                   \
+  }                                                                            \
+
+#define setup_blocks_add_blocks_direct()                                       \
+
+
+#define setup_blocks_builder(shading, texturing, dithering, sw, target)        \
+void setup_blocks_##shading##_##texturing##_##dithering##_##sw##_##target(     \
+ psx_gpu_struct *psx_gpu)                                                      \
+{                                                                              \
+  setup_blocks_load_msb_mask_##target();                                       \
+  setup_blocks_variables_##shading##_##texturing(target);                      \
+                                                                               \
+  edge_data_struct *span_edge_data = psx_gpu->span_edge_data;                  \
+  vec_4x32u *span_uvrg_offset = psx_gpu->span_uvrg_offset;                     \
+  u32 *span_b_offset = psx_gpu->span_b_offset;                                 \
+                                                                               \
+  block_struct *block = psx_gpu->blocks + psx_gpu->num_blocks;                 \
+                                                                               \
+  u32 num_spans = psx_gpu->num_spans;                                          \
+                                                                               \
+  u16 *fb_ptr;                                                                 \
+  u32 y;                                                                       \
+                                                                               \
+  u32 num_blocks = psx_gpu->num_blocks;                                        \
+  u32 span_num_blocks;                                                         \
+                                                                               \
+  while(num_spans)                                                             \
+  {                                                                            \
+    span_num_blocks = span_edge_data->num_blocks;                              \
+    if(span_num_blocks)                                                        \
+    {                                                                          \
+      y = span_edge_data->y;                                                   \
+      fb_ptr = psx_gpu->vram_ptr + span_edge_data->left_x + (y * 1024);        \
+                                                                               \
+      setup_blocks_span_initialize_##shading##_##texturing();                  \
+      setup_blocks_span_initialize_##dithering(texturing);                     \
+                                                                               \
+      setup_blocks_add_blocks_##target();                                      \
+                                                                               \
+      s32 pixel_span = span_num_blocks * 8;                                    \
+      pixel_span -= __builtin_popcount(span_edge_data->right_mask & 0xFF);     \
+      span_pixels += pixel_span;                                               \
+      span_pixel_blocks_unaligned += (pixel_span + 7) / 8;                     \
+                                                                               \
+      span_num_blocks--;                                                       \
+      while(span_num_blocks)                                                   \
+      {                                                                        \
+        setup_blocks_store_##shading##_##texturing(sw, dithering, target,      \
+         full);                                                                \
+        setup_blocks_store_draw_mask_##texturing##_##target(block, 0x00);      \
+                                                                               \
+        fb_ptr += 8;                                                           \
+        block++;                                                               \
+        span_num_blocks--;                                                     \
+      }                                                                        \
+                                                                               \
+      setup_blocks_store_##shading##_##texturing(sw, dithering, target, edge); \
+      setup_blocks_store_draw_mask_##texturing##_##target(block,               \
+       span_edge_data->right_mask);                                            \
+                                                                               \
+      block++;                                                                 \
+    }                                                                          \
+    else                                                                       \
+    {                                                                          \
+      zero_block_spans++;                                                      \
+    }                                                                          \
+                                                                               \
+    num_spans--;                                                               \
+    span_edge_data++;                                                          \
+    span_uvrg_offset++;                                                        \
+    span_b_offset++;                                                           \
+  }                                                                            \
+                                                                               \
+  psx_gpu->num_blocks = num_blocks;                                            \
+}                                                                              \
+
+void setup_blocks_shaded_textured_dithered_unswizzled_indirect(psx_gpu_struct
+ *psx_gpu);
+
+void setup_blocks_shaded_untextured_dithered_unswizzled_indirect(psx_gpu_struct
+ *psx_gpu);
+void setup_blocks_shaded_untextured_undithered_unswizzled_indirect(
+ psx_gpu_struct *psx_gpu);
+void setup_blocks_shaded_untextured_dithered_unswizzled_direct(psx_gpu_struct
+ *psx_gpu);
+void setup_blocks_shaded_untextured_undithered_unswizzled_direct(
+ psx_gpu_struct *psx_gpu);
+
+void setup_blocks_unshaded_textured_dithered_unswizzled_indirect(psx_gpu_struct
+ *psx_gpu);
+void setup_blocks_unshaded_untextured_undithered_unswizzled_indirect(
+ psx_gpu_struct *psx_gpu);
+void setup_blocks_unshaded_untextured_undithered_unswizzled_direct(
+ psx_gpu_struct *psx_gpu);
+
+void setup_blocks_shaded_textured_dithered_swizzled_indirect(psx_gpu_struct
+ *psx_gpu);
+void setup_blocks_unshaded_textured_dithered_swizzled_indirect(psx_gpu_struct
+ *psx_gpu);
+
+
+//setup_blocks_builder(unshaded, untextured, undithered, unswizzled, direct);
+
+#ifndef PANDORA_BUILD
+
+setup_blocks_builder(shaded, textured, dithered, swizzled, indirect);
+setup_blocks_builder(shaded, textured, dithered, unswizzled, indirect);
+
+setup_blocks_builder(unshaded, textured, dithered, unswizzled, indirect);
+setup_blocks_builder(unshaded, textured, dithered, swizzled, indirect);
+
+setup_blocks_builder(shaded, untextured, undithered, unswizzled, indirect);
+setup_blocks_builder(shaded, untextured, dithered, unswizzled, indirect);
+setup_blocks_builder(shaded, untextured, undithered, unswizzled, direct);
+setup_blocks_builder(shaded, untextured, dithered, unswizzled, direct);
+
+setup_blocks_builder(unshaded, untextured, undithered, unswizzled, indirect);
+setup_blocks_builder(unshaded, untextured, undithered, unswizzled, direct);
+
+#endif
+
+void texture_blocks_untextured(psx_gpu_struct *psx_gpu);
+void texture_blocks_4bpp(psx_gpu_struct *psx_gpu);
+void texture_blocks_8bpp(psx_gpu_struct *psx_gpu);
+void texture_blocks_16bpp(psx_gpu_struct *psx_gpu);
+
+#ifndef PANDORA_BUILD
+
+void texture_blocks_untextured(psx_gpu_struct *psx_gpu)
+{
+  if(psx_gpu->primitive_type != PRIMITIVE_TYPE_SPRITE)
+    texel_blocks_untextured += psx_gpu->num_blocks;
+}
+
+void texture_blocks_4bpp(psx_gpu_struct *psx_gpu)
+{
+  block_struct *block = psx_gpu->blocks;
+  u32 num_blocks = psx_gpu->num_blocks;
+  texel_blocks_4bpp += num_blocks;
+
+  vec_8x8u texels_low;
+  vec_8x8u texels_high;
+  vec_8x8u texels;
+  vec_8x16u pixels;
+
+  vec_8x16u clut_a;
+  vec_8x16u clut_b;
+  vec_16x8u clut_low;
+  vec_16x8u clut_high;
+
+  u8 *texture_ptr_8bpp = psx_gpu->texture_page_ptr;
+  u16 *clut_ptr = psx_gpu->clut_ptr;
+
+  // Can be done with one deinterleaving load on NEON
+  load_8x16b(clut_a, clut_ptr);
+  load_8x16b(clut_b, clut_ptr + 8);
+  unzip_16x8b(clut_low, clut_high, clut_a, clut_b);
+
+  if(psx_gpu->current_texture_mask & psx_gpu->dirty_textures_4bpp_mask)
+    update_texture_4bpp_cache(psx_gpu);
+
+  while(num_blocks)
+  {
+    texels.e[0] = texture_ptr_8bpp[block->uv.e[0]];
+    texels.e[1] = texture_ptr_8bpp[block->uv.e[1]];
+    texels.e[2] = texture_ptr_8bpp[block->uv.e[2]];
+    texels.e[3] = texture_ptr_8bpp[block->uv.e[3]];
+    texels.e[4] = texture_ptr_8bpp[block->uv.e[4]];
+    texels.e[5] = texture_ptr_8bpp[block->uv.e[5]];
+    texels.e[6] = texture_ptr_8bpp[block->uv.e[6]];
+    texels.e[7] = texture_ptr_8bpp[block->uv.e[7]];
+
+    tbl_16(texels_low, texels, clut_low);
+    tbl_16(texels_high, texels, clut_high);
+
+    // Can be done with an interleaving store on NEON
+    zip_8x16b(pixels, texels_low, texels_high);
+
+    block->texels = pixels;
+
+    num_blocks--;
+    block++;
+  }
+}
+
+void texture_blocks_8bpp(psx_gpu_struct *psx_gpu)
+{
+  block_struct *block = psx_gpu->blocks;
+  u32 num_blocks = psx_gpu->num_blocks;
+
+  texel_blocks_8bpp += num_blocks;
+
+  if(psx_gpu->current_texture_mask & psx_gpu->dirty_textures_8bpp_mask)
+    update_texture_8bpp_cache(psx_gpu);
+
+  vec_8x16u texels;
+  u8 *texture_ptr_8bpp = psx_gpu->texture_page_ptr;
+
+  u32 texel;
+  u32 offset;
+  u32 i;
+
+  while(num_blocks)
+  {
+    for(i = 0; i < 8; i++)
+    {
+      offset = block->uv.e[i];
+
+      texel = texture_ptr_8bpp[offset];
+      texels.e[i] = psx_gpu->clut_ptr[texel];
+    }
+
+    block->texels = texels;
+
+    num_blocks--;
+    block++;
+  }
+}
+
+void texture_blocks_16bpp(psx_gpu_struct *psx_gpu)
+{
+  block_struct *block = psx_gpu->blocks;
+  u32 num_blocks = psx_gpu->num_blocks;
+
+  texel_blocks_16bpp += num_blocks;
+
+  vec_8x16u texels;
+
+  u16 *texture_ptr_16bpp = psx_gpu->texture_page_ptr;
+  u32 offset;
+  u32 i;
+
+  while(num_blocks)
+  {
+    for(i = 0; i < 8; i++)
+    {
+      offset = block->uv.e[i];
+      offset += ((offset & 0xFF00) * 3);
+
+      texels.e[i] = texture_ptr_16bpp[offset];
+    }
+
+    block->texels = texels;
+
+    num_blocks--;
+    block++;
+  }
+}
+
+#endif
+
+
+#define shade_blocks_load_msb_mask_indirect()                                  \
+
+#define shade_blocks_load_msb_mask_direct()                                    \
+  vec_8x16u msb_mask;                                                          \
+  dup_8x16b(msb_mask, psx_gpu->mask_msb);                                      \
+
+#define shade_blocks_store_indirect(_draw_mask, _pixels)                       \
+  block->draw_mask = _draw_mask;                                               \
+  block->pixels = _pixels                                                      \
+
+#define shade_blocks_store_direct(_draw_mask, _pixels)                         \
+{                                                                              \
+  vec_8x16u fb_pixels;                                                         \
+  or_8x16b(_pixels, _pixels, msb_mask);                                        \
+  load_8x16b(fb_pixels, block->fb_ptr);                                        \
+  bif_8x16b(fb_pixels, _pixels, _draw_mask);                                   \
+  store_8x16b(fb_pixels, block->fb_ptr);                                       \
+}                                                                              \
+
+
+#define shade_blocks_textured_modulated_shaded_primitive_load()                \
+
+#define shade_blocks_textured_modulated_unshaded_primitive_load()              \
+{                                                                              \
+  u32 color = psx_gpu->triangle_color;                                         \
+  dup_8x8b(colors_r, color);                                                   \
+  dup_8x8b(colors_g, color >> 8);                                              \
+  dup_8x8b(colors_b, color >> 16);                                             \
+  if(psx_gpu->triangle_color == 0x808080)                                      \
+    false_modulated_triangles++;                                               \
+}                                                                              \
+
+#define shade_blocks_textured_modulated_shaded_block_load()                    \
+  colors_r = block->r;                                                         \
+  colors_g = block->g;                                                         \
+  colors_b = block->b                                                          \
+
+#define shade_blocks_textured_modulated_unshaded_block_load()                  \
+
+#define shade_blocks_textured_modulate_dithered(component)                     \
+  pixels_##component = block->dither_offsets;                                  \
+  mla_long_8x8b(pixels_##component, texels_##component, colors_##component)    \
+
+#define shade_blocks_textured_modulate_undithered(component)                   \
+  mul_long_8x8b(pixels_##component, texels_##component, colors_##component)    \
+
+#define shade_blocks_textured_modulated_builder(shading, dithering, target)    \
+void shade_blocks_##shading##_textured_modulated_##dithering##_##target(       \
+ psx_gpu_struct *psx_gpu)                                                      \
+{                                                                              \
+  block_struct *block = psx_gpu->blocks;                                       \
+  u32 num_blocks = psx_gpu->num_blocks;                                        \
+  vec_8x16u texels;                                                            \
+                                                                               \
+  vec_8x8u texels_r;                                                           \
+  vec_8x8u texels_g;                                                           \
+  vec_8x8u texels_b;                                                           \
+                                                                               \
+  vec_8x8u colors_r;                                                           \
+  vec_8x8u colors_g;                                                           \
+  vec_8x8u colors_b;                                                           \
+                                                                               \
+  vec_8x8u pixels_r_low;                                                       \
+  vec_8x8u pixels_g_low;                                                       \
+  vec_8x8u pixels_b_low;                                                       \
+  vec_8x16u pixels;                                                            \
+                                                                               \
+  vec_8x16u pixels_r;                                                          \
+  vec_8x16u pixels_g;                                                          \
+  vec_8x16u pixels_b;                                                          \
+                                                                               \
+  vec_8x16u draw_mask;                                                         \
+  vec_8x16u zero_mask;                                                         \
+                                                                               \
+  vec_8x8u d64_0x07;                                                           \
+  vec_8x8u d64_0x1F;                                                           \
+  vec_8x8u d64_1;                                                              \
+  vec_8x8u d64_4;                                                              \
+  vec_8x8u d64_128;                                                            \
+                                                                               \
+  vec_8x16u d128_0x8000;                                                       \
+                                                                               \
+  vec_8x16u test_mask = psx_gpu->test_mask;                                    \
+  u32 draw_mask_bits;                                                          \
+  shade_blocks_load_msb_mask_##target();                                       \
+                                                                               \
+  dup_8x8b(d64_0x07, 0x07);                                                    \
+  dup_8x8b(d64_0x1F, 0x1F);                                                    \
+  dup_8x8b(d64_1, 1);                                                          \
+  dup_8x8b(d64_4, 4);                                                          \
+  dup_8x8b(d64_128, 128);                                                      \
+                                                                               \
+  dup_8x16b(d128_0x8000, 0x8000);                                              \
+                                                                               \
+  shade_blocks_textured_modulated_##shading##_primitive_load();                \
+                                                                               \
+  while(num_blocks)                                                            \
+  {                                                                            \
+    draw_mask_bits = block->draw_mask_bits;                                    \
+    dup_8x16b(draw_mask, draw_mask_bits);                                      \
+    tst_8x16b(draw_mask, draw_mask, test_mask);                                \
+                                                                               \
+    shade_blocks_textured_modulated_##shading##_block_load();                  \
+                                                                               \
+    texels = block->texels;                                                    \
+                                                                               \
+    mov_narrow_8x16b(texels_r, texels);                                        \
+    shr_narrow_8x16b(texels_g, texels, 5);                                     \
+    shr_narrow_8x16b(texels_b, texels, 7);                                     \
+                                                                               \
+    and_8x8b(texels_r, texels_r, d64_0x1F);                                    \
+    and_8x8b(texels_g, texels_g, d64_0x1F);                                    \
+    shr_8x8b(texels_b, texels_b, 3);                                           \
+                                                                               \
+    shade_blocks_textured_modulate_##dithering(r);                             \
+    shade_blocks_textured_modulate_##dithering(g);                             \
+    shade_blocks_textured_modulate_##dithering(b);                             \
+                                                                               \
+    cmpeqz_8x16b(zero_mask, texels);                                           \
+    and_8x16b(pixels, texels, d128_0x8000);                                    \
+                                                                               \
+    shrq_narrow_signed_8x16b(pixels_r_low, pixels_r, 4);                       \
+    shrq_narrow_signed_8x16b(pixels_g_low, pixels_g, 4);                       \
+    shrq_narrow_signed_8x16b(pixels_b_low, pixels_b, 4);                       \
+                                                                               \
+    or_8x16b(zero_mask, draw_mask, zero_mask);                                 \
+                                                                               \
+    shr_8x8b(pixels_r_low, pixels_r_low, 3);                                   \
+    bic_8x8b(pixels_g_low, pixels_g_low, d64_0x07);                            \
+    bic_8x8b(pixels_b_low, pixels_b_low, d64_0x07);                            \
+                                                                               \
+    mla_long_8x8b(pixels, pixels_r_low, d64_1);                                \
+    mla_long_8x8b(pixels, pixels_g_low, d64_4);                                \
+    mla_long_8x8b(pixels, pixels_b_low, d64_128);                              \
+                                                                               \
+    shade_blocks_store_##target(zero_mask, pixels);                            \
+                                                                               \
+    num_blocks--;                                                              \
+    block++;                                                                   \
+  }                                                                            \
+}                                                                              \
+
+void shade_blocks_shaded_textured_modulated_dithered_direct(psx_gpu_struct
+ *psx_gpu);
+void shade_blocks_shaded_textured_modulated_undithered_direct(psx_gpu_struct
+ *psx_gpu);
+void shade_blocks_unshaded_textured_modulated_dithered_direct(psx_gpu_struct
+ *psx_gpu);
+void shade_blocks_unshaded_textured_modulated_undithered_direct(psx_gpu_struct
+ *psx_gpu);
+
+void shade_blocks_shaded_textured_modulated_dithered_indirect(psx_gpu_struct
+ *psx_gpu);
+void shade_blocks_shaded_textured_modulated_undithered_indirect(psx_gpu_struct
+ *psx_gpu);
+void shade_blocks_unshaded_textured_modulated_dithered_indirect(psx_gpu_struct
+ *psx_gpu);
+void shade_blocks_unshaded_textured_modulated_undithered_indirect(psx_gpu_struct
+ *psx_gpu);
+
+#ifndef PANDORA_BUILD
+
+shade_blocks_textured_modulated_builder(shaded, dithered, direct);
+shade_blocks_textured_modulated_builder(shaded, undithered, direct);
+shade_blocks_textured_modulated_builder(unshaded, dithered, direct);
+shade_blocks_textured_modulated_builder(unshaded, undithered, direct);
+
+shade_blocks_textured_modulated_builder(shaded, dithered, indirect);
+shade_blocks_textured_modulated_builder(shaded, undithered, indirect);
+shade_blocks_textured_modulated_builder(unshaded, dithered, indirect);
+shade_blocks_textured_modulated_builder(unshaded, undithered, indirect);
+
+#endif
+
+
+#define shade_blocks_textured_unmodulated_builder(target)                      \
+void shade_blocks_textured_unmodulated_##target(psx_gpu_struct *psx_gpu)       \
+{                                                                              \
+  block_struct *block = psx_gpu->blocks;                                       \
+  u32 num_blocks = psx_gpu->num_blocks;                                        \
+  vec_8x16u draw_mask;                                                         \
+  vec_8x16u test_mask = psx_gpu->test_mask;                                    \
+  u32 draw_mask_bits;                                                          \
+                                                                               \
+  vec_8x16u pixels;                                                            \
+  shade_blocks_load_msb_mask_##target();                                       \
+                                                                               \
+  while(num_blocks)                                                            \
+  {                                                                            \
+    vec_8x16u zero_mask;                                                       \
+                                                                               \
+    draw_mask_bits = block->draw_mask_bits;                                    \
+    dup_8x16b(draw_mask, draw_mask_bits);                                      \
+    tst_8x16b(draw_mask, draw_mask, test_mask);                                \
+                                                                               \
+    pixels = block->texels;                                                    \
+                                                                               \
+    cmpeqz_8x16b(zero_mask, pixels);                                           \
+    or_8x16b(zero_mask, draw_mask, zero_mask);                                 \
+                                                                               \
+    shade_blocks_store_##target(zero_mask, pixels);                            \
+                                                                               \
+    num_blocks--;                                                              \
+    block++;                                                                   \
+  }                                                                            \
+}                                                                              \
+
+void shade_blocks_textured_unmodulated_indirect(psx_gpu_struct *psx_gpu);
+void shade_blocks_textured_unmodulated_direct(psx_gpu_struct *psx_gpu);
+
+#ifndef PANDORA_BUILD
+
+shade_blocks_textured_unmodulated_builder(indirect)
+shade_blocks_textured_unmodulated_builder(direct)
+
+#endif
+
+
+void shade_blocks_unshaded_untextured_indirect(psx_gpu_struct *psx_gpu);
+void shade_blocks_unshaded_untextured_direct(psx_gpu_struct *psx_gpu);
+
+#ifndef PANDORA_BUILD
+                                                                                
+void shade_blocks_unshaded_untextured_indirect(psx_gpu_struct *psx_gpu)
+{
+}
+
+void shade_blocks_unshaded_untextured_direct(psx_gpu_struct *psx_gpu)
+{
+  block_struct *block = psx_gpu->blocks;
+  u32 num_blocks = psx_gpu->num_blocks;
+
+  vec_8x16u pixels = block->pixels;
+  shade_blocks_load_msb_mask_direct();
+
+  while(num_blocks)
+  {
+    shade_blocks_store_direct(block->draw_mask, pixels);
+
+    num_blocks--;
+    block++;
+  }
+}
+
+#endif
+
+void shade_blocks_shaded_untextured(psx_gpu_struct *psx_gpu)
+{
+}
+
+
+#define blend_blocks_mask_evaluate_on()                                        \
+  vec_8x16u mask_pixels;                                                       \
+  cmpltz_8x16b(mask_pixels, framebuffer_pixels);                               \
+  or_8x16b(draw_mask, draw_mask, mask_pixels)                                  \
+
+#define blend_blocks_mask_evaluate_off()                                       \
+
+#define blend_blocks_average()                                                 \
+{                                                                              \
+  vec_8x16u pixels_no_msb;                                                     \
+  vec_8x16u fb_pixels_no_msb;                                                  \
+                                                                               \
+  vec_8x16u d128_0x0421;                                                       \
+  vec_8x16u d128_0x8000;                                                       \
+                                                                               \
+  dup_8x16b(d128_0x0421, 0x0421);                                              \
+  dup_8x16b(d128_0x8000, 0x8000);                                              \
+                                                                               \
+  eor_8x16b(blend_pixels, pixels, framebuffer_pixels);                         \
+  bic_8x16b(pixels_no_msb, pixels, d128_0x8000);                               \
+  and_8x16b(blend_pixels, blend_pixels, d128_0x0421);                          \
+  sub_8x16b(blend_pixels, pixels_no_msb, blend_pixels);                        \
+  bic_8x16b(fb_pixels_no_msb, framebuffer_pixels, d128_0x8000);                \
+  average_8x16b(blend_pixels, fb_pixels_no_msb, blend_pixels);                 \
+}                                                                              \
+
+#define blend_blocks_add()                                                     \
+{                                                                              \
+  vec_8x16u pixels_rb, pixels_g;                                               \
+  vec_8x16u fb_rb, fb_g;                                                       \
+                                                                               \
+  vec_8x16u d128_0x7C1F;                                                       \
+  vec_8x16u d128_0x03E0;                                                       \
+                                                                               \
+  dup_8x16b(d128_0x7C1F, 0x7C1F);                                              \
+  dup_8x16b(d128_0x03E0, 0x03E0);                                              \
+                                                                               \
+  and_8x16b(pixels_rb, pixels, d128_0x7C1F);                                   \
+  and_8x16b(pixels_g, pixels, d128_0x03E0);                                    \
+                                                                               \
+  and_8x16b(fb_rb, framebuffer_pixels, d128_0x7C1F);                           \
+  and_8x16b(fb_g, framebuffer_pixels, d128_0x03E0);                            \
+                                                                               \
+  add_8x16b(fb_rb, fb_rb, pixels_rb);                                          \
+  add_8x16b(fb_g, fb_g, pixels_g);                                             \
+                                                                               \
+  min_16x8b(vector_cast(vec_16x8u, fb_rb), vector_cast(vec_16x8u, fb_rb),      \
+   vector_cast(vec_16x8u, d128_0x7C1F));                                       \
+  min_8x16b(fb_g, fb_g, d128_0x03E0);                                          \
+                                                                               \
+  or_8x16b(blend_pixels, fb_rb, fb_g);                                         \
+}                                                                              \
+
+#define blend_blocks_subtract()                                                \
+{                                                                              \
+  vec_8x16u pixels_rb, pixels_g;                                               \
+  vec_8x16u fb_rb, fb_g;                                                       \
+                                                                               \
+  vec_8x16u d128_0x7C1F;                                                       \
+  vec_8x16u d128_0x03E0;                                                       \
+                                                                               \
+  dup_8x16b(d128_0x7C1F, 0x7C1F);                                              \
+  dup_8x16b(d128_0x03E0, 0x03E0);                                              \
+                                                                               \
+  and_8x16b(pixels_rb, pixels, d128_0x7C1F);                                   \
+  and_8x16b(pixels_g, pixels, d128_0x03E0);                                    \
+                                                                               \
+  and_8x16b(fb_rb, framebuffer_pixels, d128_0x7C1F);                           \
+  and_8x16b(fb_g, framebuffer_pixels, d128_0x03E0);                            \
+                                                                               \
+  subs_16x8b(vector_cast(vec_16x8u, fb_rb),                                    \
+   vector_cast(vec_16x8u, fb_rb), vector_cast(vec_16x8u, pixels_rb));          \
+  subs_8x16b(fb_g, fb_g, pixels_g);                                            \
+                                                                               \
+  or_8x16b(blend_pixels, fb_rb, fb_g);                                         \
+}                                                                              \
+
+#define blend_blocks_add_fourth()                                              \
+{                                                                              \
+  vec_8x16u pixels_rb, pixels_g;                                               \
+  vec_8x16u pixels_fourth;                                                     \
+  vec_8x16u fb_rb, fb_g;                                                       \
+                                                                               \
+  vec_8x16u d128_0x7C1F;                                                       \
+  vec_8x16u d128_0x1C07;                                                       \
+  vec_8x16u d128_0x03E0;                                                       \
+  vec_8x16u d128_0x00E0;                                                       \
+                                                                               \
+  dup_8x16b(d128_0x7C1F, 0x7C1F);                                              \
+  dup_8x16b(d128_0x1C07, 0x1C07);                                              \
+  dup_8x16b(d128_0x03E0, 0x03E0);                                              \
+  dup_8x16b(d128_0x00E0, 0x00E0);                                              \
+                                                                               \
+  shr_8x16b(pixels_fourth, vector_cast(vec_8x16s, pixels), 2);                 \
+                                                                               \
+  and_8x16b(fb_rb, framebuffer_pixels, d128_0x7C1F);                           \
+  and_8x16b(fb_g, framebuffer_pixels, d128_0x03E0);                            \
+                                                                               \
+  and_8x16b(pixels_rb, pixels_fourth, d128_0x1C07);                            \
+  and_8x16b(pixels_g, pixels_fourth, d128_0x00E0);                             \
+                                                                               \
+  add_8x16b(fb_rb, fb_rb, pixels_rb);                                          \
+  add_8x16b(fb_g, fb_g, pixels_g);                                             \
+                                                                               \
+  min_16x8b(vector_cast(vec_16x8u, fb_rb), vector_cast(vec_16x8u, fb_rb),      \
+   vector_cast(vec_16x8u, d128_0x7C1F));                                       \
+  min_8x16b(fb_g, fb_g, d128_0x03E0);                                          \
+                                                                               \
+  or_8x16b(blend_pixels, fb_rb, fb_g);                                         \
+}                                                                              \
+
+#define blend_blocks_blended_combine_textured()                                \
+{                                                                              \
+  vec_8x16u blend_mask;                                                        \
+  cmpltz_8x16b(blend_mask, pixels);                                            \
+                                                                               \
+  or_immediate_8x16b(blend_pixels, blend_pixels, 0x8000);                      \
+  bif_8x16b(blend_pixels, pixels, blend_mask);                                 \
+}                                                                              \
+
+#define blend_blocks_blended_combine_untextured()                              \
+
+
+#define blend_blocks_body_blend(blend_mode, texturing)                         \
+{                                                                              \
+  blend_blocks_##blend_mode();                                                 \
+  blend_blocks_blended_combine_##texturing();                                  \
+}                                                                              \
+
+#define blend_blocks_body_average(texturing)                                   \
+  blend_blocks_body_blend(average, texturing)                                  \
+
+#define blend_blocks_body_add(texturing)                                       \
+  blend_blocks_body_blend(add, texturing)                                      \
+
+#define blend_blocks_body_subtract(texturing)                                  \
+  blend_blocks_body_blend(subtract, texturing)                                 \
+
+#define blend_blocks_body_add_fourth(texturing)                                \
+  blend_blocks_body_blend(add_fourth, texturing)                               \
+
+#define blend_blocks_body_unblended(texturing)                                 \
+  blend_pixels = pixels                                                        \
+
+
+#define blend_blocks_builder(texturing, blend_mode, mask_evaluate)             \
+void                                                                           \
+ blend_blocks_##texturing##_##blend_mode##_##mask_evaluate(psx_gpu_struct      \
+ *psx_gpu)                                                                     \
+{                                                                              \
+  block_struct *block = psx_gpu->blocks;                                       \
+  u32 num_blocks = psx_gpu->num_blocks;                                        \
+  vec_8x16u draw_mask;                                                         \
+  vec_8x16u pixels;                                                            \
+  vec_8x16u blend_pixels;                                                      \
+  vec_8x16u framebuffer_pixels;                                                \
+  vec_8x16u msb_mask;                                                          \
+                                                                               \
+  u16 *fb_ptr;                                                                 \
+                                                                               \
+  dup_8x16b(msb_mask, psx_gpu->mask_msb);                                      \
+                                                                               \
+  while(num_blocks)                                                            \
+  {                                                                            \
+    pixels = block->pixels;                                                    \
+    draw_mask = block->draw_mask;                                              \
+    fb_ptr = block->fb_ptr;                                                    \
+                                                                               \
+    load_8x16b(framebuffer_pixels, fb_ptr);                                    \
+                                                                               \
+    blend_blocks_mask_evaluate_##mask_evaluate();                              \
+    blend_blocks_body_##blend_mode(texturing);                                 \
+                                                                               \
+    or_8x16b(blend_pixels, blend_pixels, msb_mask);                            \
+    bif_8x16b(framebuffer_pixels, blend_pixels, draw_mask);                    \
+    store_8x16b(framebuffer_pixels, fb_ptr);                                   \
+                                                                               \
+    blend_blocks++;                                                            \
+    num_blocks--;                                                              \
+    block++;                                                                   \
+  }                                                                            \
+}                                                                              \
+
+void blend_blocks_textured_average_off(psx_gpu_struct *psx_gpu);
+void blend_blocks_textured_average_on(psx_gpu_struct *psx_gpu);
+void blend_blocks_textured_add_off(psx_gpu_struct *psx_gpu);
+void blend_blocks_textured_add_on(psx_gpu_struct *psx_gpu);
+void blend_blocks_textured_subtract_off(psx_gpu_struct *psx_gpu);
+void blend_blocks_textured_subtract_on(psx_gpu_struct *psx_gpu);
+void blend_blocks_textured_add_fourth_off(psx_gpu_struct *psx_gpu);
+void blend_blocks_textured_add_fourth_on(psx_gpu_struct *psx_gpu);
+
+void blend_blocks_untextured_average_off(psx_gpu_struct *psx_gpu);
+void blend_blocks_untextured_average_on(psx_gpu_struct *psx_gpu);
+void blend_blocks_untextured_add_off(psx_gpu_struct *psx_gpu);
+void blend_blocks_untextured_add_on(psx_gpu_struct *psx_gpu);
+void blend_blocks_untextured_subtract_off(psx_gpu_struct *psx_gpu);
+void blend_blocks_untextured_subtract_on(psx_gpu_struct *psx_gpu);
+void blend_blocks_untextured_add_fourth_off(psx_gpu_struct *psx_gpu);
+void blend_blocks_untextured_add_fourth_on(psx_gpu_struct *psx_gpu);
+
+void blend_blocks_textured_unblended_off(psx_gpu_struct *psx_gpu);
+void blend_blocks_textured_unblended_on(psx_gpu_struct *psx_gpu);
+
+#ifndef PANDORA_BUILD
+
+void blend_blocks_textured_unblended_off(psx_gpu_struct *psx_gpu)
+{
+}
+
+blend_blocks_builder(textured, average, off);
+blend_blocks_builder(textured, average, on);
+blend_blocks_builder(textured, add, off);
+blend_blocks_builder(textured, add, on);
+blend_blocks_builder(textured, subtract, off);
+blend_blocks_builder(textured, subtract, on);
+blend_blocks_builder(textured, add_fourth, off);
+blend_blocks_builder(textured, add_fourth, on);
+
+blend_blocks_builder(untextured, average, off);
+blend_blocks_builder(untextured, average, on);
+blend_blocks_builder(untextured, add, off);
+blend_blocks_builder(untextured, add, on);
+blend_blocks_builder(untextured, subtract, off);
+blend_blocks_builder(untextured, subtract, on);
+blend_blocks_builder(untextured, add_fourth, off);
+blend_blocks_builder(untextured, add_fourth, on);
+
+blend_blocks_builder(textured, unblended, on);
+
+#endif
+
+                                                                               
+#define vertex_swap(_a, _b)                                                    \
+{                                                                              \
+  vertex_struct *temp_vertex = _a;                                             \
+  _a = _b;                                                                     \
+  _b = temp_vertex;                                                            \
+  triangle_winding ^= 1;                                                       \
+}                                                                              \
+
+
+// Setup blocks parametric-variables:
+// SHADE  TEXTURE_MAP SWIZZLING
+// 0      0           x          
+// 0      1           0
+// 0      1           1
+// 1      0           x
+// 1      1           0
+// 1      1           1
+// 8 inputs, 6 combinations
+
+#define setup_blocks_switch_untextured_unshaded(dithering, target)             \
+  setup_blocks_unshaded_untextured_undithered_unswizzled_##target              \
+
+#define setup_blocks_switch_untextured_shaded(dithering, target)               \
+  setup_blocks_shaded_untextured_##dithering##_unswizzled_##target             \
+
+#define setup_blocks_switch_untextured(shading, texture_mode, dithering,       \
+ target)                                                                       \
+  setup_blocks_switch_untextured_##shading(dithering, target)                  \
+
+#define setup_blocks_switch_texture_mode_4bpp(shading)                         \
+  setup_blocks_##shading##_textured_dithered_swizzled_indirect                 \
+
+#define setup_blocks_switch_texture_mode_8bpp(shading)                         \
+  setup_blocks_##shading##_textured_dithered_swizzled_indirect                 \
+
+#define setup_blocks_switch_texture_mode_16bpp(shading)                        \
+  setup_blocks_##shading##_textured_dithered_unswizzled_indirect               \
+
+#define setup_blocks_switch_textured(shading, texture_mode, dithering, target) \
+  setup_blocks_switch_texture_mode_##texture_mode(shading)                     \
+
+#define setup_blocks_switch_blended(shading, texturing, texture_mode,          \
+ dithering, mask_evaluate)                                                     \
+  setup_blocks_switch_##texturing(shading, texture_mode, dithering, indirect)  \
+
+#define setup_blocks_switch_unblended_on(shading, texturing, texture_mode,     \
+ dithering)                                                                    \
+  setup_blocks_switch_##texturing(shading, texture_mode, dithering, indirect)  \
+
+#define setup_blocks_switch_unblended_off(shading, texturing, texture_mode,    \
+ dithering)                                                                    \
+  setup_blocks_switch_##texturing(shading, texture_mode, dithering, direct)    \
+
+#define setup_blocks_switch_unblended(shading, texturing, texture_mode,        \
+ dithering, mask_evaluate)                                                     \
+  setup_blocks_switch_unblended_##mask_evaluate(shading, texturing,            \
+   texture_mode, dithering)                                                    \
+
+#define setup_blocks_switch(shading, texturing, texture_mode, dithering,       \
+ blending, mask_evaluate)                                                      \
+  setup_blocks_switch_##blending(shading, texturing, texture_mode,             \
+   dithering, mask_evaluate)                                                   \
+
+
+// Texture blocks:
+
+#define texture_blocks_switch_untextured(texture_mode)                         \
+  texture_blocks_untextured                                                    \
+
+#define texture_blocks_switch_textured(texture_mode)                           \
+  texture_blocks_##texture_mode                                                \
+
+#define texture_blocks_switch(texturing, texture_mode)                         \
+  texture_blocks_switch_##texturing(texture_mode)                              \
+
+
+// Shade blocks parametric-variables:
+// SHADE  TEXTURE_MAP  MODULATE_TEXELS  dither_mode
+// 0      0            x                x
+// 0      1            0                0
+// 0      1            0                1
+// x      1            1                x
+// 1      0            x                0
+// 1      0            x                1
+// 1      1            0                0
+// 1      1            0                1
+// 16 inputs, 8 combinations
+
+#define shade_blocks_switch_unshaded_untextured(modulation, dithering, target) \
+  shade_blocks_unshaded_untextured_##target                                    \
+
+#define shade_blocks_switch_unshaded_textured_unmodulated(dithering, target)   \
+  shade_blocks_textured_unmodulated_##target                                   \
+
+#define shade_blocks_switch_unshaded_textured_modulated(dithering, target)     \
+  shade_blocks_unshaded_textured_modulated_##dithering##_##target              \
+
+#define shade_blocks_switch_unshaded_textured(modulation, dithering, target)   \
+  shade_blocks_switch_unshaded_textured_##modulation(dithering, target)        \
+
+#define shade_blocks_switch_unshaded(texturing, modulation, dithering, target) \
+  shade_blocks_switch_unshaded_##texturing(modulation, dithering, target)      \
+
+#define shade_blocks_switch_shaded_untextured(modulation, dithering, target)   \
+  shade_blocks_shaded_untextured                                               \
+
+#define shade_blocks_switch_shaded_textured_unmodulated(dithering, target)     \
+  shade_blocks_textured_unmodulated_##target                                   \
+
+#define shade_blocks_switch_shaded_textured_modulated(dithering, target)       \
+  shade_blocks_shaded_textured_modulated_##dithering##_##target                \
+
+#define shade_blocks_switch_shaded_textured(modulation, dithering, target)     \
+  shade_blocks_switch_shaded_textured_##modulation(dithering, target)          \
+
+#define shade_blocks_switch_shaded(texturing, modulation, dithering, target)   \
+  shade_blocks_switch_shaded_##texturing(modulation, dithering, target)        \
+
+#define shade_blocks_switch_mask_off(shading, texturing, modulation,           \
+ dithering)                                                                    \
+  shade_blocks_switch_##shading(texturing, modulation, dithering, direct)      \
+
+#define shade_blocks_switch_mask_on(shading, texturing, modulation,            \
+ dithering)                                                                    \
+  shade_blocks_switch_##shading(texturing, modulation, dithering, indirect)    \
+
+#define shade_blocks_switch_blended(shading, texturing, modulation, dithering, \
+ mask_evaluate)                                                                \
+  shade_blocks_switch_##shading(texturing, modulation, dithering, indirect)    \
+
+#define shade_blocks_switch_unblended(shading, texturing, modulation,          \
+ dithering, mask_evaluate)                                                     \
+  shade_blocks_switch_mask_##mask_evaluate(shading, texturing, modulation,     \
+   dithering)                                                                  \
+
+#define shade_blocks_switch(shading, texturing, modulation, dithering,         \
+ blending, mask_evaluate)                                                      \
+  shade_blocks_switch_##blending(shading, texturing, modulation, dithering,    \
+   mask_evaluate)                                                              \
+
+
+// Blend blocks parametric-variables:
+// TEXTURE_MAP BLEND  BM_A BM_B mask_evaluate
+// x           0      x    x    0
+// x           0      x    x    1
+// 0           1      0    0    0
+// 0           1      0    0    1
+// 0           1      0    1    0
+// 0           1      0    1    1
+// 0           1      1    0    0
+// 0           1      1    0    1
+// 0           1      1    1    0
+// 0           1      1    1    1
+// 1           1      0    0    0
+// 1           1      0    0    1
+// 1           1      0    1    0
+// 1           1      0    1    1
+// 1           1      1    0    0
+// 1           1      1    0    1
+// 1           1      1    1    0
+// 1           1      1    1    1
+// 32 inputs, 18 combinations
+
+#define blend_blocks_switch_unblended(texturing, blend_mode, mask_evaluate)    \
+  blend_blocks_textured_unblended_##mask_evaluate                              \
+
+#define blend_blocks_switch_blended(texturing, blend_mode, mask_evaluate)      \
+  blend_blocks_##texturing##_##blend_mode##_##mask_evaluate                    \
+
+#define blend_blocks_switch(texturing, blending, blend_mode, mask_evaluate)    \
+  blend_blocks_switch_##blending(texturing, blend_mode, mask_evaluate)         \
+
+
+#define render_blocks_switch_block_modulation(texture_mode, blend_mode,        \
+ mask_evaluate, shading, dithering, texturing, blending, modulation)           \
+{                                                                              \
+  setup_blocks_switch(shading, texturing, texture_mode, dithering, blending,   \
+   mask_evaluate),                                                             \
+  texture_blocks_switch(texturing, texture_mode),                              \
+  shade_blocks_switch(shading, texturing, modulation, dithering, blending,     \
+   mask_evaluate),                                                             \
+  blend_blocks_switch(texturing, blending, blend_mode, mask_evaluate)          \
+}                                                                              \
+
+#define render_blocks_switch_block_blending(texture_mode, blend_mode,          \
+ mask_evaluate, shading, dithering, texturing, blending)                       \
+  render_blocks_switch_block_modulation(texture_mode, blend_mode,              \
+   mask_evaluate, shading, dithering, texturing, blending, modulated),         \
+  render_blocks_switch_block_modulation(texture_mode, blend_mode,              \
+   mask_evaluate, shading, dithering, texturing, blending, unmodulated)        \
+
+#define render_blocks_switch_block_texturing(texture_mode, blend_mode,         \
+ mask_evaluate, shading, dithering, texturing)                                 \
+  render_blocks_switch_block_blending(texture_mode, blend_mode,                \
+   mask_evaluate, shading, dithering, texturing, unblended),                   \
+  render_blocks_switch_block_blending(texture_mode, blend_mode,                \
+   mask_evaluate, shading, dithering, texturing, blended)                      \
+
+#define render_blocks_switch_block_dithering(texture_mode, blend_mode,         \
+ mask_evaluate, shading, dithering)                                            \
+  render_blocks_switch_block_texturing(texture_mode, blend_mode,               \
+   mask_evaluate, shading, dithering, untextured),                             \
+  render_blocks_switch_block_texturing(texture_mode, blend_mode,               \
+   mask_evaluate, shading, dithering, textured)                                \
+
+#define render_blocks_switch_block_shading(texture_mode, blend_mode,           \
+ mask_evaluate, shading)                                                       \
+  render_blocks_switch_block_dithering(texture_mode, blend_mode,               \
+   mask_evaluate, shading, undithered),                                        \
+  render_blocks_switch_block_dithering(texture_mode, blend_mode,               \
+   mask_evaluate, shading, dithered)                                           \
+
+#define render_blocks_switch_block_mask_evaluate(texture_mode, blend_mode,     \
+ mask_evaluate)                                                                \
+  render_blocks_switch_block_shading(texture_mode, blend_mode, mask_evaluate,  \
+   unshaded),                                                                  \
+  render_blocks_switch_block_shading(texture_mode, blend_mode, mask_evaluate,  \
+   shaded)                                                                     \
+
+#define render_blocks_switch_block_blend_mode(texture_mode, blend_mode)        \
+  render_blocks_switch_block_mask_evaluate(texture_mode, blend_mode, off),     \
+  render_blocks_switch_block_mask_evaluate(texture_mode, blend_mode, on)       \
+
+#define render_blocks_switch_block_texture_mode(texture_mode)                  \
+  render_blocks_switch_block_blend_mode(texture_mode, average),                \
+  render_blocks_switch_block_blend_mode(texture_mode, add),                    \
+  render_blocks_switch_block_blend_mode(texture_mode, subtract),               \
+  render_blocks_switch_block_blend_mode(texture_mode, add_fourth)              \
+
+#define render_blocks_switch_block()                                           \
+  render_blocks_switch_block_texture_mode(4bpp),                               \
+  render_blocks_switch_block_texture_mode(8bpp),                               \
+  render_blocks_switch_block_texture_mode(16bpp),                              \
+  render_blocks_switch_block_texture_mode(4bpp)                                \
+
+
+render_block_handler_struct render_triangle_block_handlers[] =
+{
+  render_blocks_switch_block()
+};
+
+#undef render_blocks_switch_block_modulation
+
+#define render_blocks_switch_block_modulation(texture_mode, blend_mode,        \
+ mask_evaluate, shading, dithering, texturing, blending, modulation)           \
+  "render flags:\n"                                                            \
+  "texture mode:     " #texture_mode "\n"                                      \
+  "blend mode:       " #blend_mode "\n"                                        \
+  "mask evaluation:  " #mask_evaluate "\n"                                     \
+  #shading "\n"                                                                \
+  #dithering "\n"                                                              \
+  #texturing "\n"                                                              \
+  #blending "\n"                                                               \
+  #modulation "\n"                                                             \
+
+char *render_block_flag_strings[] =
+{                                                                               
+  render_blocks_switch_block()
+};
+
+
+#define triangle_y_direction_up   1
+#define triangle_y_direction_flat 2
+#define triangle_y_direction_down 0
+
+#define triangle_winding_positive 0
+#define triangle_winding_negative 1
+
+#define triangle_set_direction(direction_variable, value)                      \
+  u32 direction_variable = (u32)(value) >> 31;                                 \
+  if(value == 0)                                                               \
+    direction_variable = 2                                                     \
+
+#define triangle_case(direction_a, direction_b, direction_c, winding)          \
+  case (triangle_y_direction_##direction_a |                                   \
+   (triangle_y_direction_##direction_b << 2) |                                 \
+   (triangle_y_direction_##direction_c << 4) |                                 \
+   (triangle_winding_##winding << 6))                                          \
+
+psx_gpu_struct __attribute__((aligned(64))) psx_gpu_alt;
+
+void render_triangle(psx_gpu_struct *psx_gpu, vertex_struct *vertexes,
+ u32 flags)
+{
+  s32 y_top, y_bottom;
+  s32 triangle_area;
+  u32 triangle_winding = 0;
+
+  vertex_struct *a = &(vertexes[0]);
+  vertex_struct *b = &(vertexes[1]);
+  vertex_struct *c = &(vertexes[2]);
+
+  triangle_area = triangle_signed_area_x2(a->x, a->y, b->x, b->y, c->x, c->y);
+
+  triangles++;
+
+  if(triangle_area == 0)
+  {
+    trivial_rejects++;
+    return;
+  }
+
+  if(b->y < a->y)
+    vertex_swap(a, b);
+
+  if(c->y < b->y)
+  {
+    vertex_swap(b, c);
+
+    if(b->y < a->y)
+      vertex_swap(a, b);
+  }
+
+  y_bottom = c->y;
+  y_top = a->y;
+
+  if((y_bottom - y_top) >= 512)
+  {
+    trivial_rejects++;
+    return;
+  }
+
+  if(triangle_area < 0)
+  {
+    triangle_area = -triangle_area;
+    triangle_winding ^= 1;
+    vertex_swap(a, c);
+  }
+
+  if(b->x < a->x)
+    vertex_swap(a, b);
+
+  if(c->x < b->x) 
+  {
+    vertex_swap(b, c);
+
+    if(b->x < a->x)
+      vertex_swap(a, b);
+  }
+
+  if((c->x - a->x) >= 1024)
+  {
+    trivial_rejects++;
+    return;
+  }
+
+  if(invalidate_texture_cache_region_viewport(psx_gpu, a->x, y_top, c->x,
+   y_bottom) == 0)
+  {
+    trivial_rejects++;
+    return;
+  }
+
+  psx_gpu->num_spans = 0;
+  psx_gpu->triangle_area = triangle_area;
+  psx_gpu->triangle_winding = triangle_winding;
+
+  s32 y_delta_a = b->y - a->y;
+  s32 y_delta_b = c->y - b->y;
+  s32 y_delta_c = c->y - a->y;
+
+  triangle_set_direction(y_direction_a, y_delta_a);
+  triangle_set_direction(y_direction_b, y_delta_b);
+  triangle_set_direction(y_direction_c, y_delta_c);
+
+  compute_all_gradients(psx_gpu, a, b, c);
+
+  switch(y_direction_a | (y_direction_b << 2) | (y_direction_c << 4) |
+   (triangle_winding << 6))
+  {
+    triangle_case(up, up, up, negative):
+    triangle_case(up, up, flat, negative):
+    triangle_case(up, up, down, negative):
+      setup_spans_up_right(psx_gpu, a, b, c);
+      break;
+
+    triangle_case(flat, up, up, negative):
+    triangle_case(flat, up, flat, negative):
+    triangle_case(flat, up, down, negative):
+      setup_spans_up_a(psx_gpu, a, b, c);
+      break;
+
+    triangle_case(down, up, up, negative):
+      setup_spans_up_down(psx_gpu, a, c, b);
+      break;
+
+    triangle_case(down, up, flat, negative):
+      setup_spans_down_a(psx_gpu, a, c, b);
+      break;
+
+    triangle_case(down, up, down, negative):
+      setup_spans_down_right(psx_gpu, a, c, b);
+      break;
+
+    triangle_case(down, flat, up, negative):
+    triangle_case(down, flat, flat, negative):
+    triangle_case(down, flat, down, negative):
+      setup_spans_down_b(psx_gpu, a, b, c);
+      break;
+
+    triangle_case(down, down, up, negative):
+    triangle_case(down, down, flat, negative):
+    triangle_case(down, down, down, negative):
+      setup_spans_down_left(psx_gpu, a, b, c);
+      break;
+
+    triangle_case(up, up, up, positive):
+    triangle_case(up, up, flat, positive):
+    triangle_case(up, up, down, positive):
+      setup_spans_up_left(psx_gpu, a, b, c);
+      break;
+
+    triangle_case(up, flat, up, positive):
+    triangle_case(up, flat, flat, positive):
+    triangle_case(up, flat, down, positive):
+      setup_spans_up_b(psx_gpu, a, b, c);
+      break;
+
+    triangle_case(up, down, up, positive):
+      setup_spans_up_right(psx_gpu, a, c, b);
+      break;
+
+    triangle_case(up, down, flat, positive):
+      setup_spans_up_a(psx_gpu, a, c, b);
+      break;
+
+    triangle_case(up, down, down, positive):
+      setup_spans_up_down(psx_gpu, a, b, c);
+      break;
+
+    triangle_case(flat, down, up, positive):
+    triangle_case(flat, down, flat, positive):
+    triangle_case(flat, down, down, positive):
+      setup_spans_down_a(psx_gpu, a, b, c);
+      break;
+
+    triangle_case(down, down, up, positive):
+    triangle_case(down, down, flat, positive):
+    triangle_case(down, down, down, positive):
+      setup_spans_down_right(psx_gpu, a, b, c);
+      break;
+  }
+
+  spans += psx_gpu->num_spans;
+
+  u32 render_state = flags &
+   (RENDER_FLAGS_MODULATE_TEXELS | RENDER_FLAGS_BLEND | 
+   RENDER_FLAGS_TEXTURE_MAP | RENDER_FLAGS_SHADE);
+  render_state |= psx_gpu->render_state_base;
+  
+  if((psx_gpu->render_state != render_state) ||
+   (psx_gpu->primitive_type != PRIMITIVE_TYPE_TRIANGLE))
+  {
+    psx_gpu->render_state = render_state;
+    flush_render_block_buffer(psx_gpu);
+    state_changes++;
+  }
+
+  psx_gpu->primitive_type = PRIMITIVE_TYPE_TRIANGLE;
+
+  psx_gpu->render_block_handler =
+   &(render_triangle_block_handlers[render_state]);
+  ((setup_blocks_function_type *)psx_gpu->render_block_handler->setup_blocks)
+   (psx_gpu);
+}
+
+
+void texture_sprite_blocks_8bpp(psx_gpu_struct *psx_gpu);
+
+#ifndef PANDORA_BUILD
+
+void texture_sprite_blocks_8bpp(psx_gpu_struct *psx_gpu)
+{
+  block_struct *block = psx_gpu->blocks;
+  u32 num_blocks = psx_gpu->num_blocks;
+
+  vec_8x16u texels;
+  vec_8x8u texel_indexes;
+
+  u16 *clut_ptr = psx_gpu->clut_ptr;
+  u32 i;
+
+  while(num_blocks)
+  {
+    texel_indexes = block->r;
+
+    for(i = 0; i < 8; i++)
+    {
+      texels.e[i] = clut_ptr[texel_indexes.e[i]];
+    }
+
+    block->texels = texels;
+
+    num_blocks--;
+    block++;
+  }
+}
+
+#endif
+
+
+#define setup_sprite_tiled_initialize_4bpp()                                   \
+  u16 *clut_ptr = psx_gpu->clut_ptr;                                           \
+  vec_8x16u clut_a, clut_b;                                                    \
+  vec_16x8u clut_low, clut_high;                                               \
+                                                                               \
+  load_8x16b(clut_a, clut_ptr);                                                \
+  load_8x16b(clut_b, clut_ptr + 8);                                            \
+  unzip_16x8b(clut_low, clut_high, clut_a, clut_b);                            \
+                                                                               \
+  if(psx_gpu->current_texture_mask & psx_gpu->dirty_textures_4bpp_mask)        \
+    update_texture_4bpp_cache(psx_gpu)                                         \
+
+#define setup_sprite_tiled_initialize_8bpp()                                   \
+  if(psx_gpu->current_texture_mask & psx_gpu->dirty_textures_8bpp_mask)        \
+    update_texture_8bpp_cache(psx_gpu)                                         \
+
+
+#define setup_sprite_tile_fetch_texel_block_8bpp(offset)                       \
+  texture_block_ptr = psx_gpu->texture_page_ptr +                              \
+   ((texture_offset + offset) & texture_mask);                                 \
+                                                                               \
+  load_64b(texels, texture_block_ptr)                                          \
+
+
+#define setup_sprite_tile_setup_block_yes(side, offset, texture_mode)          \
+
+#define setup_sprite_tile_setup_block_no(side, offset, texture_mode)           \
+
+#define setup_sprite_tile_add_blocks(tile_num_blocks)                          \
+  num_blocks += tile_num_blocks;                                               \
+  sprite_blocks += tile_num_blocks;                                            \
+                                                                               \
+  if(num_blocks > MAX_BLOCKS)                                                  \
+  {                                                                            \
+    flush_render_block_buffer(psx_gpu);                                        \
+    num_blocks = tile_num_blocks;                                              \
+    block = psx_gpu->blocks;                                                   \
+  }                                                                            \
+
+#define setup_sprite_tile_full_4bpp(edge)                                      \
+{                                                                              \
+  vec_8x8u texels_low, texels_high;                                            \
+  vec_8x16u pixels;                                                            \
+  setup_sprite_tile_add_blocks(sub_tile_height * 2);                           \
+                                                                               \
+  while(sub_tile_height)                                                       \
+  {                                                                            \
+    setup_sprite_tile_fetch_texel_block_8bpp(0);                               \
+    tbl_16(texels_low, texels, clut_low);                                      \
+    tbl_16(texels_high, texels, clut_high);                                    \
+    zip_8x16b(pixels, texels_low, texels_high);                                \
+                                                                               \
+    block->texels = pixels;                                                    \
+    block->draw_mask_bits = left_mask_bits;                                    \
+    block->fb_ptr = fb_ptr;                                                    \
+    block++;                                                                   \
+                                                                               \
+    setup_sprite_tile_fetch_texel_block_8bpp(8);                               \
+    tbl_16(texels_low, texels, clut_low);                                      \
+    tbl_16(texels_high, texels, clut_high);                                    \
+    zip_8x16b(pixels, texels_low, texels_high);                                \
+                                                                               \
+    block->texels = pixels;                                                    \
+    block->draw_mask_bits = right_mask_bits;                                   \
+    block->fb_ptr = fb_ptr + 8;                                                \
+    block++;                                                                   \
+                                                                               \
+    fb_ptr += 1024;                                                            \
+    texture_offset += 0x10;                                                    \
+    sub_tile_height--;                                                         \
+  }                                                                            \
+  texture_offset += 0xF00;                                                     \
+  psx_gpu->num_blocks = num_blocks;                                            \
+}                                                                              \
+
+#define setup_sprite_tile_half_4bpp(edge)                                      \
+{                                                                              \
+  vec_8x8u texels_low, texels_high;                                            \
+  vec_8x16u pixels;                                                            \
+  setup_sprite_tile_add_blocks(sub_tile_height);                               \
+                                                                               \
+  while(sub_tile_height)                                                       \
+  {                                                                            \
+    setup_sprite_tile_fetch_texel_block_8bpp(0);                               \
+    tbl_16(texels_low, texels, clut_low);                                      \
+    tbl_16(texels_high, texels, clut_high);                                    \
+    zip_8x16b(pixels, texels_low, texels_high);                                \
+                                                                               \
+    block->texels = pixels;                                                    \
+    block->draw_mask_bits = edge##_mask_bits;                                  \
+    block->fb_ptr = fb_ptr;                                                    \
+    block++;                                                                   \
+                                                                               \
+    fb_ptr += 1024;                                                            \
+    texture_offset += 0x10;                                                    \
+    sub_tile_height--;                                                         \
+  }                                                                            \
+  texture_offset += 0xF00;                                                     \
+  psx_gpu->num_blocks = num_blocks;                                            \
+}                                                                              \
+
+  
+#define setup_sprite_tile_full_8bpp(edge)                                      \
+{                                                                              \
+  setup_sprite_tile_add_blocks(sub_tile_height * 2);                           \
+                                                                               \
+  while(sub_tile_height)                                                       \
+  {                                                                            \
+    setup_sprite_tile_fetch_texel_block_8bpp(0);                               \
+    block->r = texels;                                                         \
+    block->draw_mask_bits = left_mask_bits;                                    \
+    block->fb_ptr = fb_ptr;                                                    \
+    block++;                                                                   \
+                                                                               \
+    setup_sprite_tile_fetch_texel_block_8bpp(8);                               \
+    block->r = texels;                                                         \
+    block->draw_mask_bits = right_mask_bits;                                   \
+    block->fb_ptr = fb_ptr + 8;                                                \
+    block++;                                                                   \
+                                                                               \
+    fb_ptr += 1024;                                                            \
+    texture_offset += 0x10;                                                    \
+    sub_tile_height--;                                                         \
+  }                                                                            \
+  texture_offset += 0xF00;                                                     \
+  psx_gpu->num_blocks = num_blocks;                                            \
+}                                                                              \
+
+#define setup_sprite_tile_half_8bpp(edge)                                      \
+{                                                                              \
+  setup_sprite_tile_add_blocks(sub_tile_height * 2);                           \
+                                                                               \
+  while(sub_tile_height)                                                       \
+  {                                                                            \
+    setup_sprite_tile_fetch_texel_block_8bpp(0);                               \
+    block->r = texels;                                                         \
+    block->draw_mask_bits = edge##_mask_bits;                                  \
+    block->fb_ptr = fb_ptr;                                                    \
+    block++;                                                                   \
+                                                                               \
+    fb_ptr += 1024;                                                            \
+    texture_offset += 0x10;                                                    \
+    sub_tile_height--;                                                         \
+  }                                                                            \
+  texture_offset += 0xF00;                                                     \
+  psx_gpu->num_blocks = num_blocks;                                            \
+}                                                                              \
+
+  
+#define setup_sprite_tile_column_edge_pre_adjust_half_right()                  \
+  texture_offset = texture_offset_base + 8;                                    \
+  fb_ptr += 8                                                                  \
+
+#define setup_sprite_tile_column_edge_pre_adjust_half_left()                   \
+  texture_offset = texture_offset_base                                         \
+
+#define setup_sprite_tile_column_edge_pre_adjust_half(edge)                    \
+  setup_sprite_tile_column_edge_pre_adjust_half_##edge()                       \
+
+#define setup_sprite_tile_column_edge_pre_adjust_full(edge)                    \
+  texture_offset = texture_offset_base                                         \
+
+#define setup_sprite_tile_column_edge_post_adjust_half_right()                 \
+  fb_ptr -= 8                                                                  \
+
+#define setup_sprite_tile_column_edge_post_adjust_half_left()                  \
+
+#define setup_sprite_tile_column_edge_post_adjust_half(edge)                   \
+  setup_sprite_tile_column_edge_post_adjust_half_##edge()                      \
+
+#define setup_sprite_tile_column_edge_post_adjust_full(edge)                   \
+
+
+#define setup_sprite_tile_column_height_single(edge_mode, edge, texture_mode)  \
+do                                                                             \
+{                                                                              \
+  sub_tile_height = column_data;                                               \
+  setup_sprite_tile_column_edge_pre_adjust_##edge_mode(edge);                  \
+  setup_sprite_tile_##edge_mode##_##texture_mode(edge);                        \
+  setup_sprite_tile_column_edge_post_adjust_##edge_mode(edge);                 \
+} while(0)                                                                     \
+
+#define setup_sprite_tile_column_height_multi(edge_mode, edge, texture_mode)   \
+do                                                                             \
+{                                                                              \
+  u32 tiles_remaining = column_data >> 16;                                     \
+  sub_tile_height = column_data & 0xFF;                                        \
+  setup_sprite_tile_column_edge_pre_adjust_##edge_mode(edge);                  \
+  setup_sprite_tile_##edge_mode##_##texture_mode(edge);                        \
+  tiles_remaining -= 1;                                                        \
+                                                                               \
+  while(tiles_remaining)                                                       \
+  {                                                                            \
+    sub_tile_height = 16;                                                      \
+    setup_sprite_tile_##edge_mode##_##texture_mode(edge);                      \
+    tiles_remaining--;                                                         \
+  }                                                                            \
+                                                                               \
+  sub_tile_height = (column_data >> 8) & 0xFF;                                 \
+  setup_sprite_tile_##edge_mode##_##texture_mode(edge);                        \
+  setup_sprite_tile_column_edge_post_adjust_##edge_mode(edge);                 \
+} while(0)                                                                     \
+
+
+#define setup_sprite_column_data_single()                                      \
+  column_data = height                                                         \
+
+#define setup_sprite_column_data_multi()                                       \
+  column_data = 16 - offset_v;                                                 \
+  column_data |= ((height_rounded & 0xF) + 1) << 8;                            \
+  column_data |= (tile_height - 1) << 16                                       \
+
+
+#define setup_sprite_tile_column_width_single(texture_mode, multi_height,      \
+ edge_mode, edge)                                                              \
+{                                                                              \
+  setup_sprite_column_data_##multi_height();                                   \
+  left_mask_bits = left_block_mask | right_block_mask;                         \
+  right_mask_bits = left_mask_bits >> 8;                                       \
+                                                                               \
+  setup_sprite_tile_column_height_##multi_height(edge_mode, edge,              \
+   texture_mode);                                                              \
+}                                                                              \
+
+#define setup_sprite_tiled_advance_column()                                    \
+  texture_offset_base += 0x100;                                                \
+  if((texture_offset_base & 0xF00) == 0)                                       \
+    texture_offset_base -= (0x100 + 0xF00)                                     \
+
+#define setup_sprite_tile_column_width_multi(texture_mode, multi_height,       \
+ left_mode, right_mode)                                                        \
+{                                                                              \
+  setup_sprite_column_data_##multi_height();                                   \
+  s32 fb_ptr_advance_column = 16 - (1024 * height);                            \
+                                                                               \
+  tile_width -= 2;                                                             \
+  left_mask_bits = left_block_mask;                                            \
+  right_mask_bits = left_mask_bits >> 8;                                       \
+                                                                               \
+  setup_sprite_tile_column_height_##multi_height(left_mode, right,             \
+   texture_mode);                                                              \
+  fb_ptr += fb_ptr_advance_column;                                             \
+                                                                               \
+  left_mask_bits = 0x00;                                                       \
+  right_mask_bits = 0x00;                                                      \
+                                                                               \
+  while(tile_width)                                                            \
+  {                                                                            \
+    setup_sprite_tiled_advance_column();                                       \
+    setup_sprite_tile_column_height_##multi_height(full, none, texture_mode);  \
+    fb_ptr += fb_ptr_advance_column;                                           \
+    tile_width--;                                                              \
+  }                                                                            \
+                                                                               \
+  left_mask_bits = right_block_mask;                                           \
+  right_mask_bits = left_mask_bits >> 8;                                       \
+                                                                               \
+  setup_sprite_tiled_advance_column();                                         \
+  setup_sprite_tile_column_height_##multi_height(right_mode, left,             \
+   texture_mode);                                                              \
+}                                                                              \
+
+
+#define setup_sprite_tiled_builder(texture_mode)                               \
+void setup_sprite_##texture_mode(psx_gpu_struct *psx_gpu, s32 x, s32 y,        \
+ s32 u, s32 v, s32 width, s32 height, u32 color)                               \
+{                                                                              \
+  s32 offset_u = u & 0xF;                                                      \
+  s32 offset_v = v & 0xF;                                                      \
+                                                                               \
+  s32 width_rounded = offset_u + width + 15;                                   \
+  s32 height_rounded = offset_v + height + 15;                                 \
+  s32 tile_height = height_rounded / 16;                                       \
+  s32 tile_width = width_rounded / 16;                                         \
+  u32 offset_u_right = width_rounded & 0xF;                                    \
+                                                                               \
+  u32 left_block_mask = ~(0xFFFF << offset_u);                                 \
+  u32 right_block_mask = 0xFFFE << offset_u_right;                             \
+                                                                               \
+  u32 left_mask_bits;                                                          \
+  u32 right_mask_bits;                                                         \
+                                                                               \
+  u32 sub_tile_height;                                                         \
+  u32 column_data;                                                             \
+                                                                               \
+  u32 texture_mask = (psx_gpu->texture_mask_width & 0xF) |                     \
+   ((psx_gpu->texture_mask_height & 0xF) << 4) |                               \
+   ((psx_gpu->texture_mask_width >> 4) << 8) |                                 \
+   ((psx_gpu->texture_mask_height >> 4) << 12);                                \
+  u32 texture_offset = ((v & 0xF) << 4) | ((u & 0xF0) << 4) |                  \
+   ((v & 0xF0) << 8);                                                          \
+  u32 texture_offset_base = texture_offset;                                    \
+  u32 control_mask;                                                            \
+                                                                               \
+  u16 *fb_ptr = psx_gpu->vram_ptr + (y * 1024) + (x - offset_u);               \
+  u32 num_blocks = psx_gpu->num_blocks;                                        \
+  block_struct *block = psx_gpu->blocks + num_blocks;                          \
+                                                                               \
+  u16 *texture_block_ptr;                                                      \
+  vec_8x8u texels;                                                             \
+                                                                               \
+  setup_sprite_tiled_initialize_##texture_mode();                              \
+                                                                               \
+  control_mask = tile_width == 1;                                              \
+  control_mask |= (tile_height == 1) << 1;                                     \
+  control_mask |= ((left_block_mask & 0xFF) == 0xFF) << 2;                     \
+  control_mask |= (((right_block_mask >> 8) & 0xFF) == 0xFF) << 3;             \
+                                                                               \
+  sprites_##texture_mode++;                                                    \
+                                                                               \
+  switch(control_mask)                                                         \
+  {                                                                            \
+    default:                                                                   \
+    case 0x0:                                                                  \
+      setup_sprite_tile_column_width_multi(texture_mode, multi, full, full);   \
+      break;                                                                   \
+                                                                               \
+    case 0x1:                                                                  \
+      setup_sprite_tile_column_width_single(texture_mode, multi, full, none);  \
+      break;                                                                   \
+                                                                               \
+    case 0x2:                                                                  \
+      setup_sprite_tile_column_width_multi(texture_mode, single, full, full);  \
+      break;                                                                   \
+                                                                               \
+    case 0x3:                                                                  \
+      setup_sprite_tile_column_width_single(texture_mode, single, full, none); \
+      break;                                                                   \
+                                                                               \
+    case 0x4:                                                                  \
+      setup_sprite_tile_column_width_multi(texture_mode, multi, half, full);   \
+      break;                                                                   \
+                                                                               \
+    case 0x5:                                                                  \
+      setup_sprite_tile_column_width_single(texture_mode, multi, half, right); \
+      break;                                                                   \
+                                                                               \
+    case 0x6:                                                                  \
+      setup_sprite_tile_column_width_multi(texture_mode, single, half, full);  \
+      break;                                                                   \
+                                                                               \
+    case 0x7:                                                                  \
+      setup_sprite_tile_column_width_single(texture_mode, single, half, right);\
+      break;                                                                   \
+                                                                               \
+    case 0x8:                                                                  \
+      setup_sprite_tile_column_width_multi(texture_mode, multi, full, half);   \
+      break;                                                                   \
+                                                                               \
+    case 0x9:                                                                  \
+      setup_sprite_tile_column_width_single(texture_mode, multi, half, left);  \
+      break;                                                                   \
+                                                                               \
+    case 0xA:                                                                  \
+      setup_sprite_tile_column_width_multi(texture_mode, single, full, half);  \
+      break;                                                                   \
+                                                                               \
+    case 0xB:                                                                  \
+      setup_sprite_tile_column_width_single(texture_mode, single, half, left); \
+      break;                                                                   \
+                                                                               \
+    case 0xC:                                                                  \
+      setup_sprite_tile_column_width_multi(texture_mode, multi, half, half);   \
+      break;                                                                   \
+                                                                               \
+    case 0xE:                                                                  \
+      setup_sprite_tile_column_width_multi(texture_mode, single, half, half);  \
+      break;                                                                   \
+  }                                                                            \
+}                                                                              \
+
+
+void setup_sprite_4bpp(psx_gpu_struct *psx_gpu, s32 x, s32 y, s32 u, s32 v,
+ s32 width, s32 height, u32 color);
+void setup_sprite_8bpp(psx_gpu_struct *psx_gpu, s32 x, s32 y, s32 u, s32 v,
+ s32 width, s32 height, u32 color);
+void setup_sprite_16bpp(psx_gpu_struct *psx_gpu, s32 x, s32 y, s32 u, s32 v,
+ s32 width, s32 height, u32 color);
+
+#ifndef PANDORA_BUILD
+setup_sprite_tiled_builder(4bpp);
+setup_sprite_tiled_builder(8bpp);
+
+void setup_sprite_16bpp(psx_gpu_struct *psx_gpu, s32 x, s32 y, s32 u,
+ s32 v, s32 width, s32 height, u32 color)
+{
+  u32 left_offset = u & 0x7;
+  u32 width_rounded = width + left_offset + 7;
+
+  u16 *fb_ptr = psx_gpu->vram_ptr + (y * 1024) + (x - left_offset);
+  u32 right_width = width_rounded & 0x7;
+  u32 block_width = width_rounded / 8;
+  u32 fb_ptr_pitch = (1024 + 8) - (block_width * 8);
+
+  u32 left_mask_bits = ~(0xFF << left_offset);
+  u32 right_mask_bits = 0xFE << right_width;
+
+  u32 texture_offset_base = u + (v * 1024);
+  u32 texture_mask =
+   psx_gpu->texture_mask_width | (psx_gpu->texture_mask_height * 1024);
+
+  u32 blocks_remaining;
+  u32 num_blocks = psx_gpu->num_blocks;
+  block_struct *block = psx_gpu->blocks + num_blocks;
+
+  u16 *texture_page_ptr = psx_gpu->texture_page_ptr;
+  u16 *texture_block_ptr;
+
+  texture_offset_base &= ~0x7;
+
+  sprites_16bpp++;
+
+  if(block_width == 1)
+  {
+    u32 mask_bits = left_mask_bits | right_mask_bits;
+
+    while(height)
+    {
+      num_blocks++;
+      sprite_blocks++;
+
+      if(num_blocks > MAX_BLOCKS)
+      {
+        flush_render_block_buffer(psx_gpu);
+        num_blocks = 1;
+        block = psx_gpu->blocks;
+      }
+      
+      texture_block_ptr =
+       texture_page_ptr + (texture_offset_base & texture_mask);
+
+      load_128b(block->texels, texture_block_ptr);
+      block->draw_mask_bits = mask_bits;
+      block->fb_ptr = fb_ptr;
+
+      block++;
+
+      texture_offset_base += 1024;
+      fb_ptr += 1024;
+
+      height--;
+      psx_gpu->num_blocks = num_blocks;
+    }
+  }
+  else
+  {
+    u32 texture_offset;
+
+    while(height)
+    {
+      blocks_remaining = block_width - 2;
+      num_blocks += block_width;
+      sprite_blocks += block_width;
+
+      if(num_blocks > MAX_BLOCKS)
+      {
+        flush_render_block_buffer(psx_gpu);
+        num_blocks = block_width;
+        block = psx_gpu->blocks;
+      }
+
+      texture_offset = texture_offset_base;
+      texture_offset_base += 1024;
+
+      texture_block_ptr = texture_page_ptr + (texture_offset & texture_mask);
+      load_128b(block->texels, texture_block_ptr);
+
+      block->draw_mask_bits = left_mask_bits;
+      block->fb_ptr = fb_ptr;
+
+      texture_offset += 8;
+      fb_ptr += 8;
+      block++;
+
+      while(blocks_remaining)
+      {
+        texture_block_ptr = texture_page_ptr + (texture_offset & texture_mask);
+        load_128b(block->texels, texture_block_ptr);
+
+        block->draw_mask_bits = 0;
+        block->fb_ptr = fb_ptr;
+
+        texture_offset += 8;
+        fb_ptr += 8;
+        block++;
+
+        blocks_remaining--;
+      }
+
+      texture_block_ptr = texture_page_ptr + (texture_offset & texture_mask);
+      load_128b(block->texels, texture_block_ptr);
+
+      block->draw_mask_bits = right_mask_bits;
+      block->fb_ptr = fb_ptr;
+
+      fb_ptr += fb_ptr_pitch;
+      block++;
+
+      height--;
+      psx_gpu->num_blocks = num_blocks;
+    }
+  }
+}
+
+#endif
+
+void setup_sprite_untextured(psx_gpu_struct *psx_gpu, s32 x, s32 y, s32 u,
+ s32 v, s32 width, s32 height, u32 color)
+{
+  u32 right_width = ((width - 1) & 0x7) + 1;
+  u32 right_mask_bits = (0xFF << right_width);
+  u16 *fb_ptr = psx_gpu->vram_ptr + (y * 1024) + x;
+  u32 block_width = (width + 7) / 8;
+  u32 fb_ptr_pitch = 1024 - ((block_width - 1) * 8);
+  u32 blocks_remaining;
+  u32 num_blocks = psx_gpu->num_blocks;
+  block_struct *block = psx_gpu->blocks + num_blocks;
+
+  u32 color_r = color & 0xFF;
+  u32 color_g = (color >> 8) & 0xFF;
+  u32 color_b = (color >> 16) & 0xFF;
+  vec_8x16u colors;
+  vec_8x16u right_mask;
+  vec_8x16u test_mask = psx_gpu->test_mask;
+  vec_8x16u zero_mask;
+
+  sprites_untextured++;
+
+  color = (color_r >> 3) | ((color_g >> 3) << 5) | ((color_b >> 3) << 10);
+
+  dup_8x16b(colors, color);
+  dup_8x16b(zero_mask, 0x00);
+  dup_8x16b(right_mask, right_mask_bits);
+  tst_8x16b(right_mask, right_mask, test_mask);
+
+  while(height)
+  {
+    blocks_remaining = block_width - 1;
+    num_blocks += block_width;
+    sprite_blocks += block_width;
+
+    if(num_blocks > MAX_BLOCKS)
+    {
+      flush_render_block_buffer(psx_gpu);
+      num_blocks = block_width;
+      block = psx_gpu->blocks;
+    }
+
+    while(blocks_remaining)
+    {
+      block->pixels = colors;
+      block->draw_mask = zero_mask;
+      block->fb_ptr = fb_ptr;
+
+      fb_ptr += 8;
+      block++;
+      blocks_remaining--;
+    }
+
+    block->pixels = colors;
+    block->draw_mask = right_mask;
+    block->fb_ptr = fb_ptr;
+
+    block++;
+    fb_ptr += fb_ptr_pitch;
+
+    height--;
+    psx_gpu->num_blocks = num_blocks;
+  }
+}
+
+
+
+#define setup_sprite_blocks_switch_textured(texture_mode)                      \
+  setup_sprite_##texture_mode                                                  \
+
+#define setup_sprite_blocks_switch_untextured(texture_mode)                    \
+  setup_sprite_untextured                                                      \
+
+#define setup_sprite_blocks_switch(texturing, texture_mode)                    \
+  setup_sprite_blocks_switch_##texturing(texture_mode)                         \
+
+
+#define texture_sprite_blocks_switch_4bpp()                                    \
+  texture_blocks_untextured                                                    \
+
+#define texture_sprite_blocks_switch_8bpp()                                    \
+  texture_sprite_blocks_8bpp                                                   \
+
+#define texture_sprite_blocks_switch_16bpp()                                   \
+  texture_blocks_untextured                                                    \
+
+#define texture_sprite_blocks_switch_untextured(texture_mode)                  \
+  texture_blocks_untextured                                                    \
+
+#define texture_sprite_blocks_switch_textured(texture_mode)                    \
+  texture_sprite_blocks_switch_##texture_mode()                                \
+
+#define render_sprite_blocks_switch_block_modulation(texture_mode, blend_mode, \
+ mask_evaluate, shading, dithering, texturing, blending, modulation)           \
+{                                                                              \
+  setup_sprite_blocks_switch(texturing, texture_mode),                         \
+  texture_sprite_blocks_switch_##texturing(texture_mode),                      \
+  shade_blocks_switch(unshaded, texturing, modulation, undithered, blending,   \
+   mask_evaluate),                                                             \
+  blend_blocks_switch(texturing, blending, blend_mode, mask_evaluate)          \
+}                                                                              \
+
+#define render_sprite_blocks_switch_block_blending(texture_mode, blend_mode,   \
+ mask_evaluate, shading, dithering, texturing, blending)                       \
+  render_sprite_blocks_switch_block_modulation(texture_mode, blend_mode,       \
+   mask_evaluate, shading, dithering, texturing, blending, modulated),         \
+  render_sprite_blocks_switch_block_modulation(texture_mode, blend_mode,       \
+   mask_evaluate, shading, dithering, texturing, blending, unmodulated)        \
+
+#define render_sprite_blocks_switch_block_texturing(texture_mode, blend_mode,  \
+ mask_evaluate, shading, dithering, texturing)                                 \
+  render_sprite_blocks_switch_block_blending(texture_mode, blend_mode,         \
+   mask_evaluate, shading, dithering, texturing, unblended),                   \
+  render_sprite_blocks_switch_block_blending(texture_mode, blend_mode,         \
+   mask_evaluate, shading, dithering, texturing, blended)                      \
+
+#define render_sprite_blocks_switch_block_dithering(texture_mode, blend_mode,  \
+ mask_evaluate, shading, dithering)                                            \
+  render_sprite_blocks_switch_block_texturing(texture_mode, blend_mode,        \
+   mask_evaluate, shading, dithering, untextured),                             \
+  render_sprite_blocks_switch_block_texturing(texture_mode, blend_mode,        \
+   mask_evaluate, shading, dithering, textured)                                \
+
+#define render_sprite_blocks_switch_block_shading(texture_mode, blend_mode,    \
+ mask_evaluate, shading)                                                       \
+  render_sprite_blocks_switch_block_dithering(texture_mode, blend_mode,        \
+   mask_evaluate, shading, undithered),                                        \
+  render_sprite_blocks_switch_block_dithering(texture_mode, blend_mode,        \
+   mask_evaluate, shading, dithered)                                           \
+
+#define render_sprite_blocks_switch_block_mask_evaluate(texture_mode,          \
+ blend_mode, mask_evaluate)                                                    \
+  render_sprite_blocks_switch_block_shading(texture_mode, blend_mode,          \
+   mask_evaluate, unshaded),                                                   \
+  render_sprite_blocks_switch_block_shading(texture_mode, blend_mode,          \
+   mask_evaluate, shaded)                                                      \
+
+#define render_sprite_blocks_switch_block_blend_mode(texture_mode, blend_mode) \
+  render_sprite_blocks_switch_block_mask_evaluate(texture_mode, blend_mode,    \
+   off),                                                                       \
+  render_sprite_blocks_switch_block_mask_evaluate(texture_mode, blend_mode,    \
+   on)                                                                         \
+
+#define render_sprite_blocks_switch_block_texture_mode(texture_mode)           \
+  render_sprite_blocks_switch_block_blend_mode(texture_mode, average),         \
+  render_sprite_blocks_switch_block_blend_mode(texture_mode, add),             \
+  render_sprite_blocks_switch_block_blend_mode(texture_mode, subtract),        \
+  render_sprite_blocks_switch_block_blend_mode(texture_mode, add_fourth)       \
+
+#define render_sprite_blocks_switch_block()                                    \
+  render_sprite_blocks_switch_block_texture_mode(4bpp),                        \
+  render_sprite_blocks_switch_block_texture_mode(8bpp),                        \
+  render_sprite_blocks_switch_block_texture_mode(16bpp),                       \
+  render_sprite_blocks_switch_block_texture_mode(4bpp)                         \
+
+
+render_block_handler_struct render_sprite_block_handlers[] =
+{
+  render_sprite_blocks_switch_block()
+};
+
+
+void render_sprite(psx_gpu_struct *psx_gpu, s32 x, s32 y, u32 u, u32 v,
+ s32 width, s32 height, u32 flags, u32 color)
+{
+  s32 x_right = x + width - 1;
+  s32 y_bottom = y + height - 1;
+
+  if(invalidate_texture_cache_region_viewport(psx_gpu, x, y, x_right,
+   y_bottom) == 0)
+  {
+    return;
+  }
+
+  if(x < psx_gpu->viewport_start_x)
+  {
+    u32 clip = psx_gpu->viewport_start_x - x;
+    x += clip;
+    u += clip;
+    width -= clip;
+  }
+
+  if(y < psx_gpu->viewport_start_y)
+  {
+    s32 clip = psx_gpu->viewport_start_y - y;
+    y += clip;
+    v += clip;
+    height -= clip;
+  }
+
+  if(x_right > psx_gpu->viewport_end_x)
+    width -= x_right - psx_gpu->viewport_end_x;
+
+  if(y_bottom > psx_gpu->viewport_end_y)
+    height -= y_bottom - psx_gpu->viewport_end_y;
+
+  if((width <= 0) || (height <= 0))
+    return;
+
+  sprites++;
+
+  span_pixels += width * height;
+  spans += height;
+
+  u32 render_state = flags &
+   (RENDER_FLAGS_MODULATE_TEXELS | RENDER_FLAGS_BLEND |
+   RENDER_FLAGS_TEXTURE_MAP);
+  render_state |=
+   (psx_gpu->render_state_base & ~RENDER_STATE_DITHER);
+  
+  if((psx_gpu->render_state != render_state) ||
+   (psx_gpu->primitive_type != PRIMITIVE_TYPE_SPRITE))
+  {
+    psx_gpu->render_state = render_state;
+    flush_render_block_buffer(psx_gpu);
+    state_changes++;
+  }
+
+  psx_gpu->primitive_type = PRIMITIVE_TYPE_SPRITE;
+
+  color &= 0xFFFFFF;
+
+  if(psx_gpu->triangle_color != color)
+  {
+    flush_render_block_buffer(psx_gpu);
+    psx_gpu->triangle_color = color;
+  }
+
+  if(color == 0x808080)
+    render_state |= RENDER_FLAGS_MODULATE_TEXELS;
+
+  render_block_handler_struct *render_block_handler =
+   &(render_sprite_block_handlers[render_state]);
+  psx_gpu->render_block_handler = render_block_handler;
+
+  ((setup_sprite_function_type *)render_block_handler->setup_blocks)
+   (psx_gpu, x, y, u, v, width, height, color);
+}
+
+#define draw_pixel_line_mask_evaluate_yes()                                    \
+  if(*vram_ptr & 0x8000)                                                       \
+
+#define draw_pixel_line_mask_evaluate_no()                                     \
+    
+
+#define draw_pixel_line_shaded()                                               \
+{                                                                              \
+  color_r = fixed_to_int(current_r);                                           \
+  color_g = fixed_to_int(current_g);                                           \
+  color_b = fixed_to_int(current_b);                                           \
+                                                                               \
+  current_r += gradient_r;                                                     \
+  current_g += gradient_g;                                                     \
+  current_b += gradient_b;                                                     \
+}                                                                              \
+
+#define draw_pixel_line_unshaded()                                             \
+{                                                                              \
+  color_r = color & 0xFF;                                                      \
+  color_g = (color >> 8) & 0xFF;                                               \
+  color_b = (color >> 16) & 0xFF;                                              \
+}                                                                              \
+
+
+#define draw_pixel_line_dithered(_x, _y)                                       \
+{                                                                              \
+  u32 dither_xor = _x ^ _y;                                                    \
+  s32 dither_offset = (dither_xor >> 1) & 0x1;                                 \
+  dither_offset |= (_y & 0x1) << 1;                                            \
+  dither_offset |= (dither_xor & 0x1) << 2;                                    \
+  dither_offset -= 4;                                                          \
+                                                                               \
+  color_r += dither_offset;                                                    \
+  color_g += dither_offset;                                                    \
+  color_b += dither_offset;                                                    \
+                                                                               \
+  if(color_r < 0)                                                              \
+    color_r = 0;                                                               \
+                                                                               \
+  if(color_g < 0)                                                              \
+    color_g = 0;                                                               \
+                                                                               \
+  if(color_b < 0)                                                              \
+    color_b = 0;                                                               \
+                                                                               \
+  if(color_r > 255)                                                            \
+    color_r = 255;                                                             \
+                                                                               \
+  if(color_g > 255)                                                            \
+    color_g = 255;                                                             \
+                                                                               \
+  if(color_b > 255)                                                            \
+    color_b = 255;                                                             \
+}                                                                              \
+
+#define draw_pixel_line_undithered(_x, _y)                                     \
+
+
+#define draw_pixel_line_average()                                              \
+  color_r = (color_r + fb_r) / 2;                                              \
+  color_g = (color_g + fb_g) / 2;                                              \
+  color_b = (color_b + fb_b) / 2                                               \
+
+#define draw_pixel_line_add()                                                  \
+  color_r += fb_r;                                                             \
+  color_g += fb_g;                                                             \
+  color_b += fb_b;                                                             \
+                                                                               \
+  if(color_r > 31)                                                             \
+    color_r = 31;                                                              \
+                                                                               \
+  if(color_g > 31)                                                             \
+    color_g = 31;                                                              \
+                                                                               \
+  if(color_b > 31)                                                             \
+    color_b = 31                                                               \
+                                                                               \
+
+#define draw_pixel_line_subtract()                                             \
+  color_r = fb_r - color_r;                                                    \
+  color_g = fb_g - color_g;                                                    \
+  color_b = fb_b - color_b;                                                    \
+                                                                               \
+  if(color_r < 0)                                                              \
+    color_r = 0;                                                               \
+                                                                               \
+  if(color_g < 0)                                                              \
+    color_g = 0;                                                               \
+                                                                               \
+  if(color_b < 0)                                                              \
+    color_b = 0                                                                \
+
+#define draw_pixel_line_add_fourth()                                           \
+  color_r = fb_r + (color_r / 4);                                              \
+  color_g = fb_g + (color_g / 4);                                              \
+  color_b = fb_b + (color_b / 4);                                              \
+                                                                               \
+  if(color_r > 31)                                                             \
+    color_r = 31;                                                              \
+                                                                               \
+  if(color_g > 31)                                                             \
+    color_g = 31;                                                              \
+                                                                               \
+  if(color_b > 31)                                                             \
+    color_b = 31                                                               \
+
+
+#define draw_pixel_line_blended(blend_mode)                                    \
+  s32 fb_pixel = *vram_ptr;                                                    \
+  s32 fb_r = fb_pixel & 0x1F;                                                  \
+  s32 fb_g = (fb_pixel >> 5) & 0x1F;                                           \
+  s32 fb_b = (fb_pixel >> 10) & 0x1F;                                          \
+                                                                               \
+  draw_pixel_line_##blend_mode()                                               \
+
+#define draw_pixel_line_unblended(blend_mode)                                  \
+
+
+#define draw_pixel_line(_x, _y, shading, blending, dithering, mask_evaluate,   \
+ blend_mode)                                                                   \
+  if((_x >= psx_gpu->viewport_start_x) && (_y >= psx_gpu->viewport_start_y) && \
+   (_x <= psx_gpu->viewport_end_x) && (_y <= psx_gpu->viewport_end_y))         \
+  {                                                                            \
+    draw_pixel_line_mask_evaluate_##mask_evaluate()                            \
+    {                                                                          \
+      draw_pixel_line_##shading();                                             \
+      draw_pixel_line_##dithering(_x, _y);                                     \
+                                                                               \
+      color_r >>= 3;                                                           \
+      color_g >>= 3;                                                           \
+      color_b >>= 3;                                                           \
+                                                                               \
+      draw_pixel_line_##blending(blend_mode);                                  \
+                                                                               \
+      *vram_ptr = color_r | (color_g << 5) | (color_b << 10) |                 \
+       psx_gpu->mask_msb;                                                      \
+    }                                                                          \
+  }                                                                            \
+
+#define update_increment(value)                                                \
+  value++                                                                      \
+
+#define update_decrement(value)                                                \
+  value--                                                                      \
+
+#define update_vram_row_increment(value)                                       \
+  vram_ptr += 1024                                                             \
+
+#define update_vram_row_decrement(value)                                       \
+  vram_ptr -= 1024                                                             \
+
+#define compare_increment(a, b)                                                \
+  (a <= b)                                                                     \
+
+#define compare_decrement(a, b)                                                \
+  (a >= b)                                                                     \
+
+#define set_line_gradients(minor)                                              \
+{                                                                              \
+  s32 gradient_divisor = delta_##minor;                                        \
+  gradient_r = int_to_fixed(vertex_b->r - vertex_a->r) / gradient_divisor;     \
+  gradient_g = int_to_fixed(vertex_b->g - vertex_a->g) / gradient_divisor;     \
+  gradient_b = int_to_fixed(vertex_b->b - vertex_a->b) / gradient_divisor;     \
+  current_r = fixed_center(vertex_a->r);                                       \
+  current_g = fixed_center(vertex_a->g);                                       \
+  current_b = fixed_center(vertex_a->b);                                       \
+}
+
+#define draw_line_span_horizontal(direction, shading, blending, dithering,     \
+ mask_evaluate, blend_mode)                                                    \
+do                                                                             \
+{                                                                              \
+  error_step = delta_y * 2;                                                    \
+  error_wrap = delta_x * 2;                                                    \
+  error = delta_x;                                                             \
+                                                                               \
+  current_y = y_a;                                                             \
+  set_line_gradients(x);                                                       \
+                                                                               \
+  for(current_x = x_a; current_x <= x_b; current_x++)                          \
+  {                                                                            \
+    draw_pixel_line(current_x, current_y, shading, blending, dithering,        \
+     mask_evaluate, blend_mode);                                               \
+    error += error_step;                                                       \
+    vram_ptr++;                                                                \
+                                                                               \
+    if(error >= error_wrap)                                                    \
+    {                                                                          \
+      update_##direction(current_y);                                           \
+      update_vram_row_##direction();                                           \
+      error -= error_wrap;                                                     \
+    }                                                                          \
+  }                                                                            \
+} while(0)                                                                     \
+
+#define draw_line_span_vertical(direction, shading, blending, dithering,       \
+ mask_evaluate, blend_mode)                                                    \
+do                                                                             \
+{                                                                              \
+  error_step = delta_x * 2;                                                    \
+  error_wrap = delta_y * 2;                                                    \
+  error = delta_y;                                                             \
+                                                                               \
+  current_x = x_a;                                                             \
+  set_line_gradients(y);                                                       \
+                                                                               \
+  for(current_y = y_a; compare_##direction(current_y, y_b);                    \
+   update_##direction(current_y))                                              \
+  {                                                                            \
+    draw_pixel_line(current_x, current_y, shading, blending, dithering,        \
+     mask_evaluate, blend_mode);                                               \
+    error += error_step;                                                       \
+    update_vram_row_##direction();                                             \
+                                                                               \
+    if(error > error_wrap)                                                     \
+    {                                                                          \
+      vram_ptr++;                                                              \
+      current_x++;                                                             \
+      error -= error_wrap;                                                     \
+    }                                                                          \
+  }                                                                            \
+} while(0)                                                                     \
+
+
+#define render_line_body(shading, blending, dithering, mask_evaluate,          \
+ blend_mode)                                                                   \
+  if(delta_y < 0)                                                              \
+  {                                                                            \
+    delta_y *= -1;                                                             \
+                                                                               \
+    if(delta_y >= 512)                                                         \
+      return;                                                                  \
+                                                                               \
+    if(delta_x > delta_y)                                                      \
+    {                                                                          \
+      draw_line_span_horizontal(decrement, shading, blending, dithering,       \
+       mask_evaluate, blend_mode);                                             \
+    }                                                                          \
+    else                                                                       \
+    {                                                                          \
+      draw_line_span_vertical(decrement, shading, blending, dithering,         \
+       mask_evaluate, blend_mode);                                             \
+    }                                                                          \
+  }                                                                            \
+  else                                                                         \
+  {                                                                            \
+    if(delta_y >= 512)                                                         \
+      return;                                                                  \
+                                                                               \
+    if(delta_x > delta_y)                                                      \
+    {                                                                          \
+      draw_line_span_horizontal(increment, shading, blending, dithering,       \
+       mask_evaluate, blend_mode);                                             \
+    }                                                                          \
+    else                                                                       \
+    {                                                                          \
+      draw_line_span_vertical(increment, shading, blending, dithering,         \
+       mask_evaluate, blend_mode);                                             \
+    }                                                                          \
+  }                                                                            \
+
+                                                                                
+void render_line(psx_gpu_struct *psx_gpu, vertex_struct *vertexes, u32 flags,
+ u32 color)
+{
+  s32 color_r, color_g, color_b;
+  u32 triangle_winding = 0;
+
+  fixed_type gradient_r = 0;
+  fixed_type gradient_g = 0;
+  fixed_type gradient_b = 0;
+  fixed_type current_r = 0;
+  fixed_type current_g = 0;
+  fixed_type current_b = 0;
+
+  s32 y_a, y_b;
+  s32 x_a, x_b;
+
+  s32 delta_x, delta_y;
+
+  s32 current_x;
+  s32 current_y;
+
+  u32 error_step;
+  u32 error;
+  u32 error_wrap;
+
+  u16 *vram_ptr;
+
+  flush_render_block_buffer(psx_gpu);
+  psx_gpu->primitive_type = PRIMITIVE_TYPE_LINE;
+
+  vertex_struct *vertex_a = &(vertexes[0]);
+  vertex_struct *vertex_b = &(vertexes[1]);
+
+  u32 control_mask;
+
+  lines++;
+
+  if(vertex_a->x >= vertex_b->x)
+  {
+    vertex_swap(vertex_a, vertex_b);
+  }
+
+  x_a = vertex_a->x;
+  x_b = vertex_b->x;
+
+  y_a = vertex_a->y;
+  y_b = vertex_b->y;
+
+  delta_x = x_b - x_a;
+  delta_y = y_b - y_a;
+
+  if(delta_x >= 1024)
+    return;
+
+  flags &= ~RENDER_FLAGS_TEXTURE_MAP;
+
+  vram_ptr = psx_gpu->vram_ptr + (y_a * 1024) + x_a;
+
+  control_mask = 0x0;
+
+  if(flags & RENDER_FLAGS_SHADE)
+    control_mask |= 0x1;
+
+  if(flags & RENDER_FLAGS_BLEND)
+  {
+    control_mask |= 0x2;
+    control_mask |= ((psx_gpu->render_state_base >> 6) & 0x3) << 4;
+  }
+
+  if(psx_gpu->render_state_base & RENDER_STATE_DITHER)
+    control_mask |= 0x4;
+
+  if(psx_gpu->render_state_base & RENDER_STATE_MASK_EVALUATE)
+    control_mask |= 0x8;
+
+  switch(control_mask)
+  {
+    case 0x0:
+      render_line_body(unshaded, unblended, undithered, no, none);
+      break;
+
+    case 0x1:
+      render_line_body(shaded, unblended, undithered, no, none);
+      break;
+
+    case 0x2:
+      render_line_body(unshaded, blended, undithered, no, average);
+      break;
+
+    case 0x3:
+      render_line_body(shaded, blended, undithered, no, average);
+      break;
+
+    case 0x4:
+      render_line_body(unshaded, unblended, dithered, no, none);
+      break;
+
+    case 0x5:
+      render_line_body(shaded, unblended, dithered, no, none);
+      break;
+
+    case 0x6:
+      render_line_body(unshaded, blended, dithered, no, average);
+      break;
+
+    case 0x7:
+      render_line_body(shaded, blended, dithered, no, average);
+      break;
+
+    case 0x8:
+      render_line_body(unshaded, unblended, undithered, yes, none);
+      break;
+
+    case 0x9:
+      render_line_body(shaded, unblended, undithered, yes, none);
+      break;
+
+    case 0xA:
+      render_line_body(unshaded, blended, undithered, yes, average);
+      break;
+
+    case 0xB:
+      render_line_body(shaded, blended, undithered, yes, average);
+      break;
+
+    case 0xC:
+      render_line_body(unshaded, unblended, dithered, yes, none);
+      break;
+
+    case 0xD:
+      render_line_body(shaded, unblended, dithered, yes, none);
+      break;
+
+    case 0xE:
+      render_line_body(unshaded, blended, dithered, yes, average);
+      break;
+
+    case 0xF:
+      render_line_body(shaded, blended, dithered, yes, average);
+      break;
+
+    case 0x12:
+      render_line_body(unshaded, blended, undithered, no, add);
+      break;
+
+    case 0x13:
+      render_line_body(shaded, blended, undithered, no, add);
+      break;
+
+    case 0x16:
+      render_line_body(unshaded, blended, dithered, no, add);
+      break;
+
+    case 0x17:
+      render_line_body(shaded, blended, dithered, no, add);
+      break;
+
+    case 0x1A:
+      render_line_body(unshaded, blended, undithered, yes, add);
+      break;
+
+    case 0x1B:
+      render_line_body(shaded, blended, undithered, yes, add);
+      break;
+
+    case 0x1E:
+      render_line_body(unshaded, blended, dithered, yes, add);
+      break;
+
+    case 0x1F:
+      render_line_body(shaded, blended, dithered, yes, add);
+      break;
+
+    case 0x22:
+      render_line_body(unshaded, blended, undithered, no, subtract);
+      break;
+
+    case 0x23:
+      render_line_body(shaded, blended, undithered, no, subtract);
+      break;
+
+    case 0x26:
+      render_line_body(unshaded, blended, dithered, no, subtract);
+      break;
+
+    case 0x27:
+      render_line_body(shaded, blended, dithered, no, subtract);
+      break;
+
+    case 0x2A:
+      render_line_body(unshaded, blended, undithered, yes, subtract);
+      break;
+
+    case 0x2B:
+      render_line_body(shaded, blended, undithered, yes, subtract);
+      break;
+
+    case 0x2E:
+      render_line_body(unshaded, blended, dithered, yes, subtract);
+      break;
+
+    case 0x2F:
+      render_line_body(shaded, blended, dithered, yes, subtract);
+      break;
+
+    case 0x32:
+      render_line_body(unshaded, blended, undithered, no, add_fourth);
+      break;
+
+    case 0x33:
+      render_line_body(shaded, blended, undithered, no, add_fourth);
+      break;
+
+    case 0x36:
+      render_line_body(unshaded, blended, dithered, no, add_fourth);
+      break;
+
+    case 0x37:
+      render_line_body(shaded, blended, dithered, no, add_fourth);
+      break;
+
+    case 0x3A:
+      render_line_body(unshaded, blended, undithered, yes, add_fourth);
+      break;
+
+    case 0x3B:
+      render_line_body(shaded, blended, undithered, yes, add_fourth);
+      break;
+
+    case 0x3E:
+      render_line_body(unshaded, blended, dithered, yes, add_fourth);
+      break;
+
+    case 0x3F:
+      render_line_body(shaded, blended, dithered, yes, add_fourth);
+      break;
+  }
+}
+
+
+void render_block_fill(psx_gpu_struct *psx_gpu, u32 color, u32 x, u32 y,
+ u32 width, u32 height)
+{
+  invalidate_texture_cache_region(psx_gpu, x, y, x + width - 1, y + height - 1);
+
+#ifndef PANDORA_BUILD
+  u32 r = color & 0xFF;
+  u32 g = (color >> 8) & 0xFF;
+  u32 b = (color >> 16) & 0xFF;
+  u32 color_16bpp = (r >> 3) | ((g >> 3) << 5) | ((b >> 3) << 10);
+
+  u16 *vram_ptr = psx_gpu->vram_ptr + x + (y * 1024);
+  u32 draw_x, draw_y;
+
+  for(draw_y = 0; draw_y < height; draw_y++)
+  {
+    for(draw_x = 0; draw_x < width; draw_x++)
+    {
+      vram_ptr[draw_x] = color_16bpp;
+    }
+
+    vram_ptr += 1024;
+  }
+#else
+  void render_block_fill_body(psx_gpu_struct *psx_gpu, u32 color, u32 x, u32 y,
+   u32 width, u32 height);
+
+  render_block_fill_body(psx_gpu, color, x, y, width, height);
+#endif
+}
+
+void render_block_copy(psx_gpu_struct *psx_gpu, u16 *source, u32 x, u32 y,
+ u32 width, u32 height, u32 pitch)
+{
+  u16 *vram_ptr = psx_gpu->vram_ptr + x + (y * 1024);
+  u32 draw_x, draw_y;
+
+  invalidate_texture_cache_region(psx_gpu, x, y, x + width - 1, y + height - 1);
+
+  //printf("copy for %d, %d\n", width, height);
+
+  for(draw_y = 0; draw_y < height; draw_y++)
+  {
+    for(draw_x = 0; draw_x < width; draw_x++)
+    {
+      vram_ptr[draw_x] = source[draw_x];
+    }
+
+    source += pitch;
+    vram_ptr += 1024;
+  }
+}
+
+void render_block_move(psx_gpu_struct *psx_gpu, u32 source_x, u32 source_y,
+ u32 dest_x, u32 dest_y, u32 width, u32 height)
+{
+  render_block_copy(psx_gpu, psx_gpu->vram_ptr + source_x + (source_y * 1024),
+   dest_x, dest_y, width, height, 1024);
+}
+
+
+void initialize_reciprocal_table(void)
+{
+  u32 height;
+  u32 height_normalized;
+  u32 height_reciprocal;
+  s32 shift;
+
+  for(height = 1; height < 512; height++)
+  {
+    shift = __builtin_clz(height);
+    height_normalized = height << shift;
+    height_reciprocal = ((1ULL << 50) + (height_normalized - 1)) /
+     height_normalized;
+
+    shift = 32 - (50 - shift);
+
+    reciprocal_table[height] = (height_reciprocal << 12) | shift;
+  }
+}
+
+
+#define dither_table_row(a, b, c, d)                                           \
+ ((a & 0xFF) | ((b & 0xFF) << 8) | ((c & 0xFF) << 16) | ((d & 0xFF) << 24))    \
+
+void initialize_psx_gpu(psx_gpu_struct *psx_gpu)
+{
+  vec_8x16u test_mask =
+   { { { 0x01, 0x02, 0x04, 0x08, 0x10, 0x20, 0x40, 0x80 } } };
+
+  psx_gpu->test_mask = test_mask;
+
+  psx_gpu->pixel_count_mode = 0;
+  psx_gpu->pixel_compare_mode = 0;
+
+  psx_gpu->vram_pixel_counts_a = malloc(sizeof(u8) * 1024 * 512);
+  psx_gpu->vram_pixel_counts_b = malloc(sizeof(u8) * 1024 * 512);
+  memset(psx_gpu->vram_pixel_counts_a, 0, sizeof(u8) * 1024 * 512);
+  memset(psx_gpu->vram_pixel_counts_b, 0, sizeof(u8) * 1024 * 512);
+  psx_gpu->compare_vram = malloc(sizeof(u16) * 1024 * 512);
+
+  psx_gpu->dirty_textures_4bpp_mask = 0xFFFFFFFF;
+  psx_gpu->dirty_textures_8bpp_mask = 0xFFFFFFFF;
+  psx_gpu->dirty_textures_8bpp_alternate_mask = 0xFFFFFFFF;
+  psx_gpu->viewport_mask = 0;
+  psx_gpu->current_texture_page = 0;
+  psx_gpu->current_texture_mask = 0;
+  psx_gpu->last_8bpp_texture_page = 0;
+
+  psx_gpu->clut_settings = 0;
+  psx_gpu->texture_settings = 0;
+  psx_gpu->render_state = 0;
+  psx_gpu->render_state_base = 0;
+  psx_gpu->num_blocks = 0;
+
+  psx_gpu->vram_ptr = psx_gpu->_vram;
+
+  psx_gpu->texture_page_ptr = psx_gpu->vram_ptr;
+  psx_gpu->clut_ptr = psx_gpu->vram_ptr;
+
+  psx_gpu->mask_msb = 0;
+
+  memset(psx_gpu->vram_ptr, 0, sizeof(u16) * 1024 * 512);
+
+  initialize_reciprocal_table();
+
+  //    00 01 10 11
+  // 00  0  4  1  5
+  // 01  6  2  7  3
+  // 10  1  5  0  4
+  // 11  7  3  6  2
+  // (minus ones(4) * 4)
+
+  // d0: (1 3 5 7): x1 ^ y1
+  // d1: (2 3 6 7): y0
+  // d2: (4 5 6 7): x0 ^ y0
+
+
+  psx_gpu->dither_table[0] = dither_table_row(-4, 0, -3, 1);
+  psx_gpu->dither_table[1] = dither_table_row(2, -2, 3, -1);
+  psx_gpu->dither_table[2] = dither_table_row(-3, 1, -4, 0);
+  psx_gpu->dither_table[3] = dither_table_row(3, -1, 2, -2);
+
+  psx_gpu->primitive_type = PRIMITIVE_TYPE_UNKNOWN;
+}
+
+u64 get_us(void)
+{
+  struct timeval tv;
+  gettimeofday(&tv, NULL);
+  
+  return (tv.tv_sec * 1000000ULL) + tv.tv_usec;
+}
+
+#ifdef PANDORA_BUILD
+
+u32 get_counter()
+{
+  u32 counter;
+  __asm__ volatile("mrc p15, 0, %0, c9, c13, 0" : "=r"(counter));
+
+  return counter;
+}
+
+void init_counter(void)
+{
+  u32 value;
+  asm volatile("mrc p15, 0, %0, c9, c12, 0" : "=r"(value));
+  value |= 5; // master enable, ccnt reset
+  value &= ~8; // ccnt divider 0
+  asm volatile("mcr p15, 0, %0, c9, c12, 0" :: "r"(value));
+  // enable cycle counter
+  asm volatile("mcr p15, 0, %0, c9, c12, 1" :: "r"(1 << 31));
+}
+
+void triangle_benchmark(psx_gpu_struct *psx_gpu)
+{
+  u32 i;
+
+  u32 ticks;
+  u32 ticks_elapsed;
+
+  const u32 iterations = 500000;
+
+  psx_gpu->num_blocks = 64;
+  psx_gpu->clut_ptr = psx_gpu->vram_ptr;
+
+  for(i = 0; i < 64; i++)
+  {
+    memset(&(psx_gpu->blocks[i].r), 0, 16);
+  }
+
+  init_counter();
+
+  ticks = get_counter();
+
+  for(i = 0; i < iterations; i++)
+  {
+    texture_sprite_blocks_8bpp(psx_gpu);
+  }
+
+  ticks_elapsed = get_counter() - ticks;
+
+  printf("benchmark: %lf cycles\n", (double)ticks_elapsed / (iterations * 64));
+}
+
+#endif
+
diff --git a/plugins/gpu_neon/psx_gpu/psx_gpu.h b/plugins/gpu_neon/psx_gpu/psx_gpu.h
new file mode 100644
index 0000000..15d9469
--- /dev/null
+++ b/plugins/gpu_neon/psx_gpu/psx_gpu.h
@@ -0,0 +1,241 @@
+/*
+ * Copyright (C) 2011 Gilead Kutnick "Exophase" <exophase@gmail.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; either version 2 of
+ * the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ */
+
+#ifndef PSX_GPU_H
+#define PSX_GPU_H
+
+typedef enum
+{
+  PRIMITIVE_TYPE_TRIANGLE = 0,
+  PRIMITIVE_TYPE_SPRITE = 1,
+  PRIMITIVE_TYPE_LINE = 2,
+  PRIMITIVE_TYPE_UNKNOWN = 3
+} primitive_type_enum;
+
+typedef enum
+{
+  TEXTURE_MODE_4BPP  = 0,
+  TEXTURE_MODE_8BPP  = 1,
+  TEXTURE_MODE_16BPP = 2
+} texture_mode_enum;
+
+typedef enum
+{
+  BLEND_MODE_AVERAGE    = 0,
+  BLEND_MODE_ADD        = 1,
+  BLEND_MODE_SUBTRACT   = 2,
+  BLEND_MODE_ADD_FOURTH = 3
+} blend_mode_enum;
+
+typedef enum
+{
+  RENDER_FLAGS_MODULATE_TEXELS = 0x1,
+  RENDER_FLAGS_BLEND           = 0x2,
+  RENDER_FLAGS_TEXTURE_MAP     = 0x4,
+  RENDER_FLAGS_QUAD            = 0x8,
+  RENDER_FLAGS_SHADE           = 0x10,
+} render_flags_enum;
+
+typedef enum
+{
+  RENDER_STATE_DITHER          = 0x8,
+  RENDER_STATE_MASK_EVALUATE   = 0x20,
+} render_state_enum;
+
+typedef struct
+{
+  u16 left_x;
+  u16 num_blocks;
+  u16 right_mask;
+  u16 y;
+} edge_data_struct;
+
+// 64 bytes total
+typedef struct
+{
+  // 16 bytes
+  union
+  {
+    vec_8x16u uv;
+    vec_8x16u texels;
+    vec_8x16u draw_mask;
+  };
+
+  // 24 bytes
+  union
+  {
+    struct
+    {
+      vec_8x8u r;
+      vec_8x8u g;
+      vec_8x8u b;
+    };
+
+    vec_8x16u pixels;
+  };
+
+  // 8 bytes
+  u32 draw_mask_bits;
+  u16 *fb_ptr;
+
+  // 16 bytes
+  vec_8x16u dither_offsets;  
+} block_struct;
+
+#define MAX_SPANS             512
+#define MAX_BLOCKS            64
+#define MAX_BLOCKS_PER_ROW    128
+
+#define SPAN_DATA_BLOCKS_SIZE 32
+
+typedef struct render_block_handler_struct render_block_handler_struct;
+
+typedef struct
+{
+  // 144 bytes
+  vec_8x16u test_mask;
+
+  vec_4x32u uvrg;
+  vec_4x32u uvrg_dx;
+  vec_4x32u uvrg_dy;
+
+  vec_4x32u u_block_span;
+  vec_4x32u v_block_span;
+  vec_4x32u r_block_span;
+  vec_4x32u g_block_span;
+  vec_4x32u b_block_span;
+
+  // 72 bytes
+  u32 b;
+  u32 b_dy;
+
+  u32 triangle_area;
+
+  u32 texture_window_settings;
+  u32 current_texture_mask;
+  u32 viewport_mask;
+  u32 dirty_textures_4bpp_mask;
+  u32 dirty_textures_8bpp_mask;
+  u32 dirty_textures_8bpp_alternate_mask;
+
+  u32 triangle_color;
+  u32 primitive_color;
+
+  u32 dither_table[4];
+
+  struct render_block_handler_struct *render_block_handler;
+  void *texture_page_ptr;
+  u16 *clut_ptr;
+  u16 *vram_ptr;
+
+  // 26 bytes
+  u16 render_state_base;
+  u16 render_state;
+
+  u16 num_spans;
+  u16 num_blocks;
+
+  s16 offset_x;
+  s16 offset_y;
+
+  u16 clut_settings;
+  u16 texture_settings;
+
+  s16 viewport_start_x;
+  s16 viewport_start_y;
+  s16 viewport_end_x;
+  s16 viewport_end_y;
+
+  u16 mask_msb;
+
+  // 8 bytes
+  u8 triangle_winding;
+
+  u8 display_area_draw_enable;
+
+  u8 current_texture_page;
+  u8 last_8bpp_texture_page;
+
+  u8 texture_mask_width;
+  u8 texture_mask_height;
+  u8 texture_window_x;
+  u8 texture_window_y;
+
+  u8 primitive_type;
+
+  // Align up to 64 byte boundary to keep the upcoming buffers cache line
+  // aligned
+  u8 reserved_a[1];
+
+  // 8KB
+  block_struct blocks[MAX_BLOCKS_PER_ROW];
+
+  // 14336 bytes
+  vec_4x32u span_uvrg_offset[MAX_SPANS];
+  edge_data_struct span_edge_data[MAX_SPANS];
+  u32 span_b_offset[MAX_SPANS];
+
+  u16 _vram[1024 * 512];
+  u8 texture_4bpp_cache[32][256 * 256];
+  u8 texture_8bpp_even_cache[16][256 * 256];
+  u8 texture_8bpp_odd_cache[16][256 * 256];
+
+  u32 pixel_count_mode;
+  u32 pixel_compare_mode;
+
+  u8 *vram_pixel_counts_a;
+  u8 *vram_pixel_counts_b;
+  u16 *compare_vram;
+} psx_gpu_struct;
+
+typedef struct __attribute__((aligned(16)))
+{
+  u8 u;
+  u8 v;
+
+  u8 r;
+  u8 g;
+  u8 b;
+
+  u8 reserved[3];
+
+  s16 x;
+  s16 y;
+} vertex_struct;
+
+void render_block_fill(psx_gpu_struct *psx_gpu, u32 color, u32 x, u32 y,
+ u32 width, u32 height);
+void render_block_copy(psx_gpu_struct *psx_gpu, u16 *source, u32 x, u32 y,
+ u32 width, u32 height, u32 pitch);
+void render_block_move(psx_gpu_struct *psx_gpu, u32 source_x, u32 source_y,
+ u32 dest_x, u32 dest_y, u32 width, u32 height);
+
+void render_triangle(psx_gpu_struct *psx_gpu, vertex_struct *vertexes,
+ u32 flags);
+void render_sprite(psx_gpu_struct *psx_gpu, s32 x, s32 y, u32 u, u32 v,
+ s32 width, s32 height, u32 flags, u32 color);
+void render_line(psx_gpu_struct *gpu, vertex_struct *vertexes, u32 flags,
+ u32 color);
+
+u32 texture_region_mask(s32 x1, s32 y1, s32 x2, s32 y2);
+
+void flush_render_block_buffer(psx_gpu_struct *psx_gpu);
+
+void initialize_psx_gpu(psx_gpu_struct *psx_gpu);
+void gpu_parse(psx_gpu_struct *psx_gpu, u32 *list, u32 size);
+
+void triangle_benchmark(psx_gpu_struct *psx_gpu);
+
+#endif
+
diff --git a/plugins/gpu_neon/psx_gpu/psx_gpu_arm_neon.S b/plugins/gpu_neon/psx_gpu/psx_gpu_arm_neon.S
new file mode 100644
index 0000000..381f3a9
--- /dev/null
+++ b/plugins/gpu_neon/psx_gpu/psx_gpu_arm_neon.S
@@ -0,0 +1,5438 @@
+/*
+ * Copyright (C) 2011 Gilead Kutnick "Exophase" <exophase@gmail.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; either version 2 of
+ * the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ */
+
+#define MAX_SPANS                                         512
+#define MAX_BLOCKS                                        64
+#define MAX_BLOCKS_PER_ROW                                128
+
+#define psx_gpu_test_mask_offset                          0
+#define psx_gpu_uvrg_offset                               16
+#define psx_gpu_uvrg_dx_offset                            32
+#define psx_gpu_uvrg_dy_offset                            48
+#define psx_gpu_u_block_span_offset                       64
+#define psx_gpu_v_block_span_offset                       80
+#define psx_gpu_r_block_span_offset                       96
+#define psx_gpu_g_block_span_offset                       112
+#define psx_gpu_b_block_span_offset                       128
+
+#define psx_gpu_b_dx_offset                               132
+
+#define psx_gpu_b_offset                                  144
+#define psx_gpu_b_dy_offset                               148
+#define psx_gpu_triangle_area_offset                      152
+#define psx_gpu_texture_window_settings_offset            156
+#define psx_gpu_current_texture_mask_offset               160
+#define psx_gpu_viewport_mask_offset                      164
+#define psx_gpu_dirty_textures_4bpp_mask_offset           168
+#define psx_gpu_dirty_textures_8bpp_mask_offset           172
+#define psx_gpu_dirty_textures_8bpp_alternate_mask_offset 176
+#define psx_gpu_triangle_color_offset                     180
+#define psx_gpu_primitive_color_offset                    184
+#define psx_gpu_dither_table_offset                       188
+#define psx_gpu_render_block_handler_offset               204
+#define psx_gpu_texture_page_ptr_offset                   208
+#define psx_gpu_clut_ptr_offset                           212
+#define psx_gpu_vram_ptr_offset                           216
+
+#define psx_gpu_render_state_base_offset                  220
+#define psx_gpu_render_state_offset                       222
+#define psx_gpu_num_spans_offset                          224
+#define psx_gpu_num_blocks_offset                         226
+#define psx_gpu_offset_x_offset                           228
+#define psx_gpu_offset_y_offset                           230
+#define psx_gpu_clut_settings_offset                      232
+#define psx_gpu_texture_settings_offset                   234
+#define psx_gpu_viewport_start_x_offset                   236
+#define psx_gpu_viewport_start_y_offset                   238
+#define psx_gpu_viewport_end_x_offset                     240
+#define psx_gpu_viewport_end_y_offset                     242
+#define psx_gpu_mask_msb_offset                           244
+                                                          
+#define psx_gpu_triangle_winding_offset                   246
+#define psx_gpu_display_area_draw_enable_offset           247
+#define psx_gpu_current_texture_page_offset               248
+#define psx_gpu_last_8bpp_texture_page_offset             249
+#define psx_gpu_texture_mask_width_offset                 250
+#define psx_gpu_texture_mask_height_offset                251
+#define psx_gpu_texture_window_x_offset                   252
+#define psx_gpu_texture_window_y_offset                   253
+#define psx_gpu_primitive_type_offset                     254
+
+#define psx_gpu_reserved_a_offset                         255
+
+#define psx_gpu_blocks_offset                             0x0100
+#define psx_gpu_span_uvrg_offset_offset                   0x2100
+#define psx_gpu_span_edge_data_offset                     0x4100
+#define psx_gpu_span_b_offset_offset                      0x5100
+
+#define psx_gpu__vram_offset                              0x005900
+
+#define edge_data_left_x_offset                           0
+#define edge_data_num_blocks_offset                       2
+#define edge_data_right_mask_offset                       4
+#define edge_data_y_offset                                6
+
+
+#define psx_gpu                                           r0
+#define v_a                                               r1
+#define v_b                                               r2
+#define v_c                                               r3
+
+#define x0                                                r4
+#define x1                                                r5
+#define x2                                                r6
+#define x0_x1                                             r5
+#define x1_x2                                             r6
+#define y0                                                r7
+#define y1                                                r8
+#define y2                                                r9
+#define y0_y1                                             r7
+#define y1_y2                                             r8
+#define b0                                                r9
+#define b1                                                r10
+#define b2                                                r11
+#define b0_b1                                             r10
+#define b1_b2                                             r11
+
+
+#define area_r_s                                          r5
+
+#define g_bx0                                             r2
+#define g_bx                                              r3
+#define g_bx2                                             r4
+#define g_bx3                                             r5
+#define b_base                                            r6
+#define g_by                                              r8
+
+#define gs_bx                                             r7
+#define gs_by                                             r10
+
+#define ga_bx                                             g_bx
+#define ga_by                                             g_by
+
+#define gw_bx_h                                           g_bx
+#define gw_by_h                                           g_by
+
+#define gw_bx_l                                           r11
+#define gw_by_l                                           gw_bx_l
+
+#define store_a                                           r0
+#define store_b                                           r1
+#define store_inc                                         r5
+
+
+#define v0                                                q0
+#define uvrgb0                                            d0
+#define x0_y0                                             d1
+
+#define v1                                                q1
+#define uvrgb1                                            d2
+#define x1_y1                                             d3
+
+#define v2                                                q2
+#define uvrgb2                                            d4
+#define x2_y2                                             d5
+
+#define x0_ab                                             q3
+#define uvrg_xxxx0                                        q3
+#define uvrg0                                             d6
+#define xxxx0                                             d7
+
+#define x1_ab                                             q4
+#define uvrg_xxxx1                                        q4
+#define uvrg1                                             d8
+#define xxxx1                                             d9
+
+#define x2_ab                                             q5
+#define uvrg_xxxx2                                        q5
+#define uvrg2                                             d10
+#define xxxx2                                             d11
+
+#define y0_ab                                             q6
+#define yyyy_uvrg0                                        q6
+#define yyyy0                                             d12
+#define uvrg0b                                            d13
+
+#define y1_ab                                             q7
+#define yyyy_uvrg1                                        q7
+#define yyyy1                                             d14
+#define uvrg1b                                            d15
+
+#define y2_ab                                             q8
+#define yyyy_uvrg2                                        q8
+#define yyyy2                                             d16
+#define uvrg2b                                            d17
+
+#define d0_ab                                             q9
+#define d0_a                                              d18
+#define d0_b                                              d19
+
+#define d1_ab                                             q10
+#define d1_a                                              d20
+#define d1_b                                              d21
+
+#define d2_ab                                             q11
+#define d2_a                                              d22
+#define d2_b                                              d23
+
+#define d3_ab                                             q12
+#define d3_a                                              d24
+#define d3_b                                              d25
+
+#define ga_uvrg_x                                         q1
+#define ga_uvrg_y                                         q4
+
+#define dx                                                x0_x1
+#define dy                                                y0_y1
+#define db                                                b0_b1
+
+#define uvrg_base                                         q11
+
+#define gs_uvrg_x                                         q5
+#define gs_uvrg_y                                         q6
+
+#define g_uvrg_x                                          q1
+#define ga_uv_x                                           d2
+#define g_uv_x                                            d2
+#define ga_rg_x                                           d3
+#define g_rg_x                                            d3
+
+#define g_uvrg_y                                          q4
+#define ga_uv_y                                           d8
+#define g_uv_y                                            d8
+#define ga_rg_y                                           d9
+#define g_rg_y                                            d9
+
+#define gw_uv_x                                           q1
+#define gw_rg_x                                           q2
+#define gw_uv_y                                           q4
+#define gw_rg_y                                           q3
+
+#define w_mask                                            q9
+#define w_mask_l                                          d18
+
+#define r_shift                                           q10
+
+#define uvrg_dx0                                          q0
+#define uvrg_dx0l                                         d0
+#define uvrg_dx0h                                         d1
+
+#define uvrg_dx1                                          q1
+#define uvrg_dx1l                                         d2
+#define uvrg_dx1h                                         d3
+
+#define uvrg_dx2                                          q2
+#define uvrg_dx2l                                         d4
+#define uvrg_dx2h                                         d5
+
+#define uvrg_dx3                                          q3
+#define uvrg_dx3l                                         d6
+#define uvrg_dx3h                                         d7
+
+
+.align 4
+
+#define function(name)                                                         \
+  .global name;                                                                \
+  name:                                                                        \
+
+@ r0: psx_gpu
+@ r1: v_a
+@ r2: v_b
+@ r3: v_c
+
+function(compute_all_gradients)
+  // First compute the triangle area reciprocal and shift. The division will
+  // happen concurrently with much of the work which follows.
+  @ r12 = psx_gpu->triangle_area
+  ldr r12, [ psx_gpu, #psx_gpu_triangle_area_offset ]
+  stmdb sp!, { r4 - r11, lr }
+
+  @ load exponent of 62 into upper half of double
+  movw r4, #0
+  clz r14, r12                       @ r14 = shift
+
+  movt r4, #((62 + 1023) << 4)
+  mov r12, r12, lsl r14              @ r12 = triangle_area_normalized
+
+  @ load area normalized into lower half of double
+  mov r5, r12, lsr #10
+  vmov.f64 d30, r5, r4               @ d30 = (1 << 62) + ta_n
+
+  movt r4, #((1022 + 31) << 4)
+  mov r5, r12, lsl #20
+
+  add r4, r4, r12, lsr #11
+  vmov.f64 d31, r5, r4
+
+  vdiv.f64 d30, d30, d31             @ d30 = ((1 << 62) + ta_n) / ta_n
+
+  // ((x1 - x0) * (y2 - y1)) - ((x2 - x1) * (y1 - y0)) =
+  // ( d0       *  d1      ) - ( d2       *  d3      ) =
+  // ( m0                  ) - ( m1                  ) = gradient
+
+  // This is split to do 12 elements at a time over three sets: a, b, and c.
+  // Technically we only need to do 10 elements (uvrgb_x and uvrgb_y), so
+  // two of the slots are unused.
+
+  // Inputs are all 16-bit signed. The m0/m1 results are 32-bit signed, as
+  // is g.
+
+  // First type is:  uvrg bxxx xxxx 
+  // Second type is: yyyy ybyy uvrg 
+  // Since x_a and y_c are the same the same variable is used for both. 
+
+  vld1.u32 { v0 }, [ v_a, : 128 ]    @ v0 = { uvrg0, b0, x0, y0 }
+  ldrsh x0, [ v_a, #8 ]              @ load x0
+
+  vld1.u32 { v1 }, [ v_b, : 128 ]    @ v1 = { uvrg1, b1, x1, y1}
+  ldrh x1, [ v_b, #8 ]               @ load x1
+
+  vld1.u32 { v2 }, [ v_c, : 128 ]    @ v2 = { uvrg2, b2, x2, y2 }
+  ldrh x2, [ v_c, #8 ]               @ load x2
+
+  vmovl.u8 uvrg_xxxx0, uvrgb0        @ uvrg_xxxx0 = { uv0, rg0, b0-, -- }
+  ldrh y0, [ v_a, #10 ]              @ load y0
+
+  vmovl.u8 uvrg_xxxx1, uvrgb1        @ uvrg_xxxx1 = { uv1, rg1, b1-, -- }
+  ldrh y1, [ v_b, #10 ]              @ load y1
+
+  vmovl.u8 uvrg_xxxx2, uvrgb2        @ uvrg_xxxx2 = { uv2, rg2, b2-, -- }
+  ldrh y2, [ v_c, #10 ]              @ load y2
+
+  vmov.u8 uvrg0b, uvrg0              @ uvrg0b = { uv0, rg0 }
+  vdup.u16 xxxx0, x0_y0[0]           @ xxxx0 = { xx0, xx0 }
+
+  orr x1_x2, x1, x2, lsl #16         @ x1_x2 = { x1, x2 }
+  pkhbt x0_x1, x0, x1, lsl #16       @ x0_x1 = { x0, x1 }
+
+  vmov.u8 uvrg1b, uvrg1              @ uvrg1b = { uv1, rg1 }
+  vdup.u16 xxxx1, x1_y1[0]           @ xxxx1 = { xx1, xx1 }
+
+  vmov.u8 uvrg2b, uvrg2              @ uvrg2b = { uv2, rg2 }
+  vdup.u16 xxxx2, x2_y2[0]           @ xxxx2 = { xx2, xx2 }
+
+  ldrb b2, [ v_c, #4 ]               @ load b2
+  orr y0_y1, y0, y1, lsl #16         @ y0_y1 = { y0, y1 }
+
+  ldrb b1, [ v_b, #4 ]               @ load b1
+  orr y1_y2, y1, y2, lsl #16         @ y1_y2 = { y1, y2 }
+
+  vdup.u16 yyyy0, x0_y0[1]           @ yyyy0 = { yy0, yy0 }
+  vsub.s16 d0_ab, x1_ab, x0_ab
+
+  ldrb b0, [ v_a, #4 ]               @ load b0
+  orr b1_b2, b1, b2, lsl #16         @ b1_b2 = { b1, b2 }
+
+  vdup.u16 yyyy1, x1_y1[1]           @ yyyy1 = { yy1, yy1 }
+  vsub.s16 d2_ab, x2_ab, x1_ab
+
+  vdup.u16 yyyy2, x2_y2[1]           @ yyyy2 = { yy2, yy2 }
+  vsub.s16 d1_ab, y2_ab, y1_ab
+
+  orr b0_b1, b0, b1, lsl #16         @ b1_b2 = { b1, b2 }
+  ssub16 dx, x1_x2, x0_x1            @ dx = { x1 - x0, x2 - x1 }
+
+  ssub16 dy, y1_y2, y0_y1            @ dy = { y1 - y0, y2 - y1 }
+  ssub16 db, b1_b2, b0_b1            @ db = { b1 - b0, b2 - b1 }
+
+  vsub.s16 d3_ab, y1_ab, y0_ab
+  smusdx ga_by, dx, db               @ ga_by = ((x1 - x0) * (b2 - b1)) -
+                                     @         ((x2 - X1) * (b1 - b0)) 
+  vmull.s16 ga_uvrg_x, d0_a, d1_a
+  smusdx ga_bx, db, dy               @ ga_bx = ((b1 - b0) * (y2 - y1)) -
+                                     @         ((b2 - b1) * (y1 - y0))
+  vmlsl.s16 ga_uvrg_x, d2_a, d3_a
+  movs gs_bx, ga_bx, asr #31
+
+  vmull.s16 ga_uvrg_y, d0_b, d1_b
+  rsbmi ga_bx, ga_bx, #0
+
+  vmlsl.s16 ga_uvrg_y, d2_b, d3_b
+  movs gs_by, ga_by, asr #31
+
+  vshr.u64 d0, d30, #22
+  mov b_base, b0, lsl #16
+
+  rsbmi ga_by, ga_by, #0
+  vclt.s32 gs_uvrg_x, ga_uvrg_x, #0  @ gs_uvrg_x = ga_uvrg_x < 0
+
+  @ r12 = psx_gpu->triangle_winding_offset
+  ldrb r12, [ psx_gpu, #psx_gpu_triangle_winding_offset ]
+  vclt.s32 gs_uvrg_y, ga_uvrg_y, #0  @ gs_uvrg_y = ga_uvrg_y < 0
+
+  add b_base, b_base, #0x8000
+  rsb r12, r12, #0                   @ r12 = -(triangle->winding)
+
+  vdup.u32 w_mask, r12               @ w_mask = { -w, -w, -w, -w }
+  sub r14, r14, #(62 - 12)           @ r14 = shift - (62 - FIXED_BITS)
+
+  vshll.u16 uvrg_base, uvrg0, #16    @ uvrg_base = uvrg0 << 16
+  vdup.u32 r_shift, r14              @ r_shift = { shift, shift, shift, shift }
+
+  vorr.u32 uvrg_base, #0x8000
+  vabs.s32 ga_uvrg_x, ga_uvrg_x      @ ga_uvrg_x = abs(ga_uvrg_x)
+
+  vmov area_r_s, s0                  @ area_r_s = triangle_reciprocal
+  vabs.s32 ga_uvrg_y, ga_uvrg_y      @ ga_uvrg_y = abs(ga_uvrg_y)
+
+  vmull.u32 gw_rg_x, ga_rg_x, d0[0]
+  vmull.u32 gw_uv_x, ga_uv_x, d0[0]
+  vmull.u32 gw_rg_y, ga_rg_y, d0[0]
+  vmull.u32 gw_uv_y, ga_uv_y, d0[0]
+
+  vshl.u64 gw_rg_x, gw_rg_x, r_shift
+  vshl.u64 gw_uv_x, gw_uv_x, r_shift
+  vshl.u64 gw_rg_y, gw_rg_y, r_shift
+  vshl.u64 gw_uv_y, gw_uv_y, r_shift
+
+  veor.u32 gs_uvrg_x, gs_uvrg_x, w_mask
+  vmovn.u64 g_uv_x, gw_uv_x
+
+  veor.u32 gs_uvrg_y, gs_uvrg_y, w_mask
+  vmovn.u64 g_rg_x, gw_rg_x
+
+  veor.u32 g_uvrg_x, g_uvrg_x, gs_uvrg_x
+  vmovn.u64 g_uv_y, gw_uv_y
+
+  vsub.u32 g_uvrg_x, g_uvrg_x, gs_uvrg_x
+  vmovn.u64 g_rg_y, gw_rg_y
+
+  veor.u32 g_uvrg_y, g_uvrg_y, gs_uvrg_y
+  mov ga_bx, ga_bx, lsl #13
+
+  vsub.u32 g_uvrg_y, g_uvrg_y, gs_uvrg_y
+  mov ga_by, ga_by, lsl #13
+
+  vdup.u32 x0_y0, x0
+  umull gw_bx_l, gw_bx_h, ga_bx, area_r_s
+
+  vshl.u32 g_uvrg_x, g_uvrg_x, #4
+  vshl.u32 g_uvrg_y, g_uvrg_y, #4
+
+  umull gw_by_l, gw_by_h, ga_by, area_r_s
+  vmls.s32 uvrg_base, ga_uvrg_x, x0_y0[0]
+
+  eor gs_bx, gs_bx, r12
+  vadd.u32 uvrg_dx2, uvrg_dx1, uvrg_dx1
+
+  veor.u32 uvrg_dx0, uvrg_dx0, uvrg_dx0
+  eor gs_by, gs_by, r12
+
+  rsb r11, r14, #0                   @ r11 = negative shift for scalar lsr
+  add store_a, psx_gpu, #psx_gpu_uvrg_offset
+
+  sub r11, r11, #(32 - 13)
+
+  add store_b, store_a, #16
+  mov store_inc, #32
+
+  vadd.u32 uvrg_dx3, uvrg_dx2, uvrg_dx1
+  vst1.u32 { uvrg_base }, [ store_a, : 128 ], store_inc
+
+  vst1.u32 { uvrg_dx1 }, [ store_b, : 128 ], store_inc
+  mov g_bx, gw_bx_h, lsr r11
+
+  vst1.u32 { g_uvrg_y }, [ store_a, : 128 ], store_inc
+  mov g_by, gw_by_h, lsr r11
+
+  vst4.u32 { uvrg_dx0l, uvrg_dx1l, uvrg_dx2l, uvrg_dx3l },                     \
+   [ store_b, : 128 ], store_inc
+  eor g_bx, g_bx, gs_bx
+
+  vst4.u32 { uvrg_dx0h, uvrg_dx1h, uvrg_dx2h, uvrg_dx3h },                     \
+   [ store_b, : 128 ], store_inc
+  sub g_bx, g_bx, gs_bx
+
+  lsl g_bx, g_bx, #4  
+  eor g_by, g_by, gs_by
+
+  mls b_base, g_bx, x0, b_base
+  sub g_by, g_by, gs_by
+
+  lsl g_by, g_by, #4
+  mov g_bx0, #0
+
+  add g_bx2, g_bx, g_bx
+  add g_bx3, g_bx, g_bx2
+
+  stmia store_b, { g_bx0, g_bx, g_bx2, g_bx3, b_base, g_by }
+
+  ldmia sp!, { r4 - r11, pc }
+
+
+#define psx_gpu                                  r0
+#define v_a                                      r1
+#define v_b                                      r2
+#define v_c                                      r3
+
+#define temp                                     r14
+
+#define x_a                                      r4
+#define x_b                                      r5
+#define x_c                                      r6
+#define y_a                                      r1
+#define y_b                                      r2
+#define y_c                                      r3
+
+#define height_minor_a                           r7
+#define height_minor_b                           r8
+#define height_major                             r9
+#define height                                   r9
+
+#define reciprocal_table_ptr                     r10
+
+#define edge_alt_low                             r4
+#define edge_alt_high                            r5
+#define edge_dx_dy_alt                           r6
+#define edge_shift_alt                           r10
+
+#define edge_dx_dy_alt_low                       r4
+#define edge_dx_dy_alt_high                      r5
+
+#define span_edge_data                           r4
+#define span_uvrg_offset                         r5
+#define span_b_offset                            r6
+
+#define clip                                     r14
+
+#define b                                        r11
+#define b_dy                                     r12
+
+
+#define alternate_x                              q0
+#define alternate_dx_dy                          q1
+#define alternate_x_32                           q2
+
+#define alternate_x_low                          d0
+#define alternate_x_high                         d1
+#define alternate_dx_dy_low                      d2
+#define alternate_dx_dy_high                     d3
+#define alternate_x_32_low                       d4
+#define alternate_x_32_high                      d5
+
+#define left_x                                   q3
+#define right_x                                  q4
+#define left_dx_dy                               q5
+#define right_dx_dy                              q6
+#define left_edge                                q7
+#define right_edge                               q8
+
+#define left_x_low                               d6
+#define left_x_high                              d7
+#define right_x_low                              d8
+#define right_x_high                             d9
+#define left_dx_dy_low                           d10
+#define left_dx_dy_high                          d11
+#define right_dx_dy_low                          d12
+#define right_dx_dy_high                         d13
+#define left_edge_low                            d14
+#define left_edge_high                           d15
+#define right_edge_low                           d16
+#define right_edge_high                          d17
+
+#define y_mid_point                              d18
+#define c_0x0004                                 d19
+
+#define left_right_x_16                          q11
+#define span_shifts_y                            q12
+#define c_0x0001                                 q13
+
+#define span_shifts                              d24
+#define y_x4                                     d25
+#define c_0xFFFE                                 d26
+#define c_0x0007                                 d27
+
+#define left_right_x_16_low                      d22
+#define left_right_x_16_high                     d23
+
+#define uvrg                                     q14
+#define uvrg_dy                                  q15
+
+#define alternate_x_16                           d4
+
+#define v_clip                                   q3
+#define v_clip_low                               d6
+
+#define right_x_32                               q10
+#define left_x_32                                q11
+#define alternate_select                         d24
+
+#define right_x_32_low                           d20
+#define right_x_32_high                          d21
+#define left_x_32_low                            d22
+#define left_x_32_high                           d23
+
+#define edges_xy                                 q0
+#define edges_dx_dy                              d2
+#define edge_shifts                              d3
+#define edge_shifts_64                           q2
+
+#define edges_xy_left                            d0
+#define edges_xy_right                           d1
+
+#define height_reciprocals                       d6
+#define heights                                  d7
+
+#define widths                                   d8
+#define c_0x01                                   d9
+#define x_starts                                 d10
+#define x_ends                                   d11
+
+#define heights_b                                d12
+#define edges_dx_dy_64                           q10
+
+#define edges_dx_dy_64_left                      d20
+#define edges_dx_dy_64_right                     d21
+
+
+#define setup_spans_prologue()                                                 \
+  stmdb sp!, { r4 - r11, lr };                                                 \
+                                                                               \
+  ldrsh x_a, [ v_a, #8 ];                                                      \
+  ldrsh x_b, [ v_b, #8 ];                                                      \
+  ldrsh x_c, [ v_c, #8 ];                                                      \
+  ldrsh y_a, [ v_a, #10 ];                                                     \
+  ldrsh y_b, [ v_b, #10 ];                                                     \
+  ldrsh y_c, [ v_c, #10 ];                                                     \
+                                                                               \
+  add temp, psx_gpu, #psx_gpu_uvrg_offset;                                     \
+  vld1.32 { uvrg }, [ temp ];                                                  \
+  add temp, psx_gpu, #psx_gpu_uvrg_dy_offset;                                  \
+  vld1.32 { uvrg_dy }, [ temp ];                                               \
+  movw reciprocal_table_ptr, :lower16:reciprocal_table;                        \
+  movt reciprocal_table_ptr, :upper16:reciprocal_table;                        \
+                                                                               \
+  vmov.u32 c_0x01, #0x01                                                       \
+
+#define setup_spans_load_b()                                                   \
+  ldr b, [ psx_gpu, #psx_gpu_b_offset ];                                       \
+  ldr b_dy, [ psx_gpu, #psx_gpu_b_dy_offset ]                                  \
+
+#define setup_spans_prologue_b()                                               \
+  add span_uvrg_offset, psx_gpu, #psx_gpu_span_uvrg_offset_offset;             \
+  add temp, psx_gpu, #psx_gpu_viewport_start_x_offset;                         \
+                                                                               \
+  add span_edge_data, psx_gpu, #psx_gpu_span_edge_data_offset;                 \
+  vmov.u16 c_0x0004, #0x0004;                                                  \
+                                                                               \
+  add span_b_offset, psx_gpu, #psx_gpu_span_b_offset_offset;                   \
+  vmov.u16 c_0x0001, #0x0001;                                                  \
+                                                                               \
+  vld1.u16 { left_edge_low[], left_edge_high[] }, [ temp ];                    \
+  add temp, psx_gpu, #psx_gpu_viewport_end_x_offset;                           \
+                                                                               \
+  vld1.u16 { right_edge_low[], right_edge_high[] }, [ temp ];                  \
+  vadd.u16 right_edge, right_edge, c_0x0001;                                   \
+                                                                               \
+  vmov.u16 c_0x0007, #0x0007;                                                  \
+  vmvn.u16 c_0xFFFE, #0x0001                                                   \
+
+
+#define compute_edge_delta_x2()                                                \
+  ldr temp, [ reciprocal_table_ptr, height, lsl #2 ];                          \
+                                                                               \
+  vdup.u32 heights, height;                                                    \
+  vsub.u32 widths, x_ends, x_starts;                                           \
+                                                                               \
+  vdup.u32 edge_shifts, temp;                                                  \
+  vsub.u32 heights_b, heights, c_0x01;                                         \
+  vshr.u32 height_reciprocals, edge_shifts, #12;                               \
+                                                                               \
+  vmla.s32 heights_b, x_starts, heights;                                       \
+  vbic.u16 edge_shifts, #0xE0;                                                 \
+  vmul.s32 edges_dx_dy, widths, height_reciprocals;                            \
+  vmull.s32 edges_xy, heights_b, height_reciprocals                            \
+
+#define width_alt                 r6
+#define height_reciprocal_alt     r11
+#define height_b_alt              r12
+
+#define compute_edge_delta_x3(start_c, height_a, height_b)                     \
+  vmov.u32 heights, height_a, height_b;                                        \
+  ldr temp, [ reciprocal_table_ptr, height_a, lsl #2 ];                        \
+  vmov.u32 edge_shifts[0], temp;                                               \
+  ldr temp, [ reciprocal_table_ptr, height_b, lsl #2 ];                        \
+  vmov.u32 edge_shifts[1], temp;                                               \
+  ldr edge_shift_alt, [ reciprocal_table_ptr, height_minor_b, lsl #2 ];        \
+                                                                               \
+  vsub.u32 widths, x_ends, x_starts;                                           \
+  sub width_alt, x_c, start_c;                                                 \
+                                                                               \
+  vsub.u32 heights_b, heights, c_0x01;                                         \
+  sub height_b_alt, height_minor_b, #1;                                        \
+                                                                               \
+  vshr.u32 height_reciprocals, edge_shifts, #12;                               \
+  lsr height_reciprocal_alt, edge_shift_alt, #12;                              \
+                                                                               \
+  vmla.s32 heights_b, x_starts, heights;                                       \
+  mla height_b_alt, height_minor_b, start_c, height_b_alt;                     \
+                                                                               \
+  vbic.u16 edge_shifts, #0xE0;                                                 \
+  and edge_shift_alt, edge_shift_alt, #0x1F;                                   \
+                                                                               \
+  vmul.s32 edges_dx_dy, widths, height_reciprocals;                            \
+  mul edge_dx_dy_alt, width_alt, height_reciprocal_alt;                        \
+                                                                               \
+  vmull.s32 edges_xy, heights_b, height_reciprocals;                           \
+  smull edge_alt_low, edge_alt_high, height_b_alt, height_reciprocal_alt       \
+
+
+#define setup_spans_adjust_y_up()                                              \
+  vsub.u32 y_x4, y_x4, c_0x0004                                                \
+
+#define setup_spans_adjust_y_down()                                            \
+  vadd.u32 y_x4, y_x4, c_0x0004                                                \
+
+#define setup_spans_adjust_interpolants_up()                                   \
+  vsub.u32 uvrg, uvrg, uvrg_dy;                                                \
+  sub b, b, b_dy                                                               \
+
+#define setup_spans_adjust_interpolants_down()                                 \
+  vadd.u32 uvrg, uvrg, uvrg_dy;                                                \
+  add b, b, b_dy                                                               \
+
+
+#define setup_spans_clip_interpolants_increment()                              \
+  mla b, b_dy, clip, b;                                                        \
+  vmla.s32 uvrg, uvrg_dy, v_clip                                               \
+
+#define setup_spans_clip_interpolants_decrement()                              \
+  mls b, b_dy, clip, b;                                                        \
+  vmls.s32 uvrg, uvrg_dy, v_clip                                               \
+
+#define setup_spans_clip_alternate_yes()                                       \
+  smlal edge_alt_low, edge_alt_high, edge_dx_dy_alt, clip                      \
+
+#define setup_spans_clip_alternate_no()                                        \
+
+#define setup_spans_clip(direction, alternate_active)                          \
+  vdup.u32 v_clip, clip;                                                       \
+  setup_spans_clip_alternate_##alternate_active();                             \
+  setup_spans_clip_interpolants_##direction();                                 \
+  vmlal.s32 edges_xy, edges_dx_dy, v_clip_low                                  \
+
+
+#define setup_spans_adjust_edges_alternate_no(left_index, right_index)         \
+  vmovl.s32 edge_shifts_64, edge_shifts;                                       \
+  vmovl.s32 edges_dx_dy_64, edges_dx_dy;                                       \
+                                                                               \
+  vshl.s64 edges_xy, edges_xy, edge_shifts_64;                                 \
+  vshl.s64 edges_dx_dy_64, edges_dx_dy_64, edge_shifts_64;                     \
+                                                                               \
+  vmov left_x_low, edges_xy_##left_index;                                      \
+  vmov right_x_low, edges_xy_##right_index;                                    \
+                                                                               \
+  vmov left_dx_dy_low, edges_dx_dy_64_##left_index;                            \
+  vmov left_dx_dy_high, edges_dx_dy_64_##left_index;                           \
+  vmov right_dx_dy_low, edges_dx_dy_64_##right_index;                          \
+  vmov right_dx_dy_high, edges_dx_dy_64_##right_index;                         \
+                                                                               \
+  vadd.u64 left_x_high, left_x_low, left_dx_dy_low;                            \
+  vadd.u64 right_x_high, right_x_low, right_dx_dy_low;                         \
+                                                                               \
+  vadd.u64 left_dx_dy, left_dx_dy, left_dx_dy;                                 \
+  vadd.u64 right_dx_dy, right_dx_dy, right_dx_dy                               \
+
+
+#define setup_spans_adjust_edges_alternate_yes(left_index, right_index)        \
+  setup_spans_adjust_edges_alternate_no(left_index, right_index);              \
+                                                                               \
+  vdup.u16 y_mid_point, y_b;                                                   \
+  rsb temp, edge_shift_alt, #32;                                               \
+                                                                               \
+  lsl edge_alt_high, edge_alt_high, edge_shift_alt;                            \
+  orr edge_alt_high, edge_alt_high, edge_alt_low, lsr temp;                    \
+  lsl edge_alt_low, edge_alt_low, edge_shift_alt;                              \
+  vmov alternate_x_low, edge_alt_low, edge_alt_high;                           \
+                                                                               \
+  asr edge_dx_dy_alt_high, edge_dx_dy_alt, temp;                               \
+  lsl edge_dx_dy_alt_low, edge_dx_dy_alt, edge_shift_alt;                      \
+  vmov alternate_dx_dy_low, edge_dx_dy_alt_low, edge_dx_dy_alt_high;           \
+  vmov alternate_dx_dy_high, alternate_dx_dy_low;                              \
+                                                                               \
+  vadd.u64 alternate_x_high, alternate_x_low, alternate_dx_dy_low;             \
+  vadd.u64 alternate_dx_dy, alternate_dx_dy, alternate_dx_dy                   \
+
+
+#define setup_spans_y_select_up()                                              \
+  vclt.s16 alternate_select, y_x4, y_mid_point                                 \
+
+#define setup_spans_y_select_down()                                            \
+  vcgt.s16 alternate_select, y_x4, y_mid_point                                 \
+
+
+#define setup_spans_alternate_select_left()                                    \
+  vbit.u16 left_right_x_16_low, alternate_x_16, alternate_select               \
+
+#define setup_spans_alternate_select_right()                                   \
+  vbit.u16 left_right_x_16_high, alternate_x_16, alternate_select              \
+
+
+#define setup_spans_set_x4_alternate_yes(alternate, direction)                 \
+  vshrn.s64 alternate_x_32_low, alternate_x, #32;                              \
+  vshrn.s64 left_x_32_low, left_x, #32;                                        \
+  vshrn.s64 right_x_32_low, right_x, #32;                                      \
+                                                                               \
+  vadd.u64 alternate_x, alternate_x, alternate_dx_dy;                          \
+  vadd.u64 left_x, left_x, left_dx_dy;                                         \
+  vadd.u64 right_x, right_x, right_dx_dy;                                      \
+                                                                               \
+  vshrn.s64 alternate_x_32_high, alternate_x, #32;                             \
+  vshrn.s64 left_x_32_high, left_x, #32;                                       \
+  vshrn.s64 right_x_32_high, right_x, #32;                                     \
+                                                                               \
+  vadd.u64 alternate_x, alternate_x, alternate_dx_dy;                          \
+  vadd.u64 left_x, left_x, left_dx_dy;                                         \
+  vadd.u64 right_x, right_x, right_dx_dy;                                      \
+                                                                               \
+  vmovn.u32 alternate_x_16, alternate_x_32;                                    \
+  setup_spans_y_select_##direction();                                          \
+  vmovn.u32 left_right_x_16_low, left_x_32;                                    \
+                                                                               \
+  vmovn.u32 left_right_x_16_high, right_x_32;                                  \
+  setup_spans_alternate_select_##alternate();                                  \
+                                                                               \
+  vst1.u32 { uvrg }, [ span_uvrg_offset, :128 ]!;                              \
+  str b, [ span_b_offset ], #4;                                                \
+  setup_spans_adjust_interpolants_##direction();                               \
+                                                                               \
+  vmax.s16 left_right_x_16, left_right_x_16, left_edge;                        \
+                                                                               \
+  vst1.u32 { uvrg }, [ span_uvrg_offset, :128 ]!;                              \
+  str b, [ span_b_offset ], #4;                                                \
+  setup_spans_adjust_interpolants_##direction();                               \
+                                                                               \
+  vmin.s16 left_right_x_16, left_right_x_16, right_edge;                       \
+                                                                               \
+  vst1.u32 { uvrg }, [ span_uvrg_offset, :128 ]!;                              \
+  str b, [ span_b_offset ], #4;                                                \
+  setup_spans_adjust_interpolants_##direction();                               \
+                                                                               \
+  vsub.u16 left_right_x_16_high, left_right_x_16_high, left_right_x_16_low;    \
+  vadd.u16 left_right_x_16_high, left_right_x_16_high, c_0x0007;               \
+  vand.u16 span_shifts, left_right_x_16_high, c_0x0007;                        \
+                                                                               \
+  vst1.u32 { uvrg }, [ span_uvrg_offset, :128 ]!;                              \
+  str b, [ span_b_offset ], #4;                                                \
+  setup_spans_adjust_interpolants_##direction();                               \
+                                                                               \
+  vshr.u16 left_right_x_16_high, left_right_x_16_high, #3;                     \
+  vshl.u16 span_shifts, c_0xFFFE, span_shifts;                                 \
+                                                                               \
+  vst4.u16 { left_right_x_16, span_shifts_y }, [ span_edge_data ]!;            \
+                                                                               \
+  setup_spans_adjust_y_##direction()                                           \
+
+
+#define setup_spans_set_x4_alternate_no(alternate, direction)                  \
+  vshrn.s64 left_x_32_low, left_x, #32;                                        \
+  vshrn.s64 right_x_32_low, right_x, #32;                                      \
+                                                                               \
+  vadd.u64 left_x, left_x, left_dx_dy;                                         \
+  vadd.u64 right_x, right_x, right_dx_dy;                                      \
+                                                                               \
+  vshrn.s64 left_x_32_high, left_x, #32;                                       \
+  vshrn.s64 right_x_32_high, right_x, #32;                                     \
+                                                                               \
+  vadd.u64 left_x, left_x, left_dx_dy;                                         \
+  vadd.u64 right_x, right_x, right_dx_dy;                                      \
+                                                                               \
+  vmovn.u32 left_right_x_16_low, left_x_32;                                    \
+  vmovn.u32 left_right_x_16_high, right_x_32;                                  \
+                                                                               \
+  vst1.u32 { uvrg }, [ span_uvrg_offset, :128 ]!;                              \
+  str b, [ span_b_offset ], #4;                                                \
+  setup_spans_adjust_interpolants_##direction();                               \
+                                                                               \
+  vmax.s16 left_right_x_16, left_right_x_16, left_edge;                        \
+                                                                               \
+  vst1.u32 { uvrg }, [ span_uvrg_offset, :128 ]!;                              \
+  str b, [ span_b_offset ], #4;                                                \
+  setup_spans_adjust_interpolants_##direction();                               \
+                                                                               \
+  vmin.s16 left_right_x_16, left_right_x_16, right_edge;                       \
+                                                                               \
+  vst1.u32 { uvrg }, [ span_uvrg_offset, :128 ]!;                              \
+  str b, [ span_b_offset ], #4;                                                \
+  setup_spans_adjust_interpolants_##direction();                               \
+                                                                               \
+  vsub.u16 left_right_x_16_high, left_right_x_16_high, left_right_x_16_low;    \
+  vadd.u16 left_right_x_16_high, left_right_x_16_high, c_0x0007;               \
+  vand.u16 span_shifts, left_right_x_16_high, c_0x0007;                        \
+                                                                               \
+  vst1.u32 { uvrg }, [ span_uvrg_offset, :128 ]!;                              \
+  str b, [ span_b_offset ], #4;                                                \
+  setup_spans_adjust_interpolants_##direction();                               \
+                                                                               \
+  vshl.u16 span_shifts, c_0xFFFE, span_shifts;                                 \
+  vshr.u16 left_right_x_16_high, left_right_x_16_high, #3;                     \
+                                                                               \
+  vst4.u16 { left_right_x_16, span_shifts_y }, [ span_edge_data ]!;            \
+                                                                               \
+  setup_spans_adjust_y_##direction()                                           \
+
+
+#define edge_adjust_low           r11
+#define edge_adjust_high          r12
+
+#define setup_spans_alternate_adjust_yes()                                     \
+  smull edge_adjust_low, edge_adjust_high, edge_dx_dy_alt, height_minor_a;     \
+  subs edge_alt_low, edge_alt_low, edge_adjust_low;                            \
+  sbc edge_alt_high, edge_alt_high, edge_adjust_high                           \
+
+#define setup_spans_alternate_adjust_no()                                      \
+
+
+#define setup_spans_down(left_index, right_index, alternate, alternate_active) \
+  setup_spans_alternate_adjust_##alternate_active();                           \
+  setup_spans_load_b();                                                        \
+                                                                               \
+  ldrsh temp, [ psx_gpu, #psx_gpu_viewport_end_y_offset ];                     \
+  subs y_c, y_c, temp;                                                         \
+  subgt height, height, y_c;                                                   \
+  addgt height, height, #1;                                                    \
+                                                                               \
+  ldrsh temp, [ psx_gpu, #psx_gpu_viewport_start_y_offset ];                   \
+  subs clip, temp, y_a;                                                        \
+  ble 0f;                                                                      \
+                                                                               \
+  sub height, height, clip;                                                    \
+  add y_a, y_a, clip;                                                          \
+  setup_spans_clip(increment, alternate_active);                               \
+                                                                               \
+ 0:                                                                            \
+  cmp height, #0;                                                              \
+  ble 1f;                                                                      \
+                                                                               \
+  orr temp, y_a, y_a, lsl #16;                                                 \
+  add temp, temp, #(1 << 16);                                                  \
+  add y_a, temp, #2;                                                           \
+  add y_a, y_a, #(2 << 16);                                                    \
+  vmov.u32 y_x4, temp, y_a;                                                    \
+                                                                               \
+  setup_spans_adjust_edges_alternate_##alternate_active(left_index,            \
+   right_index);                                                               \
+  setup_spans_prologue_b();                                                    \
+                                                                               \
+  strh height, [ psx_gpu, #psx_gpu_num_spans_offset ];                         \
+                                                                               \
+ 2:                                                                            \
+  setup_spans_set_x4_alternate_##alternate_active(alternate, down);            \
+  subs height, height, #4;                                                     \
+  bhi 2b;                                                                      \
+                                                                               \
+ 1:                                                                            \
+
+
+#define setup_spans_alternate_pre_increment_yes()                              \
+  adds edge_alt_low, edge_alt_low, edge_dx_dy_alt;                             \
+  adc edge_alt_high, edge_alt_high, edge_dx_dy_alt, asr #31                    \
+
+#define setup_spans_alternate_pre_increment_no()                               \
+
+
+#define setup_spans_up_decrement_yes()                                         \
+  suble height, height, #1                                                     \
+
+#define setup_spans_up_decrement_no()                                          \
+
+
+#define setup_spans_up(left_index, right_index, alternate, alternate_active)   \
+  setup_spans_alternate_adjust_##alternate_active();                           \
+  setup_spans_load_b();                                                        \
+  sub y_a, y_a, #1;                                                            \
+                                                                               \
+  ldrh temp, [ psx_gpu, #psx_gpu_viewport_start_y_offset ];                    \
+  subs temp, temp, y_c;                                                        \
+  subgt height, height, temp;                                                  \
+  setup_spans_up_decrement_##alternate_active();                               \
+                                                                               \
+  ldrh temp, [ psx_gpu, #psx_gpu_viewport_end_y_offset ];                      \
+  subs clip, y_a, temp;                                                        \
+  ble 0f;                                                                      \
+                                                                               \
+  sub height, height, clip;                                                    \
+  sub y_a, y_a, clip;                                                          \
+  setup_spans_clip(decrement, alternate_active);                               \
+                                                                               \
+ 0:                                                                            \
+  cmp height, #0;                                                              \
+  ble 1f;                                                                      \
+                                                                               \
+  orr temp, y_a, y_a, lsl #16;                                                 \
+  sub temp, temp, #(1 << 16);                                                  \
+  sub y_a, temp, #2;                                                           \
+  sub y_a, y_a, #(2 << 16);                                                    \
+  vmov.u32 y_x4, temp, y_a;                                                    \
+                                                                               \
+  vaddw.s32 edges_xy, edges_xy, edges_dx_dy;                                   \
+                                                                               \
+  setup_spans_alternate_pre_increment_##alternate_active();                    \
+  setup_spans_adjust_edges_alternate_##alternate_active(left_index,            \
+   right_index);                                                               \
+  setup_spans_adjust_interpolants_up();                                        \
+  setup_spans_prologue_b();                                                    \
+                                                                               \
+  strh height, [ psx_gpu, #psx_gpu_num_spans_offset ];                         \
+                                                                               \
+ 2:                                                                            \
+  setup_spans_set_x4_alternate_##alternate_active(alternate, up);              \
+  subs height, height, #4;                                                     \
+  bhi 2b;                                                                      \
+                                                                               \
+ 1:                                                                            \
+
+
+#define setup_spans_epilogue()                                                 \
+  ldmia sp!, { r4 - r11, pc }                                                  \
+
+
+#define setup_spans_up_up(minor, major)                                        \
+  setup_spans_prologue();                                                      \
+  sub height_minor_a, y_a, y_b;                                                \
+  sub height_minor_b, y_b, y_c;                                                \
+  sub height, y_a, y_c;                                                        \
+                                                                               \
+  vdup.u32 x_starts, x_a;                                                      \
+  vmov.u32 x_ends, x_c, x_b;                                                   \
+                                                                               \
+  compute_edge_delta_x3(x_b, height_major, height_minor_a);                    \
+  setup_spans_up(major, minor, minor, yes);                                    \
+  setup_spans_epilogue()                                                       \
+
+function(setup_spans_up_left)
+  setup_spans_up_up(left, right)
+
+function(setup_spans_up_right)
+  setup_spans_up_up(right, left)
+
+
+#define setup_spans_down_down(minor, major)                                    \
+  setup_spans_prologue();                                                      \
+  sub height_minor_a, y_b, y_a;                                                \
+  sub height_minor_b, y_c, y_b;                                                \
+  sub height, y_c, y_a;                                                        \
+                                                                               \
+  vdup.u32 x_starts, x_a;                                                      \
+  vmov.u32 x_ends, x_c, x_b;                                                   \
+                                                                               \
+  compute_edge_delta_x3(x_b, height_major, height_minor_a);                    \
+  setup_spans_down(major, minor, minor, yes);                                  \
+  setup_spans_epilogue()                                                       \
+
+function(setup_spans_down_left)
+  setup_spans_down_down(left, right)
+
+function(setup_spans_down_right)
+  setup_spans_down_down(right, left)
+
+
+#define setup_spans_up_flat()                                                  \
+  sub height, y_a, y_c;                                                        \
+                                                                               \
+  compute_edge_delta_x2();                                                     \
+  setup_spans_up(left, right, none, no);                                       \
+  setup_spans_epilogue()                                                       \
+
+function(setup_spans_up_a)
+  setup_spans_prologue()
+
+  vmov.u32 x_starts, x_a, x_b
+  vdup.u32 x_ends, x_c
+
+  setup_spans_up_flat()
+
+function(setup_spans_up_b)
+  setup_spans_prologue()
+
+  vdup.u32 x_starts, x_a
+  vmov.u32 x_ends, x_b, x_c
+
+  setup_spans_up_flat()
+
+#define setup_spans_down_flat()                                                \
+  sub height, y_c, y_a;                                                        \
+                                                                               \
+  compute_edge_delta_x2();                                                     \
+  setup_spans_down(left, right, none, no);                                     \
+  setup_spans_epilogue()                                                       \
+
+function(setup_spans_down_a)
+  setup_spans_prologue()
+
+  vmov.u32 x_starts, x_a, x_b
+  vdup.u32 x_ends, x_c
+
+  setup_spans_down_flat()
+
+function(setup_spans_down_b)
+  setup_spans_prologue()
+
+  vdup.u32 x_starts, x_a
+  vmov.u32 x_ends, x_b, x_c
+
+  setup_spans_down_flat()
+
+
+#define middle_y                                          r9
+
+#define edges_xy_b                                        q11
+#define edges_dx_dy_b                                     d26
+#define edge_shifts_b                                     d27
+#define edges_dx_dy_and_shifts_b                          q13
+#define height_increment                                  d20
+
+#define edges_dx_dy_and_shifts                            q1
+
+#define edges_xy_b_left                                   d22
+#define edges_xy_b_right                                  d23
+
+#define setup_spans_up_down_load_edge_set_b()                                  \
+  vmov edges_xy, edges_xy_b;                                                   \
+  vmov edges_dx_dy_and_shifts, edges_dx_dy_and_shifts_b                        \
+
+
+function(setup_spans_up_down)
+  setup_spans_prologue()
+
+  // s32 middle_y = y_a;
+  sub height_minor_a, y_a, y_b
+  sub height_minor_b, y_c, y_a
+  sub height_major, y_c, y_b
+
+  vmov.u32 x_starts, x_a, x_c
+  vdup.u32 x_ends, x_b
+
+  compute_edge_delta_x3(x_a, height_minor_a, height_major)
+
+  mov temp, #0
+  vmov.u32 height_increment, temp, height_minor_b
+  vmlal.s32 edges_xy, edges_dx_dy, height_increment
+
+  vmov edges_xy_b_left, edge_alt_low, edge_alt_high
+  vmov edges_xy_b_right, edges_xy_right
+
+  vmov edge_shifts_b, edge_shifts
+  vmov.u32 edge_shifts_b[0], edge_shift_alt
+
+  vneg.s32 edges_dx_dy_b, edges_dx_dy
+  vmov.u32 edges_dx_dy_b[0], edge_dx_dy_alt
+
+  mov middle_y, y_a
+  
+  setup_spans_load_b()
+  sub y_a, y_a, #1
+
+  ldrh temp, [ psx_gpu, #psx_gpu_viewport_start_y_offset ]
+  subs temp, temp, y_b
+  subgt height_minor_a, height_minor_a, temp
+
+  ldrh temp, [ psx_gpu, #psx_gpu_viewport_end_y_offset ]
+  subs clip, y_a, temp
+  ble 0f
+
+  sub height_minor_a, height_minor_a, clip
+  sub y_a, y_a, clip
+  setup_spans_clip(decrement, no)
+
+ 0:                                                                
+  cmp height_minor_a, #0
+  ble 3f
+
+  orr temp, y_a, y_a, lsl #16
+  sub temp, temp, #(1 << 16)
+  sub y_a, temp, #2
+  sub y_a, y_a, #(2 << 16)
+  vmov.u32 y_x4, temp, y_a
+
+  vaddw.s32 edges_xy, edges_xy, edges_dx_dy
+
+  strh height_minor_a, [ psx_gpu, #psx_gpu_num_spans_offset ]
+
+  setup_spans_adjust_edges_alternate_no(left, right); 
+  setup_spans_adjust_interpolants_up()
+  setup_spans_up_down_load_edge_set_b()
+
+  setup_spans_prologue_b()
+
+
+ 2: 
+  setup_spans_set_x4_alternate_no(none, up)
+  subs height_minor_a, height_minor_a, #4
+  bhi 2b
+
+  add span_edge_data, span_edge_data, height_minor_a, lsl #3
+  add span_uvrg_offset, span_uvrg_offset, height_minor_a, lsl #4
+  add span_b_offset, span_b_offset, height_minor_a, lsl #2
+
+ 4:
+  add temp, psx_gpu, #psx_gpu_uvrg_offset
+  vld1.32 { uvrg }, [ temp ]
+  mov y_a, middle_y
+  
+  setup_spans_load_b()
+
+  ldrh temp, [ psx_gpu, #psx_gpu_viewport_end_y_offset ]
+  subs y_c, y_c, temp
+  subgt height_minor_b, height_minor_b, y_c
+  addgt height_minor_b, height_minor_b, #1
+
+  ldrh temp, [ psx_gpu, #psx_gpu_viewport_start_y_offset ]
+  subs clip, temp, y_a
+  ble 0f
+
+  sub height_minor_b, height_minor_b, clip
+  add y_a, y_a, clip
+  setup_spans_clip(increment, no)
+
+ 0:
+  cmp height_minor_b, #0
+  ble 1f
+
+  orr temp, y_a, y_a, lsl #16
+  add temp, temp, #(1 << 16) 
+  add y_a, temp, #2
+  add y_a, y_a, #(2 << 16)
+  vmov.u32 y_x4, temp, y_a
+
+  setup_spans_adjust_edges_alternate_no(left, right)
+
+  ldrh temp, [ psx_gpu, #psx_gpu_num_spans_offset ]
+  add temp, temp, height_minor_b
+  strh temp, [ psx_gpu, #psx_gpu_num_spans_offset ]
+
+ 2:                                                     
+  setup_spans_set_x4_alternate_no(none, down)
+  subs height_minor_b, height_minor_b, #4
+  bhi 2b
+
+ 1:
+  setup_spans_epilogue()
+
+ 3:
+  setup_spans_up_down_load_edge_set_b()
+  setup_spans_prologue_b()
+  bal 4b
+
+
+#undef span_uvrg_offset
+#undef span_edge_data
+#undef span_b_offset
+#undef left_x
+#undef b
+
+#define psx_gpu                                  r0
+#define num_spans                                r1
+#define span_uvrg_offset                         r2
+#define span_edge_data                           r3
+#define span_b_offset                            r4
+#define b_dx                                     r5
+#define span_num_blocks                          r6
+#define y                                        r7
+#define left_x                                   r8
+#define b                                        r9
+#define dither_offset_ptr                        r10
+#define block_ptr_a                              r11
+#define fb_ptr                                   r12
+#define num_blocks                               r14
+
+#define uvrg_dx_ptr                              r2
+#define texture_mask_ptr                         r3
+#define dither_shift                             r8
+#define dither_row                               r10
+
+#define c_32                                     r7
+#define b_dx4                                    r8
+#define b_dx8                                    r9
+#define block_ptr_b                              r10
+
+#define block_span_ptr                           r10
+#define right_mask                               r8
+
+#define color                                    r2
+#define color_r                                  r3
+#define color_g                                  r4
+#define color_b                                  r5
+
+#undef uvrg
+
+#define u_block                                  q0
+#define v_block                                  q1
+#define r_block                                  q2
+#define g_block                                  q3
+#define b_block                                  q4
+
+#define uv_dx4                                   d10
+#define rg_dx4                                   d11
+#define uv_dx8                                   d12
+#define rg_dx8                                   d13
+#define b_whole_8                                d14
+#define fb_mask_ptrs                             d15
+
+#define uvrg_dx4                                 q5
+#define uvrg_dx8                                 q6
+#define uv_dx8                                   d12
+#define rg_dx8                                   d13
+
+#define u_whole                                  q8
+#define v_whole                                  q9
+#define r_whole                                  q10
+#define g_whole                                  q11
+#define b_whole                                  q12
+
+#define u_whole_low                              d16
+#define u_whole_high                             d17
+#define v_whole_low                              d18
+#define v_whole_high                             d19
+#define r_whole_low                              d20
+#define r_whole_high                             d21
+#define g_whole_low                              d22
+#define g_whole_high                             d23
+#define b_whole_low                              d24
+#define b_whole_high                             d25
+
+#define dx4                                      q13
+#define dx8                                      q13
+
+#define u_whole_8                                d26
+#define v_whole_8                                d27
+#define u_whole_8b                               d24
+#define r_whole_8                                d24
+#define g_whole_8                                d25
+
+#define uv_whole_8                               q13
+#define uv_whole_8b                              q14
+
+#define dither_offsets                           q14
+#define texture_mask                             q15
+#define texture_mask_u                           d30
+#define texture_mask_v                           d31
+
+#define dither_offsets_short                     d28
+
+#define v_left_x                                 q8
+#define uvrg                                     q9
+#define block_span                               q10
+
+#define uv                                       d18
+#define rg                                       d19
+
+#define draw_mask                                q1
+#define draw_mask_edge                           q13
+#define test_mask                                q0
+
+#define uvrg_dx                                  q3
+
+#define colors                                   q2
+
+#define setup_blocks_texture_swizzled()                                        \
+  vand.u8 u_whole_8b, u_whole_8, texture_mask_u;                               \
+  vsli.u8 u_whole_8, v_whole_8, #4;                                            \
+  vsri.u8 v_whole_8, u_whole_8b, #4                                            \
+
+#define setup_blocks_texture_unswizzled()                                      \
+
+
+#define setup_blocks_shaded_textured_builder(swizzling)                        \
+.align 3;                                                                      \
+                                                                               \
+function(setup_blocks_shaded_textured_dithered_##swizzling##_indirect)         \
+  ldrh num_spans, [ psx_gpu, #psx_gpu_num_spans_offset ];                      \
+  add uvrg_dx_ptr, psx_gpu, #psx_gpu_uvrg_dx_offset;                           \
+                                                                               \
+  vld1.u32 { uvrg_dx }, [ uvrg_dx_ptr, :128 ];                                 \
+  add texture_mask_ptr, psx_gpu, #psx_gpu_texture_mask_width_offset;           \
+                                                                               \
+  cmp num_spans, #0;                                                           \
+  bxeq lr;                                                                     \
+                                                                               \
+  stmdb sp!, { r4 - r11, r14 };                                                \
+  vshl.u32 uvrg_dx4, uvrg_dx, #2;                                              \
+                                                                               \
+  ldr b_dx, [ psx_gpu, #psx_gpu_b_dx_offset ];                                 \
+  vshl.u32 uvrg_dx8, uvrg_dx, #3;                                              \
+                                                                               \
+  vld2.u8 { texture_mask_u[], texture_mask_v[] }, [ texture_mask_ptr, :16 ];   \
+  add span_uvrg_offset, psx_gpu, #psx_gpu_span_uvrg_offset_offset;             \
+                                                                               \
+  ldrh num_blocks, [ psx_gpu, #psx_gpu_num_blocks_offset ];                    \
+  add span_edge_data, psx_gpu, #psx_gpu_span_edge_data_offset;                 \
+                                                                               \
+  add span_b_offset, psx_gpu, #psx_gpu_span_b_offset_offset;                   \
+  add block_ptr_a, psx_gpu, #psx_gpu_blocks_offset;                            \
+                                                                               \
+  add block_ptr_a, block_ptr_a, num_blocks, lsl #6;                            \
+                                                                               \
+ 0:                                                                            \
+  vmov.u8 fb_mask_ptrs, #0;                                                    \
+                                                                               \
+  ldrh span_num_blocks, [ span_edge_data, #edge_data_num_blocks_offset ];      \
+  add dither_offset_ptr, psx_gpu, #psx_gpu_dither_table_offset;                \
+                                                                               \
+  ldrh y, [ span_edge_data, #edge_data_y_offset ];                             \
+  ldr fb_ptr, [ psx_gpu, #psx_gpu_vram_ptr_offset ];                           \
+                                                                               \
+  cmp span_num_blocks, #0;                                                     \
+  beq 1f;                                                                      \
+                                                                               \
+  ldrh left_x, [ span_edge_data, #edge_data_left_x_offset ];                   \
+  add num_blocks, span_num_blocks, num_blocks;                                 \
+                                                                               \
+  cmp num_blocks, #MAX_BLOCKS;                                                 \
+  bgt 2f;                                                                      \
+                                                                               \
+ 3:                                                                            \
+  ldr b, [ span_b_offset ];                                                    \
+  add fb_ptr, fb_ptr, y, lsl #11;                                              \
+                                                                               \
+  vdup.u32 v_left_x, left_x;                                                   \
+  and y, y, #0x3;                                                              \
+                                                                               \
+  ldr dither_row, [ dither_offset_ptr, y, lsl #2 ];                            \
+  add fb_ptr, fb_ptr, left_x, lsl #1;                                          \
+                                                                               \
+  mla b, b_dx, left_x, b;                                                      \
+  and dither_shift, left_x, #0x03;                                             \
+                                                                               \
+  vld1.u32 { uvrg }, [ span_uvrg_offset, :128 ];                               \
+  vshr.u32 uvrg_dx, uvrg_dx4, #2;                                              \
+                                                                               \
+  mov dither_shift, dither_shift, lsl #3;                                      \
+  vmla.u32 uvrg, uvrg_dx, v_left_x;                                            \
+                                                                               \
+  mov c_32, #32;                                                               \
+  subs span_num_blocks, span_num_blocks, #1;                                   \
+                                                                               \
+  mov dither_row, dither_row, ror dither_shift;                                \
+  mov b_dx4, b_dx, lsl #2;                                                     \
+                                                                               \
+  vdup.u32 dither_offsets_short, dither_row;                                   \
+  add block_span_ptr, psx_gpu, #psx_gpu_u_block_span_offset;                   \
+                                                                               \
+  vdup.u32 b_block, b;                                                         \
+  vshll.s8 dither_offsets, dither_offsets_short, #4;                           \
+                                                                               \
+  vdup.u32 u_block, uv[0];                                                     \
+  mov b_dx8, b_dx, lsl #3;                                                     \
+                                                                               \
+  vdup.u32 v_block, uv[1];                                                     \
+  vdup.u32 r_block, rg[0];                                                     \
+  vdup.u32 g_block, rg[1];                                                     \
+                                                                               \
+  vld1.u32 { block_span }, [ block_span_ptr, :128 ]!;                          \
+                                                                               \
+  vadd.u32 u_block, u_block, block_span;                                       \
+  vld1.u32 { block_span }, [ block_span_ptr, :128 ]!;                          \
+                                                                               \
+  vadd.u32 v_block, v_block, block_span;                                       \
+  vld1.u32 { block_span }, [ block_span_ptr, :128 ]!;                          \
+                                                                               \
+  vadd.u32 r_block, r_block, block_span;                                       \
+  vld1.u32 { block_span }, [ block_span_ptr, :128 ]!;                          \
+                                                                               \
+  vadd.u32 g_block, g_block, block_span;                                       \
+  vld1.u32 { block_span }, [ block_span_ptr, :128 ];                           \
+                                                                               \
+  vadd.u32 b_block, b_block, block_span;                                       \
+  add block_ptr_b, block_ptr_a, #16;                                           \
+                                                                               \
+  vshrn.u32 u_whole_low, u_block, #16;                                         \
+  vshrn.u32 v_whole_low, v_block, #16;                                         \
+  vshrn.u32 r_whole_low, r_block, #16;                                         \
+  vshrn.u32 g_whole_low, g_block, #16;                                         \
+                                                                               \
+  vdup.u32 dx4, uv_dx4[0];                                                     \
+  vshrn.u32 b_whole_low, b_block, #16;                                         \
+                                                                               \
+  vaddhn.u32 u_whole_high, u_block, dx4;                                       \
+  vdup.u32 dx4, uv_dx4[1];                                                     \
+                                                                               \
+  vaddhn.u32 v_whole_high, v_block, dx4;                                       \
+  vdup.u32 dx4, rg_dx4[0];                                                     \
+                                                                               \
+  vaddhn.u32 r_whole_high, r_block, dx4;                                       \
+  vdup.u32 dx4, rg_dx4[1];                                                     \
+                                                                               \
+  vaddhn.u32 g_whole_high, g_block, dx4;                                       \
+  vdup.u32 dx4, b_dx4;                                                         \
+                                                                               \
+  vaddhn.u32 b_whole_high, b_block, dx4;                                       \
+  vdup.u32 dx8, uv_dx8[0];                                                     \
+                                                                               \
+  vadd.u32 u_block, u_block, dx8;                                              \
+  vdup.u32 dx8, uv_dx8[1];                                                     \
+                                                                               \
+  vadd.u32 v_block, v_block, dx8;                                              \
+  vdup.u32 dx8, rg_dx8[0];                                                     \
+                                                                               \
+  vadd.u32 r_block, r_block, dx8;                                              \
+  vdup.u32 dx8, rg_dx8[1];                                                     \
+                                                                               \
+  vadd.u32 g_block, g_block, dx8;                                              \
+  vdup.u32 dx8, b_dx8;                                                         \
+                                                                               \
+  vadd.u32 b_block, b_block, dx8;                                              \
+  vmovn.u16 u_whole_8, u_whole;                                                \
+                                                                               \
+  vmovn.u16 v_whole_8, v_whole;                                                \
+                                                                               \
+  vmovn.u16 b_whole_8, b_whole;                                                \
+  pld [ fb_ptr ];                                                              \
+  vmov.u32 fb_mask_ptrs[1], fb_ptr;                                            \
+                                                                               \
+  vand.u8 uv_whole_8, uv_whole_8, texture_mask;                                \
+  setup_blocks_texture_##swizzling();                                          \
+                                                                               \
+  vmovn.u16 r_whole_8, r_whole;                                                \
+  beq 5f;                                                                      \
+                                                                               \
+ 4:                                                                            \
+  vmovn.u16 g_whole_8, g_whole;                                                \
+  vshrn.u32 u_whole_low, u_block, #16;                                         \
+                                                                               \
+  vst2.u8 { u_whole_8, v_whole_8 }, [ block_ptr_a, :128 ], c_32;               \
+  vshrn.u32 v_whole_low, v_block, #16;                                         \
+                                                                               \
+  vst1.u32 { r_whole_8, g_whole_8 }, [ block_ptr_b, :128 ], c_32;              \
+  vshrn.u32 r_whole_low, r_block, #16;                                         \
+                                                                               \
+  vst1.u32 { b_whole_8, fb_mask_ptrs }, [ block_ptr_a, :128 ], c_32;           \
+  vshrn.u32 g_whole_low, g_block, #16;                                         \
+                                                                               \
+  vdup.u32 dx4, uv_dx4[0];                                                     \
+  vshrn.u32 b_whole_low, b_block, #16;                                         \
+                                                                               \
+  vaddhn.u32 u_whole_high, u_block, dx4;                                       \
+  vdup.u32 dx4, uv_dx4[1];                                                     \
+                                                                               \
+  vaddhn.u32 v_whole_high, v_block, dx4;                                       \
+  vdup.u32 dx4, rg_dx4[0];                                                     \
+                                                                               \
+  vaddhn.u32 r_whole_high, r_block, dx4;                                       \
+  vdup.u32 dx4, rg_dx4[1];                                                     \
+                                                                               \
+  vaddhn.u32 g_whole_high, g_block, dx4;                                       \
+  vdup.u32 dx4, b_dx4;                                                         \
+                                                                               \
+  vaddhn.u32 b_whole_high, b_block, dx4;                                       \
+  vdup.u32 dx8, uv_dx8[0];                                                     \
+                                                                               \
+  vadd.u32 u_block, u_block, dx8;                                              \
+  vdup.u32 dx8, uv_dx8[1];                                                     \
+                                                                               \
+  vadd.u32 v_block, v_block, dx8;                                              \
+  vdup.u32 dx8, rg_dx8[0];                                                     \
+                                                                               \
+  vadd.u32 r_block, r_block, dx8;                                              \
+  vdup.u32 dx8, rg_dx8[1];                                                     \
+                                                                               \
+  vadd.u32 g_block, g_block, dx8;                                              \
+  vdup.u32 dx8, b_dx8;                                                         \
+                                                                               \
+  vadd.u32 b_block, b_block, dx8;                                              \
+  vmovn.u16 u_whole_8, u_whole;                                                \
+                                                                               \
+  add fb_ptr, fb_ptr, #16;                                                     \
+  vmovn.u16 v_whole_8, v_whole;                                                \
+                                                                               \
+  vst1.u32 { dither_offsets }, [ block_ptr_b, :128 ], c_32;                    \
+  vmovn.u16 b_whole_8, b_whole;                                                \
+                                                                               \
+  pld [ fb_ptr ];                                                              \
+                                                                               \
+  vmov.u32 fb_mask_ptrs[1], fb_ptr;                                            \
+  subs span_num_blocks, span_num_blocks, #1;                                   \
+                                                                               \
+  vand.u8 uv_whole_8, uv_whole_8, texture_mask;                                \
+  setup_blocks_texture_##swizzling();                                          \
+                                                                               \
+  vmovn.u16 r_whole_8, r_whole;                                                \
+  bne 4b;                                                                      \
+                                                                               \
+ 5:                                                                            \
+  vmovn.u16 g_whole_8, g_whole;                                                \
+  ldrh right_mask, [ span_edge_data, #edge_data_right_mask_offset ];           \
+                                                                               \
+  vld1.u32 { test_mask }, [ psx_gpu, :128 ];                                   \
+  vdup.u8 draw_mask, right_mask;                                               \
+                                                                               \
+  vmov.u32 fb_mask_ptrs[0], right_mask;                                        \
+  vtst.u16 draw_mask, draw_mask, test_mask;                                    \
+  vzip.u8 u_whole_8, v_whole_8;                                                \
+                                                                               \
+  vbic.u16 uv_whole_8, uv_whole_8, draw_mask;                                  \
+  vst1.u32 { r_whole_8, g_whole_8 }, [ block_ptr_b, :128 ], c_32;              \
+  vst1.u32 { uv_whole_8 }, [ block_ptr_a, :128 ], c_32;                        \
+  vst1.u32 { dither_offsets }, [ block_ptr_b, :128 ], c_32;                    \
+  vst1.u32 { b_whole_8, fb_mask_ptrs }, [ block_ptr_a, :128 ], c_32;           \
+                                                                               \
+ 1:                                                                            \
+  add span_uvrg_offset, span_uvrg_offset, #16;                                 \
+  add span_b_offset, span_b_offset, #4;                                        \
+                                                                               \
+  add span_edge_data, span_edge_data, #8;                                      \
+  subs num_spans, num_spans, #1;                                               \
+                                                                               \
+  strh num_blocks, [ psx_gpu, #psx_gpu_num_blocks_offset ];                    \
+  bne 0b;                                                                      \
+                                                                               \
+  ldmia sp!, { r4 - r11, pc };                                                 \
+                                                                               \
+ 2:                                                                            \
+  /* TODO: Load from psx_gpu instead of saving/restoring these               */\
+  vpush { texture_mask };                                                      \
+  vpush { uvrg_dx4 };                                                          \
+                                                                               \
+  stmdb sp!, { r0 - r3, r12, r14 };                                            \
+  bl flush_render_block_buffer;                                                \
+  ldmia sp!, { r0 - r3, r12, r14 };                                            \
+                                                                               \
+  vpop { uvrg_dx4 };                                                           \
+  vpop { texture_mask };                                                       \
+                                                                               \
+  vadd.u32 uvrg_dx8, uvrg_dx4, uvrg_dx4;                                       \
+  vmov.u8 fb_mask_ptrs, #0;                                                    \
+                                                                               \
+  mov num_blocks, span_num_blocks;                                             \
+  add block_ptr_a, psx_gpu, #psx_gpu_blocks_offset;                            \
+  bal 3b                                                                       \
+
+
+setup_blocks_shaded_textured_builder(swizzled)
+setup_blocks_shaded_textured_builder(unswizzled)
+
+
+#define setup_blocks_unshaded_textured_builder(swizzling)                      \
+.align 3;                                                                      \
+                                                                               \
+function(setup_blocks_unshaded_textured_dithered_##swizzling##_indirect)       \
+  ldrh num_spans, [ psx_gpu, #psx_gpu_num_spans_offset ];                      \
+  add uvrg_dx_ptr, psx_gpu, #psx_gpu_uvrg_dx_offset;                           \
+                                                                               \
+  vld1.u32 { uvrg_dx }, [ uvrg_dx_ptr, :128 ];                                 \
+  add texture_mask_ptr, psx_gpu, #psx_gpu_texture_mask_width_offset;           \
+                                                                               \
+  cmp num_spans, #0;                                                           \
+  bxeq lr;                                                                     \
+                                                                               \
+  stmdb sp!, { r4 - r11, r14 };                                                \
+  vshl.u32 uvrg_dx4, uvrg_dx, #2;                                              \
+                                                                               \
+  vshl.u32 uvrg_dx8, uvrg_dx, #3;                                              \
+                                                                               \
+  vld2.u8 { texture_mask_u[], texture_mask_v[] }, [ texture_mask_ptr, :16 ];   \
+  add span_uvrg_offset, psx_gpu, #psx_gpu_span_uvrg_offset_offset;             \
+                                                                               \
+  ldrh num_blocks, [ psx_gpu, #psx_gpu_num_blocks_offset ];                    \
+  add span_edge_data, psx_gpu, #psx_gpu_span_edge_data_offset;                 \
+                                                                               \
+  add block_ptr_a, psx_gpu, #psx_gpu_blocks_offset;                            \
+                                                                               \
+  add block_ptr_a, block_ptr_a, num_blocks, lsl #6;                            \
+                                                                               \
+ 0:                                                                            \
+  vmov.u8 fb_mask_ptrs, #0;                                                    \
+                                                                               \
+  ldrh span_num_blocks, [ span_edge_data, #edge_data_num_blocks_offset ];      \
+  add dither_offset_ptr, psx_gpu, #psx_gpu_dither_table_offset;                \
+                                                                               \
+  ldrh y, [ span_edge_data, #edge_data_y_offset ];                             \
+  ldr fb_ptr, [ psx_gpu, #psx_gpu_vram_ptr_offset ];                           \
+                                                                               \
+  cmp span_num_blocks, #0;                                                     \
+  beq 1f;                                                                      \
+                                                                               \
+  ldrh left_x, [ span_edge_data, #edge_data_left_x_offset ];                   \
+  add num_blocks, span_num_blocks, num_blocks;                                 \
+                                                                               \
+  cmp num_blocks, #MAX_BLOCKS;                                                 \
+  bgt 2f;                                                                      \
+                                                                               \
+ 3:                                                                            \
+  add fb_ptr, fb_ptr, y, lsl #11;                                              \
+                                                                               \
+  vdup.u32 v_left_x, left_x;                                                   \
+  and y, y, #0x3;                                                              \
+                                                                               \
+  ldr dither_row, [ dither_offset_ptr, y, lsl #2 ];                            \
+  add fb_ptr, fb_ptr, left_x, lsl #1;                                          \
+                                                                               \
+  and dither_shift, left_x, #0x03;                                             \
+                                                                               \
+  vld1.u32 { uvrg }, [ span_uvrg_offset, :128 ];                               \
+  vshr.u32 uvrg_dx, uvrg_dx4, #2;                                              \
+                                                                               \
+  mov dither_shift, dither_shift, lsl #3;                                      \
+  vmla.u32 uvrg, uvrg_dx, v_left_x;                                            \
+                                                                               \
+  mov c_32, #32;                                                               \
+  subs span_num_blocks, span_num_blocks, #1;                                   \
+                                                                               \
+  mov dither_row, dither_row, ror dither_shift;                                \
+                                                                               \
+  vdup.u32 dither_offsets_short, dither_row;                                   \
+  add block_span_ptr, psx_gpu, #psx_gpu_u_block_span_offset;                   \
+                                                                               \
+  vshll.s8 dither_offsets, dither_offsets_short, #4;                           \
+                                                                               \
+  vdup.u32 u_block, uv[0];                                                     \
+                                                                               \
+  vdup.u32 v_block, uv[1];                                                     \
+  vld1.u32 { block_span }, [ block_span_ptr, :128 ]!;                          \
+                                                                               \
+  vadd.u32 u_block, u_block, block_span;                                       \
+  vld1.u32 { block_span }, [ block_span_ptr, :128 ]!;                          \
+                                                                               \
+  vadd.u32 v_block, v_block, block_span;                                       \
+  add block_ptr_b, block_ptr_a, #16;                                           \
+                                                                               \
+  vshrn.u32 u_whole_low, u_block, #16;                                         \
+  vshrn.u32 v_whole_low, v_block, #16;                                         \
+                                                                               \
+  vdup.u32 dx4, uv_dx4[0];                                                     \
+                                                                               \
+  vaddhn.u32 u_whole_high, u_block, dx4;                                       \
+  vdup.u32 dx4, uv_dx4[1];                                                     \
+                                                                               \
+  vaddhn.u32 v_whole_high, v_block, dx4;                                       \
+  vdup.u32 dx8, uv_dx8[0];                                                     \
+                                                                               \
+  vadd.u32 u_block, u_block, dx8;                                              \
+  vdup.u32 dx8, uv_dx8[1];                                                     \
+                                                                               \
+  vadd.u32 v_block, v_block, dx8;                                              \
+  vmovn.u16 u_whole_8, u_whole;                                                \
+                                                                               \
+  vmovn.u16 v_whole_8, v_whole;                                                \
+                                                                               \
+  pld [ fb_ptr ];                                                              \
+  vmov.u32 fb_mask_ptrs[1], fb_ptr;                                            \
+                                                                               \
+  vand.u8 uv_whole_8, uv_whole_8, texture_mask;                                \
+  setup_blocks_texture_##swizzling();                                          \
+                                                                               \
+  beq 5f;                                                                      \
+                                                                               \
+ 4:                                                                            \
+  vshrn.u32 u_whole_low, u_block, #16;                                         \
+                                                                               \
+  vst2.u8 { u_whole_8, v_whole_8 }, [ block_ptr_a, :128 ], c_32;               \
+  vshrn.u32 v_whole_low, v_block, #16;                                         \
+                                                                               \
+  add block_ptr_b, block_ptr_b, #32;                                           \
+  vst1.u32 { b_whole_8, fb_mask_ptrs }, [ block_ptr_a, :128 ], c_32;           \
+                                                                               \
+  vdup.u32 dx4, uv_dx4[0];                                                     \
+  vaddhn.u32 u_whole_high, u_block, dx4;                                       \
+  vdup.u32 dx4, uv_dx4[1];                                                     \
+                                                                               \
+  vaddhn.u32 v_whole_high, v_block, dx4;                                       \
+  vdup.u32 dx8, uv_dx8[0];                                                     \
+                                                                               \
+  vadd.u32 u_block, u_block, dx8;                                              \
+  vdup.u32 dx8, uv_dx8[1];                                                     \
+                                                                               \
+  vadd.u32 v_block, v_block, dx8;                                              \
+  vmovn.u16 u_whole_8, u_whole;                                                \
+                                                                               \
+  add fb_ptr, fb_ptr, #16;                                                     \
+  vmovn.u16 v_whole_8, v_whole;                                                \
+                                                                               \
+  vst1.u32 { dither_offsets }, [ block_ptr_b, :128 ], c_32;                    \
+  pld [ fb_ptr ];                                                              \
+                                                                               \
+  vmov.u32 fb_mask_ptrs[1], fb_ptr;                                            \
+  subs span_num_blocks, span_num_blocks, #1;                                   \
+                                                                               \
+  vand.u8 uv_whole_8, uv_whole_8, texture_mask;                                \
+  setup_blocks_texture_##swizzling();                                          \
+                                                                               \
+  bne 4b;                                                                      \
+                                                                               \
+ 5:                                                                            \
+  ldrh right_mask, [ span_edge_data, #edge_data_right_mask_offset ];           \
+                                                                               \
+  vld1.u32 { test_mask }, [ psx_gpu, :128 ];                                   \
+  vdup.u8 draw_mask, right_mask;                                               \
+                                                                               \
+  vmov.u32 fb_mask_ptrs[0], right_mask;                                        \
+  vtst.u16 draw_mask, draw_mask, test_mask;                                    \
+  vzip.u8 u_whole_8, v_whole_8;                                                \
+                                                                               \
+  vbic.u16 uv_whole_8, uv_whole_8, draw_mask;                                  \
+  add block_ptr_b, block_ptr_b, #32;                                           \
+  vst1.u32 { uv_whole_8 }, [ block_ptr_a, :128 ], c_32;                        \
+  vst1.u32 { dither_offsets }, [ block_ptr_b, :128 ], c_32;                    \
+  vst1.u32 { b_whole_8, fb_mask_ptrs }, [ block_ptr_a, :128 ], c_32;           \
+                                                                               \
+ 1:                                                                            \
+  add span_uvrg_offset, span_uvrg_offset, #16;                                 \
+  add span_edge_data, span_edge_data, #8;                                      \
+  subs num_spans, num_spans, #1;                                               \
+                                                                               \
+  strh num_blocks, [ psx_gpu, #psx_gpu_num_blocks_offset ];                    \
+  bne 0b;                                                                      \
+                                                                               \
+  ldmia sp!, { r4 - r11, pc };                                                 \
+                                                                               \
+ 2:                                                                            \
+  /* TODO: Load from psx_gpu instead of saving/restoring these               */\
+  vpush { texture_mask };                                                      \
+  vpush { uvrg_dx4 };                                                          \
+                                                                               \
+  stmdb sp!, { r0 - r3, r12, r14 };                                            \
+  bl flush_render_block_buffer;                                                \
+  ldmia sp!, { r0 - r3, r12, r14 };                                            \
+                                                                               \
+  vpop { uvrg_dx4 };                                                           \
+  vpop { texture_mask };                                                       \
+                                                                               \
+  vadd.u32 uvrg_dx8, uvrg_dx4, uvrg_dx4;                                       \
+  vmov.u8 fb_mask_ptrs, #0;                                                    \
+                                                                               \
+  mov num_blocks, span_num_blocks;                                             \
+  add block_ptr_a, psx_gpu, #psx_gpu_blocks_offset;                            \
+  bal 3b                                                                       \
+
+
+setup_blocks_unshaded_textured_builder(swizzled)
+setup_blocks_unshaded_textured_builder(unswizzled)
+
+
+.align 3
+
+function(setup_blocks_unshaded_untextured_undithered_unswizzled_indirect)
+  ldrh num_spans, [ psx_gpu, #psx_gpu_num_spans_offset ]
+  veor.u32 draw_mask, draw_mask, draw_mask
+
+  cmp num_spans, #0
+  bxeq lr
+
+  stmdb sp!, { r4 - r11, r14 }
+  vld1.u32 { test_mask }, [ psx_gpu, :128 ]
+
+  ldr color, [ psx_gpu, #psx_gpu_triangle_color_offset ]
+
+  ubfx color_r, color, #3, #5
+  ubfx color_g, color, #11, #5
+  ubfx color_b, color, #19, #5
+
+  orr color, color_r, color_b, lsl #10
+  orr color, color, color_g, lsl #5
+
+  vdup.u16 colors, color
+
+  ldrh num_blocks, [ psx_gpu, #psx_gpu_num_blocks_offset ]
+  add span_edge_data, psx_gpu, #psx_gpu_span_edge_data_offset
+
+  add block_ptr_a, psx_gpu, #psx_gpu_blocks_offset
+  add block_ptr_a, block_ptr_a, num_blocks, lsl #6
+
+ 0:
+  ldrh span_num_blocks, [ span_edge_data, #edge_data_num_blocks_offset ]
+  ldrh y, [ span_edge_data, #edge_data_y_offset ]
+
+  ldr fb_ptr, [ psx_gpu, #psx_gpu_vram_ptr_offset ]
+
+  cmp span_num_blocks, #0
+  beq 1f
+
+  ldrh left_x, [ span_edge_data, #edge_data_left_x_offset ]
+  add num_blocks, span_num_blocks, num_blocks
+
+  cmp num_blocks, #MAX_BLOCKS
+  bgt 2f
+
+ 3:
+  add fb_ptr, fb_ptr, y, lsl #11
+  and y, y, #0x3
+
+  add fb_ptr, fb_ptr, left_x, lsl #1
+  mov c_32, #32
+
+  subs span_num_blocks, span_num_blocks, #1
+
+  add block_ptr_b, block_ptr_a, #16
+  pld [ fb_ptr ]
+
+  vmov.u32 fb_mask_ptrs[1], fb_ptr
+  beq 5f
+
+ 4:
+  vst1.u32 { draw_mask }, [ block_ptr_a, :128 ], c_32
+  vst1.u32 { colors }, [ block_ptr_b, :128 ], c_32
+  vst1.u32 { b_whole_8, fb_mask_ptrs }, [ block_ptr_a, :128 ], c_32
+
+  add fb_ptr, fb_ptr, #16
+  add block_ptr_b, block_ptr_b, #32
+
+  pld [ fb_ptr ]
+
+  vmov.u32 fb_mask_ptrs[1], fb_ptr
+  subs span_num_blocks, span_num_blocks, #1
+
+  bne 4b
+
+ 5:
+  ldrh right_mask, [ span_edge_data, #edge_data_right_mask_offset ]
+
+  vdup.u8 draw_mask_edge, right_mask
+  vtst.u16 draw_mask_edge, draw_mask_edge, test_mask
+
+  vst1.u32 { colors }, [ block_ptr_b, :128 ], c_32
+  vst1.u32 { draw_mask_edge }, [ block_ptr_a, :128 ], c_32
+  add block_ptr_b, block_ptr_b, #32
+  vst1.u32 { b_whole_8, fb_mask_ptrs }, [ block_ptr_a, :128 ], c_32
+
+ 1:
+  add span_edge_data, span_edge_data, #8
+  subs num_spans, num_spans, #1
+
+  strh num_blocks, [ psx_gpu, #psx_gpu_num_blocks_offset ]
+  bne 0b
+
+  ldmia sp!, { r4 - r11, pc }
+                                                                           
+ 2:
+  vpush { colors }
+
+  stmdb sp!, { r0 - r3, r12, r14 }
+  bl flush_render_block_buffer
+  ldmia sp!, { r0 - r3, r12, r14 }
+
+  vpop { colors }
+
+  vld1.u32 { test_mask }, [ psx_gpu, :128 ]
+  veor.u32 draw_mask, draw_mask, draw_mask
+
+  mov num_blocks, span_num_blocks
+  add block_ptr_a, psx_gpu, #psx_gpu_blocks_offset
+  bal 3b
+
+
+#define mask_msb_scalar                                   r14
+
+#define msb_mask                                          q15
+
+#define pixels_low                                        d16
+
+#define msb_mask_low                                      d30
+#define msb_mask_high                                     d31
+
+
+.align 3
+
+function(setup_blocks_unshaded_untextured_undithered_unswizzled_direct)
+  ldrh num_spans, [ psx_gpu, #psx_gpu_num_spans_offset ]
+
+  cmp num_spans, #0
+  bxeq lr
+
+  stmdb sp!, { r4 - r11, r14 }
+
+  ldr color, [ psx_gpu, #psx_gpu_triangle_color_offset ]
+
+  ubfx color_r, color, #3, #5
+  ubfx color_g, color, #11, #5
+
+  ldrh mask_msb_scalar, [ psx_gpu, #psx_gpu_mask_msb_offset ]
+  ubfx color_b, color, #19, #5
+
+  orr color, color_r, color_b, lsl #10
+  orr color, color, color_g, lsl #5
+  orr color, color, mask_msb_scalar
+
+  vdup.u16 colors, color
+
+  add span_edge_data, psx_gpu, #psx_gpu_span_edge_data_offset
+
+ 0:
+  ldrh span_num_blocks, [ span_edge_data, #edge_data_num_blocks_offset ]
+  ldrh y, [ span_edge_data, #edge_data_y_offset ]
+
+  ldr fb_ptr, [ psx_gpu, #psx_gpu_vram_ptr_offset ]
+
+  cmp span_num_blocks, #0
+  beq 1f
+
+  ldrh left_x, [ span_edge_data, #edge_data_left_x_offset ]
+
+  add fb_ptr, fb_ptr, y, lsl #11
+  subs span_num_blocks, span_num_blocks, #1
+
+  add fb_ptr, fb_ptr, left_x, lsl #1
+  beq 3f
+
+ 2:
+  vst1.u32 { colors }, [ fb_ptr ]!
+  subs span_num_blocks, span_num_blocks, #1
+
+  bne 2b
+
+ 3:
+  ldrb right_mask, [ span_edge_data, #edge_data_right_mask_offset ]
+  eor right_mask, right_mask, #0xFF
+
+ 4:
+  strh color, [ fb_ptr ], #2
+  movs right_mask, right_mask, lsr #1
+  bne 4b
+
+ 1:
+  add span_edge_data, span_edge_data, #8
+  subs num_spans, num_spans, #1
+
+  strh num_blocks, [ psx_gpu, #psx_gpu_num_blocks_offset ]
+  bne 0b
+
+  ldmia sp!, { r4 - r11, pc }
+                                                                           
+
+
+#undef c_64
+
+#define c_64                                              r7
+#define rg_dx_ptr                                         r2
+
+
+#undef r_block
+#undef g_block
+#undef b_block
+#undef r_whole
+#undef g_whole
+#undef b_whole
+#undef r_whole_low
+#undef r_whole_high
+#undef g_whole_low
+#undef g_whole_high
+#undef b_whole_low
+#undef b_whole_high
+#undef r_whole_8
+#undef g_whole_8
+#undef b_whole_8
+#undef dither_offsets
+#undef rg_dx4
+#undef rg_dx8
+#undef dx4
+#undef dx8
+#undef v_left_x
+#undef uvrg
+#undef block_span
+#undef rg
+#undef draw_mask
+#undef test_mask
+
+#define r_block                                           q0
+#define g_block                                           q1
+#define b_block                                           q2
+
+#define r_whole                                           q3
+#define g_whole                                           q4
+#define b_whole                                           q5
+
+#define r_whole_low                                       d6
+#define r_whole_high                                      d7
+#define g_whole_low                                       d8
+#define g_whole_high                                      d9
+#define b_whole_low                                       d10
+#define b_whole_high                                      d11
+
+#define gb_whole_8                                        q6
+
+#define g_whole_8                                         d12
+#define b_whole_8                                         d13
+
+#define r_whole_8                                         d14
+
+#define pixels                                            q8
+
+#define rg_dx4                                            d18
+#define rg_dx8                                            d19
+
+#define dx4                                               q10
+#define dx8                                               q10
+
+#define v_left_x                                          d6
+#define uvrg                                              q4
+#define block_span                                        q5
+
+#define rg                                                d9
+
+#define d64_1                                             d22
+#define d64_128                                           d23
+
+#define d128_4                                            q12
+#define d128_0x7                                          q13
+
+#define d64_4                                             d24
+
+#define dither_offsets                                    q14
+#define draw_mask                                         q15
+
+#define dither_offsets_low                                d28
+
+#define rg_dx                                             d0
+#define test_mask                                         q10
+
+
+#define setup_blocks_shaded_untextured_dither_a_dithered()                     \
+  vqadd.u8 r_whole_8, r_whole_8, dither_offsets_low;                           \
+  vqadd.u8 gb_whole_8, gb_whole_8, dither_offsets;                             \
+
+#define setup_blocks_shaded_untextured_dither_b_dithered()                     \
+  vqsub.u8 r_whole_8, r_whole_8, d64_4;                                        \
+  vqsub.u8 gb_whole_8, gb_whole_8, d128_4                                      \
+
+#define setup_blocks_shaded_untextured_dither_a_undithered()                   \
+
+#define setup_blocks_shaded_untextured_dither_b_undithered()                   \
+
+
+#define setup_blocks_shaded_untextured_indirect_builder(dithering)             \
+.align 3;                                                                      \
+                                                                               \
+function(setup_blocks_shaded_untextured_##dithering##_unswizzled_indirect)     \
+  ldrh num_spans, [ psx_gpu, #psx_gpu_num_spans_offset ];                      \
+  add rg_dx_ptr, psx_gpu, #(psx_gpu_uvrg_dx_offset + 8);                       \
+                                                                               \
+  vld1.u32 { rg_dx }, [ rg_dx_ptr, :64 ];                                      \
+                                                                               \
+  cmp num_spans, #0;                                                           \
+  bxeq lr;                                                                     \
+                                                                               \
+  stmdb sp!, { r4 - r11, r14 };                                                \
+  vshl.u32 rg_dx4, rg_dx, #2;                                                  \
+                                                                               \
+  ldr b_dx, [ psx_gpu, #psx_gpu_b_dx_offset ];                                 \
+  vshl.u32 rg_dx8, rg_dx, #3;                                                  \
+                                                                               \
+  add span_uvrg_offset, psx_gpu, #psx_gpu_span_uvrg_offset_offset;             \
+                                                                               \
+  ldrh num_blocks, [ psx_gpu, #psx_gpu_num_blocks_offset ];                    \
+  add span_edge_data, psx_gpu, #psx_gpu_span_edge_data_offset;                 \
+                                                                               \
+  add span_b_offset, psx_gpu, #psx_gpu_span_b_offset_offset;                   \
+  add block_ptr_a, psx_gpu, #psx_gpu_blocks_offset;                            \
+                                                                               \
+  add block_ptr_a, block_ptr_a, num_blocks, lsl #6;                            \
+  vmov.u8 d64_1, #1;                                                           \
+                                                                               \
+  vmov.u8 d128_4, #4;                                                          \
+  vmov.u8 d64_128, #128;                                                       \
+                                                                               \
+  vmov.u8 d128_0x7, #0x7;                                                      \
+                                                                               \
+ 0:                                                                            \
+  ldrh span_num_blocks, [ span_edge_data, #edge_data_num_blocks_offset ];      \
+  add dither_offset_ptr, psx_gpu, #psx_gpu_dither_table_offset;                \
+                                                                               \
+  ldrh y, [ span_edge_data, #edge_data_y_offset ];                             \
+  ldr fb_ptr, [ psx_gpu, #psx_gpu_vram_ptr_offset ];                           \
+                                                                               \
+  cmp span_num_blocks, #0;                                                     \
+  beq 1f;                                                                      \
+                                                                               \
+  ldrh left_x, [ span_edge_data, #edge_data_left_x_offset ];                   \
+  add num_blocks, span_num_blocks, num_blocks;                                 \
+                                                                               \
+  cmp num_blocks, #MAX_BLOCKS;                                                 \
+  bgt 2f;                                                                      \
+                                                                               \
+ 3:                                                                            \
+  ldr b, [ span_b_offset ];                                                    \
+  add fb_ptr, fb_ptr, y, lsl #11;                                              \
+                                                                               \
+  vdup.u32 v_left_x, left_x;                                                   \
+  and y, y, #0x3;                                                              \
+                                                                               \
+  ldr dither_row, [ dither_offset_ptr, y, lsl #2 ];                            \
+  add fb_ptr, fb_ptr, left_x, lsl #1;                                          \
+                                                                               \
+  mla b, b_dx, left_x, b;                                                      \
+  and dither_shift, left_x, #0x03;                                             \
+                                                                               \
+  vld1.u32 { uvrg }, [ span_uvrg_offset, :128 ];                               \
+  vshr.u32 rg_dx, rg_dx4, #2;                                                  \
+                                                                               \
+  mov dither_shift, dither_shift, lsl #3;                                      \
+  vmla.u32 rg, rg_dx, v_left_x;                                                \
+                                                                               \
+  mov c_64, #64;                                                               \
+  subs span_num_blocks, span_num_blocks, #1;                                   \
+                                                                               \
+  mov dither_row, dither_row, ror dither_shift;                                \
+  mov b_dx4, b_dx, lsl #2;                                                     \
+                                                                               \
+  vdup.u32 dither_offsets, dither_row;                                         \
+  add block_span_ptr, psx_gpu, #psx_gpu_r_block_span_offset;                   \
+                                                                               \
+  vdup.u32 b_block, b;                                                         \
+  vadd.u8 dither_offsets, dither_offsets, d128_4;                              \
+                                                                               \
+  mov b_dx8, b_dx, lsl #3;                                                     \
+  vdup.u32 r_block, rg[0];                                                     \
+  vdup.u32 g_block, rg[1];                                                     \
+                                                                               \
+  vld1.u32 { block_span }, [ block_span_ptr, :128 ]!;                          \
+                                                                               \
+  vadd.u32 r_block, r_block, block_span;                                       \
+  vld1.u32 { block_span }, [ block_span_ptr, :128 ]!;                          \
+                                                                               \
+  vadd.u32 g_block, g_block, block_span;                                       \
+  vld1.u32 { block_span }, [ block_span_ptr, :128 ];                           \
+                                                                               \
+  vadd.u32 b_block, b_block, block_span;                                       \
+  add block_ptr_b, block_ptr_a, #16;                                           \
+                                                                               \
+  vshrn.u32 r_whole_low, r_block, #16;                                         \
+  vshrn.u32 g_whole_low, g_block, #16;                                         \
+  vshrn.u32 b_whole_low, b_block, #16;                                         \
+  vdup.u32 dx4, rg_dx4[0];                                                     \
+                                                                               \
+  vaddhn.u32 r_whole_high, r_block, dx4;                                       \
+  vdup.u32 dx4, rg_dx4[1];                                                     \
+                                                                               \
+  vaddhn.u32 g_whole_high, g_block, dx4;                                       \
+  vdup.u32 dx4, b_dx4;                                                         \
+                                                                               \
+  vaddhn.u32 b_whole_high, b_block, dx4;                                       \
+  vdup.u32 dx8, rg_dx8[0];                                                     \
+                                                                               \
+  vadd.u32 r_block, r_block, dx8;                                              \
+  vdup.u32 dx8, rg_dx8[1];                                                     \
+                                                                               \
+  vadd.u32 g_block, g_block, dx8;                                              \
+  vdup.u32 dx8, b_dx8;                                                         \
+                                                                               \
+  vadd.u32 b_block, b_block, dx8;                                              \
+                                                                               \
+  vmovn.u16 r_whole_8, r_whole;                                                \
+  vmovn.u16 g_whole_8, g_whole;                                                \
+  vmovn.u16 b_whole_8, b_whole;                                                \
+                                                                               \
+  beq 5f;                                                                      \
+  veor.u32 draw_mask, draw_mask, draw_mask;                                    \
+                                                                               \
+ 4:                                                                            \
+  setup_blocks_shaded_untextured_dither_a_##dithering();                       \
+  vshrn.u32 r_whole_low, r_block, #16;                                         \
+                                                                               \
+  setup_blocks_shaded_untextured_dither_b_##dithering();                       \
+  vshrn.u32 g_whole_low, g_block, #16;                                         \
+                                                                               \
+  vshrn.u32 b_whole_low, b_block, #16;                                         \
+  str fb_ptr, [ block_ptr_a, #44 ];                                            \
+                                                                               \
+  vdup.u32 dx4, rg_dx4[0];                                                     \
+  vshr.u8 r_whole_8, r_whole_8, #3;                                            \
+  vbic.u8 gb_whole_8, gb_whole_8, d128_0x7;                                    \
+                                                                               \
+  vaddhn.u32 r_whole_high, r_block, dx4;                                       \
+  vdup.u32 dx4, rg_dx4[1];                                                     \
+                                                                               \
+  vaddhn.u32 g_whole_high, g_block, dx4;                                       \
+  vdup.u32 dx4, b_dx4;                                                         \
+                                                                               \
+  vaddhn.u32 b_whole_high, b_block, dx4;                                       \
+  vdup.u32 dx8, rg_dx8[0];                                                     \
+                                                                               \
+  vmull.u8 pixels, r_whole_8, d64_1;                                           \
+  vmlal.u8 pixels, g_whole_8, d64_4;                                           \
+  vmlal.u8 pixels, b_whole_8, d64_128;                                         \
+                                                                               \
+  vadd.u32 r_block, r_block, dx8;                                              \
+  vdup.u32 dx8, rg_dx8[1];                                                     \
+                                                                               \
+  vadd.u32 g_block, g_block, dx8;                                              \
+  vdup.u32 dx8, b_dx8;                                                         \
+                                                                               \
+  vadd.u32 b_block, b_block, dx8;                                              \
+  add fb_ptr, fb_ptr, #16;                                                     \
+                                                                               \
+  vmovn.u16 r_whole_8, r_whole;                                                \
+  vmovn.u16 g_whole_8, g_whole;                                                \
+  vmovn.u16 b_whole_8, b_whole;                                                \
+                                                                               \
+  vst1.u32 { draw_mask }, [ block_ptr_a, :128 ], c_64;                         \
+  vst1.u32 { pixels }, [ block_ptr_b, :128 ], c_64;                            \
+                                                                               \
+  pld [ fb_ptr ];                                                              \
+                                                                               \
+  subs span_num_blocks, span_num_blocks, #1;                                   \
+  bne 4b;                                                                      \
+                                                                               \
+ 5:                                                                            \
+  str fb_ptr, [ block_ptr_a, #44 ];                                            \
+  setup_blocks_shaded_untextured_dither_a_##dithering();                       \
+                                                                               \
+  ldrh right_mask, [ span_edge_data, #edge_data_right_mask_offset ];           \
+  setup_blocks_shaded_untextured_dither_b_##dithering();                       \
+                                                                               \
+  vshr.u8 r_whole_8, r_whole_8, #3;                                            \
+  vdup.u8 draw_mask, right_mask;                                               \
+                                                                               \
+  vbic.u8 gb_whole_8, gb_whole_8, d128_0x7;                                    \
+  vld1.u32 { test_mask }, [ psx_gpu, :128 ];                                   \
+                                                                               \
+  vtst.u16 draw_mask, draw_mask, test_mask;                                    \
+                                                                               \
+  vmull.u8 pixels, r_whole_8, d64_1;                                           \
+  vmlal.u8 pixels, g_whole_8, d64_4;                                           \
+  vmlal.u8 pixels, b_whole_8, d64_128;                                         \
+                                                                               \
+  vst1.u32 { draw_mask }, [ block_ptr_a, :128 ], c_64;                         \
+  vst1.u32 { pixels }, [ block_ptr_b, :128 ], c_64;                            \
+                                                                               \
+ 1:                                                                            \
+  add span_uvrg_offset, span_uvrg_offset, #16;                                 \
+  add span_b_offset, span_b_offset, #4;                                        \
+                                                                               \
+  add span_edge_data, span_edge_data, #8;                                      \
+  subs num_spans, num_spans, #1;                                               \
+                                                                               \
+  strh num_blocks, [ psx_gpu, #psx_gpu_num_blocks_offset ];                    \
+  bne 0b;                                                                      \
+                                                                               \
+  ldmia sp!, { r4 - r11, pc };                                                 \
+                                                                               \
+ 2:                                                                            \
+  /* TODO: Load from psx_gpu instead of saving/restoring these               */\
+  vpush { rg_dx4 };                                                            \
+                                                                               \
+  stmdb sp!, { r0 - r3, r12, r14 };                                            \
+  bl flush_render_block_buffer;                                                \
+  ldmia sp!, { r0 - r3, r12, r14 };                                            \
+                                                                               \
+  vpop { rg_dx4 };                                                             \
+                                                                               \
+  vmov.u8 d64_1, #1;                                                           \
+  vmov.u8 d128_4, #4;                                                          \
+  vmov.u8 d64_128, #128;                                                       \
+  vmov.u8 d128_0x7, #0x7;                                                      \
+                                                                               \
+  vadd.u32 rg_dx8, rg_dx4, rg_dx4;                                             \
+                                                                               \
+  mov num_blocks, span_num_blocks;                                             \
+  add block_ptr_a, psx_gpu, #psx_gpu_blocks_offset;                            \
+  bal 3b                                                                       \
+
+
+setup_blocks_shaded_untextured_indirect_builder(undithered)
+setup_blocks_shaded_untextured_indirect_builder(dithered)
+
+
+#undef draw_mask
+
+#define mask_msb_ptr                                      r14
+
+#define draw_mask                                         q0
+#define pixels_low                                        d16
+
+
+
+#define setup_blocks_shaded_untextured_direct_builder(dithering)               \
+.align 3;                                                                      \
+                                                                               \
+function(setup_blocks_shaded_untextured_##dithering##_unswizzled_direct)       \
+  ldrh num_spans, [ psx_gpu, #psx_gpu_num_spans_offset ];                      \
+  add rg_dx_ptr, psx_gpu, #(psx_gpu_uvrg_dx_offset + 8);                       \
+                                                                               \
+  vld1.u32 { rg_dx }, [ rg_dx_ptr, :64 ];                                      \
+                                                                               \
+  cmp num_spans, #0;                                                           \
+  bxeq lr;                                                                     \
+                                                                               \
+  stmdb sp!, { r4 - r11, r14 };                                                \
+  vshl.u32 rg_dx4, rg_dx, #2;                                                  \
+                                                                               \
+  ldr b_dx, [ psx_gpu, #psx_gpu_b_dx_offset ];                                 \
+  vshl.u32 rg_dx8, rg_dx, #3;                                                  \
+                                                                               \
+  add span_uvrg_offset, psx_gpu, #psx_gpu_span_uvrg_offset_offset;             \
+  add span_edge_data, psx_gpu, #psx_gpu_span_edge_data_offset;                 \
+                                                                               \
+  add span_b_offset, psx_gpu, #psx_gpu_span_b_offset_offset;                   \
+  vmov.u8 d64_1, #1;                                                           \
+                                                                               \
+  vmov.u8 d128_4, #4;                                                          \
+  vmov.u8 d64_128, #128;                                                       \
+                                                                               \
+  vmov.u8 d128_0x7, #0x7;                                                      \
+  add mask_msb_ptr, psx_gpu, #psx_gpu_mask_msb_offset;                         \
+  vld1.u16 { msb_mask_low[], msb_mask_high[] }, [ mask_msb_ptr, :16 ];         \
+                                                                               \
+ 0:                                                                            \
+  ldrh span_num_blocks, [ span_edge_data, #edge_data_num_blocks_offset ];      \
+  add dither_offset_ptr, psx_gpu, #psx_gpu_dither_table_offset;                \
+                                                                               \
+  ldrh y, [ span_edge_data, #edge_data_y_offset ];                             \
+  ldr fb_ptr, [ psx_gpu, #psx_gpu_vram_ptr_offset ];                           \
+                                                                               \
+  cmp span_num_blocks, #0;                                                     \
+  beq 1f;                                                                      \
+                                                                               \
+  ldrh left_x, [ span_edge_data, #edge_data_left_x_offset ];                   \
+  add fb_ptr, fb_ptr, y, lsl #11;                                              \
+                                                                               \
+  ldr b, [ span_b_offset ];                                                    \
+  vdup.u32 v_left_x, left_x;                                                   \
+  and y, y, #0x3;                                                              \
+                                                                               \
+  ldr dither_row, [ dither_offset_ptr, y, lsl #2 ];                            \
+  add fb_ptr, fb_ptr, left_x, lsl #1;                                          \
+                                                                               \
+  mla b, b_dx, left_x, b;                                                      \
+  and dither_shift, left_x, #0x03;                                             \
+                                                                               \
+  vld1.u32 { uvrg }, [ span_uvrg_offset, :128 ];                               \
+  vshr.u32 rg_dx, rg_dx4, #2;                                                  \
+                                                                               \
+  mov dither_shift, dither_shift, lsl #3;                                      \
+  vmla.u32 rg, rg_dx, v_left_x;                                                \
+                                                                               \
+  subs span_num_blocks, span_num_blocks, #1;                                   \
+                                                                               \
+  mov dither_row, dither_row, ror dither_shift;                                \
+  mov b_dx4, b_dx, lsl #2;                                                     \
+                                                                               \
+  vdup.u32 dither_offsets, dither_row;                                         \
+  add block_span_ptr, psx_gpu, #psx_gpu_r_block_span_offset;                   \
+                                                                               \
+  vdup.u32 b_block, b;                                                         \
+  vadd.u8 dither_offsets, dither_offsets, d128_4;                              \
+                                                                               \
+  mov b_dx8, b_dx, lsl #3;                                                     \
+  vdup.u32 r_block, rg[0];                                                     \
+  vdup.u32 g_block, rg[1];                                                     \
+                                                                               \
+  vld1.u32 { block_span }, [ block_span_ptr, :128 ]!;                          \
+                                                                               \
+  vadd.u32 r_block, r_block, block_span;                                       \
+  vld1.u32 { block_span }, [ block_span_ptr, :128 ]!;                          \
+                                                                               \
+  vadd.u32 g_block, g_block, block_span;                                       \
+  vld1.u32 { block_span }, [ block_span_ptr, :128 ];                           \
+                                                                               \
+  vadd.u32 b_block, b_block, block_span;                                       \
+  add block_ptr_b, block_ptr_a, #16;                                           \
+                                                                               \
+  vshrn.u32 r_whole_low, r_block, #16;                                         \
+  vshrn.u32 g_whole_low, g_block, #16;                                         \
+  vshrn.u32 b_whole_low, b_block, #16;                                         \
+  vdup.u32 dx4, rg_dx4[0];                                                     \
+                                                                               \
+  vaddhn.u32 r_whole_high, r_block, dx4;                                       \
+  vdup.u32 dx4, rg_dx4[1];                                                     \
+                                                                               \
+  vaddhn.u32 g_whole_high, g_block, dx4;                                       \
+  vdup.u32 dx4, b_dx4;                                                         \
+                                                                               \
+  vaddhn.u32 b_whole_high, b_block, dx4;                                       \
+  vdup.u32 dx8, rg_dx8[0];                                                     \
+                                                                               \
+  vadd.u32 r_block, r_block, dx8;                                              \
+  vdup.u32 dx8, rg_dx8[1];                                                     \
+                                                                               \
+  vadd.u32 g_block, g_block, dx8;                                              \
+  vdup.u32 dx8, b_dx8;                                                         \
+                                                                               \
+  vadd.u32 b_block, b_block, dx8;                                              \
+                                                                               \
+  vmovn.u16 r_whole_8, r_whole;                                                \
+  vmovn.u16 g_whole_8, g_whole;                                                \
+  vmovn.u16 b_whole_8, b_whole;                                                \
+                                                                               \
+  beq 3f;                                                                      \
+                                                                               \
+ 2:                                                                            \
+  setup_blocks_shaded_untextured_dither_a_##dithering();                       \
+  vshrn.u32 r_whole_low, r_block, #16;                                         \
+                                                                               \
+  setup_blocks_shaded_untextured_dither_b_##dithering();                       \
+  vshrn.u32 g_whole_low, g_block, #16;                                         \
+                                                                               \
+  vshrn.u32 b_whole_low, b_block, #16;                                         \
+                                                                               \
+  vdup.u32 dx4, rg_dx4[0];                                                     \
+  vshr.u8 r_whole_8, r_whole_8, #3;                                            \
+  vbic.u8 gb_whole_8, gb_whole_8, d128_0x7;                                    \
+                                                                               \
+  vaddhn.u32 r_whole_high, r_block, dx4;                                       \
+  vdup.u32 dx4, rg_dx4[1];                                                     \
+                                                                               \
+  vmov pixels, msb_mask;                                                       \
+  vaddhn.u32 g_whole_high, g_block, dx4;                                       \
+  vdup.u32 dx4, b_dx4;                                                         \
+                                                                               \
+  vaddhn.u32 b_whole_high, b_block, dx4;                                       \
+  vdup.u32 dx8, rg_dx8[0];                                                     \
+                                                                               \
+  vmlal.u8 pixels, r_whole_8, d64_1;                                           \
+  vmlal.u8 pixels, g_whole_8, d64_4;                                           \
+  vmlal.u8 pixels, b_whole_8, d64_128;                                         \
+                                                                               \
+  vadd.u32 r_block, r_block, dx8;                                              \
+  vdup.u32 dx8, rg_dx8[1];                                                     \
+                                                                               \
+  vadd.u32 g_block, g_block, dx8;                                              \
+  vdup.u32 dx8, b_dx8;                                                         \
+                                                                               \
+  vadd.u32 b_block, b_block, dx8;                                              \
+                                                                               \
+  vmovn.u16 r_whole_8, r_whole;                                                \
+  vmovn.u16 g_whole_8, g_whole;                                                \
+  vmovn.u16 b_whole_8, b_whole;                                                \
+                                                                               \
+  vst1.u32 { pixels }, [ fb_ptr ]!;                                            \
+  subs span_num_blocks, span_num_blocks, #1;                                   \
+  bne 2b;                                                                      \
+                                                                               \
+ 3:                                                                            \
+  setup_blocks_shaded_untextured_dither_a_##dithering();                       \
+                                                                               \
+  ldrb right_mask, [ span_edge_data, #edge_data_right_mask_offset ];           \
+  setup_blocks_shaded_untextured_dither_b_##dithering();                       \
+                                                                               \
+  vshr.u8 r_whole_8, r_whole_8, #3;                                            \
+  vmov pixels, msb_mask;                                                       \
+  vbic.u8 gb_whole_8, gb_whole_8, d128_0x7;                                    \
+  eor right_mask, right_mask, #0xFF;                                           \
+                                                                               \
+  vmlal.u8 pixels, r_whole_8, d64_1;                                           \
+  vmlal.u8 pixels, g_whole_8, d64_4;                                           \
+  vmlal.u8 pixels, b_whole_8, d64_128;                                         \
+                                                                               \
+ 4:                                                                            \
+  vst1.u16 { pixels_low[0] }, [ fb_ptr ]!;                                     \
+  vext.16 pixels, pixels, #1;                                                  \
+  movs right_mask, right_mask, lsr #1;                                         \
+  bne 4b;                                                                      \
+                                                                               \
+ 1:                                                                            \
+  add span_uvrg_offset, span_uvrg_offset, #16;                                 \
+  add span_b_offset, span_b_offset, #4;                                        \
+                                                                               \
+  add span_edge_data, span_edge_data, #8;                                      \
+  subs num_spans, num_spans, #1;                                               \
+                                                                               \
+  bne 0b;                                                                      \
+                                                                               \
+  ldmia sp!, { r4 - r11, pc }                                                  \
+
+setup_blocks_shaded_untextured_direct_builder(undithered)
+setup_blocks_shaded_untextured_direct_builder(dithered)
+
+
+#undef psx_gpu
+#undef num_blocks
+#undef triangle
+#undef c_64
+
+#define psx_gpu                                  r0
+#define block_ptr                                r1
+#define num_blocks                               r2
+#define uv_01                                    r3
+#define uv_23                                    r4
+#define uv_45                                    r5
+#define uv_67                                    r6
+#define uv_0                                     r7
+#define uv_1                                     r3
+#define uv_2                                     r8
+#define uv_3                                     r4
+#define uv_4                                     r9
+#define uv_5                                     r5
+#define uv_6                                     r10
+#define uv_7                                     r6
+#define texture_ptr                              r11
+
+#define pixel_0                                  r7
+#define pixel_1                                  r3
+#define pixel_2                                  r8
+#define pixel_3                                  r4
+#define pixel_4                                  r9
+#define pixel_5                                  r5
+#define pixel_6                                  r10
+#define pixel_7                                  r6
+
+#define pixels_a                                 r7
+#define pixels_b                                 r9
+#define pixels_c                                 r8
+#define pixels_d                                 r10
+
+#define c_64                                     r0
+
+#define clut_ptr                                 r12
+#define current_texture_mask                     r5
+#define dirty_textures_mask                      r6
+
+#define texels                                   d0
+
+#define clut_low_a                               d2
+#define clut_low_b                               d3
+#define clut_high_a                              d4
+#define clut_high_b                              d5
+
+#define clut_a                                   q1
+#define clut_b                                   q2
+
+#define texels_low                               d6
+#define texels_high                              d7
+
+.align 3
+
+function(texture_blocks_untextured)
+  bx lr
+
+
+.align 3
+
+function(texture_blocks_4bpp)
+  stmdb sp!, { r3 - r11, r14 }
+  add block_ptr, psx_gpu, #psx_gpu_blocks_offset
+
+  ldr texture_ptr, [ psx_gpu, #psx_gpu_texture_page_ptr_offset ]
+  ldrh num_blocks, [ psx_gpu, #psx_gpu_num_blocks_offset ]
+
+  ldr clut_ptr, [ psx_gpu, #psx_gpu_clut_ptr_offset ]
+  vld1.u32 { clut_a, clut_b }, [ clut_ptr, :128 ]
+
+  ldr current_texture_mask, [ psx_gpu, #psx_gpu_current_texture_mask_offset ]
+  vuzp.u8 clut_a, clut_b
+
+  ldr dirty_textures_mask, [ psx_gpu, #psx_gpu_dirty_textures_4bpp_mask_offset ]
+  tst dirty_textures_mask, current_texture_mask
+
+  bne 1f
+  mov c_64, #64
+
+0:
+  ldm block_ptr, { uv_01, uv_23, uv_45, uv_67 }
+
+  uxtah uv_0, texture_ptr, uv_01
+  uxtah uv_1, texture_ptr, uv_01, ror #16
+
+  uxtah uv_2, texture_ptr, uv_23
+  uxtah uv_3, texture_ptr, uv_23, ror #16
+
+  uxtah uv_4, texture_ptr, uv_45
+  ldrb pixel_0, [ uv_0 ]
+
+  uxtah uv_5, texture_ptr, uv_45, ror #16
+  ldrb pixel_1, [ uv_1 ]
+
+  uxtah uv_6, texture_ptr, uv_67
+  ldrb pixel_2, [ uv_2 ]
+
+  uxtah uv_7, texture_ptr, uv_67, ror #16
+  ldrb pixel_3, [ uv_3 ]
+
+  ldrb pixel_4, [ uv_4 ]
+  subs num_blocks, num_blocks, #1
+
+  ldrb pixel_5, [ uv_5 ]
+  orr pixels_a, pixel_0, pixel_1, lsl #8
+
+  ldrb pixel_6, [ uv_6 ]
+  orr pixels_b, pixel_4, pixel_5, lsl #8
+
+  ldrb pixel_7, [ uv_7 ]
+  orr pixels_a, pixels_a, pixel_2, lsl #16
+
+  orr pixels_b, pixels_b, pixel_6, lsl #16
+  orr pixels_a, pixels_a, pixel_3, lsl #24
+
+  orr pixels_b, pixels_b, pixel_7, lsl #24
+  vmov.u32 texels, pixels_a, pixels_b
+
+  vtbl.8 texels_low, { clut_low_a, clut_low_b }, texels
+  vtbl.8 texels_high, { clut_high_a, clut_high_b }, texels
+
+  vst2.u8 { texels_low, texels_high }, [ block_ptr, :128 ], c_64
+  bne 0b
+
+  ldmia sp!, { r3 - r11, pc }
+
+1:
+  stmdb sp!, { r1 - r2 }  
+  bl update_texture_4bpp_cache
+
+  mov c_64, #64
+  ldmia sp!, { r1 - r2 }
+  bal 0b
+
+
+.align 3
+
+function(texture_blocks_8bpp)
+  stmdb sp!, { r3 - r11, r14 }
+  add block_ptr, psx_gpu, #psx_gpu_blocks_offset
+
+  ldr texture_ptr, [ psx_gpu, #psx_gpu_texture_page_ptr_offset ]
+  ldrh num_blocks, [ psx_gpu, #psx_gpu_num_blocks_offset ]
+
+  ldr clut_ptr, [ psx_gpu, #psx_gpu_clut_ptr_offset ]
+  ldr current_texture_mask, [ psx_gpu, #psx_gpu_current_texture_mask_offset ]
+
+  ldr dirty_textures_mask, [ psx_gpu, #psx_gpu_dirty_textures_8bpp_mask_offset ]
+  tst dirty_textures_mask, current_texture_mask
+
+  bne 1f
+  nop
+
+0:
+  ldm block_ptr, { uv_01, uv_23, uv_45, uv_67 }
+
+  uxtah uv_0, texture_ptr, uv_01
+  uxtah uv_1, texture_ptr, uv_01, ror #16
+
+  uxtah uv_2, texture_ptr, uv_23
+  uxtah uv_3, texture_ptr, uv_23, ror #16
+
+  uxtah uv_4, texture_ptr, uv_45
+  ldrb pixel_0, [ uv_0 ]
+
+  uxtah uv_5, texture_ptr, uv_45, ror #16
+  ldrb pixel_1, [ uv_1 ]
+
+  uxtah uv_6, texture_ptr, uv_67
+  ldrb pixel_2, [ uv_2 ]
+
+  uxtah uv_7, texture_ptr, uv_67, ror #16
+  ldrb pixel_3, [ uv_3 ]
+
+  ldrb pixel_4, [ uv_4 ]
+  add pixel_0, pixel_0, pixel_0
+
+  ldrb pixel_5, [ uv_5 ]
+  add pixel_1, pixel_1, pixel_1
+
+  ldrb pixel_6, [ uv_6 ]
+  add pixel_2, pixel_2, pixel_2
+
+  ldrb pixel_7, [ uv_7 ]
+  add pixel_3, pixel_3, pixel_3
+
+  ldrh pixel_0, [ clut_ptr, pixel_0 ]
+  add pixel_4, pixel_4, pixel_4
+
+  ldrh pixel_1, [ clut_ptr, pixel_1 ]
+  add pixel_5, pixel_5, pixel_5
+
+  ldrh pixel_2, [ clut_ptr, pixel_2 ]
+  add pixel_6, pixel_6, pixel_6
+
+  ldrh pixel_3, [ clut_ptr, pixel_3 ]
+  add pixel_7, pixel_7, pixel_7
+
+  ldrh pixel_4, [ clut_ptr, pixel_4 ]
+  orr pixels_a, pixel_0, pixel_1, lsl #16
+
+  ldrh pixel_5, [ clut_ptr, pixel_5 ]
+  orr pixels_c, pixel_2, pixel_3, lsl #16
+
+  ldrh pixel_6, [ clut_ptr, pixel_6 ]
+  subs num_blocks, num_blocks, #1
+
+  ldrh pixel_7, [ clut_ptr, pixel_7 ]
+  orr pixels_b, pixel_4, pixel_5, lsl #16
+
+  orr pixels_d, pixel_6, pixel_7, lsl #16
+  stm block_ptr, { pixels_a, pixels_c, pixels_b, pixels_d } 
+
+  add block_ptr, block_ptr, #64
+  bne 0b
+
+  ldmia sp!, { r3 - r11, pc }
+
+1:
+  stmdb sp!, { r1 - r2, r12 }
+
+  bl update_texture_8bpp_cache
+
+  ldmia sp!, { r1 - r2, r12 }
+  bal 0b
+
+
+#undef uv_0
+#undef uv_1
+#undef uv_2
+#undef uv_3
+#undef uv_4
+#undef uv_5
+#undef uv_6
+#undef uv_7
+
+#undef pixel_0
+#undef pixel_1
+#undef pixel_2
+#undef pixel_3
+#undef pixel_4
+#undef pixel_5
+#undef pixel_6
+#undef pixel_7
+
+#undef texture_ptr
+
+#undef pixels_a
+#undef pixels_b
+#undef pixels_c
+#undef pixels_d
+
+#define psx_gpu                                  r0
+#define block_ptr                                r1
+#define num_blocks                               r2
+
+#define uv_0                                     r3
+#define uv_1                                     r4
+#define u_0                                      r3
+#define u_1                                      r4
+#define v_0                                      r5
+#define v_1                                      r6
+
+#define uv_2                                     r5
+#define uv_3                                     r6
+#define u_2                                      r5
+#define u_3                                      r6
+#define v_2                                      r7
+#define v_3                                      r8
+
+#define uv_4                                     r7
+#define uv_5                                     r8
+#define u_4                                      r7
+#define u_5                                      r8
+#define v_4                                      r9
+#define v_5                                      r10
+
+#define uv_6                                     r9
+#define uv_7                                     r10
+#define u_6                                      r9
+#define u_7                                      r10
+#define v_6                                      r11
+#define v_7                                      r0
+
+#define pixel_0                                  r3
+#define pixel_1                                  r4
+#define pixel_2                                  r5
+#define pixel_3                                  r6
+#define pixel_4                                  r7
+#define pixel_5                                  r8
+#define pixel_6                                  r9
+#define pixel_7                                  r10
+
+#define pixels_a                                 r3
+#define pixels_b                                 r5
+#define pixels_c                                 r7
+#define pixels_d                                 r9
+
+#define texture_ptr                              r12
+
+
+.align 3
+
+function(texture_blocks_16bpp)
+  stmdb sp!, { r3 - r11, r14 }
+  add block_ptr, psx_gpu, #psx_gpu_blocks_offset
+
+  ldrh num_blocks, [ psx_gpu, #psx_gpu_num_blocks_offset ]
+  ldr texture_ptr, [ psx_gpu, #psx_gpu_texture_page_ptr_offset ]
+
+0:
+  ldrh uv_0, [ block_ptr ]
+  subs num_blocks, num_blocks, #1
+
+  ldrh uv_1, [ block_ptr, #2 ]
+
+  and v_0, uv_0, #0xFF00
+  and v_1, uv_1, #0xFF00
+
+  and u_0, uv_0, #0xFF
+  and u_1, uv_1, #0xFF
+
+  add uv_0, u_0, v_0, lsl #2
+  ldrh uv_2, [ block_ptr, #4 ]
+
+  add uv_1, u_1, v_1, lsl #2
+  ldrh uv_3, [ block_ptr, #6 ]
+
+  add uv_0, uv_0, uv_0
+  add uv_1, uv_1, uv_1
+
+  and v_2, uv_2, #0xFF00
+  and v_3, uv_3, #0xFF00
+
+  and u_2, uv_2, #0xFF
+  and u_3, uv_3, #0xFF
+
+  add uv_2, u_2, v_2, lsl #2
+  ldrh uv_4, [ block_ptr, #8 ]
+
+  add uv_3, u_3, v_3, lsl #2
+  ldrh uv_5, [ block_ptr, #10 ]
+
+  add uv_2, uv_2, uv_2
+  add uv_3, uv_3, uv_3
+
+  and v_4, uv_4, #0xFF00
+  and v_5, uv_5, #0xFF00
+
+  and u_4, uv_4, #0xFF
+  and u_5, uv_5, #0xFF
+
+  add uv_4, u_4, v_4, lsl #2
+  ldrh uv_6, [ block_ptr, #12 ]
+
+  add uv_5, u_5, v_5, lsl #2
+  ldrh uv_7, [ block_ptr, #14 ]
+
+  add uv_4, uv_4, uv_4
+  ldrh pixel_0, [ texture_ptr, uv_0 ]
+
+  add uv_5, uv_5, uv_5
+  ldrh pixel_1, [ texture_ptr, uv_1 ]
+
+  and v_6, uv_6, #0xFF00
+  ldrh pixel_2, [ texture_ptr, uv_2 ]
+
+  and v_7, uv_7, #0xFF00
+  ldrh pixel_3, [ texture_ptr, uv_3 ]
+
+  and u_6, uv_6, #0xFF
+  ldrh pixel_4, [ texture_ptr, uv_4 ]
+
+  and u_7, uv_7, #0xFF
+  ldrh pixel_5, [ texture_ptr, uv_5 ]
+
+  add uv_6, u_6, v_6, lsl #2
+  add uv_7, u_7, v_7, lsl #2
+
+  add uv_6, uv_6, uv_6
+  add uv_7, uv_7, uv_7
+
+  orr pixels_a, pixel_0, pixel_1, lsl #16
+  orr pixels_b, pixel_2, pixel_3, lsl #16
+
+  ldrh pixel_6, [ texture_ptr, uv_6 ]
+  orr pixels_c, pixel_4, pixel_5, lsl #16
+
+  ldrh pixel_7, [ texture_ptr, uv_7 ]
+  orr pixels_d, pixel_6, pixel_7, lsl #16
+
+  stm block_ptr, { pixels_a, pixels_b, pixels_c, pixels_d }
+  add block_ptr, block_ptr, #64
+
+  bne 0b
+
+  ldmia sp!, { r3 - r11, pc }
+
+
+#undef num_blocks
+
+#undef test_mask
+#undef texels
+#undef pixels_b
+#undef pixels
+#undef d64_1
+#undef d64_4
+#undef d64_128
+#undef draw_mask
+#undef msb_mask
+#undef msb_mask_low
+#undef msb_mask_high
+#undef fb_pixels
+
+#undef c_32
+#undef fb_ptr
+#undef mask_msb_ptr
+
+#define psx_gpu                                  r0
+#define num_blocks                               r1
+#define color_ptr                                r2
+#define mask_msb_ptr                             r2
+
+#define block_ptr_load_a                         r0
+#define block_ptr_store                          r3
+#define block_ptr_load_b                         r12
+#define c_32                                     r2
+
+#define c_48                                     r4
+#define fb_ptr                                   r14
+#define draw_mask_bits_scalar                    r5
+
+#define d128_0x07                                q0
+#define d128_0x1F                                q1
+#define d128_0x8000                              q2
+#define test_mask                                q3
+#define texels                                   q4
+#define colors_rg                                q5
+#define colors_b_dm_bits                         q6
+#define texels_rg                                q7
+#define pixels_r                                 q8
+#define pixels_g                                 q9
+#define pixels_b                                 q10
+#define pixels                                   q11
+#define zero_mask                                q4
+#define draw_mask                                q12
+#define msb_mask                                 q13
+
+#define fb_pixels                                q8
+
+#define pixels_gb_low                            q9
+
+#define colors_r                                 d10
+#define colors_g                                 d11
+#define colors_b                                 d12
+#define draw_mask_bits                           d13
+#define texels_r                                 d14
+#define texels_g                                 d15
+#define pixels_r_low                             d16
+#define pixels_g_low                             d18
+#define pixels_b_low                             d19
+#define msb_mask_low                             d26
+#define msb_mask_high                            d27
+
+#define d64_1                                    d28
+#define d64_4                                    d29
+#define d64_128                                  d30
+#define texels_b                                 d31
+
+#define shade_blocks_textured_modulated_prologue_indirect()                    \
+  mov c_48, #48;                                                               \
+  add block_ptr_store, psx_gpu, #psx_gpu_blocks_offset                         \
+
+#define shade_blocks_textured_modulated_prologue_direct()                      \
+  add mask_msb_ptr, psx_gpu, #psx_gpu_mask_msb_offset;                         \
+  vld1.u16 { msb_mask_low[], msb_mask_high[] }, [ mask_msb_ptr, :16 ]          \
+
+#define shade_blocks_textured_modulated_prologue_shaded()                      \
+
+#define shade_blocks_textured_modulated_prologue_unshaded()                    \
+  add color_ptr, psx_gpu, #psx_gpu_triangle_color_offset;                      \
+  vld1.u32 { colors_r[] }, [ color_ptr, :32 ];                                 \
+  vdup.u8 colors_g, colors_r[1];                                               \
+  vdup.u8 colors_b, colors_r[2];                                               \
+  vdup.u8 colors_r, colors_r[0]                                                \
+
+
+#define shade_blocks_textured_modulated_load_dithered(target)                  \
+  vld1.u32 { target }, [ block_ptr_load_b, :128 ]                              \
+
+#define shade_blocks_textured_modulated_load_last_dithered(target)             \
+  vld1.u32 { target }, [ block_ptr_load_b, :128 ], c_32                        \
+
+#define shade_blocks_textured_modulated_load_undithered(target)                \
+
+#define shade_blocks_textured_modulated_load_last_undithered(target)           \
+  add block_ptr_load_b, block_ptr_load_b, #32                                  \
+
+#define shade_blocks_textured_modulate_dithered(channel)                       \
+  vmlal.u8 pixels_##channel, texels_##channel, colors_##channel                \
+
+#define shade_blocks_textured_modulate_undithered(channel)                     \
+  vmull.u8 pixels_##channel, texels_##channel, colors_##channel                \
+
+
+#define shade_blocks_textured_modulated_store_draw_mask_indirect(offset)       \
+  vst1.u32 { draw_mask }, [ block_ptr_store, :128 ]!                           \
+
+#define shade_blocks_textured_modulated_store_draw_mask_direct(offset)         \
+  ldr fb_ptr, [ block_ptr_load_b, #(offset - 64) ];                            \
+  vld1.u32 { fb_pixels }, [ fb_ptr ];                                          \
+  vbit.u16 pixels, fb_pixels, draw_mask                                        \
+
+#define shade_blocks_textured_modulated_store_pixels_indirect()                \
+  vst1.u32 { pixels }, [ block_ptr_store, :128 ], c_48                         \
+
+#define shade_blocks_textured_modulated_store_pixels_direct()                  \
+  vst1.u32 { pixels }, [ fb_ptr ]                                              \
+
+
+#define shade_blocks_textured_modulated_load_rg_shaded()                       \
+  vld1.u32 { colors_r, colors_g }, [ block_ptr_load_b, :128 ], c_32            \
+
+#define shade_blocks_textured_modulated_load_rg_unshaded()                     \
+  add block_ptr_load_b, block_ptr_load_b, #32                                  \
+
+#define shade_blocks_textured_modulated_load_bdm_shaded()                      \
+  vld1.u32 { colors_b, draw_mask_bits }, [ block_ptr_load_a, :128 ], c_32      \
+
+#define shade_blocks_textured_modulated_load_bdm_unshaded()                    \
+  ldr draw_mask_bits_scalar, [ block_ptr_load_a, #8 ];                         \
+  add block_ptr_load_a, block_ptr_load_a, #32                                  \
+
+#define shade_blocks_textured_modulated_expand_draw_mask_shaded()              \
+  vdup.u16 draw_mask, draw_mask_bits[0]                                        \
+
+#define shade_blocks_textured_modulated_expand_draw_mask_unshaded()            \
+  vdup.u16 draw_mask, draw_mask_bits_scalar                                    \
+
+
+#define shade_blocks_textured_modulated_apply_msb_mask_indirect()              \
+
+#define shade_blocks_textured_modulated_apply_msb_mask_direct()                \
+  vorr.u16 pixels, pixels, msb_mask                                            \
+
+
+#define shade_blocks_textured_modulated_builder(shading, dithering, target)    \
+.align 3;                                                                      \
+                                                                               \
+function(shade_blocks_##shading##_textured_modulated_##dithering##_##target)   \
+  stmdb sp!, { r4 - r5, lr };                                                  \
+  ldrh num_blocks, [ psx_gpu, #psx_gpu_num_blocks_offset ];                    \
+                                                                               \
+  vld1.u32 { test_mask }, [ psx_gpu, :128 ];                                   \
+                                                                               \
+  shade_blocks_textured_modulated_prologue_##target();                         \
+  shade_blocks_textured_modulated_prologue_##shading();                        \
+                                                                               \
+  add block_ptr_load_a, psx_gpu, #psx_gpu_blocks_offset;                       \
+  mov c_32, #32;                                                               \
+                                                                               \
+  add block_ptr_load_b, block_ptr_load_a, #16;                                 \
+  vmov.u8 d64_1, #1;                                                           \
+  vmov.u8 d64_4, #4;                                                           \
+  vmov.u8 d64_128, #128;                                                       \
+                                                                               \
+  vld1.u32 { texels }, [ block_ptr_load_a, :128 ], c_32;                       \
+  vmov.u8 d128_0x07, #0x07;                                                    \
+                                                                               \
+  shade_blocks_textured_modulated_load_rg_##shading();                         \
+  vmov.u8 d128_0x1F, #0x1F;                                                    \
+                                                                               \
+  shade_blocks_textured_modulated_load_bdm_##shading();                        \
+  vmov.u16 d128_0x8000, #0x8000;                                               \
+                                                                               \
+  vmovn.u16 texels_r, texels;                                                  \
+  vshrn.u16 texels_g, texels, #5;                                              \
+                                                                               \
+  vshrn.u16 texels_b, texels, #7;                                              \
+  shade_blocks_textured_modulated_expand_draw_mask_##shading();                \
+                                                                               \
+  shade_blocks_textured_modulated_load_##dithering(pixels_r);                  \
+  vtst.u16 draw_mask, draw_mask, test_mask;                                    \
+                                                                               \
+  shade_blocks_textured_modulated_load_##dithering(pixels_g);                  \
+  vand.u8 texels_rg, texels_rg, d128_0x1F;                                     \
+                                                                               \
+  shade_blocks_textured_modulated_load_last_##dithering(pixels_b);             \
+  vshr.u8 texels_b, texels_b, #3;                                              \
+                                                                               \
+  shade_blocks_textured_modulate_##dithering(r);                               \
+  shade_blocks_textured_modulate_##dithering(g);                               \
+  shade_blocks_textured_modulate_##dithering(b);                               \
+                                                                               \
+  vand.u16 pixels, texels, d128_0x8000;                                        \
+  vceq.u16 zero_mask, texels, #0;                                              \
+                                                                               \
+  vqshrun.s16 pixels_r_low, pixels_r, #4;                                      \
+  vqshrun.s16 pixels_g_low, pixels_g, #4;                                      \
+  vqshrun.s16 pixels_b_low, pixels_b, #4;                                      \
+                                                                               \
+  shade_blocks_textured_modulated_apply_msb_mask_##target();                   \
+  vorr.u16 draw_mask, draw_mask, zero_mask;                                    \
+  vshr.u8 pixels_r_low, pixels_r_low, #3;                                      \
+  vbic.u8 pixels_gb_low, pixels_gb_low, d128_0x07;                             \
+                                                                               \
+  subs num_blocks, num_blocks, #1;                                             \
+  beq 1f;                                                                      \
+                                                                               \
+ .align 3;                                                                     \
+                                                                               \
+ 0:                                                                            \
+  vld1.u32 { texels }, [ block_ptr_load_a, :128 ], c_32;                       \
+  shade_blocks_textured_modulated_load_rg_##shading();                         \
+  vshrn.u16 texels_g, texels, #5;                                              \
+                                                                               \
+  shade_blocks_textured_modulated_load_bdm_##shading();                        \
+  vshrn.u16 texels_b, texels, #7;                                              \
+                                                                               \
+  vmovn.u16 texels_r, texels;                                                  \
+  vmlal.u8 pixels, pixels_r_low, d64_1;                                        \
+                                                                               \
+  vmlal.u8 pixels, pixels_g_low, d64_4;                                        \
+  vmlal.u8 pixels, pixels_b_low, d64_128;                                      \
+  shade_blocks_textured_modulated_store_draw_mask_##target(-4);                \
+                                                                               \
+  shade_blocks_textured_modulated_load_##dithering(pixels_r);                  \
+  shade_blocks_textured_modulated_expand_draw_mask_##shading();                \
+                                                                               \
+  shade_blocks_textured_modulated_load_##dithering(pixels_g);                  \
+  vand.u8 texels_rg, texels_rg, d128_0x1F;                                     \
+                                                                               \
+  shade_blocks_textured_modulated_load_last_##dithering(pixels_b);             \
+  vtst.u16 draw_mask, draw_mask, test_mask;                                    \
+                                                                               \
+  shade_blocks_textured_modulated_store_pixels_##target();                     \
+  vshr.u8 texels_b, texels_b, #3;                                              \
+                                                                               \
+  shade_blocks_textured_modulate_##dithering(r);                               \
+  shade_blocks_textured_modulate_##dithering(g);                               \
+  shade_blocks_textured_modulate_##dithering(b);                               \
+                                                                               \
+  vand.u16 pixels, texels, d128_0x8000;                                        \
+  vceq.u16 zero_mask, texels, #0;                                              \
+                                                                               \
+  subs num_blocks, num_blocks, #1;                                             \
+                                                                               \
+  vqshrun.s16 pixels_r_low, pixels_r, #4;                                      \
+  vqshrun.s16 pixels_g_low, pixels_g, #4;                                      \
+  vqshrun.s16 pixels_b_low, pixels_b, #4;                                      \
+                                                                               \
+  shade_blocks_textured_modulated_apply_msb_mask_##target();                   \
+  vorr.u16 draw_mask, draw_mask, zero_mask;                                    \
+  vshr.u8 pixels_r_low, pixels_r_low, #3;                                      \
+  vbic.u8 pixels_gb_low, pixels_gb_low, d128_0x07;                             \
+                                                                               \
+  bne 0b;                                                                      \
+                                                                               \
+ 1:                                                                            \
+  vmlal.u8 pixels, pixels_r_low, d64_1;                                        \
+  vmlal.u8 pixels, pixels_g_low, d64_4;                                        \
+  vmlal.u8 pixels, pixels_b_low, d64_128;                                      \
+                                                                               \
+  shade_blocks_textured_modulated_store_draw_mask_##target(28);                \
+  shade_blocks_textured_modulated_store_pixels_##target();                     \
+                                                                               \
+  ldmia sp!, { r4 - r5, pc }                                                   \
+
+
+shade_blocks_textured_modulated_builder(shaded, dithered, direct);
+shade_blocks_textured_modulated_builder(shaded, undithered, direct);
+shade_blocks_textured_modulated_builder(unshaded, dithered, direct);
+shade_blocks_textured_modulated_builder(unshaded, undithered, direct);
+
+shade_blocks_textured_modulated_builder(shaded, dithered, indirect);
+shade_blocks_textured_modulated_builder(shaded, undithered, indirect);
+shade_blocks_textured_modulated_builder(unshaded, dithered, indirect);
+shade_blocks_textured_modulated_builder(unshaded, undithered, indirect);
+
+
+#undef c_64
+#undef fb_ptr
+#undef color_ptr
+
+#undef color_r
+#undef color_g
+#undef color_b
+
+#undef test_mask
+#undef pixels
+#undef draw_mask
+#undef zero_mask
+#undef fb_pixels
+#undef msb_mask
+#undef msb_mask_low
+#undef msb_mask_high
+
+#define psx_gpu                                  r0
+#define num_blocks                               r1
+#define mask_msb_ptr                             r2
+#define color_ptr                                r3
+
+#define block_ptr_load                           r0
+#define draw_mask_store_ptr                      r3
+#define draw_mask_bits_ptr                       r12
+#define draw_mask_ptr                            r12
+#define pixel_store_ptr                          r14
+
+#define fb_ptr_cmp                               r4
+
+#define fb_ptr                                   r3
+#define fb_ptr_next                              r14
+
+#define c_64                                     r2
+
+#define test_mask                                q0
+#define pixels                                   q1
+#define draw_mask                                q2
+#define zero_mask                                q3
+#define draw_mask_combined                       q4
+#define fb_pixels                                q5
+#define fb_pixels_next                           q6
+#define msb_mask                                 q7
+
+#define draw_mask_low                            d4
+#define draw_mask_high                           d5
+#define msb_mask_low                             d14
+#define msb_mask_high                            d15
+
+.align 3
+function(shade_blocks_textured_unmodulated_indirect)
+  str r14, [ sp, #-4 ]
+  add draw_mask_bits_ptr, psx_gpu, #(psx_gpu_blocks_offset + 40)
+
+  ldrh num_blocks, [ psx_gpu, #psx_gpu_num_blocks_offset ]
+  add pixel_store_ptr, psx_gpu, #(psx_gpu_blocks_offset + 16)
+
+  vld1.u32 { test_mask }, [ psx_gpu, :128 ]
+  add draw_mask_store_ptr, psx_gpu, #psx_gpu_blocks_offset
+
+  mov c_64, #64
+  add block_ptr_load, psx_gpu, #psx_gpu_blocks_offset
+
+  vld1.u32 { pixels }, [ block_ptr_load, :128 ], c_64
+  vld1.u16 { draw_mask_low[], draw_mask_high[] },                              \
+   [ draw_mask_bits_ptr, :16 ], c_64
+  vceq.u16 zero_mask, pixels, #0
+
+  vtst.u16 draw_mask, draw_mask, test_mask
+  vst1.u32 { pixels }, [ pixel_store_ptr, :128 ], c_64
+
+  subs num_blocks, num_blocks, #1
+  beq 1f
+
+ 0:
+  vld1.u32 { pixels }, [ block_ptr_load, :128 ], c_64
+  vorr.u16 draw_mask_combined, draw_mask, zero_mask
+
+  vld1.u16 { draw_mask_low[], draw_mask_high[] },                              \
+   [ draw_mask_bits_ptr, :16 ], c_64
+  vceq.u16 zero_mask, pixels, #0
+
+  vtst.u16 draw_mask, draw_mask, test_mask
+  vst1.u32 { pixels }, [ pixel_store_ptr, :128 ], c_64
+
+  vst1.u32 { draw_mask_combined }, [ draw_mask_store_ptr, :128 ], c_64
+  subs num_blocks, num_blocks, #1
+
+  bne 0b
+
+ 1:
+  vorr.u16 draw_mask_combined, draw_mask, zero_mask
+  vst1.u32 { draw_mask_combined }, [ draw_mask_store_ptr, :128 ], c_64
+
+  ldr pc, [ sp, #-4 ]
+
+
+.align 3
+
+function(shade_blocks_textured_unmodulated_direct)
+  stmdb sp!, { r4, r14 }
+  add draw_mask_bits_ptr, psx_gpu, #(psx_gpu_blocks_offset + 40)
+
+  ldrh num_blocks, [ psx_gpu, #psx_gpu_num_blocks_offset ]
+  add mask_msb_ptr, psx_gpu, #psx_gpu_mask_msb_offset
+
+  vld1.u16 { msb_mask_low[], msb_mask_high[] }, [ mask_msb_ptr, :16 ]
+  mov c_64, #64
+
+  vld1.u32 { test_mask }, [ psx_gpu, :128 ]
+  add block_ptr_load, psx_gpu, #psx_gpu_blocks_offset
+
+  vld1.u16 { draw_mask_low[], draw_mask_high[] },                              \
+   [ draw_mask_bits_ptr, :16 ], c_64
+  ldr fb_ptr_next, [ block_ptr_load, #44 ]
+
+  vld1.u32 { pixels }, [ block_ptr_load, :128 ], c_64
+  vld1.u16 { fb_pixels_next }, [ fb_ptr_next ]
+  vceq.u16 zero_mask, pixels, #0
+  vtst.u16 draw_mask, draw_mask, test_mask
+
+  subs num_blocks, num_blocks, #1
+  beq 1f
+
+ 0:
+  mov fb_ptr, fb_ptr_next
+  ldr fb_ptr_next, [ block_ptr_load, #44 ]
+
+  vorr.u16 pixels, pixels, msb_mask
+
+  vorr.u16 draw_mask_combined, draw_mask, zero_mask
+  vmov fb_pixels, fb_pixels_next
+
+  vld1.u16 { draw_mask_low[], draw_mask_high[] },                              \
+   [ draw_mask_bits_ptr, :16 ], c_64
+  vbif.u16 fb_pixels, pixels, draw_mask_combined
+
+  vld1.u32 { pixels }, [ block_ptr_load, :128 ], c_64
+
+  sub fb_ptr_cmp, fb_ptr_next, fb_ptr
+  add fb_ptr_cmp, fb_ptr_cmp, #14
+  cmp fb_ptr_cmp, #28
+  bls 4f
+
+  vld1.u16 { fb_pixels_next }, [ fb_ptr_next ]
+  vceq.u16 zero_mask, pixels, #0
+
+  vst1.u16 { fb_pixels }, [ fb_ptr ]
+  vtst.u16 draw_mask, draw_mask, test_mask
+
+ 3:
+  subs num_blocks, num_blocks, #1
+  bne 0b
+
+ 1:
+  vorr.u16 draw_mask_combined, draw_mask, zero_mask
+  vbif.u16 fb_pixels_next, pixels, draw_mask_combined
+
+  vst1.u16 { fb_pixels_next }, [ fb_ptr_next ]
+
+  ldmia sp!, { r4, pc }
+
+ 4:
+  vst1.u16 { fb_pixels }, [ fb_ptr ]
+  vceq.u16 zero_mask, pixels, #0
+
+  vld1.u16 { fb_pixels_next }, [ fb_ptr_next ]
+  vtst.u16 draw_mask, draw_mask, test_mask
+
+  bal 3b
+
+
+function(shade_blocks_unshaded_untextured_indirect)
+  bx lr
+
+.align 3
+
+function(shade_blocks_unshaded_untextured_direct)
+  stmdb sp!, { r4, r14 }
+  add draw_mask_ptr, psx_gpu, #psx_gpu_blocks_offset
+
+  ldrh num_blocks, [ psx_gpu, #psx_gpu_num_blocks_offset ]
+  add mask_msb_ptr, psx_gpu, #psx_gpu_mask_msb_offset
+
+  vld1.u16 { msb_mask_low[], msb_mask_high[] }, [ mask_msb_ptr, :16 ]
+  add color_ptr, psx_gpu, #(psx_gpu_blocks_offset + 16)
+
+  add block_ptr_load, psx_gpu, #(psx_gpu_blocks_offset + 44)
+  vld1.u16 { pixels }, [ color_ptr, :128 ]
+
+  mov c_64, #64
+  vld1.u16 { draw_mask }, [ draw_mask_ptr, :128 ], c_64
+
+  vorr.u16 pixels, pixels, msb_mask
+  subs num_blocks, num_blocks, #1
+
+  ldr fb_ptr_next, [ block_ptr_load ], #64
+
+  vld1.u16 { fb_pixels_next }, [ fb_ptr_next ]
+  beq 1f
+
+ 0:
+  vmov fb_pixels, fb_pixels_next
+  mov fb_ptr, fb_ptr_next
+  ldr fb_ptr_next, [ block_ptr_load ], #64
+
+  vbif.u16 fb_pixels, pixels, draw_mask
+  vld1.u16 { draw_mask }, [ draw_mask_ptr, :128 ], c_64
+
+  sub fb_ptr_cmp, fb_ptr_next, fb_ptr
+  add fb_ptr_cmp, fb_ptr_cmp, #14
+  cmp fb_ptr_cmp, #28
+  bls 4f
+
+  vld1.u16 { fb_pixels_next }, [ fb_ptr_next ]
+  vst1.u16 { fb_pixels }, [ fb_ptr ]
+
+ 3:
+  subs num_blocks, num_blocks, #1
+  bne 0b
+
+ 1:
+  vbif.u16 fb_pixels_next, pixels, draw_mask
+  vst1.u16 { fb_pixels_next }, [ fb_ptr_next ]
+
+  ldmia sp!, { r4, pc }
+
+ 4:
+  vst1.u16 { fb_pixels }, [ fb_ptr ]
+  vld1.u16 { fb_pixels_next }, [ fb_ptr_next ]
+  bal 3b
+
+
+#undef draw_mask_ptr
+#undef c_64
+#undef fb_ptr
+#undef fb_ptr_next
+#undef fb_ptr_cmp
+
+#define psx_gpu                                  r0
+#define num_blocks                               r1
+#define msb_mask_ptr                             r2
+#define pixel_ptr                                r3
+#define draw_mask_ptr                            r0
+#define c_64                                     r2
+#define fb_ptr                                   r12
+#define fb_ptr_next                              r14
+#define fb_ptr_cmp                               r4
+
+#undef msb_mask
+#undef draw_mask
+#undef pixels
+#undef fb_pixels
+#undef d128_0x8000
+#undef msb_mask_low
+#undef msb_mask_high
+#undef draw_mask_next
+#undef pixels_g
+#undef blend_pixels
+#undef fb_pixels_next
+
+#define msb_mask                                 q0
+#define draw_mask                                q1
+#define pixels                                   q2
+#define fb_pixels                                q3
+#define blend_pixels                             q4
+#define pixels_no_msb                            q5
+#define blend_mask                               q6
+#define fb_pixels_no_msb                         q7
+#define d128_0x8000                              q8
+#define d128_0x0421                              q9
+#define fb_pixels_next                           q10
+#define blend_pixels_next                        q11
+#define pixels_next                              q12
+#define draw_mask_next                           q13
+#define write_mask                               q14
+
+#define pixels_rb                                q5
+#define pixels_mg                                q7
+#define pixels_g                                 q7
+#define d128_0x7C1F                              q8
+#define d128_0x03E0                              q9
+#define fb_pixels_rb                             q10
+#define fb_pixels_g                              q11
+#define fb_pixels_masked                         q11
+#define d128_0x83E0                              q15
+#define pixels_fourth                            q7
+#define d128_0x1C07                              q12
+#define d128_0x00E0                              q13
+#define d128_0x80E0                              q13
+
+#define msb_mask_low                             d0
+#define msb_mask_high                            d1
+
+#define blend_blocks_average_set_blend_mask_textured(source)                   \
+  vclt.s16 blend_mask, source, #0                                              \
+
+#define blend_blocks_average_set_stp_bit_textured()                            \
+  vorr.u16 blend_pixels, #0x8000                                               \
+
+#define blend_blocks_average_combine_textured(source)                          \
+  vbif.u16 blend_pixels, source, blend_mask                                    \
+  
+#define blend_blocks_average_set_blend_mask_untextured(source)                 \
+
+#define blend_blocks_average_set_stp_bit_untextured()                          \
+
+#define blend_blocks_average_combine_untextured(source)                        \
+
+#define blend_blocks_average_mask_set_on()                                     \
+  vclt.s16 write_mask, fb_pixels_next, #0                                      \
+
+#define blend_blocks_average_mask_copy_on()                                    \
+  vorr.u16 draw_mask, draw_mask_next, write_mask                               \
+
+#define blend_blocks_average_mask_copy_b_on()                                  \
+  vorr.u16 draw_mask_next, draw_mask_next, write_mask                          \
+
+#define blend_blocks_average_mask_set_off()                                    \
+
+#define blend_blocks_average_mask_copy_off()                                   \
+  vmov draw_mask, draw_mask_next                                               \
+
+#define blend_blocks_average_mask_copy_b_off()                                 \
+
+#define blend_blocks_average_builder(texturing, mask_evaluate)                 \
+.align 3;                                                                      \
+                                                                               \
+function(blend_blocks_##texturing##_average_##mask_evaluate)                   \
+  stmdb sp!, { r4, r14 };                                                      \
+  add mask_msb_ptr, psx_gpu, #psx_gpu_mask_msb_offset;                         \
+  ldrh num_blocks, [ psx_gpu, #psx_gpu_num_blocks_offset ];                    \
+                                                                               \
+  add pixel_ptr, psx_gpu, #(psx_gpu_blocks_offset + 16);                       \
+  vld1.u16 { msb_mask_low[], msb_mask_high[] }, [ mask_msb_ptr, :16 ];         \
+                                                                               \
+  add draw_mask_ptr, psx_gpu, #psx_gpu_blocks_offset;                          \
+  mov c_64, #64;                                                               \
+                                                                               \
+  vmov.u16 d128_0x8000, #0x8000;                                               \
+  vld1.u32 { draw_mask_next }, [ draw_mask_ptr, :128 ], c_64;                  \
+  ldr fb_ptr_next, [ pixel_ptr, #28 ];                                         \
+                                                                               \
+  vmov.u16 d128_0x0421, #0x0400;                                               \
+  vld1.u32 { pixels_next }, [ pixel_ptr, :128 ], c_64;                         \
+                                                                               \
+  vorr.u16 d128_0x0421, #0x0021;                                               \
+  vld1.u16 { fb_pixels_next }, [ fb_ptr_next ];                                \
+                                                                               \
+  veor.u16 blend_pixels_next, pixels_next, fb_pixels_next;                     \
+  vbic.u16 pixels_no_msb, pixels_next, d128_0x8000;                            \
+  vand.u16 blend_pixels_next, blend_pixels_next, d128_0x0421;                  \
+  vsub.u16 blend_pixels_next, pixels_no_msb, blend_pixels_next;                \
+  blend_blocks_average_mask_set_##mask_evaluate();                             \
+  vbic.u16 fb_pixels_no_msb, fb_pixels_next, d128_0x8000;                      \
+                                                                               \
+  subs num_blocks, num_blocks, #1;                                             \
+  beq 1f;                                                                      \
+                                                                               \
+ 0:                                                                            \
+  mov fb_ptr, fb_ptr_next;                                                     \
+  ldr fb_ptr_next, [ pixel_ptr, #28 ];                                         \
+                                                                               \
+  vmov pixels, pixels_next;                                                    \
+  vld1.u32 { pixels_next }, [ pixel_ptr, :128 ], c_64;                         \
+                                                                               \
+  vhadd.u16 blend_pixels, fb_pixels_no_msb, blend_pixels_next;                 \
+                                                                               \
+  blend_blocks_average_mask_copy_##mask_evaluate();                            \
+  vld1.u32 { draw_mask_next }, [ draw_mask_ptr, :128 ], c_64;                  \
+                                                                               \
+  blend_blocks_average_set_blend_mask_##texturing(pixels);                     \
+  blend_blocks_average_set_stp_bit_##texturing();                              \
+  vmov fb_pixels, fb_pixels_next;                                              \
+  blend_blocks_average_combine_##texturing(pixels);                            \
+                                                                               \
+  sub fb_ptr_cmp, fb_ptr_next, fb_ptr;                                         \
+  add fb_ptr_cmp, fb_ptr_cmp, #14;                                             \
+  cmp fb_ptr_cmp, #28;                                                         \
+  bls 2f;                                                                      \
+                                                                               \
+  vld1.u16 { fb_pixels_next }, [ fb_ptr_next ];                                \
+  veor.u16 blend_pixels_next, pixels_next, fb_pixels_next;                     \
+                                                                               \
+  vorr.u16 blend_pixels, blend_pixels, msb_mask;                               \
+  vbic.u16 pixels_no_msb, pixels_next, d128_0x8000;                            \
+                                                                               \
+  vand.u16 blend_pixels_next, blend_pixels_next, d128_0x0421;                  \
+  vbif.u16 fb_pixels, blend_pixels, draw_mask;                                 \
+                                                                               \
+  vbic.u16 fb_pixels_no_msb, fb_pixels_next, d128_0x8000;                      \
+  vsub.u16 blend_pixels_next, pixels_no_msb, blend_pixels_next;                \
+  blend_blocks_average_mask_set_##mask_evaluate();                             \
+  vst1.u16 { fb_pixels }, [ fb_ptr ];                                          \
+                                                                               \
+ 3:                                                                            \
+  subs num_blocks, num_blocks, #1;                                             \
+  bne 0b;                                                                      \
+                                                                               \
+ 1:                                                                            \
+  blend_blocks_average_mask_copy_b_##mask_evaluate();                          \
+  vhadd.u16 blend_pixels, fb_pixels_no_msb, blend_pixels_next;                 \
+                                                                               \
+  blend_blocks_average_set_blend_mask_##texturing(pixels_next);                \
+  blend_blocks_average_set_stp_bit_##texturing();                              \
+  blend_blocks_average_combine_##texturing(pixels_next);                       \
+                                                                               \
+  vorr.u16 blend_pixels, blend_pixels, msb_mask;                               \
+  vbif.u16 fb_pixels_next, blend_pixels, draw_mask_next;                       \
+  vst1.u16 { fb_pixels_next }, [ fb_ptr_next ];                                \
+                                                                               \
+  ldmia sp!, { r4, pc };                                                       \
+                                                                               \
+ 2:                                                                            \
+  vorr.u16 blend_pixels, blend_pixels, msb_mask;                               \
+  vbif.u16 fb_pixels, blend_pixels, draw_mask;                                 \
+  vst1.u16 { fb_pixels }, [ fb_ptr ];                                          \
+                                                                               \
+  vld1.u16 { fb_pixels_next }, [ fb_ptr_next ];                                \
+  veor.u16 blend_pixels_next, pixels_next, fb_pixels_next;                     \
+  vbic.u16 pixels_no_msb, pixels_next, d128_0x8000;                            \
+  vand.u16 blend_pixels_next, blend_pixels_next, d128_0x0421;                  \
+  vsub.u16 blend_pixels_next, pixels_no_msb, blend_pixels_next;                \
+  vbic.u16 fb_pixels_no_msb, fb_pixels_next, d128_0x8000;                      \
+                                                                               \
+  bal 3b                                                                       \
+
+blend_blocks_average_builder(textured, off)
+blend_blocks_average_builder(untextured, off)
+blend_blocks_average_builder(textured, on)
+blend_blocks_average_builder(untextured, on)
+
+
+#define blend_blocks_add_mask_set_on()                                         \
+  vclt.s16 write_mask, fb_pixels, #0                                           \
+
+#define blend_blocks_add_mask_copy_on()                                        \
+  vorr.u16 draw_mask, draw_mask, write_mask                                    \
+
+#define blend_blocks_add_mask_set_off()                                        \
+
+#define blend_blocks_add_mask_copy_off()                                       \
+
+
+#define blend_blocks_add_textured_builder(mask_evaluate)                       \
+.align 3;                                                                      \
+                                                                               \
+function(blend_blocks_textured_add_##mask_evaluate)                            \
+  stmdb sp!, { r4, r14 };                                                      \
+  add mask_msb_ptr, psx_gpu, #psx_gpu_mask_msb_offset;                         \
+  ldrh num_blocks, [ psx_gpu, #psx_gpu_num_blocks_offset ];                    \
+                                                                               \
+  add pixel_ptr, psx_gpu, #(psx_gpu_blocks_offset + 16);                       \
+  vld1.u16 { msb_mask_low[], msb_mask_high[] }, [ mask_msb_ptr, :16 ];         \
+                                                                               \
+  add draw_mask_ptr, psx_gpu, #psx_gpu_blocks_offset;                          \
+  mov c_64, #64;                                                               \
+                                                                               \
+  vmov.u16 d128_0x7C1F, #0x7C00;                                               \
+  vmov.u16 d128_0x03E0, #0x0300;                                               \
+  vmov.u16 d128_0x83E0, #0x8000;                                               \
+  vorr.u16 d128_0x03E0, #0x00E0;                                               \
+  vorr.u16 d128_0x7C1F, #0x001F;                                               \
+  vorr.u16 d128_0x83E0, d128_0x83E0, d128_0x03E0;                              \
+                                                                               \
+  vld1.u32 { draw_mask }, [ draw_mask_ptr, :128 ], c_64;                       \
+  ldr fb_ptr_next, [ pixel_ptr, #28 ];                                         \
+  vld1.u32 { pixels }, [ pixel_ptr, :128 ], c_64;                              \
+  vclt.s16 blend_mask, pixels, #0;                                             \
+  vld1.u16 { fb_pixels }, [ fb_ptr_next ];                                     \
+  blend_blocks_add_mask_set_##mask_evaluate();                                 \
+  vand.u16 pixels_rb, pixels, d128_0x7C1F;                                     \
+                                                                               \
+  blend_blocks_add_mask_copy_##mask_evaluate();                                \
+  vorr.u16 pixels, pixels, msb_mask;                                           \
+  vand.u16 fb_pixels_masked, fb_pixels, blend_mask;                            \
+  vand.u16 pixels_mg, pixels, d128_0x83E0;                                     \
+  vand.u16 fb_pixels_rb, fb_pixels_masked, d128_0x7C1F;                        \
+  vand.u16 fb_pixels_g, fb_pixels_masked, d128_0x03E0;                         \
+  vadd.u16 fb_pixels_rb, fb_pixels_rb, pixels_rb;                              \
+  vadd.u16 fb_pixels_g, fb_pixels_g, pixels_mg;                                \
+  vmin.u8 fb_pixels_rb, fb_pixels_rb, d128_0x7C1F;                             \
+  vmin.u16 fb_pixels_g, fb_pixels_g, d128_0x83E0;                              \
+                                                                               \
+  subs num_blocks, num_blocks, #1;                                             \
+  beq 1f;                                                                      \
+                                                                               \
+ 0:                                                                            \
+  mov fb_ptr, fb_ptr_next;                                                     \
+                                                                               \
+  ldr fb_ptr_next, [ pixel_ptr, #28 ];                                         \
+                                                                               \
+  vld1.u32 { pixels }, [ pixel_ptr, :128 ], c_64;                              \
+  vclt.s16 blend_mask, pixels, #0;                                             \
+                                                                               \
+  vorr.u16 pixels, pixels, msb_mask;                                           \
+  vorr.u16 blend_pixels, fb_pixels_rb, fb_pixels_g;                            \
+  vand.u16 pixels_mg, pixels, d128_0x83E0;                                     \
+                                                                               \
+  vbit.u16 blend_pixels, fb_pixels, draw_mask;                                 \
+  vld1.u32 { draw_mask }, [ draw_mask_ptr, :128 ], c_64;                       \
+                                                                               \
+  sub fb_ptr_cmp, fb_ptr_next, fb_ptr;                                         \
+  add fb_ptr_cmp, fb_ptr_cmp, #14;                                             \
+  cmp fb_ptr_cmp, #28;                                                         \
+  bls 2f;                                                                      \
+                                                                               \
+  vld1.u16 { fb_pixels }, [ fb_ptr_next ];                                     \
+  blend_blocks_add_mask_set_##mask_evaluate();                                 \
+  vand.u16 fb_pixels_masked, fb_pixels, blend_mask;                            \
+  blend_blocks_add_mask_copy_##mask_evaluate();                                \
+  vand.u16 pixels_rb, pixels, d128_0x7C1F;                                     \
+  vand.u16 fb_pixels_rb, fb_pixels_masked, d128_0x7C1F;                        \
+  vst1.u16 { blend_pixels }, [ fb_ptr ];                                       \
+                                                                               \
+ 3:                                                                            \
+  vand.u16 fb_pixels_g, fb_pixels_masked, d128_0x03E0;                         \
+  vadd.u16 fb_pixels_rb, fb_pixels_rb, pixels_rb;                              \
+  vadd.u16 fb_pixels_g, fb_pixels_g, pixels_mg;                                \
+  vmin.u8 fb_pixels_rb, fb_pixels_rb, d128_0x7C1F;                             \
+  vmin.u16 fb_pixels_g, fb_pixels_g, d128_0x83E0;                              \
+                                                                               \
+  subs num_blocks, num_blocks, #1;                                             \
+  bne 0b;                                                                      \
+                                                                               \
+ 1:                                                                            \
+  vorr.u16 blend_pixels, fb_pixels_rb, fb_pixels_g;                            \
+  vbit.u16 blend_pixels, fb_pixels, draw_mask;                                 \
+  vst1.u16 { blend_pixels }, [ fb_ptr_next ];                                  \
+                                                                               \
+  ldmia sp!, { r4, pc };                                                       \
+                                                                               \
+ 2:                                                                            \
+  vst1.u16 { blend_pixels }, [ fb_ptr ];                                       \
+  vand.u16 pixels_rb, pixels, d128_0x7C1F;                                     \
+                                                                               \
+  vld1.u16 { fb_pixels }, [ fb_ptr_next ];                                     \
+  blend_blocks_add_mask_set_##mask_evaluate();                                 \
+  vand.u16 fb_pixels_masked, fb_pixels, blend_mask;                            \
+  blend_blocks_add_mask_copy_##mask_evaluate();                                \
+  vand.u16 fb_pixels_rb, fb_pixels_masked, d128_0x7C1F;                        \
+  bal 3b                                                                       \
+
+
+#define blend_blocks_add_untextured_builder(mask_evaluate)                     \
+.align 3;                                                                      \
+                                                                               \
+function(blend_blocks_untextured_add_##mask_evaluate)                          \
+  stmdb sp!, { r4, r14 };                                                      \
+  add mask_msb_ptr, psx_gpu, #psx_gpu_mask_msb_offset;                         \
+  ldrh num_blocks, [ psx_gpu, #psx_gpu_num_blocks_offset ];                    \
+                                                                               \
+  add pixel_ptr, psx_gpu, #(psx_gpu_blocks_offset + 16);                       \
+  vld1.u16 { msb_mask_low[], msb_mask_high[] }, [ mask_msb_ptr, :16 ];         \
+                                                                               \
+  add draw_mask_ptr, psx_gpu, #psx_gpu_blocks_offset;                          \
+  mov c_64, #64;                                                               \
+                                                                               \
+  vmov.u16 d128_0x7C1F, #0x7C00;                                               \
+  vmov.u16 d128_0x03E0, #0x0300;                                               \
+  vorr.u16 d128_0x7C1F, #0x001F;                                               \
+  vorr.u16 d128_0x03E0, #0x00E0;                                               \
+                                                                               \
+  vld1.u32 { draw_mask }, [ draw_mask_ptr, :128 ], c_64;                       \
+  ldr fb_ptr_next, [ pixel_ptr, #28 ];                                         \
+  vld1.u32 { pixels }, [ pixel_ptr, :128 ], c_64;                              \
+  vld1.u16 { fb_pixels }, [ fb_ptr_next ];                                     \
+  blend_blocks_add_mask_set_##mask_evaluate();                                 \
+  vand.u16 pixels_rb, pixels, d128_0x7C1F;                                     \
+                                                                               \
+  blend_blocks_add_mask_copy_##mask_evaluate();                                \
+  vand.u16 pixels_g, pixels, d128_0x03E0;                                      \
+  vand.u16 fb_pixels_rb, fb_pixels, d128_0x7C1F;                               \
+  vand.u16 fb_pixels_g, fb_pixels, d128_0x03E0;                                \
+  vadd.u16 fb_pixels_rb, fb_pixels_rb, pixels_rb;                              \
+  vadd.u16 fb_pixels_g, fb_pixels_g, pixels_g;                                 \
+  vmin.u8 fb_pixels_rb, fb_pixels_rb, d128_0x7C1F;                             \
+  vmin.u16 fb_pixels_g, fb_pixels_g, d128_0x03E0;                              \
+                                                                               \
+  subs num_blocks, num_blocks, #1;                                             \
+  beq 1f;                                                                      \
+                                                                               \
+ 0:                                                                            \
+  mov fb_ptr, fb_ptr_next;                                                     \
+                                                                               \
+  ldr fb_ptr_next, [ pixel_ptr, #28 ];                                         \
+                                                                               \
+  vld1.u32 { pixels }, [ pixel_ptr, :128 ], c_64;                              \
+                                                                               \
+  vorr.u16 blend_pixels, fb_pixels_rb, fb_pixels_g;                            \
+  vorr.u16 blend_pixels, blend_pixels, msb_mask;                               \
+  vand.u16 pixels_g, pixels, d128_0x03E0;                                      \
+                                                                               \
+  vbit.u16 blend_pixels, fb_pixels, draw_mask;                                 \
+  vld1.u32 { draw_mask }, [ draw_mask_ptr, :128 ], c_64;                       \
+                                                                               \
+  sub fb_ptr_cmp, fb_ptr_next, fb_ptr;                                         \
+  add fb_ptr_cmp, fb_ptr_cmp, #14;                                             \
+  cmp fb_ptr_cmp, #28;                                                         \
+  bls 2f;                                                                      \
+                                                                               \
+  vld1.u16 { fb_pixels }, [ fb_ptr_next ];                                     \
+  blend_blocks_add_mask_set_##mask_evaluate();                                 \
+  blend_blocks_add_mask_copy_##mask_evaluate();                                \
+  vand.u16 pixels_rb, pixels, d128_0x7C1F;                                     \
+  vand.u16 fb_pixels_rb, fb_pixels, d128_0x7C1F;                               \
+  vst1.u16 { blend_pixels }, [ fb_ptr ];                                       \
+                                                                               \
+ 3:                                                                            \
+  vand.u16 fb_pixels_g, fb_pixels, d128_0x03E0;                                \
+  vadd.u16 fb_pixels_rb, fb_pixels_rb, pixels_rb;                              \
+  vadd.u16 fb_pixels_g, fb_pixels_g, pixels_g;                                 \
+  vmin.u8 fb_pixels_rb, fb_pixels_rb, d128_0x7C1F;                             \
+  vmin.u16 fb_pixels_g, fb_pixels_g, d128_0x03E0;                              \
+                                                                               \
+  subs num_blocks, num_blocks, #1;                                             \
+  bne 0b;                                                                      \
+                                                                               \
+ 1:                                                                            \
+  vorr.u16 blend_pixels, fb_pixels_rb, fb_pixels_g;                            \
+  vorr.u16 blend_pixels, blend_pixels, msb_mask;                               \
+  vbit.u16 blend_pixels, fb_pixels, draw_mask;                                 \
+  vst1.u16 { blend_pixels }, [ fb_ptr_next ];                                  \
+                                                                               \
+  ldmia sp!, { r4, pc };                                                       \
+                                                                               \
+ 2:                                                                            \
+  vst1.u16 { blend_pixels }, [ fb_ptr ];                                       \
+  vand.u16 pixels_rb, pixels, d128_0x7C1F;                                     \
+                                                                               \
+  vld1.u16 { fb_pixels }, [ fb_ptr_next ];                                     \
+  blend_blocks_add_mask_set_##mask_evaluate();                                 \
+  blend_blocks_add_mask_copy_##mask_evaluate();                                \
+  vand.u16 fb_pixels_rb, fb_pixels, d128_0x7C1F;                               \
+  bal 3b                                                                       \
+
+
+blend_blocks_add_textured_builder(off)
+blend_blocks_add_textured_builder(on)
+blend_blocks_add_untextured_builder(off)
+blend_blocks_add_untextured_builder(on)
+
+#define blend_blocks_subtract_set_blend_mask_textured()                        \
+  vclt.s16 blend_mask, pixels_next, #0                                         \
+
+#define blend_blocks_subtract_combine_textured()                               \
+  vbif.u16 blend_pixels, pixels, blend_mask                                    \
+
+#define blend_blocks_subtract_set_stb_textured()                               \
+  vorr.u16 blend_pixels, #0x8000                                               \
+
+#define blend_blocks_subtract_msb_mask_textured()                              \
+  vorr.u16 pixels, pixels_next, msb_mask                                       \
+
+#define blend_blocks_subtract_set_blend_mask_untextured()                      \
+
+#define blend_blocks_subtract_combine_untextured()                             \
+
+#define blend_blocks_subtract_set_stb_untextured()                             \
+  vorr.u16 blend_pixels, blend_pixels, msb_mask                                \
+
+#define blend_blocks_subtract_msb_mask_untextured()                            \
+
+
+#define blend_blocks_subtract_mask_set_on()                                    \
+  vclt.s16 write_mask, fb_pixels, #0                                           \
+
+#define blend_blocks_subtract_mask_copy_on()                                   \
+  vorr.u16 draw_mask, draw_mask_next, write_mask                               \
+
+#define blend_blocks_subtract_mask_set_off()                                   \
+
+#define blend_blocks_subtract_mask_copy_off()                                  \
+  vmov draw_mask, draw_mask_next                                               \
+
+
+#define blend_blocks_subtract_builder(texturing, mask_evaluate)                \
+.align 3;                                                                      \
+                                                                               \
+function(blend_blocks_##texturing##_subtract_##mask_evaluate)                  \
+  stmdb sp!, { r4, r14 };                                                      \
+  add mask_msb_ptr, psx_gpu, #psx_gpu_mask_msb_offset;                         \
+  ldrh num_blocks, [ psx_gpu, #psx_gpu_num_blocks_offset ];                    \
+                                                                               \
+  add pixel_ptr, psx_gpu, #(psx_gpu_blocks_offset + 16);                       \
+  vld1.u16 { msb_mask_low[], msb_mask_high[] }, [ mask_msb_ptr, :16 ];         \
+                                                                               \
+  add draw_mask_ptr, psx_gpu, #psx_gpu_blocks_offset;                          \
+  mov c_64, #64;                                                               \
+                                                                               \
+  vmov.u16 d128_0x7C1F, #0x7C00;                                               \
+  vmov.u16 d128_0x03E0, #0x0300;                                               \
+  vorr.u16 d128_0x7C1F, #0x001F;                                               \
+  vorr.u16 d128_0x03E0, #0x00E0;                                               \
+                                                                               \
+  vld1.u32 { draw_mask_next }, [ draw_mask_ptr, :128 ], c_64;                  \
+  ldr fb_ptr_next, [ pixel_ptr, #28 ];                                         \
+  vld1.u32 { pixels_next }, [ pixel_ptr, :128 ], c_64;                         \
+  blend_blocks_subtract_set_blend_mask_##texturing();                          \
+  vld1.u16 { fb_pixels }, [ fb_ptr_next ];                                     \
+  blend_blocks_subtract_mask_set_##mask_evaluate();                            \
+  vand.u16 pixels_rb, pixels_next, d128_0x7C1F;                                \
+                                                                               \
+  vand.u16 pixels_g, pixels_next, d128_0x03E0;                                 \
+  vand.u16 fb_pixels_rb, fb_pixels, d128_0x7C1F;                               \
+  vand.u16 fb_pixels_g, fb_pixels, d128_0x03E0;                                \
+  vqsub.u8 fb_pixels_rb, fb_pixels_rb, pixels_rb;                              \
+  vqsub.u16 fb_pixels_g, fb_pixels_g, pixels_g;                                \
+                                                                               \
+  subs num_blocks, num_blocks, #1;                                             \
+  beq 1f;                                                                      \
+                                                                               \
+ 0:                                                                            \
+  blend_blocks_subtract_mask_copy_##mask_evaluate();                           \
+  mov fb_ptr, fb_ptr_next;                                                     \
+  ldr fb_ptr_next, [ pixel_ptr, #28 ];                                         \
+                                                                               \
+  vld1.u32 { draw_mask_next }, [ draw_mask_ptr, :128 ], c_64;                  \
+  blend_blocks_subtract_msb_mask_##texturing();                                \
+                                                                               \
+  vld1.u32 { pixels_next }, [ pixel_ptr, :128 ], c_64;                         \
+  vorr.u16 blend_pixels, fb_pixels_rb, fb_pixels_g;                            \
+  vand.u16 pixels_rb, pixels_next, d128_0x7C1F;                                \
+  blend_blocks_subtract_set_stb_##texturing();                                 \
+  vand.u16 pixels_g, pixels_next, d128_0x03E0;                                 \
+  blend_blocks_subtract_combine_##texturing();                                 \
+  blend_blocks_subtract_set_blend_mask_##texturing();                          \
+  vbit.u16 blend_pixels, fb_pixels, draw_mask;                                 \
+                                                                               \
+  sub fb_ptr_cmp, fb_ptr_next, fb_ptr;                                         \
+  add fb_ptr_cmp, fb_ptr_cmp, #14;                                             \
+  cmp fb_ptr_cmp, #28;                                                         \
+  bls 2f;                                                                      \
+                                                                               \
+  vld1.u16 { fb_pixels }, [ fb_ptr_next ];                                     \
+  blend_blocks_subtract_mask_set_##mask_evaluate();                            \
+  vand.u16 fb_pixels_rb, fb_pixels, d128_0x7C1F;                               \
+  vand.u16 fb_pixels_g, fb_pixels, d128_0x03E0;                                \
+  vqsub.u8 fb_pixels_rb, fb_pixels_rb, pixels_rb;                              \
+  vst1.u16 { blend_pixels }, [ fb_ptr ];                                       \
+  vqsub.u16 fb_pixels_g, fb_pixels_g, pixels_g;                                \
+                                                                               \
+ 3:                                                                            \
+  subs num_blocks, num_blocks, #1;                                             \
+  bne 0b;                                                                      \
+                                                                               \
+ 1:                                                                            \
+  blend_blocks_subtract_mask_copy_##mask_evaluate();                           \
+                                                                               \
+  blend_blocks_subtract_msb_mask_##texturing();                                \
+  vorr.u16 blend_pixels, fb_pixels_rb, fb_pixels_g;                            \
+  blend_blocks_subtract_set_stb_##texturing();                                 \
+  blend_blocks_subtract_combine_##texturing();                                 \
+  vbit.u16 blend_pixels, fb_pixels, draw_mask;                                 \
+  vst1.u16 { blend_pixels }, [ fb_ptr_next ];                                  \
+                                                                               \
+  ldmia sp!, { r4, pc };                                                       \
+                                                                               \
+ 2:                                                                            \
+  vst1.u16 { blend_pixels }, [ fb_ptr ];                                       \
+  vld1.u16 { fb_pixels }, [ fb_ptr_next ];                                     \
+  blend_blocks_subtract_mask_set_##mask_evaluate();                            \
+  vand.u16 fb_pixels_rb, fb_pixels, d128_0x7C1F;                               \
+  vand.u16 fb_pixels_g, fb_pixels, d128_0x03E0;                                \
+  vqsub.u8 fb_pixels_rb, fb_pixels_rb, pixels_rb;                              \
+  vqsub.u16 fb_pixels_g, fb_pixels_g, pixels_g;                                \
+  bal 3b                                                                       \
+
+
+blend_blocks_subtract_builder(textured, off)
+blend_blocks_subtract_builder(textured, on)
+blend_blocks_subtract_builder(untextured, off)
+blend_blocks_subtract_builder(untextured, on)
+
+
+#define blend_blocks_add_fourth_textured_builder(mask_evaluate)                \
+.align 3;                                                                      \
+                                                                               \
+function(blend_blocks_textured_add_fourth_##mask_evaluate)                     \
+  stmdb sp!, { r4, r14 };                                                      \
+  add mask_msb_ptr, psx_gpu, #psx_gpu_mask_msb_offset;                         \
+  ldrh num_blocks, [ psx_gpu, #psx_gpu_num_blocks_offset ];                    \
+                                                                               \
+  add pixel_ptr, psx_gpu, #(psx_gpu_blocks_offset + 16);                       \
+  vld1.u16 { msb_mask_low[], msb_mask_high[] }, [ mask_msb_ptr, :16 ];         \
+                                                                               \
+  add draw_mask_ptr, psx_gpu, #psx_gpu_blocks_offset;                          \
+  mov c_64, #64;                                                               \
+                                                                               \
+  vmov.u16 d128_0x7C1F, #0x7C00;                                               \
+  vmov.u16 d128_0x03E0, #0x0300;                                               \
+  vmov.u16 d128_0x83E0, #0x8300;                                               \
+  vmov.u16 d128_0x1C07, #0x1C00;                                               \
+  vmov.u16 d128_0x80E0, #0x8000;                                               \
+  vorr.u16 d128_0x7C1F, #0x001F;                                               \
+  vorr.u16 d128_0x03E0, #0x00E0;                                               \
+  vorr.u16 d128_0x83E0, #0x00E0;                                               \
+  vorr.u16 d128_0x1C07, #0x0007;                                               \
+  vorr.u16 d128_0x80E0, #0x00E0;                                               \
+                                                                               \
+  vld1.u32 { draw_mask }, [ draw_mask_ptr, :128 ], c_64;                       \
+  ldr fb_ptr_next, [ pixel_ptr, #28 ];                                         \
+  vld1.u32 { pixels }, [ pixel_ptr, :128 ], c_64;                              \
+  vclt.s16 blend_mask, pixels, #0;                                             \
+  vld1.u16 { fb_pixels }, [ fb_ptr_next ];                                     \
+  blend_blocks_add_mask_set_##mask_evaluate();                                 \
+  vshr.s16 pixels_fourth, pixels, #2;                                          \
+                                                                               \
+  blend_blocks_add_mask_copy_##mask_evaluate();                                \
+  vorr.u16 pixels, pixels, msb_mask;                                           \
+  vand.u16 pixels_rb, pixels_fourth, d128_0x1C07;                              \
+  vand.u16 fb_pixels_masked, fb_pixels, blend_mask;                            \
+  vand.u16 pixels_mg, pixels_fourth, d128_0x80E0;                              \
+  vand.u16 fb_pixels_rb, fb_pixels_masked, d128_0x7C1F;                        \
+  vand.u16 fb_pixels_g, fb_pixels_masked, d128_0x03E0;                         \
+  vadd.u16 fb_pixels_rb, fb_pixels_rb, pixels_rb;                              \
+  vadd.u16 fb_pixels_g, fb_pixels_g, pixels_mg;                                \
+  vmin.u8 fb_pixels_rb, fb_pixels_rb, d128_0x7C1F;                             \
+  vmin.u16 fb_pixels_g, fb_pixels_g, d128_0x83E0;                              \
+                                                                               \
+  subs num_blocks, num_blocks, #1;                                             \
+  beq 1f;                                                                      \
+                                                                               \
+ 0:                                                                            \
+  mov fb_ptr, fb_ptr_next;                                                     \
+                                                                               \
+  ldr fb_ptr_next, [ pixel_ptr, #28 ];                                         \
+                                                                               \
+  vld1.u32 { pixels }, [ pixel_ptr, :128 ], c_64;                              \
+  vclt.s16 blend_mask, pixels, #0;                                             \
+                                                                               \
+  vshr.s16 pixels_fourth, pixels, #2;                                          \
+  vorr.u16 pixels, pixels, msb_mask;                                           \
+  vorr.u16 blend_pixels, fb_pixels_rb, fb_pixels_g;                            \
+  vand.u16 pixels_rb, pixels_fourth, d128_0x1C07;                              \
+                                                                               \
+  vbit.u16 blend_pixels, fb_pixels, draw_mask;                                 \
+  vld1.u32 { draw_mask }, [ draw_mask_ptr, :128 ], c_64;                       \
+                                                                               \
+  sub fb_ptr_cmp, fb_ptr_next, fb_ptr;                                         \
+  add fb_ptr_cmp, fb_ptr_cmp, #14;                                             \
+  cmp fb_ptr_cmp, #28;                                                         \
+  bls 2f;                                                                      \
+                                                                               \
+  vld1.u16 { fb_pixels }, [ fb_ptr_next ];                                     \
+  blend_blocks_add_mask_set_##mask_evaluate();                                 \
+  vand.u16 fb_pixels_masked, fb_pixels, blend_mask;                            \
+  blend_blocks_add_mask_copy_##mask_evaluate();                                \
+  vand.u16 pixels_mg, pixels_fourth, d128_0x80E0;                              \
+  vand.u16 fb_pixels_rb, fb_pixels_masked, d128_0x7C1F;                        \
+  vst1.u16 { blend_pixels }, [ fb_ptr ];                                       \
+                                                                               \
+ 3:                                                                            \
+  vand.u16 fb_pixels_g, fb_pixels_masked, d128_0x03E0;                         \
+  vadd.u16 fb_pixels_rb, fb_pixels_rb, pixels_rb;                              \
+  vadd.u16 fb_pixels_g, fb_pixels_g, pixels_mg;                                \
+  vmin.u8 fb_pixels_rb, fb_pixels_rb, d128_0x7C1F;                             \
+  vmin.u16 fb_pixels_g, fb_pixels_g, d128_0x83E0;                              \
+                                                                               \
+  subs num_blocks, num_blocks, #1;                                             \
+  bne 0b;                                                                      \
+                                                                               \
+ 1:                                                                            \
+  vorr.u16 blend_pixels, fb_pixels_rb, fb_pixels_g;                            \
+  vbit.u16 blend_pixels, fb_pixels, draw_mask;                                 \
+  vst1.u16 { blend_pixels }, [ fb_ptr_next ];                                  \
+                                                                               \
+  ldmia sp!, { r4, pc };                                                       \
+                                                                               \
+ 2:                                                                            \
+  vst1.u16 { blend_pixels }, [ fb_ptr ];                                       \
+  vand.u16 pixels_mg, pixels_fourth, d128_0x80E0;                              \
+                                                                               \
+  vld1.u16 { fb_pixels }, [ fb_ptr_next ];                                     \
+  blend_blocks_add_mask_set_##mask_evaluate();                                 \
+  vand.u16 fb_pixels_masked, fb_pixels, blend_mask;                            \
+  blend_blocks_add_mask_copy_##mask_evaluate();                                \
+  vand.u16 fb_pixels_rb, fb_pixels_masked, d128_0x7C1F;                        \
+  bal 3b                                                                       \
+
+
+#define blend_blocks_add_fourth_untextured_builder(mask_evaluate)              \
+.align 3;                                                                      \
+                                                                               \
+function(blend_blocks_untextured_add_fourth_##mask_evaluate)                   \
+  stmdb sp!, { r4, r14 };                                                      \
+  add mask_msb_ptr, psx_gpu, #psx_gpu_mask_msb_offset;                         \
+  ldrh num_blocks, [ psx_gpu, #psx_gpu_num_blocks_offset ];                    \
+                                                                               \
+  add pixel_ptr, psx_gpu, #(psx_gpu_blocks_offset + 16);                       \
+  vld1.u16 { msb_mask_low[], msb_mask_high[] }, [ mask_msb_ptr, :16 ];         \
+                                                                               \
+  add draw_mask_ptr, psx_gpu, #psx_gpu_blocks_offset;                          \
+  mov c_64, #64;                                                               \
+                                                                               \
+  vmov.u16 d128_0x7C1F, #0x7C00;                                               \
+  vmov.u16 d128_0x03E0, #0x0300;                                               \
+  vmov.u16 d128_0x83E0, #0x8300;                                               \
+  vmov.u16 d128_0x1C07, #0x1C00;                                               \
+  vmov.u16 d128_0x00E0, #0x00E0;                                               \
+  vorr.u16 d128_0x7C1F, #0x001F;                                               \
+  vorr.u16 d128_0x03E0, #0x00E0;                                               \
+  vorr.u16 d128_0x83E0, #0x00E0;                                               \
+  vorr.u16 d128_0x1C07, #0x0007;                                               \
+                                                                               \
+  vld1.u32 { draw_mask }, [ draw_mask_ptr, :128 ], c_64;                       \
+  ldr fb_ptr_next, [ pixel_ptr, #28 ];                                         \
+  vld1.u32 { pixels }, [ pixel_ptr, :128 ], c_64;                              \
+  vld1.u16 { fb_pixels }, [ fb_ptr_next ];                                     \
+  blend_blocks_add_mask_set_##mask_evaluate();                                 \
+  vshr.s16 pixels_fourth, pixels, #2;                                          \
+  vand.u16 pixels_rb, pixels_fourth, d128_0x1C07;                              \
+                                                                               \
+  blend_blocks_add_mask_copy_##mask_evaluate();                                \
+  vand.u16 pixels_g, pixels_fourth, d128_0x00E0;                               \
+  vand.u16 fb_pixels_rb, fb_pixels, d128_0x7C1F;                               \
+  vand.u16 fb_pixels_g, fb_pixels, d128_0x03E0;                                \
+  vadd.u16 fb_pixels_rb, fb_pixels_rb, pixels_rb;                              \
+  vadd.u16 fb_pixels_g, fb_pixels_g, pixels_g;                                 \
+  vmin.u8 fb_pixels_rb, fb_pixels_rb, d128_0x7C1F;                             \
+  vmin.u16 fb_pixels_g, fb_pixels_g, d128_0x03E0;                              \
+                                                                               \
+  subs num_blocks, num_blocks, #1;                                             \
+  beq 1f;                                                                      \
+                                                                               \
+ 0:                                                                            \
+  mov fb_ptr, fb_ptr_next;                                                     \
+                                                                               \
+  ldr fb_ptr_next, [ pixel_ptr, #28 ];                                         \
+                                                                               \
+  vld1.u32 { pixels }, [ pixel_ptr, :128 ], c_64;                              \
+                                                                               \
+  vorr.u16 blend_pixels, fb_pixels_rb, fb_pixels_g;                            \
+  vshr.s16 pixels_fourth, pixels, #2;                                          \
+  vorr.u16 blend_pixels, blend_pixels, msb_mask;                               \
+  vand.u16 pixels_rb, pixels_fourth, d128_0x1C07;                              \
+                                                                               \
+  vbit.u16 blend_pixels, fb_pixels, draw_mask;                                 \
+  vld1.u32 { draw_mask }, [ draw_mask_ptr, :128 ], c_64;                       \
+                                                                               \
+  sub fb_ptr_cmp, fb_ptr_next, fb_ptr;                                         \
+  add fb_ptr_cmp, fb_ptr_cmp, #14;                                             \
+  cmp fb_ptr_cmp, #28;                                                         \
+  bls 2f;                                                                      \
+                                                                               \
+  vld1.u16 { fb_pixels }, [ fb_ptr_next ];                                     \
+  blend_blocks_add_mask_set_##mask_evaluate();                                 \
+  blend_blocks_add_mask_copy_##mask_evaluate();                                \
+  vand.u16 pixels_g, pixels_fourth, d128_0x00E0;                               \
+  vand.u16 fb_pixels_rb, fb_pixels, d128_0x7C1F;                               \
+  vst1.u16 { blend_pixels }, [ fb_ptr ];                                       \
+                                                                               \
+ 3:                                                                            \
+  vand.u16 fb_pixels_g, fb_pixels, d128_0x03E0;                                \
+  vadd.u16 fb_pixels_rb, fb_pixels_rb, pixels_rb;                              \
+  vadd.u16 fb_pixels_g, fb_pixels_g, pixels_g;                                 \
+  vmin.u8 fb_pixels_rb, fb_pixels_rb, d128_0x7C1F;                             \
+  vmin.u16 fb_pixels_g, fb_pixels_g, d128_0x03E0;                              \
+                                                                               \
+  subs num_blocks, num_blocks, #1;                                             \
+  bne 0b;                                                                      \
+                                                                               \
+ 1:                                                                            \
+  vorr.u16 blend_pixels, fb_pixels_rb, fb_pixels_g;                            \
+  vorr.u16 blend_pixels, blend_pixels, msb_mask;                               \
+  vbit.u16 blend_pixels, fb_pixels, draw_mask;                                 \
+  vst1.u16 { blend_pixels }, [ fb_ptr_next ];                                  \
+                                                                               \
+  ldmia sp!, { r4, pc };                                                       \
+                                                                               \
+ 2:                                                                            \
+  vst1.u16 { blend_pixels }, [ fb_ptr ];                                       \
+  vand.u16 pixels_g, pixels_fourth, d128_0x00E0;                               \
+                                                                               \
+  vld1.u16 { fb_pixels }, [ fb_ptr_next ];                                     \
+  blend_blocks_add_mask_set_##mask_evaluate();                                 \
+  blend_blocks_add_mask_copy_##mask_evaluate();                                \
+  vand.u16 fb_pixels_rb, fb_pixels, d128_0x7C1F;                               \
+  bal 3b                                                                       \
+
+
+blend_blocks_add_fourth_textured_builder(off)
+blend_blocks_add_fourth_textured_builder(on)
+blend_blocks_add_fourth_untextured_builder(off)
+blend_blocks_add_fourth_untextured_builder(on)
+
+// TODO: Optimize this more. Need a scene that actually uses it for
+// confirmation..
+
+.align 3
+
+function(blend_blocks_textured_unblended_on)         
+  stmdb sp!, { r4, r14 }
+  add mask_msb_ptr, psx_gpu, #psx_gpu_mask_msb_offset
+  ldrh num_blocks, [ psx_gpu, #psx_gpu_num_blocks_offset ]
+
+  add pixel_ptr, psx_gpu, #(psx_gpu_blocks_offset + 16)
+  vld1.u16 { msb_mask_low[], msb_mask_high[] }, [ mask_msb_ptr, :16 ]
+
+  add draw_mask_ptr, psx_gpu, #psx_gpu_blocks_offset
+  mov c_64, #64
+
+  ldr fb_ptr, [ pixel_ptr, #28 ]
+  vld1.u16 { fb_pixels }, [ fb_ptr ]
+  vld1.u32 { draw_mask }, [ draw_mask_ptr, :128 ], c_64
+  vclt.s16 write_mask, fb_pixels, #0
+  vld1.u32 { pixels }, [ pixel_ptr, :128 ], c_64
+
+  subs num_blocks, num_blocks, #1
+  beq 1f
+
+ 0:
+  vorr.u16 draw_mask, draw_mask, write_mask
+  vbif.u16 fb_pixels, pixels, draw_mask
+  vst1.u16 { fb_pixels }, [ fb_ptr ]
+
+  ldr fb_ptr, [ pixel_ptr, #28 ]
+  vld1.u16 { fb_pixels }, [ fb_ptr ]
+  vld1.u32 { draw_mask }, [ draw_mask_ptr, :128 ], c_64
+  vclt.s16 write_mask, fb_pixels, #0
+  vld1.u32 { pixels }, [ pixel_ptr, :128 ], c_64
+
+  subs num_blocks, num_blocks, #1
+  bne 0b
+ 
+ 1:
+  vorr.u16 draw_mask, draw_mask, write_mask
+  vbif.u16 fb_pixels, pixels, draw_mask
+  vst1.u16 { fb_pixels }, [ fb_ptr ]
+
+  ldmia sp!, { r4, pc }
+
+
+function(blend_blocks_textured_unblended_off)
+  bx lr
+
+
+function(warmup)
+  mov r3, #64
+  cmp r0, #0
+  bxeq lr
+
+ 0:
+  vld1.u32 { u_whole_8, v_whole_8 }, [ r1, :128 ], r3
+
+  subs r0, r0, #1
+  bne 0b
+
+  bx lr
+
+#undef color
+#undef y
+#undef height
+
+#define psx_gpu                                           r0
+#define color                                             r1
+#define x                                                 r2
+#define y                                                 r3
+
+#define vram_ptr                                          r0
+#define width                                             r3
+#define height                                            r12
+
+#define parameter_width_offset                            0
+#define parameter_height_offset                           4
+
+#define color_r                                           r14
+#define color_g                                           r4
+#define color_b                                           r5  
+
+#define left_unaligned                                    r14
+#define right_unaligned                                   r4
+#define pitch                                             r5
+#define num_unaligned                                     r2
+#define num_width                                         r6
+
+#undef colors
+
+#define colors                                            q0
+
+.align 3
+
+function(render_block_fill_body)
+  ldr vram_ptr, [ psx_gpu, #psx_gpu_vram_ptr_offset ]
+  ldr height, [ sp, #parameter_height_offset ]
+
+  add vram_ptr, vram_ptr, y, lsl #11
+  ldr width, [ sp, #parameter_width_offset ]
+
+  add vram_ptr, vram_ptr, x, lsl #1
+  stmdb sp!, { r4 - r6, r14 }
+
+  ubfx color_r, color, #3, #5
+  ubfx color_g, color, #11, #5
+
+  ubfx color_b, color, #19, #5
+  orr color, color_r, color_g, lsl #5
+
+  orr color, color, color_b, lsl #10
+  add left_unaligned, x, #0x7
+
+  bic left_unaligned, left_unaligned, #0x7
+  vdup.u16 colors, color
+
+  sub left_unaligned, left_unaligned, x
+  mov pitch, #2048
+
+  sub pitch, pitch, width, lsl #1
+  sub width, width, left_unaligned
+
+  and right_unaligned, width, #0x7
+  bic width, width, #0x7
+
+ 0:
+  mov num_width, width, lsr #3
+
+  movs num_unaligned, left_unaligned
+  beq 2f
+
+ 1:
+  strh color, [ vram_ptr ], #2
+
+  subs num_unaligned, num_unaligned, #1
+  bne 1b
+
+ 2:
+  vst1.u32 { colors }, [ vram_ptr, :128 ]!
+  subs num_width, num_width, #1
+  bne 2b
+
+  movs num_unaligned, right_unaligned
+  beq 4f
+
+ 3:
+  strh color, [ vram_ptr ], #2
+  
+  subs num_unaligned, num_unaligned, #1
+  bne 3b
+
+ 4:
+  add vram_ptr, vram_ptr, pitch
+  subs height, height, #1
+  bne 0b
+  
+  ldmia sp!, { r4 - r6, pc }
+
+
+#undef x
+#undef y
+#undef width
+#undef height
+#undef fb_ptr
+#undef texture_mask
+#undef num_blocks
+#undef temp
+#undef dirty_textures_mask
+#undef clut_ptr
+#undef current_texture_mask
+
+#define psx_gpu                                           r0
+#define x                                                 r1
+#define y                                                 r2
+#define u                                                 r3
+#define v                                                 r4
+#define width                                             r5
+#define height                                            r6
+#define offset_u                                          r8
+#define offset_v                                          r9
+#define offset_u_right                                    r10
+#define width_rounded                                     r11
+#define height_rounded                                    r12
+
+#define texture_offset_base                               r1
+#define tile_width                                        r2
+#define tile_height                                       r3
+#define num_blocks                                        r4
+#define block                                             r5
+#define sub_tile_height                                   r6
+#define fb_ptr                                            r7
+#define texture_mask                                      r8
+#define column_data                                       r9
+#define texture_offset                                    r10
+#define tiles_remaining                                   r11
+#define fb_ptr_advance_column                             r12
+#define texture_block_ptr                                 r14
+
+#define texture_page_ptr                                  r3
+#define left_block_mask                                   r4
+#define right_block_mask                                  r5
+#define texture_mask_rev                                  r10
+#define control_mask                                      r11
+
+#define dirty_textures_mask                               r4
+#define clut_ptr                                          r5
+#define current_texture_mask                              r6
+
+
+#undef texels
+#undef clut_low_a
+#undef clut_low_b
+#undef clut_high_a
+#undef clut_high_b
+#undef clut_a
+#undef clut_b
+#undef texels_low
+#undef texels_high
+
+#define texels                                            d0
+#define draw_masks_fb_ptrs                                q1
+
+#define draw_mask_fb_ptr_left                             d2
+#define draw_mask_fb_ptr_right                            d3
+
+#define clut_low_a                                        d4
+#define clut_low_b                                        d5
+#define clut_high_a                                       d6
+#define clut_high_b                                       d7
+
+#define block_masks                                       d8
+#define block_masks_shifted                               d9
+
+#define clut_a                                            q2
+#define clut_b                                            q3
+
+#define texels_low                                        d10
+#define texels_high                                       d11
+
+
+setup_sprite_flush_blocks_single:
+  vpush { q1 - q4 }
+
+  stmdb sp!, { r0 - r3, r12, r14 }
+  bl flush_render_block_buffer
+  ldmia sp!, { r0 - r3, r12, r14 }
+
+  vpop { q1 - q4 }
+
+  add block, psx_gpu, #psx_gpu_blocks_offset
+
+  mov num_blocks, sub_tile_height
+  bx lr
+
+
+setup_sprite_flush_blocks_double:
+  vpush { q1 - q4 }
+
+  stmdb sp!, { r0 - r3, r12, r14 }
+  bl flush_render_block_buffer
+  ldmia sp!, { r0 - r3, r12, r14 }
+
+  vpop { q1 - q4 }
+
+  add block, psx_gpu, #psx_gpu_blocks_offset
+
+  mov num_blocks, sub_tile_height, lsl #1
+  bx lr
+
+
+setup_sprite_update_texture_4bpp_cache:
+  stmdb sp!, { r0 - r3, r14 }
+  bl update_texture_4bpp_cache
+  ldmia sp!, { r0 - r3, pc }
+
+
+setup_sprite_update_texture_8bpp_cache:
+  stmdb sp!, { r0 - r3, r14 }
+  bl update_texture_8bpp_cache
+  ldmia sp!, { r0 - r3, pc }
+
+
+#define setup_sprite_tiled_initialize_4bpp()                                   \
+  ldr dirty_textures_mask,                                                     \
+   [ psx_gpu, #psx_gpu_dirty_textures_4bpp_mask_offset ];                      \
+  ldr clut_ptr, [ psx_gpu, #psx_gpu_clut_ptr_offset ];                         \
+                                                                               \
+  ldr current_texture_mask, [ psx_gpu, #psx_gpu_current_texture_mask_offset ]; \
+  vld1.u32 { clut_a, clut_b }, [ clut_ptr, :128 ];                             \
+                                                                               \
+  tst current_texture_mask, dirty_textures_mask;                               \
+  vuzp.u8 clut_a, clut_b;                                                      \
+                                                                               \
+  blne setup_sprite_update_texture_4bpp_cache                                  \
+
+#define setup_sprite_tiled_initialize_8bpp()                                   \
+  ldr dirty_textures_mask,                                                     \
+   [ psx_gpu, #psx_gpu_dirty_textures_8bpp_mask_offset ];                      \
+  ldr current_texture_mask, [ psx_gpu, #psx_gpu_current_texture_mask_offset ]; \
+                                                                               \
+  tst current_texture_mask, dirty_textures_mask;                               \
+  blne setup_sprite_update_texture_8bpp_cache                                  \
+
+
+#define setup_sprite_tile_setup_block_no(side, offset, texture_mode)           \
+
+#define setup_sprite_block_count_single()                                      \
+  sub_tile_height                                                              \
+
+#define setup_sprite_block_count_double()                                      \
+  sub_tile_height, lsl #1                                                      \
+
+#define setup_sprite_tile_add_blocks(type)                                     \
+  add num_blocks, num_blocks, setup_sprite_block_count_##type();               \
+  cmp num_blocks, #MAX_BLOCKS;                                                 \
+                                                                               \
+  blgt setup_sprite_flush_blocks_##type                                        \
+
+
+#define setup_sprite_tile_full_4bpp(edge)                                      \
+  setup_sprite_tile_add_blocks(double);                                        \
+                                                                               \
+ 4:                                                                            \
+  and texture_block_ptr, texture_offset, texture_mask;                         \
+  vmov.u32 draw_mask_fb_ptr_left[1], fb_ptr;                                   \
+                                                                               \
+  pld [ fb_ptr ];                                                              \
+  add texture_block_ptr, texture_page_ptr, texture_block_ptr;                  \
+  vld1.u32 { texels }, [ texture_block_ptr, :64 ];                             \
+                                                                               \
+  vtbl.8 texels_low, { clut_low_a, clut_low_b }, texels;                       \
+  vtbl.8 texels_high, { clut_high_a, clut_high_b }, texels;                    \
+                                                                               \
+  vst2.u8 { texels_low, texels_high }, [ block, :128 ];                        \
+  add texture_block_ptr, texture_offset, #8;                                   \
+                                                                               \
+  and texture_block_ptr, texture_block_ptr, texture_mask;                      \
+  add block, block, #40;                                                       \
+                                                                               \
+  add texture_block_ptr, texture_page_ptr, texture_block_ptr;                  \
+  add fb_ptr, fb_ptr, #16;                                                     \
+                                                                               \
+  vst1.u32 { draw_mask_fb_ptr_left }, [ block, :64 ];                          \
+  add block, block, #24;                                                       \
+                                                                               \
+  vld1.u32 { texels }, [ texture_block_ptr, :64 ];                             \
+  vtbl.8 texels_low, { clut_low_a, clut_low_b }, texels;                       \
+                                                                               \
+  pld [ fb_ptr ];                                                              \
+  vmov.u32 draw_mask_fb_ptr_right[1], fb_ptr;                                  \
+  vtbl.8 texels_high, { clut_high_a, clut_high_b }, texels;                    \
+                                                                               \
+  vst2.u8 { texels_low, texels_high }, [ block, :128 ];                        \
+  add block, block, #40;                                                       \
+                                                                               \
+  add texture_offset, texture_offset, #0x10;                                   \
+  add fb_ptr, fb_ptr, #(2048 - 16);                                            \
+                                                                               \
+  vst1.u32 { draw_mask_fb_ptr_right }, [ block, :64 ];                         \
+  add block, block, #24;                                                       \
+                                                                               \
+  subs sub_tile_height, sub_tile_height, #1;                                   \
+  bne 4b;                                                                      \
+                                                                               \
+  add texture_offset, texture_offset, #0xF00;                                  \
+  strh num_blocks, [ psx_gpu, #psx_gpu_num_blocks_offset ]                     \
+
+  
+#define setup_sprite_tile_half_4bpp(edge)                                      \
+  setup_sprite_tile_add_blocks(single);                                        \
+                                                                               \
+ 4:                                                                            \
+  and texture_block_ptr, texture_offset, texture_mask;                         \
+  vmov.u32 draw_mask_fb_ptr_##edge[1], fb_ptr;                                 \
+                                                                               \
+  pld [ fb_ptr ];                                                              \
+  add texture_block_ptr, texture_page_ptr, texture_block_ptr;                  \
+  vld1.u32 { texels }, [ texture_block_ptr, :64 ];                             \
+                                                                               \
+  vtbl.8 texels_low, { clut_low_a, clut_low_b }, texels;                       \
+  vtbl.8 texels_high, { clut_high_a, clut_high_b }, texels;                    \
+                                                                               \
+  vst2.u8 { texels_low, texels_high }, [ block, :128 ];                        \
+  add block, block, #40;                                                       \
+                                                                               \
+  add texture_block_ptr, texture_page_ptr, texture_block_ptr;                  \
+  vst1.u32 { draw_mask_fb_ptr_##edge }, [ block, :64 ];                        \
+                                                                               \
+  add block, block, #24;                                                       \
+  add texture_offset, texture_offset, #0x10;                                   \
+                                                                               \
+  add fb_ptr, fb_ptr, #2048;                                                   \
+  subs sub_tile_height, sub_tile_height, #1;                                   \
+                                                                               \
+  bne 4b;                                                                      \
+                                                                               \
+  add texture_offset, texture_offset, #0xF00;                                  \
+  strh num_blocks, [ psx_gpu, #psx_gpu_num_blocks_offset ]                     \
+ 
+ 
+#define setup_sprite_tile_full_8bpp(edge)                                      \
+  setup_sprite_tile_add_blocks(double);                                        \
+  add block, block, #16;                                                       \
+                                                                               \
+ 4:                                                                            \
+  and texture_block_ptr, texture_offset, texture_mask;                         \
+  vmov.u32 draw_mask_fb_ptr_left[1], fb_ptr;                                   \
+                                                                               \
+  pld [ fb_ptr ];                                                              \
+  add texture_block_ptr, texture_page_ptr, texture_block_ptr;                  \
+  vld1.u32 { texels }, [ texture_block_ptr, :64 ];                             \
+                                                                               \
+  add texture_block_ptr, texture_offset, #8;                                   \
+  vst1.u32 { texels }, [ block, :64 ];                                         \
+                                                                               \
+  and texture_block_ptr, texture_block_ptr, texture_mask;                      \
+  add block, block, #24;                                                       \
+                                                                               \
+  add texture_block_ptr, texture_page_ptr, texture_block_ptr;                  \
+                                                                               \
+  add fb_ptr, fb_ptr, #16;                                                     \
+  vst1.u32 { draw_mask_fb_ptr_left }, [ block, :64 ];                          \
+                                                                               \
+  add block, block, #40;                                                       \
+  vld1.u32 { texels }, [ texture_block_ptr, :64 ];                             \
+  pld [ fb_ptr ];                                                              \
+                                                                               \
+  vmov.u32 draw_mask_fb_ptr_right[1], fb_ptr;                                  \
+  vst1.u32 { texels }, [ block, :64 ];                                         \
+  add block, block, #24;                                                       \
+                                                                               \
+  add texture_offset, texture_offset, #0x10;                                   \
+  add fb_ptr, fb_ptr, #(2048 - 16);                                            \
+                                                                               \
+  vst1.u32 { draw_mask_fb_ptr_right }, [ block, :64 ];                         \
+  add block, block, #40;                                                       \
+                                                                               \
+  subs sub_tile_height, sub_tile_height, #1;                                   \
+  bne 4b;                                                                      \
+                                                                               \
+  sub block, block, #16;                                                       \
+  add texture_offset, texture_offset, #0xF00;                                  \
+  strh num_blocks, [ psx_gpu, #psx_gpu_num_blocks_offset ]                     \
+
+  
+#define setup_sprite_tile_half_8bpp(edge)                                      \
+  setup_sprite_tile_add_blocks(single);                                        \
+  add block, block, #16;                                                       \
+                                                                               \
+ 4:                                                                            \
+  and texture_block_ptr, texture_offset, texture_mask;                         \
+  vmov.u32 draw_mask_fb_ptr_##edge[1], fb_ptr;                                 \
+  pld [ fb_ptr ];                                                              \
+                                                                               \
+  add texture_block_ptr, texture_page_ptr, texture_block_ptr;                  \
+  vld1.u32 { texels }, [ texture_block_ptr, :64 ];                             \
+                                                                               \
+  vst1.u32 { texels }, [ block, :64 ];                                         \
+  add block, block, #24;                                                       \
+                                                                               \
+  vst1.u32 { draw_mask_fb_ptr_##edge }, [ block, :64 ];                        \
+  add block, block, #40;                                                       \
+                                                                               \
+  add texture_offset, texture_offset, #0x10;                                   \
+  add fb_ptr, fb_ptr, #2048;                                                   \
+                                                                               \
+  subs sub_tile_height, sub_tile_height, #1;                                   \
+  bne 4b;                                                                      \
+                                                                               \
+  sub block, block, #16;                                                       \
+  add texture_offset, texture_offset, #0xF00;                                  \
+  strh num_blocks, [ psx_gpu, #psx_gpu_num_blocks_offset ]                     \
+
+ 
+#define setup_sprite_tile_column_edge_pre_adjust_half_right()                  \
+  add texture_offset, texture_offset_base, #8;                                 \
+  add fb_ptr, fb_ptr, #16                                                      \
+
+#define setup_sprite_tile_column_edge_pre_adjust_half_left()                   \
+  mov texture_offset, texture_offset_base                                      \
+
+#define setup_sprite_tile_column_edge_pre_adjust_half(edge)                    \
+  setup_sprite_tile_column_edge_pre_adjust_half_##edge()                       \
+
+#define setup_sprite_tile_column_edge_pre_adjust_full(edge)                    \
+  mov texture_offset, texture_offset_base                                      \
+
+#define setup_sprite_tile_column_edge_post_adjust_half_right()                 \
+  sub fb_ptr, fb_ptr, #16                                                      \
+
+#define setup_sprite_tile_column_edge_post_adjust_half_left()                  \
+
+#define setup_sprite_tile_column_edge_post_adjust_half(edge)                   \
+  setup_sprite_tile_column_edge_post_adjust_half_##edge()                      \
+
+#define setup_sprite_tile_column_edge_post_adjust_full(edge)                   \
+
+
+#define setup_sprite_tile_column_height_single(edge_mode, edge, texture_mode)  \
+  mov sub_tile_height, column_data;                                            \
+  setup_sprite_tile_column_edge_pre_adjust_##edge_mode(edge);                  \
+  setup_sprite_tile_##edge_mode##_##texture_mode(edge);                        \
+  setup_sprite_tile_column_edge_post_adjust_##edge_mode(edge)                  \
+
+#define setup_sprite_tile_column_height_multi(edge_mode, edge, texture_mode)   \
+  and sub_tile_height, column_data, #0xFF;                                     \
+  mov tiles_remaining, column_data, lsr #16;                                   \
+  setup_sprite_tile_column_edge_pre_adjust_##edge_mode(edge);                  \
+  setup_sprite_tile_##edge_mode##_##texture_mode(edge);                        \
+                                                                               \
+  subs tiles_remaining, tiles_remaining, #1;                                   \
+  beq 2f;                                                                      \
+                                                                               \
+ 3:                                                                            \
+  mov sub_tile_height, #16;                                                    \
+  setup_sprite_tile_##edge_mode##_##texture_mode(edge);                        \
+  subs tiles_remaining, tiles_remaining, #1;                                   \
+  bne 3b;                                                                      \
+                                                                               \
+ 2:                                                                            \
+  uxtb sub_tile_height, column_data, ror #8;                                   \
+  setup_sprite_tile_##edge_mode##_##texture_mode(edge);                        \
+  setup_sprite_tile_column_edge_post_adjust_##edge_mode(edge)                  \
+
+
+#define setup_sprite_column_data_single()                                      \
+  mov column_data, height;                                                     \
+  ldr texture_page_ptr, [ psx_gpu, #psx_gpu_texture_page_ptr_offset ]          \
+
+#define setup_sprite_column_data_multi()                                       \
+  and height_rounded, height_rounded, #0xF;                                    \
+  rsb column_data, offset_v, #16;                                              \
+                                                                               \
+  add height_rounded, height_rounded, #1;                                      \
+  sub tile_height, tile_height, #1;                                            \
+                                                                               \
+  orr column_data, column_data, tile_height, lsl #16;                          \
+  ldr texture_page_ptr, [ psx_gpu, #psx_gpu_texture_page_ptr_offset ];         \
+                                                                               \
+  orr column_data, column_data, height_rounded, lsl #8                         \
+
+#define setup_sprite_tile_column_width_single(texture_mode, multi_height,      \
+ edge_mode, edge)                                                              \
+ setup_sprite_##texture_mode##_single_##multi_height##_##edge_mode##_##edge:   \
+  setup_sprite_column_data_##multi_height();                                   \
+  vext.32 block_masks_shifted, block_masks, block_masks, #1;                   \
+  vorr.u32 block_masks, block_masks, block_masks_shifted;                      \
+  vdup.u8 draw_mask_fb_ptr_left, block_masks[0];                               \
+  vdup.u8 draw_mask_fb_ptr_right, block_masks[1];                              \
+                                                                               \
+  setup_sprite_tile_column_height_##multi_height(edge_mode, edge,              \
+   texture_mode);                                                              \
+  ldmia sp!, { r4 - r11, pc }                                                  \
+
+#define setup_sprite_tiled_advance_column()                                    \
+  add texture_offset_base, texture_offset_base, #0x100;                        \
+  tst texture_offset_base, #0xF00;                                             \
+  subeq texture_offset_base, texture_offset_base, #(0x100 + 0xF00)             \
+
+#define setup_sprite_tile_column_width_multi(tm, multi_height, left_mode,      \
+ right_mode)                                                                   \
+ setup_sprite_##tm##_multi_##multi_height##_##left_mode##_##right_mode:        \
+  setup_sprite_column_data_##multi_height();                                   \
+  mov fb_ptr_advance_column, #32;                                              \
+                                                                               \
+  sub fb_ptr_advance_column, height, lsl #11;                                  \
+  vdup.u8 draw_mask_fb_ptr_left, block_masks[0];                               \
+                                                                               \
+  vdup.u8 draw_mask_fb_ptr_right, block_masks[1];                              \
+  setup_sprite_tile_column_height_##multi_height(left_mode, right, tm);        \
+                                                                               \
+  subs tile_width, tile_width, #2;                                             \
+  add fb_ptr, fb_ptr, fb_ptr_advance_column;                                   \
+                                                                               \
+  vmov.u8 draw_masks_fb_ptrs, #0;                                              \
+  beq 1f;                                                                      \
+                                                                               \
+ 0:                                                                            \
+  setup_sprite_tiled_advance_column();                                         \
+  setup_sprite_tile_column_height_##multi_height(full, none, tm);              \
+  add fb_ptr, fb_ptr, fb_ptr_advance_column;                                   \
+  subs tile_width, tile_width, #1;                                             \
+  bne 0b;                                                                      \
+                                                                               \
+ 1:                                                                            \
+  vdup.u8 draw_mask_fb_ptr_left, block_masks[4];                               \
+  vdup.u8 draw_mask_fb_ptr_right, block_masks[5];                              \
+                                                                               \
+  setup_sprite_tiled_advance_column();                                         \
+  setup_sprite_tile_column_height_##multi_height(right_mode, left, tm);        \
+  ldmia sp!, { r4 - r11, pc }                                                  \
+
+
+// r0: psx_gpu
+// r1: x
+// r2: y
+// r3: u
+// [ sp ]: v
+// [ sp + 4 ]: width
+// [ sp + 8 ]: height
+// [ sp + 12 ]: color (unused)
+
+#define setup_sprite_tiled_builder(texture_mode)                               \
+                                                                               \
+setup_sprite_tile_column_width_multi(texture_mode,  multi,  full, full);       \
+setup_sprite_tile_column_width_single(texture_mode, multi,  full, none);       \
+setup_sprite_tile_column_width_multi(texture_mode,  single, full, full);       \
+setup_sprite_tile_column_width_single(texture_mode, single, full, none);       \
+setup_sprite_tile_column_width_multi(texture_mode,  multi,  half, full);       \
+setup_sprite_tile_column_width_single(texture_mode, multi,  half, right);      \
+setup_sprite_tile_column_width_multi(texture_mode,  single, half, full);       \
+setup_sprite_tile_column_width_single(texture_mode, single, half, right);      \
+setup_sprite_tile_column_width_multi(texture_mode,  multi,  full, half);       \
+setup_sprite_tile_column_width_single(texture_mode, multi,  half, left);       \
+setup_sprite_tile_column_width_multi(texture_mode,  single, full, half);       \
+setup_sprite_tile_column_width_single(texture_mode, single, half, left);       \
+setup_sprite_tile_column_width_multi(texture_mode,  multi,  half, half);       \
+setup_sprite_tile_column_width_multi(texture_mode,  single, half, half);       \
+                                                                               \
+.align 4;                                                                      \
+                                                                               \
+function(setup_sprite_##texture_mode)                                          \
+  stmdb sp!, { r4 - r11, r14 };                                                \
+  setup_sprite_tiled_initialize_##texture_mode();                              \
+                                                                               \
+  ldr v, [ sp, #36 ];                                                          \
+  and offset_u, u, #0xF;                                                       \
+                                                                               \
+  ldr width, [ sp, #40 ];                                                      \
+  ldr fb_ptr, [ psx_gpu, #psx_gpu_vram_ptr_offset ];                           \
+                                                                               \
+  ldr height, [ sp, #44 ];                                                     \
+  add fb_ptr, fb_ptr, y, lsl #11;                                              \
+                                                                               \
+  add fb_ptr, fb_ptr, x, lsl #1;                                               \
+  and offset_v, v, #0xF;                                                       \
+                                                                               \
+  sub fb_ptr, fb_ptr, offset_u, lsl #1;                                        \
+  add width_rounded, offset_u, width;                                          \
+                                                                               \
+  add height_rounded, offset_v, height;                                        \
+  add width_rounded, width_rounded, #15;                                       \
+                                                                               \
+  add height_rounded, height_rounded, #15;                                     \
+  mov tile_width, width_rounded, lsr #4;                                       \
+                                                                               \
+  /* texture_offset_base = VH-VL-00-00                                       */\
+  mov texture_offset_base, v, lsl #8;                                          \
+  and offset_u_right, width_rounded, #0xF;                                     \
+                                                                               \
+  /* texture_offset_base = VH-UH-UL-00                                       */\
+  bfi texture_offset_base, u, #4, #8;                                          \
+  movw right_block_mask, #0xFFFE;                                              \
+                                                                               \
+  /* texture_offset_base = VH-UH-VL-00                                       */\
+  bfi texture_offset_base, v, #4, #4;                                          \
+  movw left_block_mask, #0xFFFF;                                               \
+                                                                               \
+  mov tile_height, height_rounded, lsr #4;                                     \
+  mvn left_block_mask, left_block_mask, lsl offset_u;                          \
+                                                                               \
+  /* texture_mask = HH-HL-WH-WL                                              */\
+  ldrh texture_mask, [ psx_gpu, #psx_gpu_texture_mask_width_offset ];          \
+  mov right_block_mask, right_block_mask, lsl offset_u_right;                  \
+                                                                               \
+  /* texture_mask_rev = WH-WL-HH-HL                                          */\
+  rev16 texture_mask_rev, texture_mask;                                        \
+  vmov block_masks, left_block_mask, right_block_mask;                         \
+                                                                               \
+  /* texture_mask = HH-HL-HL-WL                                              */\
+  bfi texture_mask, texture_mask_rev, #4, #4;                                  \
+  /* texture_mask_rev = 00-00-00-WH                                          */\
+  mov texture_mask_rev, texture_mask_rev, lsr #12;                             \
+                                                                               \
+  /* texture_mask = HH-WH-HL-WL                                              */\
+  bfi texture_mask, texture_mask_rev, #8, #4;                                  \
+  and left_block_mask, left_block_mask, #0xFF;                                 \
+                                                                               \
+  mov control_mask, #0;                                                        \
+  cmp left_block_mask, #0xFF;                                                  \
+                                                                               \
+  uxtb right_block_mask, right_block_mask, ror #8;                             \
+  orreq control_mask, control_mask, #0x4;                                      \
+                                                                               \
+  ldrh num_blocks, [ psx_gpu, #psx_gpu_num_blocks_offset ];                    \
+  cmp right_block_mask, #0xFF;                                                 \
+                                                                               \
+  orreq control_mask, control_mask, #0x8;                                      \
+  cmp tile_width, #1;                                                          \
+                                                                               \
+  add block, psx_gpu, #psx_gpu_blocks_offset;                                  \
+  orreq control_mask, control_mask, #0x1;                                      \
+                                                                               \
+  cmp tile_height, #1;                                                         \
+  add block, block, num_blocks, lsl #6;                                        \
+                                                                               \
+  orreq control_mask, control_mask, #0x2;                                      \
+  ldr pc, [ pc, control_mask, lsl #2 ];                                        \
+  nop;                                                                         \
+                                                                               \
+ .word setup_sprite_##texture_mode##_multi_multi_full_full;                    \
+ .word setup_sprite_##texture_mode##_single_multi_full_none;                   \
+ .word setup_sprite_##texture_mode##_multi_single_full_full;                   \
+ .word setup_sprite_##texture_mode##_single_single_full_none;                  \
+ .word setup_sprite_##texture_mode##_multi_multi_half_full;                    \
+ .word setup_sprite_##texture_mode##_single_multi_half_right;                  \
+ .word setup_sprite_##texture_mode##_multi_single_half_full;                   \
+ .word setup_sprite_##texture_mode##_single_single_half_right;                 \
+ .word setup_sprite_##texture_mode##_multi_multi_full_half;                    \
+ .word setup_sprite_##texture_mode##_single_multi_half_left;                   \
+ .word setup_sprite_##texture_mode##_multi_single_full_half;                   \
+ .word setup_sprite_##texture_mode##_single_single_half_left;                  \
+ .word setup_sprite_##texture_mode##_multi_multi_half_half;                    \
+ .word 0x00000000;                                                             \
+ .word setup_sprite_##texture_mode##_multi_single_half_half                    \
+
+
+setup_sprite_tiled_builder(4bpp);
+setup_sprite_tiled_builder(8bpp);
+
+
+#undef block_ptr
+#undef num_blocks
+#undef clut_ptr
+
+#define psx_gpu                                           r0
+#define block_ptr                                         r0
+#define num_blocks                                        r1
+#define clut_ptr                                          r2
+#define texel_shift_mask                                  r3
+#define block_pixels_a                                    r4
+#define block_pixels_b                                    r5
+#define texel_0                                           r6
+#define texel_2                                           r7
+#define texel_4                                           r8
+#define texel_6                                           r9
+#define texel_1                                           r10
+#define texel_3                                           r11
+#define texel_5                                           r12
+#define texel_7                                           r14
+#define texels_01                                         r6
+#define texels_23                                         r7
+#define texels_45                                         r8
+#define texels_67                                         r9
+
+function(texture_sprite_blocks_8bpp)
+  stmdb sp!, { r4 - r11, r14 }
+  movw texel_shift_mask, #(0xFF << 1)
+
+  ldrh num_blocks, [ psx_gpu, #psx_gpu_num_blocks_offset ]
+  ldr clut_ptr, [ psx_gpu, #psx_gpu_clut_ptr_offset ]
+
+  add block_ptr, psx_gpu, #psx_gpu_blocks_offset
+  ldr block_pixels_a, [ block_ptr, #16 ]
+
+ 0:
+  and texel_0, texel_shift_mask, block_pixels_a, lsl #1
+  ldr block_pixels_b, [ block_ptr, #20 ]
+
+  and texel_1, texel_shift_mask, block_pixels_a, lsr #7
+  ldrh texel_0, [ clut_ptr, texel_0 ]
+
+  and texel_2, texel_shift_mask, block_pixels_a, lsr #15
+  ldrh texel_1, [ clut_ptr, texel_1 ]
+
+  and texel_3, texel_shift_mask, block_pixels_a, lsr #23
+  ldr block_pixels_a, [ block_ptr, #(64 + 16) ]
+
+  ldrh texel_2, [ clut_ptr, texel_2 ]
+  and texel_4, texel_shift_mask, block_pixels_b, lsl #1
+
+  ldrh texel_3, [ clut_ptr, texel_3 ]
+  and texel_5, texel_shift_mask, block_pixels_b, lsr #7
+
+  ldrh texel_4, [ clut_ptr, texel_4 ]
+  and texel_6, texel_shift_mask, block_pixels_b, lsr #15
+
+  ldrh texel_5, [ clut_ptr, texel_5 ]
+  and texel_7, texel_shift_mask, block_pixels_b, lsr #23
+
+  ldrh texel_6, [ clut_ptr, texel_6 ]
+  orr texels_01, texel_0, texel_1, lsl #16
+
+  ldrh texel_7, [ clut_ptr, texel_7 ]
+  orr texels_23, texel_2, texel_3, lsl #16
+
+  orr texels_45, texel_4, texel_5, lsl #16
+  str texels_01, [ block_ptr, #0 ]
+
+  orr texels_67, texel_6, texel_7, lsl #16
+  str texels_23, [ block_ptr, #4 ]
+
+  subs num_blocks, num_blocks, #1
+  str texels_45, [ block_ptr, #8 ]
+
+  str texels_67, [ block_ptr, #12 ]
+  add block_ptr, block_ptr, #64
+
+  bne 0b
+
+  ldmia sp!, { r4 - r11, pc }
+
+
+#undef width_rounded
+#undef texture_mask
+#undef num_blocks
+#undef texture_offset
+
+#define psx_gpu                                           r0
+#define x                                                 r1
+#define y                                                 r2
+#define u                                                 r3
+#define v                                                 r4
+#define width                                             r5
+#define height                                            r6
+#define left_offset                                       r8
+#define width_rounded                                     r9
+#define right_width                                       r10
+#define block_width                                       r11
+
+#define texture_offset_base                               r1
+#define texture_mask                                      r2
+#define texture_page_ptr                                  r3
+#define num_blocks                                        r4
+#define block                                             r5
+#define fb_ptr                                            r7
+#define texture_offset                                    r8
+#define blocks_remaining                                  r9
+#define fb_ptr_pitch                                      r12
+#define texture_block_ptr                                 r14
+
+#define texture_mask_width                                r2
+#define texture_mask_height                               r3
+#define left_mask_bits                                    r4
+#define right_mask_bits                                   r5
+
+
+#undef block_masks
+#undef block_masks_shifted
+#undef texels
+
+#define block_masks                                       d0
+#define block_masks_shifted                               d1
+#define draw_mask_fb_ptr                                  d2
+#define texels                                            q2
+
+
+setup_sprites_16bpp_flush_single:
+  vpush { d0 - d2 }
+
+  stmdb sp!, { r0 - r3, r12, r14 }
+  bl flush_render_block_buffer
+  ldmia sp!, { r0 - r3, r12, r14 }
+
+  vpop { d0 - d2 }
+
+  add block, psx_gpu, #psx_gpu_blocks_offset
+  mov num_blocks, #1
+
+  bx lr
+
+setup_sprites_16bpp_flush_row:
+  vpush { d0 - d2 }
+
+  stmdb sp!, { r0 - r3, r12, r14 }
+  bl flush_render_block_buffer
+  ldmia sp!, { r0 - r3, r12, r14 }
+
+  vpop { d0 - d2 }
+
+  add block, psx_gpu, #psx_gpu_blocks_offset
+  mov num_blocks, block_width
+
+  bx lr
+
+function(setup_sprite_16bpp)
+  stmdb sp!, { r4 - r11, r14 }
+  ldr fb_ptr, [ psx_gpu, #psx_gpu_vram_ptr_offset ]
+
+  ldr v, [ sp, #36 ]
+  add fb_ptr, fb_ptr, y, lsl #11
+
+  ldr width, [ sp, #40 ]
+  add fb_ptr, fb_ptr, x, lsl #1
+
+  ldr height, [ sp, #44 ]
+  and left_offset, u, #0x7
+
+  add texture_offset_base, u, u
+  add width_rounded, width, #7
+
+  add texture_offset_base, v, lsl #11
+  mov left_mask_bits, #0xFF
+  
+  ldrb texture_mask_width, [ psx_gpu, #psx_gpu_texture_mask_width_offset ]
+  add width_rounded, width_rounded, left_offset
+
+  ldrb texture_mask_height, [ psx_gpu, #psx_gpu_texture_mask_height_offset ]
+  sub fb_ptr, fb_ptr, left_offset, lsl #1
+
+  add texture_mask, texture_mask_width, texture_mask_width
+  mov right_mask_bits, #0xFE
+
+  and right_width, width_rounded, #0x7
+  mvn left_mask_bits, left_mask_bits, lsl left_offset
+
+  add texture_mask, texture_mask_height, lsl #11
+  mov block_width, width_rounded, lsr #3
+
+  mov right_mask_bits, right_mask_bits, lsl right_width
+  movw fb_ptr_pitch, #(2048 + 16)
+
+  sub fb_ptr_pitch, fb_ptr_pitch, block_width, lsl #4
+  vmov block_masks, left_mask_bits, right_mask_bits
+
+  ldrh num_blocks, [ psx_gpu, #psx_gpu_num_blocks_offset ]
+  add block, psx_gpu, #psx_gpu_blocks_offset
+
+  bic texture_offset_base, texture_offset_base, #0x7
+  cmp block_width, #1
+
+  ldr texture_page_ptr, [ psx_gpu, #psx_gpu_texture_page_ptr_offset ]
+  add block, block, num_blocks, lsl #6
+
+  bne 0f
+
+  vext.32 block_masks_shifted, block_masks, block_masks, #1
+  vorr.u32 block_masks, block_masks, block_masks_shifted
+  vdup.u8 draw_mask_fb_ptr, block_masks[0]
+
+ 1:
+  add num_blocks, num_blocks, #1
+  cmp num_blocks, #MAX_BLOCKS
+  blgt setup_sprites_16bpp_flush_single
+
+  and texture_block_ptr, texture_offset_base, texture_mask
+  subs height, height, #1
+
+  add texture_block_ptr, texture_page_ptr, texture_block_ptr
+  vld1.u32 { texels }, [ texture_block_ptr, :128 ]
+
+  vst1.u32 { texels }, [ block, :128 ]
+  add block, block, #40
+
+  vmov.u32 draw_mask_fb_ptr[1], fb_ptr
+  pld [ fb_ptr ]
+
+  vst1.u32 { draw_mask_fb_ptr }, [ block, :64 ]
+
+  add block, block, #24
+  add texture_offset_base, texture_offset_base, #2048
+  add fb_ptr, fb_ptr, #2048
+  strh num_blocks, [ psx_gpu, #psx_gpu_num_blocks_offset ]
+  bne 1b
+
+  ldmia sp!, { r4 - r11, pc }
+
+ 0:
+  add num_blocks, num_blocks, block_width
+  mov texture_offset, texture_offset_base
+
+  cmp num_blocks, #MAX_BLOCKS
+  blgt setup_sprites_16bpp_flush_row
+
+  add texture_offset_base, texture_offset_base, #2048
+  and texture_block_ptr, texture_offset, texture_mask
+
+  add texture_block_ptr, texture_page_ptr, texture_block_ptr
+  vld1.u32 { texels }, [ texture_block_ptr, :128 ]  
+
+  vst1.u32 { texels }, [ block, :128 ]
+  add block, block, #40
+
+  vdup.u8 draw_mask_fb_ptr, block_masks[0]
+  vmov.u32 draw_mask_fb_ptr[1], fb_ptr
+  pld [ fb_ptr ]
+
+  vst1.u32 { draw_mask_fb_ptr }, [ block, :64 ]
+  subs blocks_remaining, block_width, #2
+
+  add texture_offset, texture_offset, #16
+  add fb_ptr, fb_ptr, #16
+
+  vmov.u8 draw_mask_fb_ptr, #0
+
+  add block, block, #24
+  beq 2f
+
+ 1:
+  and texture_block_ptr, texture_offset, texture_mask
+  subs blocks_remaining, blocks_remaining, #1
+
+  add texture_block_ptr, texture_page_ptr, texture_block_ptr
+  vld1.u32 { texels }, [ texture_block_ptr, :128 ]
+
+  vst1.u32 { texels }, [ block, :128 ]
+  add block, block, #40
+
+  vmov.u32 draw_mask_fb_ptr[1], fb_ptr
+  pld [ fb_ptr ]
+
+  vst1.u32 { draw_mask_fb_ptr }, [ block, :64 ]
+  
+  add texture_offset, texture_offset, #16
+  add fb_ptr, fb_ptr, #16
+
+  add block, block, #24
+  bne 1b
+
+ 2:
+  and texture_block_ptr, texture_offset, texture_mask
+  add texture_block_ptr, texture_page_ptr, texture_block_ptr
+
+  vld1.u32 { texels }, [ texture_block_ptr, :128 ]
+  vdup.u8 draw_mask_fb_ptr, block_masks[4]
+
+  vst1.u32 { texels }, [ block, :128 ]
+  add block, block, #40
+
+  vmov.u32 draw_mask_fb_ptr[1], fb_ptr
+  vst1.u32 { draw_mask_fb_ptr }, [ block, :64 ]
+  
+  add block, block, #24
+  subs height, height, #1
+
+  add fb_ptr, fb_ptr, fb_ptr_pitch
+  strh num_blocks, [ psx_gpu, #psx_gpu_num_blocks_offset ]
+
+  bne 0b
+
+  ldmia sp!, { r4 - r11, pc }
+
+
+#undef texture_page_ptr
+#undef vram_ptr
+#undef dirty_textures_mask
+#undef current_texture_mask
+
+#define psx_gpu                                           r0
+#define current_texture_page                              r1
+#define texture_page_ptr                                  r2
+#define vram_ptr_a                                        r3
+#define current_texture_page_x                            r12
+#define current_texture_page_y                            r4
+#define dirty_textures_mask                               r5
+#define tile_y                                            r6
+#define tile_x                                            r7
+#define sub_y                                             r8
+#define current_texture_mask                              r9
+#define c_4096                                            r10
+#define vram_ptr_b                                        r11
+
+#define texel_block_a                                     d0
+#define texel_block_b                                     d1
+#define texel_block_expanded_a                            q1
+#define texel_block_expanded_b                            q2
+#define texel_block_expanded_ab                           q2
+#define texel_block_expanded_c                            q3
+#define texel_block_expanded_d                            q4
+#define texel_block_expanded_cd                           q3
+
+function(update_texture_4bpp_cache)
+  stmdb sp!, { r4 - r11, r14 }
+  vpush { q0 - q3 }
+
+  ldrb current_texture_page, [ psx_gpu, #psx_gpu_current_texture_page_offset ]
+
+  ldr texture_page_ptr, [ psx_gpu, #psx_gpu_texture_page_ptr_offset ]
+  ldr vram_ptr_a, [ psx_gpu, #psx_gpu_vram_ptr_offset ]
+
+  and current_texture_page_x, current_texture_page, #0xF
+  ldr current_texture_mask, [ psx_gpu, #psx_gpu_current_texture_mask_offset ]
+
+  mov current_texture_page_y, current_texture_page, lsr #4
+  ldr dirty_textures_mask, [ psx_gpu, #psx_gpu_dirty_textures_4bpp_mask_offset ]
+
+  add vram_ptr_a, vram_ptr_a, current_texture_page_y, lsl #19
+  mov tile_y, #16
+
+  add vram_ptr_a, vram_ptr_a, current_texture_page_x, lsl #7  
+  bic dirty_textures_mask, current_texture_mask
+  
+  mov tile_x, #16
+  str dirty_textures_mask, [ psx_gpu, #psx_gpu_dirty_textures_4bpp_mask_offset ]
+
+  mov sub_y, #8
+  movw c_4096, #4096
+
+  add vram_ptr_b, vram_ptr_a, #2048
+
+ 0:
+  vld1.u32 { texel_block_a }, [ vram_ptr_a, :64 ], c_4096
+  vld1.u32 { texel_block_b }, [ vram_ptr_b, :64 ], c_4096
+
+  vmovl.u8 texel_block_expanded_a, texel_block_a
+  vshll.u8 texel_block_expanded_b, texel_block_a, #4
+  vmovl.u8 texel_block_expanded_c, texel_block_b
+  vshll.u8 texel_block_expanded_d, texel_block_b, #4
+
+  vbic.u16 texel_block_expanded_a, #0x00F0
+  vbic.u16 texel_block_expanded_b, #0x00F0
+  vbic.u16 texel_block_expanded_c, #0x00F0
+  vbic.u16 texel_block_expanded_d, #0x00F0
+
+  vorr.u16 texel_block_expanded_ab, texel_block_expanded_a,                    \
+   texel_block_expanded_b
+  vorr.u16 texel_block_expanded_cd, texel_block_expanded_c,                    \
+   texel_block_expanded_d
+
+  vst1.u32 { texel_block_expanded_ab, texel_block_expanded_cd },               \
+   [ texture_page_ptr, :256 ]!
+
+  subs sub_y, sub_y, #1
+  bne 0b
+
+  mov sub_y, #8
+  add vram_ptr_a, vram_ptr_a, #8
+  add vram_ptr_b, vram_ptr_b, #8
+
+  sub vram_ptr_a, vram_ptr_a, #(16 * 2048)
+  sub vram_ptr_b, vram_ptr_b, #(16 * 2048)
+
+  subs tile_x, tile_x, #1
+  bne 0b
+
+  mov tile_x, #16
+  add vram_ptr_a, vram_ptr_a, #(16 * 2048)
+  add vram_ptr_b, vram_ptr_b, #(16 * 2048)
+
+  sub vram_ptr_a, vram_ptr_a, #(8 * 16)
+  sub vram_ptr_b, vram_ptr_b, #(8 * 16)
+
+  subs tile_y, tile_y, #1
+  bne 0b
+
+  vpop { q0 - q3 }
+  ldmia sp!, { r4 - r11, pc }
+
+
+#undef current_texture_page
+
+#define psx_gpu                                           r0
+#define texture_page                                      r1
+#define texture_page_ptr                                  r2
+#define vram_ptr_a                                        r3
+#define texture_page_x                                    r12
+#define texture_page_y                                    r4
+#define current_texture_page                              r5
+#define tile_y                                            r6
+#define tile_x                                            r7
+#define sub_y                                             r8
+#define c_4096                                            r10
+#define vram_ptr_b                                        r11
+
+
+#undef texels_a
+#undef texels_b
+
+#define texels_a                                          q0
+#define texels_b                                          q1
+#define texels_c                                          q2
+#define texels_d                                          q3
+
+
+function(update_texture_8bpp_cache_slice)
+  stmdb sp!, { r4 - r11, r14 }
+  vpush { q0 - q3 }
+
+  ldrb current_texture_page, [ psx_gpu, #psx_gpu_current_texture_page_offset ]
+  ldr vram_ptr_a, [ psx_gpu, #psx_gpu_vram_ptr_offset ]
+
+  ldr texture_page_ptr, [ psx_gpu, #psx_gpu_texture_page_ptr_offset ]
+  mov tile_y, #16
+
+  and texture_page_x, texture_page, #0xF
+  mov texture_page_y, texture_page, lsr #4
+
+  add vram_ptr_a, vram_ptr_a, texture_page_x, lsl #7  
+  mov tile_x, #8
+
+  add vram_ptr_a, vram_ptr_a, texture_page_y, lsl #19
+  eor current_texture_page, current_texture_page, texture_page
+
+  ands current_texture_page, current_texture_page, #0x1
+  mov sub_y, #4
+
+  addne texture_page_ptr, texture_page_ptr, #(8 * 16 * 16)
+  movw c_4096, #4096
+
+  add vram_ptr_b, vram_ptr_a, #2048
+
+ 0:
+  vld1.u32 { texels_a }, [ vram_ptr_a, :128 ], c_4096
+  vld1.u32 { texels_b }, [ vram_ptr_b, :128 ], c_4096
+  vld1.u32 { texels_c }, [ vram_ptr_a, :128 ], c_4096
+  vld1.u32 { texels_d }, [ vram_ptr_b, :128 ], c_4096
+
+  vst1.u32 { texels_a, texels_b }, [ texture_page_ptr, :256 ]!
+  vst1.u32 { texels_c, texels_d }, [ texture_page_ptr, :256 ]!
+
+  subs sub_y, sub_y, #1
+  bne 0b
+
+  mov sub_y, #4
+
+  add vram_ptr_a, vram_ptr_a, #16
+  add vram_ptr_b, vram_ptr_b, #16
+
+  sub vram_ptr_a, vram_ptr_a, #(16 * 2048)
+  sub vram_ptr_b, vram_ptr_b, #(16 * 2048)
+
+  subs tile_x, tile_x, #1
+  bne 0b
+
+  mov tile_x, #8
+
+  add vram_ptr_a, vram_ptr_a, #(16 * 2048)
+  add vram_ptr_b, vram_ptr_b, #(16 * 2048)
+
+  sub vram_ptr_a, vram_ptr_a, #(8 * 16)
+  sub vram_ptr_b, vram_ptr_b, #(8 * 16)
+
+  subs tile_y, tile_y, #1
+  add texture_page_ptr, texture_page_ptr, #(8 * 16 * 16)
+
+  bne 0b
+
+  vpop { q0 - q3 }
+  ldmia sp!, { r4 - r11, pc }
+
diff --git a/plugins/gpu_neon/psx_gpu/psx_gpu_main.c b/plugins/gpu_neon/psx_gpu/psx_gpu_main.c
new file mode 100644
index 0000000..97f62ca
--- /dev/null
+++ b/plugins/gpu_neon/psx_gpu/psx_gpu_main.c
@@ -0,0 +1,352 @@
+/*
+ * Copyright (C) 2011 Gilead Kutnick "Exophase" <exophase@gmail.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; either version 2 of
+ * the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ */
+
+#include <stdio.h>
+#include <stdlib.h>
+
+#include "SDL.h"
+#include "common.h"
+
+extern u32 span_pixels;
+extern u32 span_pixel_blocks;
+extern u32 span_pixel_blocks_unaligned;
+extern u32 spans;
+extern u32 triangles;
+extern u32 sprites;
+extern u32 sprites_4bpp;
+extern u32 sprites_8bpp;
+extern u32 sprites_16bpp;
+extern u32 sprites_untextured;
+extern u32 sprite_blocks;
+extern u32 lines;
+extern u32 texels_4bpp;
+extern u32 texels_8bpp;
+extern u32 texels_16bpp;
+extern u32 texel_blocks_4bpp;
+extern u32 texel_blocks_8bpp;
+extern u32 texel_blocks_16bpp;
+extern u32 texel_blocks_untextured;
+extern u32 blend_blocks;
+extern u32 untextured_pixels;
+extern u32 blend_pixels;
+extern u32 transparent_pixels;
+extern u32 render_buffer_flushes;
+extern u32 state_changes;
+extern u32 trivial_rejects;
+extern u32 left_split_triangles;
+extern u32 flat_triangles;
+extern u32 clipped_triangles;
+extern u32 zero_block_spans;
+extern u32 texture_cache_loads;
+extern u32 false_modulated_triangles;
+extern u32 false_modulated_sprites;
+
+static u32 mismatches;
+
+typedef struct
+{
+	u16 vram[1024 * 512];
+	u32 gpu_register[15];
+	u32 status;
+} gpu_dump_struct;
+
+static gpu_dump_struct state;
+
+psx_gpu_struct __attribute__((aligned(256))) _psx_gpu;
+
+#define percent_of(numerator, denominator)                                     \
+  ((((double)(numerator)) / (denominator)) * 100.0)                            \
+
+void clear_stats(void)
+{
+  triangles = 0;
+  sprites = 0;
+  sprites_4bpp = 0;
+  sprites_8bpp = 0;
+  sprites_16bpp = 0;
+  sprites_untextured = 0;
+  sprite_blocks = 0;
+  lines = 0;
+  span_pixels = 0;
+  span_pixel_blocks = 0;
+  span_pixel_blocks_unaligned = 0;
+  spans = 0;
+  texels_4bpp = 0;
+  texels_8bpp = 0;
+  texels_16bpp = 0;
+  texel_blocks_untextured = 0;
+  texel_blocks_4bpp = 0;
+  texel_blocks_8bpp = 0;
+  texel_blocks_16bpp = 0;
+  blend_blocks = 0;
+  untextured_pixels = 0;
+  blend_pixels = 0;
+  transparent_pixels = 0;
+  render_buffer_flushes = 0;
+  state_changes = 0;
+  trivial_rejects = 0;
+  left_split_triangles = 0;
+  flat_triangles = 0;
+  clipped_triangles = 0;
+  zero_block_spans = 0;
+  texture_cache_loads = 0;
+  false_modulated_triangles = 0;
+  false_modulated_sprites = 0;
+}
+
+void update_screen(psx_gpu_struct *psx_gpu, SDL_Surface *screen)
+{
+  u32 x, y;
+
+  for(y = 0; y < 512; y++)
+  {
+    for(x = 0; x < 1024; x++)
+    {
+      u32 pixel = psx_gpu->vram_ptr[(y * 1024) + x];
+      ((u32 *)screen->pixels)[(y * 1024) + x] =
+       ((pixel & 0x1F) << (16 + 3)) |
+       (((pixel >> 5) & 0x1F) << (8 + 3)) |
+       (((pixel >> 10) & 0x1F) << 3);
+    }
+  }
+
+  SDL_Flip(screen);
+}
+
+#ifdef PANDORA_BUILD
+
+#include <fcntl.h>
+#include <linux/fb.h>
+#include <sys/mman.h>
+#include <sys/ioctl.h>
+  
+#endif
+
+int main(int argc, char *argv[])
+{
+  psx_gpu_struct *psx_gpu = &_psx_gpu;
+  SDL_Surface *screen;
+  SDL_Event event;
+
+  u32 *list;
+  int size;
+  FILE *state_file;
+  FILE *list_file;
+  u32 no_display = 0;
+
+  if((argc != 3) && (argc != 4))
+  {
+    printf("usage:\n%s <state> <list>\n", argv[0]);
+    return 1;
+  }
+
+  if((argc == 4) && !strcmp(argv[3], "-n"))
+    no_display = 1;
+  
+  state_file = fopen(argv[1], "rb");
+  fread(&state, 1, sizeof(gpu_dump_struct), state_file);
+  fclose(state_file);
+  
+  list_file = fopen(argv[2], "rb");
+  
+  fseek(list_file, 0, SEEK_END);
+  size = ftell(list_file);
+  fseek(list_file, 0, SEEK_SET);
+  //size = 0;
+  
+  list = malloc(size);
+  fread(list, 1, size, list_file);
+  fclose(list_file);
+ 
+  if(no_display == 0) 
+  {
+    SDL_Init(SDL_INIT_EVERYTHING);
+    screen = SDL_SetVideoMode(1024, 512, 32, 0);
+  }
+  
+  initialize_psx_gpu(psx_gpu);
+
+#ifdef PANDORA_BUILD
+  system("ofbset -fb /dev/fb1 -mem 6291456 -en 0");
+  u32 fbdev_handle = open("/dev/fb1", O_RDWR);
+  psx_gpu->vram_ptr = (mmap((void *)0x50000000, 1024 * 1024 * 2, PROT_READ | PROT_WRITE,
+   MAP_SHARED | 0xA0000000, fbdev_handle, 0));
+  psx_gpu->vram_ptr += 64;
+#endif
+
+
+
+#ifdef PANDORA_BUILD
+  //triangle_benchmark(psx_gpu);
+  //return 0;
+#endif
+
+#ifdef FULL_COMPARE_MODE
+  psx_gpu->pixel_count_mode = 1; 
+  psx_gpu->pixel_compare_mode = 0;
+  memcpy(psx_gpu->vram_ptr, state.vram, 1024 * 512 * 2);
+  //render_block_fill(psx_gpu, 0, 0, 0, 1024, 512);
+  gpu_parse(psx_gpu, list, size);
+
+  psx_gpu->pixel_count_mode = 0;
+  psx_gpu->pixel_compare_mode = 1;
+  memcpy(psx_gpu->compare_vram, state.vram, 1024 * 512 * 2); 
+  memcpy(psx_gpu->vram_ptr, state.vram, 1024 * 512 * 2);
+  //render_block_fill(psx_gpu, 0, 0, 0, 1024, 512);
+  clear_stats();
+  gpu_parse(psx_gpu, list, size);
+  flush_render_block_buffer(psx_gpu);
+#else
+  memcpy(psx_gpu->vram_ptr, state.vram, 1024 * 512 * 2);
+
+  psx_gpu->pixel_count_mode = 0;
+  psx_gpu->pixel_compare_mode = 0;
+
+  clear_stats();
+
+#ifdef PANDORA_BUILD
+  init_counter();
+#endif
+
+  gpu_parse(psx_gpu, list, size);
+  flush_render_block_buffer(psx_gpu);
+
+  clear_stats();
+
+#ifdef PANDORA_BUILD
+  u32 cycles = get_counter();
+#endif
+
+  gpu_parse(psx_gpu, list, size);
+  flush_render_block_buffer(psx_gpu);
+
+  printf("%s: ", argv[1]);
+#ifdef PANDORA_BUILD
+  u32 cycles_elapsed = get_counter() - cycles;
+
+  printf("%d\n", cycles_elapsed);
+#endif
+
+#if 1
+  u32 i;
+
+  for(i = 0; i < 1024 * 512; i++)
+  {
+    if((psx_gpu->vram_ptr[i] & 0x7FFF) != (state.vram[i] & 0x7FFF))
+    {
+      printf("(%d %d %d) vs (%d %d %d) at (%d %d)\n",
+       psx_gpu->vram_ptr[i] & 0x1F,
+       (psx_gpu->vram_ptr[i] >> 5) & 0x1F,
+       (psx_gpu->vram_ptr[i] >> 10) & 0x1F,
+       state.vram[i] & 0x1F,
+       (state.vram[i] >> 5) & 0x1F,
+       (state.vram[i] >> 10) & 0x1F, i % 1024, i / 1024);
+
+      mismatches++;
+    }
+    else
+    {
+      psx_gpu->vram_ptr[i] =
+       ((psx_gpu->vram_ptr[i] & 0x1F) / 4) |
+       ((((psx_gpu->vram_ptr[i] >> 5) & 0x1F) / 4) << 5) |
+       ((((psx_gpu->vram_ptr[i] >> 10) & 0x1F) / 4) << 10);
+    }
+  }
+#endif
+#endif
+
+#if 0
+  printf("\n");
+  printf("  %d pixels, %d pixel blocks (%d unaligned), %d spans\n"
+   "   (%lf pixels per block (%lf unaligned, r %lf), %lf pixels per span),\n"
+   "   %lf blocks per span (%lf per non-zero span), %lf overdraw)\n\n",
+   span_pixels, span_pixel_blocks, span_pixel_blocks_unaligned, spans,
+   (double)span_pixels / span_pixel_blocks,
+   (double)span_pixels / span_pixel_blocks_unaligned,
+   (double)span_pixel_blocks / span_pixel_blocks_unaligned,
+   (double)span_pixels / spans,
+   (double)span_pixel_blocks / spans, 
+   (double)span_pixel_blocks / (spans - zero_block_spans),
+   (double)span_pixels / 
+   ((psx_gpu->viewport_end_x - psx_gpu->viewport_start_x) * 
+   (psx_gpu->viewport_end_y - psx_gpu->viewport_start_y)));
+
+  printf("  %d triangles (%d false modulated)\n"
+   "   (%d trivial rejects, %lf%% flat, %lf%% left split, %lf%% clipped)\n"
+   "   (%lf pixels per triangle, %lf rows per triangle)\n\n",
+   triangles, false_modulated_triangles, trivial_rejects,
+   percent_of(flat_triangles, triangles),
+   percent_of(left_split_triangles, triangles),
+   percent_of(clipped_triangles, triangles),
+   (double)span_pixels / triangles,
+   (double)spans / triangles);
+
+  printf("  Block data:\n");
+  printf("   %7d 4bpp texel blocks  (%lf%%)\n", texel_blocks_4bpp,
+   percent_of(texel_blocks_4bpp, span_pixel_blocks));
+  printf("   %7d 8bpp texel blocks  (%lf%%)\n", texel_blocks_8bpp,
+   percent_of(texel_blocks_8bpp, span_pixel_blocks));
+  printf("   %7d 16bpp texel blocks (%lf%%)\n", texel_blocks_16bpp,
+   percent_of(texel_blocks_16bpp, span_pixel_blocks));
+  printf("   %7d untextured blocks  (%lf%%)\n", texel_blocks_untextured,
+   percent_of(texel_blocks_untextured, span_pixel_blocks));
+  printf("   %7d sprite blocks      (%lf%%)\n", sprite_blocks,  
+   percent_of(sprite_blocks, span_pixel_blocks));
+  printf("   %7d blended blocks     (%lf%%)\n", blend_blocks,
+   percent_of(blend_blocks, span_pixel_blocks));
+  printf("\n");
+  printf("  %lf blocks per render buffer flush\n", (double)span_pixel_blocks /
+   render_buffer_flushes);
+  printf("  %d zero block spans\n", zero_block_spans);
+  printf("  %d state changes, %d texture cache loads\n", state_changes,
+   texture_cache_loads);
+  if(sprites)
+  {
+    printf("  %d sprites\n"
+     "    4bpp:       %lf%%\n"
+     "    8bpp:       %lf%%\n"
+     "    16bpp:      %lf%%\n"
+     "    untextured: %lf%%\n",
+     sprites, percent_of(sprites_4bpp, sprites),
+     percent_of(sprites_8bpp, sprites), percent_of(sprites_16bpp, sprites),
+     percent_of(sprites_untextured, sprites));
+  }
+  printf("\n");
+  printf("  %d mismatches\n\n\n", mismatches);
+#endif
+
+  fflush(stdout);
+
+  if(no_display == 0)
+  {
+    while(1)
+    {
+      update_screen(psx_gpu, screen);
+  
+      if(SDL_PollEvent(&event))
+      {
+        if((event.type == SDL_QUIT) ||
+         ((event.type == SDL_KEYDOWN) &&
+         (event.key.keysym.sym == SDLK_ESCAPE)))
+        {
+          break;
+        }      
+      }
+  
+      SDL_Delay(20);
+    }
+  }
+
+  return (mismatches != 0);
+}
diff --git a/plugins/gpu_neon/psx_gpu/psx_gpu_parse.c b/plugins/gpu_neon/psx_gpu/psx_gpu_parse.c
new file mode 100644
index 0000000..98f2d85
--- /dev/null
+++ b/plugins/gpu_neon/psx_gpu/psx_gpu_parse.c
@@ -0,0 +1,670 @@
+/*
+ * Copyright (C) 2011 Gilead Kutnick "Exophase" <exophase@gmail.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; either version 2 of
+ * the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ */
+
+#include <stdio.h>
+
+#include "common.h"
+
+const u8 command_lengths[256] =
+{
+	0,  0,  2,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,   // 00
+	0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,   // 10
+	3,  3,  3,  3,  6,  6,  6,  6,  4,  4,  4,  4,  8,  8,  8,  8,   // 20
+	5,  5,  5,  5,  8,  8,  8,  8,  7,  7,  7,  7,  11, 11, 11, 11,  // 30
+	2,  2,  2,  2,  2,  2,  2,  2,  3,  3,  3,  3,  3,  3,  3,  3,   // 40
+	3,  3,  3,  3,  3,  3,  3,  3,  4,  4,  4,  4,  4,  4,  4,  4,   // 50
+	2,  2,  2,  2,  3,  3,  3,  3,  1,  1,  1,  1,  1,  1,  1,  1,   // 60
+	1,  1,  1,  1,  2,  2,  2,  2,  1,  1,  1,  1,  2,  2,  2,  2,   // 70
+	3,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,   // 80
+	0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,   // 90
+	2,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,   // a0
+	0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,   // b0
+	2,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,   // c0
+	0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,   // d0
+	0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,   // e0
+	0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0    // f0
+};
+
+void update_texture_ptr(psx_gpu_struct *psx_gpu)
+{
+  u8 *texture_ptr;
+
+  switch((psx_gpu->render_state_base >> 8) & 0x3)
+  {
+    default:
+    case TEXTURE_MODE_4BPP:
+#ifdef TEXTURE_CACHE_4BPP
+      texture_ptr = psx_gpu->texture_4bpp_cache[psx_gpu->current_texture_page];
+      texture_ptr += psx_gpu->texture_window_x & 0xF;
+      texture_ptr += (psx_gpu->texture_window_y & 0xF) << 4;
+      texture_ptr += (psx_gpu->texture_window_x >> 4) << 8;
+      texture_ptr += (psx_gpu->texture_window_y >> 4) << 12;
+#else
+      texture_ptr = (u8 *)(psx_gpu->vram_ptr);
+      texture_ptr += (psx_gpu->current_texture_page & 0xF) * 128;
+      texture_ptr += ((psx_gpu->current_texture_page >> 4) * 256) * 2048;
+      texture_ptr += psx_gpu->texture_window_x / 2;
+      texture_ptr += (psx_gpu->texture_window_y) * 2048;
+#endif
+      break;
+
+    case TEXTURE_MODE_8BPP:
+#ifdef TEXTURE_CACHE_8BPP
+      if(psx_gpu->current_texture_page & 0x1)
+      {
+        texture_ptr =
+         psx_gpu->texture_8bpp_odd_cache[psx_gpu->current_texture_page >> 1];
+      }
+      else
+      {
+        texture_ptr =
+         psx_gpu->texture_8bpp_even_cache[psx_gpu->current_texture_page >> 1];
+      }
+      
+      texture_ptr += (psx_gpu->texture_window_y & 0xF) << 4;
+      texture_ptr += (psx_gpu->texture_window_x >> 4) << 8;
+      texture_ptr += (psx_gpu->texture_window_y >> 4) << 12;
+#else
+      texture_ptr = (u8 *)(psx_gpu->vram_ptr);
+      texture_ptr += (psx_gpu->current_texture_page & 0xF) * 128;
+      texture_ptr += ((psx_gpu->current_texture_page >> 4) * 256) * 2048;
+      texture_ptr += psx_gpu->texture_window_x;
+      texture_ptr += (psx_gpu->texture_window_y) * 2048;
+#endif
+      break;
+
+    case TEXTURE_MODE_16BPP:
+      texture_ptr = (u8 *)(psx_gpu->vram_ptr);
+      texture_ptr += (psx_gpu->current_texture_page & 0xF) * 128;
+      texture_ptr += ((psx_gpu->current_texture_page >> 4) * 256) * 2048;
+      texture_ptr += psx_gpu->texture_window_x * 2;
+      texture_ptr += (psx_gpu->texture_window_y) * 2048;
+      break;
+  }
+
+  psx_gpu->texture_page_ptr = texture_ptr;  
+}
+
+void set_texture(psx_gpu_struct *psx_gpu, u32 texture_settings)
+{
+  if(psx_gpu->texture_settings != texture_settings)
+  {
+    u32 new_texture_page = texture_settings & 0x1F;
+    u32 texture_mode = (texture_settings >> 7) & 0x3;
+    u32 render_state_base = psx_gpu->render_state_base;
+
+    if(psx_gpu->current_texture_page != new_texture_page)
+      flush_render_block_buffer(psx_gpu);
+
+    render_state_base &= ~(0xF << 6);
+    render_state_base |= ((texture_settings >> 5) & 0xF) << 6;
+
+    psx_gpu->render_state_base = render_state_base;
+
+    psx_gpu->current_texture_mask = 0x1 << new_texture_page;
+
+    if(texture_mode == TEXTURE_MODE_8BPP)
+    {     
+      // In 8bpp mode 256x256 takes up two pages. If it's on the right edge it
+      // wraps back around to the left edge.
+      u32 adjacent_texture_page = ((texture_settings + 1) & 0xF) | (texture_settings & 0x10);
+      psx_gpu->current_texture_mask |= 0x1 << adjacent_texture_page;
+
+      if((psx_gpu->last_8bpp_texture_page ^ new_texture_page) & 0x1)
+      {
+        u32 dirty_textures_8bpp_alternate_mask =
+         psx_gpu->dirty_textures_8bpp_alternate_mask;
+        psx_gpu->dirty_textures_8bpp_alternate_mask =
+         psx_gpu->dirty_textures_8bpp_mask;
+        psx_gpu->dirty_textures_8bpp_mask = dirty_textures_8bpp_alternate_mask;
+      }
+
+      psx_gpu->last_8bpp_texture_page = new_texture_page;
+    }
+
+    psx_gpu->current_texture_page = new_texture_page;
+    psx_gpu->texture_settings = texture_settings;
+
+    update_texture_ptr(psx_gpu);
+  }
+}
+
+void set_clut(psx_gpu_struct *psx_gpu, u32 clut_settings)
+{
+  if(psx_gpu->clut_settings != clut_settings)
+  {
+    flush_render_block_buffer(psx_gpu);
+    psx_gpu->clut_settings = clut_settings;
+    psx_gpu->clut_ptr = psx_gpu->vram_ptr + ((clut_settings & 0x7FFF) * 16);
+  }
+}
+
+void set_triangle_color(psx_gpu_struct *psx_gpu, u32 triangle_color)
+{
+  if(psx_gpu->triangle_color != triangle_color)
+  {
+    flush_render_block_buffer(psx_gpu);
+    psx_gpu->triangle_color = triangle_color;
+  }
+}
+
+#define sign_extend_12bit(value)                                               \
+  (((s32)((value) << 20)) >> 20)                                               \
+
+#define get_vertex_data_xy(vertex_number, offset16)                            \
+  vertexes[vertex_number].x =                                                  \
+   sign_extend_12bit(list_s16[offset16]) + psx_gpu->offset_x;                  \
+  vertexes[vertex_number].y =                                                  \
+   sign_extend_12bit(list_s16[(offset16) + 1]) + psx_gpu->offset_y;            \
+
+#define get_vertex_data_uv(vertex_number, offset16)                            \
+  vertexes[vertex_number].u = list_s16[offset16] & 0xFF;                       \
+  vertexes[vertex_number].v = (list_s16[offset16] >> 8) & 0xFF                 \
+
+#define get_vertex_data_rgb(vertex_number, offset32)                           \
+  vertexes[vertex_number].r = list[offset32] & 0xFF;                           \
+  vertexes[vertex_number].g = (list[offset32] >> 8) & 0xFF;                    \
+  vertexes[vertex_number].b = (list[offset32] >> 16) & 0xFF                    \
+
+#define get_vertex_data_xy_uv(vertex_number, offset16)                         \
+  get_vertex_data_xy(vertex_number, offset16);                                 \
+  get_vertex_data_uv(vertex_number, (offset16) + 2)                            \
+
+#define get_vertex_data_xy_rgb(vertex_number, offset16)                        \
+  get_vertex_data_rgb(vertex_number, (offset16) / 2);                          \
+  get_vertex_data_xy(vertex_number, (offset16) + 2);                           \
+
+#define get_vertex_data_xy_uv_rgb(vertex_number, offset16)                     \
+  get_vertex_data_rgb(vertex_number, (offset16) / 2);                          \
+  get_vertex_data_xy(vertex_number, (offset16) + 2);                           \
+  get_vertex_data_uv(vertex_number, (offset16) + 4);                           \
+
+#define set_vertex_color_constant(vertex_number, color)                        \
+  vertexes[vertex_number].r = color & 0xFF;                                    \
+  vertexes[vertex_number].g = (color >> 8) & 0xFF;                             \
+  vertexes[vertex_number].b = (color >> 16) & 0xFF                             \
+
+#define get_vertex_data_xy_rgb_constant(vertex_number, offset16, color)        \
+  get_vertex_data_xy(vertex_number, offset16);                                 \
+  set_vertex_color_constant(vertex_number, color)                              \
+
+vertex_struct vertexes[4] __attribute__((aligned(32)));
+
+void gpu_parse(psx_gpu_struct *psx_gpu, u32 *list, u32 size)
+{
+  u32 current_command, command_length;
+  
+  u32 *list_end = list + (size / 4);
+
+  for(; list < list_end; list += 1 + command_length)
+  {
+  	s16 *list_s16 = (void *)list;
+  	current_command = *list >> 24;
+  	command_length = command_lengths[current_command];
+  
+  	switch(current_command)
+  	{
+  		case 0x00:
+  			break;
+  
+  		case 0x02:
+        render_block_fill(psx_gpu, list[0] & 0xFFFFFF, list_s16[2], list_s16[3],
+         list_s16[4] & 0x3FF, list_s16[5] & 0x3FF);
+  			break;
+  
+  		case 0x20 ... 0x23:
+      {
+        set_triangle_color(psx_gpu, list[0] & 0xFFFFFF);
+  
+        get_vertex_data_xy(0, 2);
+        get_vertex_data_xy(1, 4);
+        get_vertex_data_xy(2, 6);
+  
+        render_triangle(psx_gpu, vertexes, current_command);
+  			break;
+      }
+  
+  		case 0x24 ... 0x27:
+      {
+        set_clut(psx_gpu, list_s16[5]);
+        set_texture(psx_gpu, list_s16[9]);
+        set_triangle_color(psx_gpu, list[0] & 0xFFFFFF);
+  
+        get_vertex_data_xy_uv(0, 2);
+        get_vertex_data_xy_uv(1, 6);
+        get_vertex_data_xy_uv(2, 10);
+  
+        render_triangle(psx_gpu, vertexes, current_command);
+  			break;
+      }
+  
+  		case 0x28 ... 0x2B:
+      {
+        set_triangle_color(psx_gpu, list[0] & 0xFFFFFF);
+  
+        get_vertex_data_xy(0, 2);
+        get_vertex_data_xy(1, 4);
+        get_vertex_data_xy(2, 6);
+        get_vertex_data_xy(3, 8);
+  
+        render_triangle(psx_gpu, vertexes, current_command);
+        render_triangle(psx_gpu, &(vertexes[1]), current_command);
+  			break;
+      }
+  
+  		case 0x2C ... 0x2F:
+      {
+        set_clut(psx_gpu, list_s16[5]);
+        set_texture(psx_gpu, list_s16[9]);
+        set_triangle_color(psx_gpu, list[0] & 0xFFFFFF);
+  
+        get_vertex_data_xy_uv(0, 2);   
+        get_vertex_data_xy_uv(1, 6);   
+        get_vertex_data_xy_uv(2, 10);  
+        get_vertex_data_xy_uv(3, 14);
+  
+        render_triangle(psx_gpu, vertexes, current_command);
+        render_triangle(psx_gpu, &(vertexes[1]), current_command);
+  			break;
+      }
+  
+  		case 0x30 ... 0x33:
+      {
+        get_vertex_data_xy_rgb(0, 0);
+        get_vertex_data_xy_rgb(1, 4);
+        get_vertex_data_xy_rgb(2, 8);
+  
+        render_triangle(psx_gpu, vertexes, current_command);
+  			break;
+      }
+  
+  		case 0x34:
+  		case 0x35:
+  		case 0x36:
+  		case 0x37:
+      {
+        set_clut(psx_gpu, list_s16[5]);
+        set_texture(psx_gpu, list_s16[11]);
+  
+        get_vertex_data_xy_uv_rgb(0, 0);
+        get_vertex_data_xy_uv_rgb(1, 6);
+        get_vertex_data_xy_uv_rgb(2, 12);
+
+        render_triangle(psx_gpu, vertexes, current_command);
+  			break;
+      }
+  
+  		case 0x38:
+  		case 0x39:
+  		case 0x3A:
+  		case 0x3B:
+      {
+        get_vertex_data_xy_rgb(0, 0);
+        get_vertex_data_xy_rgb(1, 4);
+        get_vertex_data_xy_rgb(2, 8);
+        get_vertex_data_xy_rgb(3, 12);
+  
+        render_triangle(psx_gpu, vertexes, current_command);
+        render_triangle(psx_gpu, &(vertexes[1]), current_command);
+  			break;
+      }
+  
+  		case 0x3C:
+  		case 0x3D:
+  		case 0x3E:
+  		case 0x3F:
+      {
+        set_clut(psx_gpu, list_s16[5]);
+        set_texture(psx_gpu, list_s16[11]);
+  
+        get_vertex_data_xy_uv_rgb(0, 0);
+        get_vertex_data_xy_uv_rgb(1, 6);
+        get_vertex_data_xy_uv_rgb(2, 12);
+        get_vertex_data_xy_uv_rgb(3, 18);
+  
+        render_triangle(psx_gpu, vertexes, current_command);
+        render_triangle(psx_gpu, &(vertexes[1]), current_command);
+  			break;
+      }
+  
+  		case 0x40 ... 0x47:
+      {
+        vertexes[0].x = list_s16[2] + psx_gpu->offset_x;
+        vertexes[0].y = list_s16[3] + psx_gpu->offset_y;
+        vertexes[1].x = list_s16[4] + psx_gpu->offset_x;
+        vertexes[1].y = list_s16[5] + psx_gpu->offset_y;
+
+        render_line(psx_gpu, vertexes, current_command, list[0]);
+  			break;
+      }
+  
+  		case 0x48 ... 0x4F:
+      {
+        u32 num_vertexes = 1;
+        u32 *list_position = &(list[2]);
+        u32 xy = list[1];
+
+        vertexes[1].x = (xy & 0xFFFF) + psx_gpu->offset_x;
+        vertexes[1].y = (xy >> 16) + psx_gpu->offset_y;
+      
+        while(1)
+        {
+          xy = *list_position;
+          if(xy == 0x55555555)
+            break;
+
+          vertexes[0] = vertexes[1];
+
+          vertexes[1].x = (xy & 0xFFFF) + psx_gpu->offset_x;
+          vertexes[1].y = (xy >> 16) + psx_gpu->offset_y;
+
+          list_position++;
+          num_vertexes++;
+
+          render_line(psx_gpu, vertexes, current_command, list[0]);
+        }
+
+        if(num_vertexes > 2)
+          command_length += (num_vertexes - 2);
+
+  			break;
+      }
+  
+  		case 0x50 ... 0x57:
+      {
+        vertexes[0].r = list[0] & 0xFF;
+        vertexes[0].g = (list[0] >> 8) & 0xFF;
+        vertexes[0].b = (list[0] >> 16) & 0xFF;
+        vertexes[0].x = list_s16[2] + psx_gpu->offset_x;
+        vertexes[0].y = list_s16[3] + psx_gpu->offset_y;
+
+        vertexes[1].r = list[2] & 0xFF;
+        vertexes[1].g = (list[2] >> 8) & 0xFF;
+        vertexes[1].b = (list[2] >> 16) & 0xFF;
+        vertexes[1].x = list_s16[6] + psx_gpu->offset_x;
+        vertexes[1].y = list_s16[7] + psx_gpu->offset_y;
+
+        render_line(psx_gpu, vertexes, current_command, 0);
+  			break;
+      }
+ 
+      case 0x58 ... 0x5F:
+      {
+        u32 num_vertexes = 1;
+        u32 *list_position = &(list[2]);
+        u32 color = list[0];
+        u32 xy = list[1];
+
+        vertexes[1].r = color & 0xFF;
+        vertexes[1].g = (color >> 8) & 0xFF;
+        vertexes[1].b = (color >> 16) & 0xFF;
+        vertexes[1].x = (xy & 0xFFFF) + psx_gpu->offset_x;
+        vertexes[1].y = (xy >> 16) + psx_gpu->offset_y;
+      
+        while(1)
+        {
+          color = list_position[0];
+          if(color == 0x55555555)
+            break;
+
+          xy = list_position[1];
+
+          vertexes[0] = vertexes[1];
+
+          vertexes[1].r = color & 0xFF;
+          vertexes[1].g = (color >> 8) & 0xFF;
+          vertexes[1].b = (color >> 16) & 0xFF;
+          vertexes[1].x = (xy & 0xFFFF) + psx_gpu->offset_x;
+          vertexes[1].y = (xy >> 16) + psx_gpu->offset_y;
+
+          list_position += 2;
+          num_vertexes++;
+
+          render_line(psx_gpu, vertexes, current_command, 0);
+        }
+
+        if(num_vertexes > 2)
+          command_length += ((num_vertexes * 2) - 2);
+
+  			break;
+      }
+  
+  		case 0x60 ... 0x63:
+      {        
+        u32 x = list_s16[2] + psx_gpu->offset_x;
+        u32 y = list_s16[3] + psx_gpu->offset_y;
+        u32 width = list_s16[4] & 0x3FF;
+        u32 height = list_s16[5] & 0x1FF;
+
+        psx_gpu->primitive_color = list[0] & 0xFFFFFF;
+
+        render_sprite(psx_gpu, x, y, 0, 0, width, height, current_command, list[0]);
+  			break;
+      }
+  
+  		case 0x64 ... 0x67:
+      {        
+        u32 x = list_s16[2] + psx_gpu->offset_x;
+        u32 y = list_s16[3] + psx_gpu->offset_y;
+        u32 uv = list_s16[4];
+        u32 width = list_s16[6] & 0x3FF;
+        u32 height = list_s16[7] & 0x1FF;
+
+        psx_gpu->primitive_color = list[0] & 0xFFFFFF;
+        set_clut(psx_gpu, list_s16[5]);
+
+        render_sprite(psx_gpu, x, y, uv & 0xFF, (uv >> 8) & 0xFF, width, height,
+         current_command, list[0]);
+  			break;
+      }
+  
+  		case 0x68:
+  		case 0x69:
+  		case 0x6A:
+  		case 0x6B:
+      {
+        s32 x = list_s16[2] + psx_gpu->offset_x;
+        s32 y = list_s16[3] + psx_gpu->offset_y;
+
+        psx_gpu->primitive_color = list[0] & 0xFFFFFF;
+
+        render_sprite(psx_gpu, x, y, 0, 0, 1, 1, current_command, list[0]);
+  			break;
+      }
+  
+  		case 0x70:
+  		case 0x71:
+  		case 0x72:
+  		case 0x73:
+      {        
+        s32 x = list_s16[2] + psx_gpu->offset_x;
+        s32 y = list_s16[3] + psx_gpu->offset_y;
+
+        psx_gpu->primitive_color = list[0] & 0xFFFFFF;
+
+        render_sprite(psx_gpu, x, y, 0, 0, 8, 8, current_command, list[0]);
+  			break;
+      }
+  
+  		case 0x74:
+  		case 0x75:
+  		case 0x76:
+  		case 0x77:
+      {        
+        s32 x = list_s16[2] + psx_gpu->offset_x;
+        s32 y = list_s16[3] + psx_gpu->offset_y;
+        u32 uv = list_s16[4];
+
+        psx_gpu->primitive_color = list[0] & 0xFFFFFF;
+        set_clut(psx_gpu, list_s16[5]);
+
+        render_sprite(psx_gpu, x, y, uv & 0xFF, (uv >> 8) & 0xFF, 8, 8,
+         current_command, list[0]);
+  			break;
+      }
+  
+  		case 0x78:
+  		case 0x79:
+  		case 0x7A:
+  		case 0x7B:
+      {        
+        s32 x = list_s16[2] + psx_gpu->offset_x;
+        s32 y = list_s16[3] + psx_gpu->offset_y;
+
+        psx_gpu->primitive_color = list[0] & 0xFFFFFF;
+        render_sprite(psx_gpu, x, y, 0, 0, 16, 16, current_command, list[0]);
+  			break;
+      }
+  
+  		case 0x7C:
+  		case 0x7D:
+  		case 0x7E:
+  		case 0x7F:
+      {        
+        s32 x = list_s16[2] + psx_gpu->offset_x;
+        s32 y = list_s16[3] + psx_gpu->offset_y;
+        u32 uv = list_s16[4];
+
+        psx_gpu->primitive_color = list[0] & 0xFFFFFF;
+        set_clut(psx_gpu, list_s16[5]);
+
+        render_sprite(psx_gpu, x, y, uv & 0xFF, (uv >> 8) & 0xFF, 16, 16,
+         current_command, list[0]);
+  			break;
+      }
+  
+  		case 0x80:          //  vid -> vid
+        render_block_move(psx_gpu, list_s16[2] & 0x3FF, list_s16[3] & 0x1FF,
+         list_s16[4] & 0x3FF, list_s16[5] & 0x1FF, list_s16[6], list_s16[7]);
+  			break;
+  
+  		case 0xA0:          //  sys -> vid
+      {
+        u32 load_x = list_s16[2];
+        u32 load_y = list_s16[3];
+        u32 load_width = list_s16[4];
+        u32 load_height = list_s16[5];
+        u32 load_size = load_width * load_height;
+  
+        command_length += load_size / 2;
+  
+        render_block_copy(psx_gpu, (u16 *)&(list_s16[6]), load_x, load_y,
+         load_width, load_height, load_width);
+  			break;
+      }
+  
+  		case 0xC0:          //  vid -> sys
+  			break;
+  
+  		case 0xE1:
+        set_texture(psx_gpu, list[0] & 0x1FF);
+        if(list[0] & (1 << 9))
+          psx_gpu->render_state_base |= RENDER_STATE_DITHER;
+        else
+          psx_gpu->render_state_base &= ~RENDER_STATE_DITHER;
+
+        psx_gpu->display_area_draw_enable = (list[0] >> 10) & 0x1;
+  			break;
+  
+  		case 0xE2:
+      {
+        // TODO: Clean
+        u32 texture_window_settings = list[0];
+        u32 tmp, x, y, w, h;
+
+        if(texture_window_settings != psx_gpu->texture_window_settings)
+        {
+          tmp = (texture_window_settings & 0x1F) | 0x20;
+          for(w = 8; (tmp & 1) == 0; tmp >>= 1, w <<= 1);
+
+          tmp = ((texture_window_settings >> 5) & 0x1f) | 0x20;
+          for (h = 8; (tmp & 1) == 0; tmp >>= 1, h <<= 1);
+
+          tmp = 32 - (w >> 3);
+          x = ((texture_window_settings >> 10) & tmp) << 3;
+
+          tmp = 32 - (h >> 3);
+          y = ((texture_window_settings >> 15) & tmp) << 3;
+
+          flush_render_block_buffer(psx_gpu);
+          
+          psx_gpu->texture_window_x = x;
+          psx_gpu->texture_window_y = y;
+          psx_gpu->texture_mask_width = w - 1;
+          psx_gpu->texture_mask_height = h - 1;
+
+          update_texture_ptr(psx_gpu);
+        }
+        break;
+  		}
+  
+  		case 0xE3:
+        psx_gpu->viewport_start_x = list[0] & 0x3FF;
+        psx_gpu->viewport_start_y = (list[0] >> 10) & 0x1FF;
+
+#ifdef TEXTURE_CACHE_4BPP
+        psx_gpu->viewport_mask =
+         texture_region_mask(psx_gpu->viewport_start_x,
+         psx_gpu->viewport_start_y, psx_gpu->viewport_end_x,
+         psx_gpu->viewport_end_y);
+#endif
+  			break;
+  
+  		case 0xE4:
+        psx_gpu->viewport_end_x = list[0] & 0x3FF;
+        psx_gpu->viewport_end_y = (list[0] >> 10) & 0x1FF;
+
+#ifdef TEXTURE_CACHE_4BPP
+        psx_gpu->viewport_mask =
+         texture_region_mask(psx_gpu->viewport_start_x,
+         psx_gpu->viewport_start_y, psx_gpu->viewport_end_x,
+         psx_gpu->viewport_end_y);
+#endif
+  			break;
+  
+  		case 0xE5:
+      {
+        s32 offset_x = list[0] << 21;
+        s32 offset_y = list[0] << 10;
+        psx_gpu->offset_x = offset_x >> 21;
+        psx_gpu->offset_y = offset_y >> 21; 
+  
+  			break;
+  		}
+
+  		case 0xE6:
+      {
+        u32 mask_settings = list[0];
+        u16 mask_msb = mask_settings << 15;
+
+        if(list[0] & 0x2)
+          psx_gpu->render_state_base |= RENDER_STATE_MASK_EVALUATE;
+        else
+          psx_gpu->render_state_base &= ~RENDER_STATE_MASK_EVALUATE;
+
+        if(mask_msb != psx_gpu->mask_msb)
+        {
+          flush_render_block_buffer(psx_gpu);
+          psx_gpu->mask_msb = mask_msb;
+        }
+
+  			break;
+      }
+  
+  		default:
+  			break;
+  	}
+  }
+}
+
diff --git a/plugins/gpu_neon/psx_gpu/psx_gpu_standard.c b/plugins/gpu_neon/psx_gpu/psx_gpu_standard.c
new file mode 100644
index 0000000..68d29cf
--- /dev/null
+++ b/plugins/gpu_neon/psx_gpu/psx_gpu_standard.c
@@ -0,0 +1,1166 @@
+/*
+ * Copyright (C) 2011 Gilead Kutnick "Exophase" <exophase@gmail.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; either version 2 of
+ * the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <malloc.h>
+#include <math.h>
+
+#include "common.h"
+
+typedef s32 fixed_type;
+
+#define EDGE_STEP_BITS 32
+#define FIXED_BITS     12
+
+#define fixed_center(value)                                                    \
+  ((((fixed_type)value) << FIXED_BITS) + (1 << (FIXED_BITS - 1)))              \
+
+#define int_to_fixed(value)                                                    \
+  (((fixed_type)value) << FIXED_BITS)                                          \
+
+#define fixed_to_int(value)                                                    \
+  ((value) >> FIXED_BITS)                                                      \
+
+#define fixed_mul(_a, _b)                                                      \
+  (((s64)(_a) * (_b)) >> FIXED_BITS)                                           \
+
+#define fixed_to_double(value)                                                 \
+  ((value) / (double)(1 << FIXED_BITS))                                        \
+
+#define double_to_fixed(value)                                                 \
+  (fixed_type)(((value) * (double)(1 << FIXED_BITS)))                          \
+
+typedef struct
+{
+  fixed_type current_value;
+  fixed_type step_dx;
+  fixed_type step_dy;
+  fixed_type gradient_area_x;
+  fixed_type gradient_area_y;
+} interpolant_struct;
+
+typedef struct
+{
+  s32 base_x;
+
+  s64 left_x;
+  s64 left_dx_dy;
+
+  s64 right_x;
+  s64 right_dx_dy;
+
+  u32 triangle_area;
+  u32 triangle_winding;
+
+  interpolant_struct u;
+  interpolant_struct v;
+  interpolant_struct r;
+  interpolant_struct g;
+  interpolant_struct b;
+} _span_struct;
+
+
+u32 span_pixels = 0;
+u32 span_pixel_blocks = 0;
+u32 spans = 0;
+u32 triangles = 0;
+
+u32 texels_4bpp = 0;
+u32 texels_8bpp = 0;
+u32 texels_16bpp = 0;
+u32 untextured_pixels = 0;
+u32 blend_pixels = 0;
+u32 transparent_pixels = 0;
+
+u32 state_changes = 0;
+u32 render_buffer_flushes = 0;
+u32 trivial_rejects = 0;
+
+void flush_render_block_buffer(psx_gpu_struct *psx_gpu)
+{
+
+}
+
+
+u32 fixed_reciprocal(u32 denominator, u32 *_shift)
+{
+  u32 shift = __builtin_clz(denominator);
+  u32 denominator_normalized = denominator << shift;
+
+  // Implement with a DP divide
+  u32 reciprocal =
+   (double)((1ULL << 62) + (denominator_normalized - 1)) / 
+   (double)denominator_normalized;
+
+  *_shift = 62 - shift;
+  return reciprocal;
+}
+
+fixed_type fixed_reciprocal_multiply(s32 numerator, u32 reciprocal,
+ u32 reciprocal_sign, u32 shift)
+{
+  u32 numerator_sign = (u32)numerator >> 31;
+  u32 flip_sign = numerator_sign ^ reciprocal_sign;
+  u32 flip_sign_mask = ~(flip_sign - 1);
+  fixed_type value;
+
+  numerator = abs(numerator);
+
+  value = ((u64)numerator * reciprocal) >> shift;
+
+  value ^= flip_sign_mask;
+  value -= flip_sign_mask;
+
+  return value;
+}
+
+s32 triangle_signed_area_x2(s32 x0, s32 y0, s32 x1, s32 y1, s32 x2, s32 y2)
+{
+	return ((x1 - x0) * (y2 - y1)) - ((x2 - x1) * (y1 - y0));
+}
+
+u32 fetch_texel_4bpp(psx_gpu_struct *psx_gpu, u32 u, u32 v)
+{
+  u8 *texture_ptr_8bpp = psx_gpu->texture_page_ptr;
+  u32 texel = texture_ptr_8bpp[(v * 2048) + (u / 2)];
+
+  if(u & 1)
+    texel >>= 4;
+  else
+    texel &= 0xF;
+
+  texels_4bpp++;
+
+  return psx_gpu->clut_ptr[texel];
+}
+
+u32 fetch_texel_8bpp(psx_gpu_struct *psx_gpu, u32 u, u32 v)
+{
+  u8 *texture_ptr_8bpp = psx_gpu->texture_page_ptr;
+  u32 texel = texture_ptr_8bpp[(v * 2048) + u];
+
+  texels_8bpp++;
+
+  return psx_gpu->clut_ptr[texel];
+}
+
+u32 fetch_texel_16bpp(psx_gpu_struct *psx_gpu, u32 u, u32 v)
+{
+  u16 *texture_ptr_16bpp = psx_gpu->texture_page_ptr;
+
+  texels_16bpp++;
+
+  return texture_ptr_16bpp[(v * 1024) + u];
+}
+
+u32 fetch_texel(psx_gpu_struct *psx_gpu, u32 u, u32 v)
+{
+  u &= psx_gpu->texture_mask_width;
+  v &= psx_gpu->texture_mask_height;
+
+  switch(psx_gpu->texture_mode)
+  {
+    case TEXTURE_MODE_4BPP:
+      return fetch_texel_4bpp(psx_gpu, u, v);
+
+    case TEXTURE_MODE_8BPP:
+      return fetch_texel_8bpp(psx_gpu, u, v);
+
+    case TEXTURE_MODE_16BPP:
+      return fetch_texel_16bpp(psx_gpu, u, v);
+  }
+
+  return 0;
+}
+
+void draw_pixel(psx_gpu_struct *psx_gpu, s32 r, s32 g, s32 b, u32 texel,
+ u32 x, u32 y, u32 flags)
+{
+  u32 pixel;
+
+  if(r > 31)
+    r = 31;
+
+  if(g > 31)
+    g = 31;
+
+  if(b > 31)
+    b = 31;
+
+  if(flags & RENDER_FLAGS_BLEND)
+  {
+    if(((flags & RENDER_FLAGS_TEXTURE_MAP) == 0) || (texel & 0x8000))
+    {
+      s32 fb_pixel = psx_gpu->vram[(y * 1024) + x];
+      s32 fb_r = fb_pixel & 0x1F;
+      s32 fb_g = (fb_pixel >> 5) & 0x1F;
+      s32 fb_b = (fb_pixel >> 10) & 0x1F;
+
+      blend_pixels++;
+
+      switch(psx_gpu->blend_mode)
+      {
+        case BLEND_MODE_AVERAGE:
+          r = (r + fb_r) / 2;
+          g = (g + fb_g) / 2;
+          b = (b + fb_b) / 2;
+          break;
+
+        case BLEND_MODE_ADD:
+          r += fb_r;
+          g += fb_g;
+          b += fb_b;
+
+          if(r > 31)
+            r = 31;
+
+          if(g > 31)
+            g = 31;
+
+          if(b > 31)
+            b = 31;
+
+          break;
+
+        case BLEND_MODE_SUBTRACT:
+          r = fb_r - r;
+          g = fb_g - g;
+          b = fb_b - b;
+
+          if(r < 0)
+            r = 0;
+
+          if(g < 0)
+            g = 0;
+
+          if(b < 0)
+            b = 0;
+
+          break;
+
+        case BLEND_MODE_ADD_FOURTH:
+          r = fb_r + (r / 4);
+          g = fb_g + (g / 4);
+          b = fb_b + (b / 4);
+
+          if(r > 31)
+            r = 31;
+
+          if(g > 31)
+            g = 31;
+
+          if(b > 31)
+            b = 31;
+
+          break;      
+      }
+    }
+  }
+
+  pixel = r | (g << 5) | (b << 10);
+
+  if(psx_gpu->mask_apply || (texel & 0x8000))
+    pixel |= 0x8000;
+
+  psx_gpu->vram[(y * 1024) + x] = pixel;
+}
+
+s32 dither_table[4][4] =
+{
+  { -4,  0, -3,  1 },
+  {  2, -2,  3, -1 },
+  { -3,  1, -4,  0 },
+  {  3, -1,  2, -2 },
+};
+
+void render_span(psx_gpu_struct *psx_gpu, _span_struct *span, s32 y,
+ u32 flags)
+{
+  s32 left_x = span->left_x >> EDGE_STEP_BITS;
+  s32 right_x = span->right_x >> EDGE_STEP_BITS;
+  s32 current_x = left_x;
+  s32 delta_x;
+
+  fixed_type current_u = span->u.current_value;
+  fixed_type current_v = span->v.current_value;
+  fixed_type current_r = span->r.current_value;
+  fixed_type current_g = span->g.current_value;
+  fixed_type current_b = span->b.current_value;
+
+  if(y < psx_gpu->viewport_start_y)
+    return;
+
+  if(y > psx_gpu->viewport_end_y)
+    return;
+
+  if(right_x < psx_gpu->viewport_start_x)
+    return;
+
+  if(current_x > psx_gpu->viewport_end_x)
+    return;
+
+  spans++;
+
+  if(current_x < psx_gpu->viewport_start_x)
+    current_x = psx_gpu->viewport_start_x;  
+
+  if(right_x > psx_gpu->viewport_end_x + 1)
+    right_x = psx_gpu->viewport_end_x + 1;
+
+  delta_x = current_x - span->base_x;
+
+  current_u += delta_x * span->u.step_dx;
+  current_v += delta_x * span->v.step_dx;
+  current_r += delta_x * span->r.step_dx;
+  current_g += delta_x * span->g.step_dx;
+  current_b += delta_x * span->b.step_dx;
+
+  span_pixels += right_x - current_x;
+  span_pixel_blocks += ((right_x / 8) - (current_x / 8)) + 1;
+
+  while(current_x < right_x)
+  {
+    s32 color_r, color_g, color_b;
+    u32 texel = 0;
+
+    if(psx_gpu->mask_evaluate &&
+     (psx_gpu->vram[(y * 1024) + current_x] & 0x8000))
+    {
+      goto skip_pixel;
+    }
+
+    if(flags & RENDER_FLAGS_SHADE)
+    {
+      color_r = fixed_to_int(current_r);
+      color_g = fixed_to_int(current_g);
+      color_b = fixed_to_int(current_b);
+    }
+    else
+    {
+      color_r = psx_gpu->primitive_color & 0xFF;
+      color_g = (psx_gpu->primitive_color >> 8) & 0xFF;
+      color_b = (psx_gpu->primitive_color >> 16) & 0xFF;
+    }      
+
+    if(flags & RENDER_FLAGS_TEXTURE_MAP)
+    {
+      u32 texel_r, texel_g, texel_b;
+      u32 u = fixed_to_int(current_u);
+      u32 v = fixed_to_int(current_v);
+
+      texel = fetch_texel(psx_gpu, u, v);
+
+      if(texel == 0)
+      {
+        transparent_pixels++;
+        goto skip_pixel;
+      }
+
+      texel_r = texel & 0x1F;
+      texel_g = (texel >> 5) & 0x1F;
+      texel_b = (texel >> 10) & 0x1F;
+
+      if((flags & RENDER_FLAGS_MODULATE_TEXELS) == 0)
+      {
+        color_r *= texel_r;
+        color_g *= texel_g;
+        color_b *= texel_b;
+      }
+      else
+      {
+        color_r = texel_r << 7;
+        color_g = texel_g << 7;
+        color_b = texel_b << 7;
+      }
+
+      color_r >>= 4;
+      color_g >>= 4;
+      color_b >>= 4;
+    }
+    else
+    {
+      untextured_pixels++;
+    }
+
+    if(psx_gpu->dither_mode && ((flags & RENDER_FLAGS_SHADE) ||
+     ((flags & RENDER_FLAGS_TEXTURE_MAP) &&
+     ((flags & RENDER_FLAGS_MODULATE_TEXELS) == 0))))
+    {
+      s32 dither_offset = dither_table[y % 4][current_x % 4];
+      color_r += dither_offset;
+      color_g += dither_offset;
+      color_b += dither_offset;
+
+      if(color_r < 0)
+        color_r = 0;
+  
+      if(color_g < 0)
+        color_g = 0;
+  
+      if(color_b < 0)
+        color_b = 0;
+    }
+
+    color_r >>= 3;
+    color_g >>= 3;
+    color_b >>= 3;
+
+    draw_pixel(psx_gpu, color_r, color_g, color_b, texel, current_x, y, flags);
+
+  skip_pixel:
+  
+    current_u += span->u.step_dx;
+    current_v += span->v.step_dx;
+    current_r += span->r.step_dx;
+    current_g += span->g.step_dx;
+    current_b += span->b.step_dx;
+
+    current_x++;
+  }
+}
+
+void increment_span(_span_struct *span)
+{
+  span->left_x += span->left_dx_dy;
+  span->right_x += span->right_dx_dy;
+
+  span->u.current_value += span->u.step_dy;
+  span->v.current_value += span->v.step_dy;
+  span->r.current_value += span->r.step_dy;
+  span->g.current_value += span->g.step_dy;
+  span->b.current_value += span->b.step_dy;
+}
+
+void decrement_span(_span_struct *span)
+{
+  span->left_x += span->left_dx_dy;
+  span->right_x += span->right_dx_dy;
+
+  span->u.current_value -= span->u.step_dy;
+  span->v.current_value -= span->v.step_dy;
+  span->r.current_value -= span->r.step_dy;
+  span->g.current_value -= span->g.step_dy;
+  span->b.current_value -= span->b.step_dy;
+}
+
+
+#define compute_gradient_area_x(interpolant)                                   \
+{                                                                              \
+  span.interpolant.gradient_area_x =                                           \
+   triangle_signed_area_x2(a->interpolant, a->y, b->interpolant, b->y,         \
+   c->interpolant, c->y);                                                      \
+}                                                                              \
+
+#define compute_gradient_area_y(interpolant)                                   \
+{                                                                              \
+  span.interpolant.gradient_area_y =                                           \
+   triangle_signed_area_x2(a->x, a->interpolant,  b->x, b->interpolant,        \
+   c->x, c->interpolant);                                                      \
+}                                                                              \
+
+#define compute_all_gradient_areas()                                           \
+  compute_gradient_area_x(u);                                                  \
+  compute_gradient_area_x(v);                                                  \
+  compute_gradient_area_x(r);                                                  \
+  compute_gradient_area_x(g);                                                  \
+  compute_gradient_area_x(b);                                                  \
+  compute_gradient_area_y(u);                                                  \
+  compute_gradient_area_y(v);                                                  \
+  compute_gradient_area_y(r);                                                  \
+  compute_gradient_area_y(g);                                                  \
+  compute_gradient_area_y(b)                                                   \
+
+#define set_interpolant_base(interpolant, base_vertex)                         \
+  span->interpolant.step_dx =                                                  \
+   fixed_reciprocal_multiply(span->interpolant.gradient_area_x, reciprocal,    \
+   span->triangle_winding, shift);                                             \
+  span->interpolant.step_dy =                                                  \
+   fixed_reciprocal_multiply(span->interpolant.gradient_area_y, reciprocal,    \
+   span->triangle_winding, shift);                                             \
+  span->interpolant.current_value = fixed_center(base_vertex->interpolant)     \
+
+#define set_interpolant_bases(base_vertex)                                     \
+{                                                                              \
+  u32 shift;                                                                   \
+  u32 reciprocal = fixed_reciprocal(span->triangle_area, &shift);              \
+  shift -= FIXED_BITS;                                                         \
+  set_interpolant_base(u, base_vertex);                                        \
+  set_interpolant_base(v, base_vertex);                                        \
+  set_interpolant_base(r, base_vertex);                                        \
+  set_interpolant_base(g, base_vertex);                                        \
+  set_interpolant_base(b, base_vertex);                                        \
+  span->base_x = span->left_x >> EDGE_STEP_BITS;                               \
+}                                                                              \
+
+#define compute_edge_delta(edge, start, end, height)                           \
+{                                                                              \
+  s32 x_start = start->x;                                                      \
+  s32 x_end = end->x;                                                          \
+  s32 width = x_end - x_start;                                                 \
+                                                                               \
+  s32 shift = __builtin_clz(height);                                           \
+  u32 height_normalized = height << shift;                                     \
+  u32 height_reciprocal = ((1ULL << 50) + (height_normalized - 1)) /           \
+   height_normalized;                                                          \
+                                                                               \
+  shift -= (50 - EDGE_STEP_BITS);                                              \
+                                                                               \
+  span->edge##_x =                                                             \
+   ((((s64)x_start * height) + (height - 1)) * height_reciprocal) << shift;    \
+  span->edge##_dx_dy = ((s64)width * height_reciprocal) << shift;              \
+}                                                                              \
+
+
+#define render_spans_up(height)                                                \
+  do                                                                           \
+  {                                                                            \
+    decrement_span(span);                                                      \
+    render_span(psx_gpu, span, current_y, flags);                              \
+    current_y--;                                                               \
+    height--;                                                                  \
+  } while(height)                                                              \
+
+#define render_spans_down(height)                                              \
+  do                                                                           \
+  {                                                                            \
+    render_span(psx_gpu, span, current_y, flags);                              \
+    increment_span(span);                                                      \
+    current_y++;                                                               \
+    height--;                                                                  \
+  } while(height)                                                              \
+
+#define render_spans_up_up(minor, major)                                       \
+  s32 current_y = bottom->y - 1;                                               \
+  s32 height_minor_a = bottom->y - middle->y;                                  \
+  s32 height_minor_b = middle->y - top->y;                                     \
+  s32 height_major = height_minor_a + height_minor_b;                          \
+                                                                               \
+  compute_edge_delta(major, bottom, top, height_major);                        \
+  compute_edge_delta(minor, bottom, middle, height_minor_a);                   \
+  set_interpolant_bases(bottom);                                               \
+                                                                               \
+  render_spans_up(height_minor_a);                                             \
+                                                                               \
+  compute_edge_delta(minor, middle, top, height_minor_b);                      \
+  render_spans_up(height_minor_b)                                              \
+
+void render_spans_up_left(psx_gpu_struct *psx_gpu, _span_struct *span,
+ vertex_struct *bottom, vertex_struct *middle, vertex_struct *top, u32 flags)
+{
+  render_spans_up_up(left, right);
+}
+
+void render_spans_up_right(psx_gpu_struct *psx_gpu, _span_struct *span,
+ vertex_struct *bottom, vertex_struct *middle, vertex_struct *top, u32 flags)
+{
+  render_spans_up_up(right, left);
+}
+
+#define render_spans_down_down(minor, major)                                   \
+  s32 current_y = top->y;                                                      \
+  s32 height_minor_a = middle->y - top->y;                                     \
+  s32 height_minor_b = bottom->y - middle->y;                                  \
+  s32 height_major = height_minor_a + height_minor_b;                          \
+                                                                               \
+  compute_edge_delta(minor, top, middle, height_minor_a);                      \
+  compute_edge_delta(major, top, bottom, height_major);                        \
+  set_interpolant_bases(top);                                                  \
+                                                                               \
+  render_spans_down(height_minor_a);                                           \
+                                                                               \
+  compute_edge_delta(minor, middle, bottom, height_minor_b);                   \
+  render_spans_down(height_minor_b)                                            \
+
+void render_spans_down_left(psx_gpu_struct *psx_gpu, _span_struct *span,
+ vertex_struct *top, vertex_struct *middle, vertex_struct *bottom, u32 flags)
+{
+  render_spans_down_down(left, right);
+}
+
+void render_spans_down_right(psx_gpu_struct *psx_gpu, _span_struct *span,
+ vertex_struct *top, vertex_struct *middle, vertex_struct *bottom, u32 flags)
+{
+  render_spans_down_down(right, left);
+}
+
+#define render_spans_up_flat(bottom_left, bottom_right, top_left, top_right)   \
+  s32 current_y = bottom_left->y - 1;                                          \
+  s32 height = bottom_left->y - top_left->y;                                   \
+                                                                               \
+  compute_edge_delta(left, bottom_left, top_left, height);                     \
+  compute_edge_delta(right, bottom_right, top_right, height);                  \
+  set_interpolant_bases(bottom_left);                                          \
+  render_spans_up(height)                                                      \
+
+void render_spans_up_a(psx_gpu_struct *psx_gpu, _span_struct *span,
+ vertex_struct *bottom_left, vertex_struct *bottom_right, vertex_struct *top,
+ u32 flags)
+{
+  render_spans_up_flat(bottom_left, bottom_right, top, top);
+}
+
+void render_spans_up_b(psx_gpu_struct *psx_gpu, _span_struct *span,
+ vertex_struct *bottom, vertex_struct *top_left, vertex_struct *top_right,
+ u32 flags)
+{
+  render_spans_up_flat(bottom, bottom, top_left, top_right);
+}
+
+#define render_spans_down_flat(top_left, top_right, bottom_left, bottom_right) \
+  s32 current_y = top_left->y;                                                 \
+  s32 height = bottom_left->y - top_left->y;                                   \
+                                                                               \
+  compute_edge_delta(left, top_left, bottom_left, height);                     \
+  compute_edge_delta(right, top_right, bottom_right, height);                  \
+  set_interpolant_bases(top_left);                                             \
+  render_spans_down(height)                                                    \
+
+void render_spans_down_a(psx_gpu_struct *psx_gpu, _span_struct *span,
+ vertex_struct *top_left, vertex_struct *top_right, vertex_struct *bottom,
+ u32 flags)
+{
+  render_spans_down_flat(top_left, top_right, bottom, bottom);
+}
+
+void render_spans_down_b(psx_gpu_struct *psx_gpu, _span_struct *span,
+ vertex_struct *top, vertex_struct *bottom_left, vertex_struct *bottom_right,
+ u32 flags)
+{
+  render_spans_down_flat(top, top, bottom_left, bottom_right);
+}
+
+void render_spans_up_down(psx_gpu_struct *psx_gpu, _span_struct *span,
+ vertex_struct *middle, vertex_struct *top, vertex_struct *bottom, u32 flags)
+{
+  s32 middle_y = middle->y;
+  s32 current_y = middle_y - 1;
+  s32 height_minor_a = middle->y - top->y;
+  s32 height_minor_b = bottom->y - middle->y;
+  s32 height_major = height_minor_a + height_minor_b;
+
+  u64 right_x_mid;
+
+  compute_edge_delta(left, middle, top, height_minor_a);
+  compute_edge_delta(right, bottom, top, height_major);
+  set_interpolant_bases(middle);
+
+  right_x_mid = span->right_x + (span->right_dx_dy * height_minor_b);
+  span->right_x = right_x_mid;
+
+  render_spans_up(height_minor_a);  
+
+  compute_edge_delta(left, middle, bottom, height_minor_b);
+  set_interpolant_bases(middle);
+
+  span->right_dx_dy *= -1;
+  span->right_x = right_x_mid;
+  current_y = middle_y;
+
+  render_spans_down(height_minor_b);
+}
+
+#define vertex_swap(_a, _b)                                                    \
+{                                                                              \
+  vertex_struct *temp_vertex = _a;                                             \
+  _a = _b;                                                                     \
+  _b = temp_vertex;                                                            \
+  triangle_winding ^= 1;                                                       \
+}                                                                              \
+
+
+#define triangle_y_direction_up   1
+#define triangle_y_direction_flat 2
+#define triangle_y_direction_down 0
+
+#define triangle_winding_positive 0
+#define triangle_winding_negative 1
+
+#define triangle_set_direction(direction_variable, value)                      \
+  u32 direction_variable = (u32)(value) >> 31;                                 \
+  if(value == 0)                                                               \
+    direction_variable = 2                                                     \
+
+#define triangle_case(direction_a, direction_b, direction_c, winding)          \
+  case (triangle_y_direction_##direction_a |                                   \
+   (triangle_y_direction_##direction_b << 2) |                                 \
+   (triangle_y_direction_##direction_c << 4) |                                 \
+   (triangle_winding_##winding << 6))                                          \
+
+
+void render_triangle(psx_gpu_struct *psx_gpu, vertex_struct *vertexes,
+ u32 flags)
+{
+  s32 triangle_area;
+  u32 triangle_winding = 0;
+  _span_struct span;
+
+  vertex_struct *a = &(vertexes[0]);
+  vertex_struct *b = &(vertexes[1]);
+  vertex_struct *c = &(vertexes[2]);
+
+  triangle_area = triangle_signed_area_x2(a->x, a->y, b->x, b->y, c->x, c->y);
+
+  triangles++;
+
+  if(triangle_area == 0)
+    return;
+
+  if(b->y < a->y)
+    vertex_swap(a, b);
+
+  if(c->y < b->y)
+  {
+    vertex_swap(b, c);
+
+    if(b->y < a->y)
+      vertex_swap(a, b);
+  }
+
+  if((c->y - a->y) >= 512)
+    return;
+
+  if(triangle_area < 0)
+  {
+    triangle_area = -triangle_area;
+    triangle_winding ^= 1;
+    vertex_swap(a, c);
+  }
+
+  if(b->x < a->x)
+    vertex_swap(a, b);
+
+  if(c->x < b->x) 
+  {
+    vertex_swap(b, c);
+
+    if(b->x < a->x)
+      vertex_swap(a, b);
+  }
+
+  if((c->x - a->x) >= 1024)
+    return;
+
+  s32 y_delta_a = b->y - a->y;
+  s32 y_delta_b = c->y - b->y;
+  s32 y_delta_c = c->y - a->y;
+
+  triangle_set_direction(y_direction_a, y_delta_a);
+  triangle_set_direction(y_direction_b, y_delta_b);
+  triangle_set_direction(y_direction_c, y_delta_c);
+
+  compute_all_gradient_areas();
+  span.triangle_area = triangle_area;
+  span.triangle_winding = triangle_winding;
+
+  switch(y_direction_a | (y_direction_b << 2) | (y_direction_c << 4) |
+   (triangle_winding << 6))
+  {
+    triangle_case(up, up, up, negative):
+    triangle_case(up, up, flat, negative):
+    triangle_case(up, up, down, negative):
+      render_spans_up_right(psx_gpu, &span, a, b, c, flags);
+      break;
+
+    triangle_case(flat, up, up, negative):
+    triangle_case(flat, up, flat, negative):
+    triangle_case(flat, up, down, negative):
+      render_spans_up_a(psx_gpu, &span, a, b, c, flags);
+      break;
+
+    triangle_case(down, up, up, negative):
+      render_spans_up_down(psx_gpu, &span, a, c, b, flags);
+      break;
+
+    triangle_case(down, up, flat, negative):
+      render_spans_down_a(psx_gpu, &span, a, c, b, flags);
+      break;
+
+    triangle_case(down, up, down, negative):
+      render_spans_down_right(psx_gpu, &span, a, c, b, flags);
+      break;
+
+    triangle_case(down, flat, up, negative):
+    triangle_case(down, flat, flat, negative):
+    triangle_case(down, flat, down, negative):
+      render_spans_down_b(psx_gpu, &span, a, b, c, flags);
+      break;
+
+    triangle_case(down, down, up, negative):
+    triangle_case(down, down, flat, negative):
+    triangle_case(down, down, down, negative):
+      render_spans_down_left(psx_gpu, &span, a, b, c, flags);
+      break;
+
+    triangle_case(up, up, up, positive):
+    triangle_case(up, up, flat, positive):
+    triangle_case(up, up, down, positive):
+      render_spans_up_left(psx_gpu, &span, a, b, c, flags);
+      break;
+
+    triangle_case(up, flat, up, positive):
+    triangle_case(up, flat, flat, positive):
+    triangle_case(up, flat, down, positive):
+      render_spans_up_b(psx_gpu, &span, a, b, c, flags);
+      break;
+
+    triangle_case(up, down, up, positive):
+      render_spans_up_right(psx_gpu, &span, a, c, b, flags);
+      break;
+
+    triangle_case(up, down, flat, positive):
+      render_spans_up_a(psx_gpu, &span, a, c, b, flags);
+      break;
+
+    triangle_case(up, down, down, positive):
+      render_spans_up_down(psx_gpu, &span, a, b, c, flags);
+      break;
+
+    triangle_case(flat, down, up, positive):
+    triangle_case(flat, down, flat, positive):
+    triangle_case(flat, down, down, positive):
+      render_spans_down_a(psx_gpu, &span, a, b, c, flags);
+      break;
+
+    triangle_case(down, down, up, positive):
+    triangle_case(down, down, flat, positive):
+    triangle_case(down, down, down, positive):
+      render_spans_down_right(psx_gpu, &span, a, b, c, flags);
+      break;
+  }
+  
+}
+
+
+void render_sprite(psx_gpu_struct *psx_gpu, s32 x, s32 y, u32 u, u32 v,
+ s32 width, s32 height, u32 flags)
+{
+  // TODO: Flip/mirror
+  s32 current_x, current_y;
+  u32 current_u, current_v;
+  u32 primitive_color = psx_gpu->primitive_color;
+  u32 sprite_r, sprite_g, sprite_b;
+  s32 color_r = 0;
+  s32 color_g = 0; 
+  s32 color_b = 0;
+  u32 texel = 0;
+
+  sprite_r = primitive_color & 0xFF;
+  sprite_g = (primitive_color >> 8) & 0xFF;
+  sprite_b = (primitive_color >> 16) & 0xFF;
+
+  static u32 sprites = 0;
+
+  sprites++;
+
+  for(current_y = y, current_v = v; 
+   current_y < y + height; current_y++, current_v++)
+  {
+    for(current_x = x, current_u = u;
+     current_x < x + width; current_x++, current_u++)
+    {
+      if((current_x >= psx_gpu->viewport_start_x) &&
+       (current_y >= psx_gpu->viewport_start_y) &&
+       (current_x <= psx_gpu->viewport_end_x) &&
+       (current_y <= psx_gpu->viewport_end_y))
+      { 
+        if(psx_gpu->mask_evaluate &&
+         (psx_gpu->vram[(y * 1024) + current_x] & 0x8000))
+        {
+          continue;
+        }
+
+        if(flags & RENDER_FLAGS_TEXTURE_MAP)
+        {
+          texel = fetch_texel(psx_gpu, current_u, current_v);
+          if(texel == 0)
+            continue;
+
+          color_r = texel & 0x1F;
+          color_g = (texel >> 5) & 0x1F;
+          color_b = (texel >> 10) & 0x1F;
+
+          if((flags & RENDER_FLAGS_MODULATE_TEXELS) == 0)
+          {
+            color_r *= sprite_r;
+            color_g *= sprite_g;
+            color_b *= sprite_b;
+
+            color_r >>= 7;
+            color_g >>= 7;
+            color_b >>= 7;
+          }
+        }
+        else
+        {
+          color_r = sprite_r >> 3;
+          color_g = sprite_g >> 3;
+          color_b = sprite_b >> 3;
+        }
+
+        draw_pixel(psx_gpu, color_r, color_g, color_b, texel, current_x,
+         current_y, flags);
+      }
+    }
+  }
+}
+
+
+#define draw_pixel_line(_x, _y)                                                \
+  if((_x >= psx_gpu->viewport_start_x) && (_y >= psx_gpu->viewport_start_y) && \
+   (_x <= psx_gpu->viewport_end_x) && (_y <= psx_gpu->viewport_end_y))         \
+  {                                                                            \
+    if(flags & RENDER_FLAGS_SHADE)                                             \
+    {                                                                          \
+      color_r = fixed_to_int(current_r);                                       \
+      color_g = fixed_to_int(current_g);                                       \
+      color_b = fixed_to_int(current_b);                                       \
+                                                                               \
+      current_r += gradient_r;                                                 \
+      current_g += gradient_g;                                                 \
+      current_b += gradient_b;                                                 \
+    }                                                                          \
+    else                                                                       \
+    {                                                                          \
+      color_r = primitive_color & 0xFF;                                        \
+      color_g = (primitive_color >> 8) & 0xFF;                                 \
+      color_b = (primitive_color >> 16) & 0xFF;                                \
+    }                                                                          \
+                                                                               \
+    if(psx_gpu->dither_mode)                                                   \
+    {                                                                          \
+      s32 dither_offset = dither_table[_y % 4][_x % 4];                        \
+                                                                               \
+      color_r += dither_offset;                                                \
+      color_g += dither_offset;                                                \
+      color_b += dither_offset;                                                \
+                                                                               \
+      if(color_r < 0)                                                          \
+        color_r = 0;                                                           \
+                                                                               \
+      if(color_g < 0)                                                          \
+        color_g = 0;                                                           \
+                                                                               \
+      if(color_b < 0)                                                          \
+        color_b = 0;                                                           \
+    }                                                                          \
+    color_r >>= 3;                                                             \
+    color_g >>= 3;                                                             \
+    color_b >>= 3;                                                             \
+                                                                               \
+    span_pixels++;                                                             \
+                                                                               \
+    draw_pixel(psx_gpu, color_r, color_g, color_b, 0, _x, _y, flags);          \
+  }                                                                            \
+
+#define update_increment(value)                                                \
+  value++                                                                      \
+
+#define update_decrement(value)                                                \
+  value--                                                                      \
+
+#define compare_increment(a, b)                                                \
+  (a <= b)                                                                     \
+
+#define compare_decrement(a, b)                                                \
+  (a >= b)                                                                     \
+
+#define set_line_gradients(minor)                                              \
+{                                                                              \
+  s32 gradient_divisor = delta_##minor;                                        \
+  gradient_r = int_to_fixed(vertex_b->r - vertex_a->r) / gradient_divisor;     \
+  gradient_g = int_to_fixed(vertex_b->g - vertex_a->g) / gradient_divisor;     \
+  gradient_b = int_to_fixed(vertex_b->b - vertex_a->b) / gradient_divisor;     \
+  current_r = fixed_center(vertex_a->r);                                       \
+  current_g = fixed_center(vertex_a->g);                                       \
+  current_b = fixed_center(vertex_a->b);                                       \
+}
+
+#define draw_line_span_horizontal(direction)                                   \
+do                                                                             \
+{                                                                              \
+  error_step = delta_y * 2;                                                    \
+  error_wrap = delta_x * 2;                                                    \
+  error = delta_x;                                                             \
+                                                                               \
+  current_y = y_a;                                                             \
+  set_line_gradients(x);                                                       \
+                                                                               \
+  for(current_x = x_a; current_x <= x_b; current_x++)                          \
+  {                                                                            \
+    draw_pixel_line(current_x, current_y);                                     \
+    error += error_step;                                                       \
+                                                                               \
+    if(error >= error_wrap)                                                    \
+    {                                                                          \
+      update_##direction(current_y);                                           \
+      error -= error_wrap;                                                     \
+    }                                                                          \
+  }                                                                            \
+} while(0)                                                                     \
+
+#define draw_line_span_vertical(direction)                                     \
+do                                                                             \
+{                                                                              \
+  error_step = delta_x * 2;                                                    \
+  error_wrap = delta_y * 2;                                                    \
+  error = delta_y;                                                             \
+                                                                               \
+  current_x = x_a;                                                             \
+  set_line_gradients(y);                                                       \
+                                                                               \
+  for(current_y = y_a; compare_##direction(current_y, y_b);                    \
+   update_##direction(current_y))                                              \
+  {                                                                            \
+    draw_pixel_line(current_x, current_y);                                     \
+    error += error_step;                                                       \
+                                                                               \
+    if(error > error_wrap)                                                     \
+    {                                                                          \
+      current_x++;                                                             \
+      error -= error_wrap;                                                     \
+    }                                                                          \
+  }                                                                            \
+} while(0)                                                                     \
+                                                                               
+void render_line(psx_gpu_struct *psx_gpu, vertex_struct *vertexes, u32 flags)
+{
+  u32 primitive_color = psx_gpu->primitive_color;
+  s32 color_r, color_g, color_b;
+
+  fixed_type gradient_r = 0;
+  fixed_type gradient_g = 0;
+  fixed_type gradient_b = 0;
+  fixed_type current_r = 0;
+  fixed_type current_g = 0;
+  fixed_type current_b = 0;
+
+  s32 y_a, y_b;
+  s32 x_a, x_b;
+
+  s32 delta_x, delta_y;
+  u32 triangle_winding = 0;
+
+  s32 current_x;
+  s32 current_y;
+
+  u32 error_step;
+  u32 error;
+  u32 error_wrap;
+
+  vertex_struct *vertex_a = &(vertexes[0]);
+  vertex_struct *vertex_b = &(vertexes[1]);
+
+  if(vertex_a->x >= vertex_b->x)
+  {
+    vertex_swap(vertex_a, vertex_b);
+  }
+
+  x_a = vertex_a->x;
+  x_b = vertex_b->x;
+
+  y_a = vertex_a->y;
+  y_b = vertex_b->y;
+
+  delta_x = x_b - x_a;
+  delta_y = y_b - y_a;
+
+  if(delta_x >= 1024)
+    return;
+
+  flags &= ~RENDER_FLAGS_TEXTURE_MAP;
+
+  if(delta_y < 0)
+  {
+    delta_y *= -1;
+
+    if(delta_y >= 512)
+      return;
+
+    if(delta_x > delta_y)
+      draw_line_span_horizontal(decrement);
+    else
+      draw_line_span_vertical(decrement);
+  }
+  else
+  {
+    if(delta_y >= 512)
+      return;
+
+    if(delta_x > delta_y)
+      draw_line_span_horizontal(increment);
+    else
+      draw_line_span_vertical(increment);
+  }
+}
+
+
+void render_block_fill(psx_gpu_struct *psx_gpu, u32 color, u32 x, u32 y,
+ u32 width, u32 height)
+{
+  u32 r = color & 0xFF;
+  u32 g = (color >> 8) & 0xFF;
+  u32 b = (color >> 16) & 0xFF;
+  u32 color_16bpp = (r >> 3) | ((g >> 3) << 5) | ((b >> 3) << 10);
+
+  u16 *vram_ptr = psx_gpu->vram + x + (y * 1024);
+  u32 draw_x, draw_y;
+
+  for(draw_y = 0; draw_y < height; draw_y++)
+  {
+    for(draw_x = 0; draw_x < width; draw_x++)
+    {
+      vram_ptr[draw_x] = color_16bpp;
+    }
+
+    vram_ptr += 1024;
+  }
+}
+
+void render_block_copy(psx_gpu_struct *psx_gpu, u16 *source, u32 x, u32 y,
+ u32 width, u32 height, u32 pitch)
+{
+  u16 *vram_ptr = psx_gpu->vram + x + (y * 1024);
+  u32 draw_x, draw_y;
+
+  for(draw_y = 0; draw_y < height; draw_y++)
+  {
+    for(draw_x = 0; draw_x < width; draw_x++)
+    {
+      vram_ptr[draw_x] = source[draw_x];
+    }
+
+    source += pitch;
+    vram_ptr += 1024;
+  }
+}
+
+void render_block_move(psx_gpu_struct *psx_gpu, u32 source_x, u32 source_y,
+ u32 dest_x, u32 dest_y, u32 width, u32 height)
+{
+  render_block_copy(psx_gpu, psx_gpu->vram + source_x + (source_y * 1024),
+   dest_x, dest_y, width, height, 1024);
+}
+
+void initialize_psx_gpu(psx_gpu_struct *psx_gpu)
+{
+  psx_gpu->pixel_count_mode = 0;
+  psx_gpu->pixel_compare_mode = 0;
+
+  psx_gpu->vram_pixel_counts_a = malloc(sizeof(u8) * 1024 * 512);
+  psx_gpu->vram_pixel_counts_b = malloc(sizeof(u8) * 1024 * 512);
+  memset(psx_gpu->vram_pixel_counts_a, 0, sizeof(u8) * 1024 * 512);
+  memset(psx_gpu->vram_pixel_counts_b, 0, sizeof(u8) * 1024 * 512);
+  psx_gpu->compare_vram = malloc(sizeof(u16) * 1024 * 512);
+}
diff --git a/plugins/gpu_neon/psx_gpu/vector_ops.h b/plugins/gpu_neon/psx_gpu/vector_ops.h
new file mode 100644
index 0000000..c11955d
--- /dev/null
+++ b/plugins/gpu_neon/psx_gpu/vector_ops.h
@@ -0,0 +1,602 @@
+/*
+ * Copyright (C) 2011 Gilead Kutnick "Exophase" <exophase@gmail.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; either version 2 of
+ * the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ */
+
+#ifndef VECTOR_OPS
+#define VECTOR_OPS
+
+#define build_vector_type_pair(sign, size, count, count_x2)                    \
+typedef struct                                                                 \
+{                                                                              \
+  sign##size e[count];                                                         \
+} vec_##count##x##size##sign;                                                  \
+                                                                               \
+typedef struct                                                                 \
+{                                                                              \
+  union                                                                        \
+  {                                                                            \
+    sign##size e[count_x2];                                                    \
+    struct                                                                     \
+    {                                                                          \
+      vec_##count##x##size##sign low;                                          \
+      vec_##count##x##size##sign high;                                         \
+    };                                                                         \
+  };                                                                           \
+} vec_##count_x2##x##size##sign                                                \
+
+#define build_vector_types(sign)                                               \
+  build_vector_type_pair(sign, 8, 8, 16);                                      \
+  build_vector_type_pair(sign, 16, 4, 8);                                      \
+  build_vector_type_pair(sign, 32, 2, 4);                                      \
+  build_vector_type_pair(sign, 64, 1, 2)                                       \
+
+build_vector_types(u);
+build_vector_types(s);
+
+
+#define foreach_element(iterations, operation)                                 \
+{                                                                              \
+  u32 _i;                                                                      \
+  for(_i = 0; _i < iterations; _i++)                                           \
+  {                                                                            \
+    operation;                                                                 \
+  }                                                                            \
+}                                                                              \
+
+#define load_64b(dest, source)                                                 \
+ *((u64 *)(dest).e) = *((u64 *)(source))                                       \
+
+#define load_128b(dest, source)                                                \
+ *((u64 *)(dest).e) = *((u64 *)(source));                                      \
+ *((u64 *)(dest).e + 1) = *(((u64 *)(source)) + 1)                             \
+
+#define load_8x16b(dest, source)                                               \
+  foreach_element(8, (dest).e[_i] = ((u16 *)(source))[_i])                     \
+
+#define store_64b(source, dest)                                                \
+ *((u64 *)(dest)) = *((u64 *)(source).e)                                       \
+
+#define store_128b(source, dest)                                               \
+ *((u64 *)(dest)) = *((u64 *)(source).e);                                      \
+ *(((u64 *)(dest)) + 1) = *((u64 *)(source).e + 1)                             \
+
+#define store_8x16b(source, dest)                                              \
+  foreach_element(8, ((u16 *)dest)[_i] = (source).e[_i])                       \
+
+
+#define split_8x16b(dest, source)                                              \
+  foreach_element(8,                                                           \
+  {                                                                            \
+    (dest).e[_i * 2] = (source).e[_i];                                         \
+    (dest).e[(_i * 2) + 1] = (source).e[_i] >> 8;                              \
+  })                                                                           \
+
+#define merge_16x8b(dest, source)                                              \
+  foreach_element(8,                                                           \
+    (dest).e[_i] = (source).e[_i * 2] | ((source).e[(_i * 2) + 1] << 8))       \
+
+#define vector_cast(vec_to, source)                                            \
+  (*((volatile vec_to *)(&(source))))                                          \
+
+#define vector_cast_high(vec_to, source)                                       \
+  (*((volatile vec_to *)((u8 *)source.e + (sizeof(source.e) / 2))))            \
+
+
+#define dup_8x8b(dest, value)                                                  \
+  foreach_element(8, (dest).e[_i] = value)                                     \
+
+#define dup_16x8b(dest, value)                                                 \
+  foreach_element(16, (dest).e[_i] = value)                                    \
+
+#define dup_4x16b(dest, value)                                                 \
+  foreach_element(4, (dest).e[_i] = value)                                     \
+
+#define dup_8x16b(dest, value)                                                 \
+  foreach_element(8, (dest).e[_i] = value)                                     \
+
+#define dup_2x32b(dest, value)                                                 \
+  foreach_element(2, (dest).e[_i] = value)                                     \
+
+#define dup_4x32b(dest, value)                                                 \
+  foreach_element(4, (dest).e[_i] = value)                                     \
+
+#define shr_narrow_8x16b(dest, source, shift)                                  \
+  foreach_element(8, (dest).e[_i] = (u16)(source).e[_i] >> (shift))            \
+
+#define shr_narrow_2x64b(dest, source, shift)                                  \
+  foreach_element(2, (dest).e[_i] = (source).e[_i] >> (shift))                 \
+
+#define shr_8x8b(dest, source, shift)                                          \
+  foreach_element(8, (dest).e[_i] = (u8)(source).e[_i] >> (shift))             \
+
+#define shl_8x8b(dest, source, shift)                                          \
+  foreach_element(8, (dest).e[_i] = (source).e[_i] << (shift))                 \
+
+#define shr_8x16b(dest, source, shift)                                         \
+  foreach_element(8, (dest).e[_i] = (u16)(source).e[_i] >> (shift))            \
+
+#define shr_2x32b(dest, source, shift)                                         \
+  foreach_element(2, (dest).e[_i] = (u32)(source).e[_i] >> (shift))            \
+
+#define shr_4x16b(dest, source, shift)                                         \
+  foreach_element(4, (dest).e[_i] = (source).e[_i] >> (shift))                 \
+
+#define shl_4x16b(dest, source, shift)                                         \
+  foreach_element(4, (dest).e[_i] = (u32)(source).e[_i] << (shift))            \
+
+#define shr_4x32b(dest, source, shift)                                         \
+  foreach_element(4, (dest).e[_i] = (u32)(source).e[_i] >> (shift))            \
+
+#define shr_narrow_4x32b(dest, source, shift)                                  \
+  foreach_element(4, (dest).e[_i] = (u32)(source).e[_i] >> (shift))            \
+
+#define shl_8x16b(dest, source, shift)                                         \
+  foreach_element(8, (dest).e[_i] = (source).e[_i] << (shift))                 \
+
+#define shl_4x32b(dest, source, shift)                                         \
+  foreach_element(4, (dest).e[_i] = (source).e[_i] << (shift))                 \
+
+#define shl_2x32b(dest, source, shift)                                         \
+  foreach_element(2, (dest).e[_i] = (source).e[_i] << (shift))                 \
+
+#define shl_1x64b(dest, source, shift)                                         \
+  ((dest).e[0] = (source).e[0] << (shift))                                     \
+
+#define shl_2x64b(dest, source, shift)                                         \
+  foreach_element(2, (dest).e[_i] = (source).e[_i] << (shift))                 \
+
+#define shl_variable_2x64b(dest, source_a, source_b)                           \
+  foreach_element(2,                                                           \
+   (dest).e[_i] = (source_a).e[_i] << ((source_b).e[_i] & 0xFF))               \
+
+#define shl_variable_8x16b(dest, source_a, source_b)                           \
+  foreach_element(8,                                                           \
+   (dest).e[_i] = (source_a).e[_i] << ((source_b).e[_i] & 0xFF))               \
+
+#define shl_variable_4x16b(dest, source_a, source_b)                           \
+  foreach_element(4,                                                           \
+   (dest).e[_i] = (source_a).e[_i] << ((source_b).e[_i] & 0xFF))               \
+
+#define shr_1x64b(dest, source, shift)                                         \
+  ((dest).e[0] = (source).e[0] >> (shift))                                     \
+
+#define shl_long_8x8b(dest, source, shift)                                     \
+  foreach_element(8, (dest).e[_i] = (source).e[_i] << (shift))                 \
+
+#define shl_long_4x16b(dest, source, shift)                                    \
+  foreach_element(4, (dest).e[_i] = (source).e[_i] << (shift))                 \
+
+#define shrq_narrow_signed_8x16b(dest, source, shift)                          \
+  foreach_element(8,                                                           \
+  {                                                                            \
+    s32 result = ((s16)(source).e[_i]) >> shift;                               \
+    if(result < 0)                                                             \
+      result = 0;                                                              \
+    if(result > 0xFF)                                                          \
+      result = 0xFF;                                                           \
+    (dest).e[_i] = result;                                                     \
+  })                                                                           \
+
+#define shl_reg_4x32b(dest, source_a, source_b)                                \
+  foreach_element(4,                                                           \
+  {                                                                            \
+    s8 shift  = (source_b).e[_i];                                              \
+    if(shift < 0)                                                              \
+      dest.e[_i] = (source_a).e[_i] >> (-shift);                               \
+    else                                                                       \
+      dest.e[_i] = (source_a).e[_i] << shift;                                  \
+  })                                                                           \
+
+#define shl_reg_2x32b(dest, source_a, source_b)                                \
+  foreach_element(2,                                                           \
+  {                                                                            \
+    s8 shift  = (source_b).e[_i];                                              \
+    if(shift < 0)                                                              \
+      dest.e[_i] = (source_a).e[_i] >> (-shift);                               \
+    else                                                                       \
+      dest.e[_i] = (source_a).e[_i] << shift;                                  \
+  })                                                                           \
+
+#define shl_reg_2x64b(dest, source_a, source_b)                                \
+  foreach_element(2,                                                           \
+  {                                                                            \
+    s8 shift  = (source_b).e[_i];                                              \
+    if(shift < 0)                                                              \
+      dest.e[_i] = (source_a).e[_i] >> (-shift);                               \
+    else                                                                       \
+      dest.e[_i] = (source_a).e[_i] << shift;                                  \
+  })                                                                           \
+
+
+#define sri_8x8b(dest, source, shift)                                          \
+  foreach_element(8, (dest).e[_i] = ((dest).e[_i] & ~(0xFF >> (shift))) |      \
+   ((u8)(source).e[_i] >> (shift)))                                            \
+
+#define sli_8x8b(dest, source, shift)                                          \
+  foreach_element(8, (dest).e[_i] = ((dest).e[_i] & ~(0xFF << (shift))) |      \
+   ((source).e[_i] << (shift)))                                                \
+
+
+
+#define mov_narrow_8x16b(dest, source)                                         \
+  foreach_element(8, (dest).e[_i] = (source).e[_i])                            \
+
+#define mov_narrow_4x32b(dest, source)                                         \
+  foreach_element(4, (dest).e[_i] = (source).e[_i])                            \
+
+#define mov_narrow_2x64b(dest, source)                                         \
+  foreach_element(2, (dest).e[_i] = (source).e[_i])                            \
+
+#define mov_wide_8x8b(dest, source)                                            \
+  foreach_element(8, (dest).e[_i] = (source).e[_i])                            \
+
+#define mov_wide_2x32b(dest, source)                                           \
+  foreach_element(2, (dest).e[_i] = (source).e[_i])                            \
+
+#define mvn_4x16b(dest, source)                                                \
+  foreach_element(4, (dest).e[_i] = ~((source).e[_i]))                         \
+
+#define add_4x16b(dest, source_a, source_b)                                    \
+  foreach_element(4, (dest).e[_i] = (source_a).e[_i] + (source_b).e[_i])       \
+
+#define add_4x32b(dest, source_a, source_b)                                    \
+  foreach_element(4, (dest).e[_i] = (source_a).e[_i] + (source_b).e[_i])       \
+
+#define add_2x32b(dest, source_a, source_b)                                    \
+  foreach_element(2, (dest).e[_i] = (source_a).e[_i] + (source_b).e[_i])       \
+
+#define add_8x16b(dest, source_a, source_b)                                    \
+  foreach_element(8, (dest).e[_i] = (source_a).e[_i] + (source_b).e[_i])       \
+
+#define add_16x8b(dest, source_a, source_b)                                    \
+  foreach_element(16, (dest).e[_i] = (source_a).e[_i] + (source_b).e[_i])      \
+
+#define add_8x8b(dest, source_a, source_b)                                     \
+  foreach_element(8, (dest).e[_i] = (source_a).e[_i] + (source_b).e[_i])       \
+
+#define add_1x64b(dest, source_a, source_b)                                    \
+  (dest).e[0] = (source_a).e[0] + (source_b).e[0]                              \
+
+#define add_2x64b(dest, source_a, source_b)                                    \
+  foreach_element(2, (dest).e[_i] = (source_a).e[_i] + (source_b).e[_i])       \
+
+#define add_high_narrow_2x64b(dest, source_a, source_b)                        \
+  foreach_element(2,                                                           \
+   ((dest).e[_i] = (source_a).e[_i] + (source_b).e[_i]) >> 32)                 \
+
+#define add_high_narrow_4x32b(dest, source_a, source_b)                        \
+  foreach_element(4,                                                           \
+   ((dest).e[_i] = ((source_a).e[_i] + (source_b).e[_i]) >> 16))               \
+
+#define sub_4x16b(dest, source_a, source_b)                                    \
+  foreach_element(4, (dest).e[_i] = (source_a).e[_i] - (source_b).e[_i])       \
+
+#define sub_4x32b(dest, source_a, source_b)                                    \
+  foreach_element(4, (dest).e[_i] = (source_a).e[_i] - (source_b).e[_i])       \
+
+#define sub_2x32b(dest, source_a, source_b)                                    \
+  foreach_element(2, (dest).e[_i] = (source_a).e[_i] - (source_b).e[_i])       \
+
+#define sub_wide_8x8b(dest, source_a, source_b)                                \
+  foreach_element(8, (dest).e[_i] = (source_a).e[_i] - (source_b).e[_i])       \
+
+#define add_wide_8x8b(dest, source_a, source_b)                                \
+  foreach_element(8, (dest).e[_i] = (source_a).e[_i] + (source_b).e[_i])       \
+
+#define add_wide_2x32b(dest, source_a, source_b)                               \
+  foreach_element(2, (dest).e[_i] = (source_a).e[_i] + (source_b).e[_i])       \
+
+#define addq_8x8b(dest, source_a, source_b)                                    \
+  foreach_element(8,                                                           \
+  {                                                                            \
+    u32 result = (source_a).e[_i] + (source_b).e[_i];                          \
+    if(result > 0xFF)                                                          \
+      result = 0xFF;                                                           \
+    (dest).e[_i] = result;                                                     \
+  })                                                                           \
+
+#define subq_8x8b(dest, source_a, source_b)                                    \
+  foreach_element(8,                                                           \
+  {                                                                            \
+    u32 result = (source_a).e[_i] - (source_b).e[_i];                          \
+    if(result > 0xFF)                                                          \
+      result = 0;                                                              \
+    (dest).e[_i] = result;                                                     \
+  })                                                                           \
+
+#define subs_long_8x8b(dest, source_a, source_b)                               \
+  subs_8x8b(dest, source_a, source_b)                                          \
+
+#define subs_16x8b(dest, source_a, source_b)                                   \
+  foreach_element(16,                                                          \
+  {                                                                            \
+    u32 result = (source_a).e[_i] - (source_b).e[_i];                          \
+    if(result > 0xFF)                                                          \
+      result = 0;                                                              \
+    (dest).e[_i] = result;                                                     \
+  })                                                                           \
+
+#define subs_8x16b(dest, source_a, source_b)                                   \
+  foreach_element(8,                                                           \
+  {                                                                            \
+    s32 result = (source_a).e[_i] - (source_b).e[_i];                          \
+    if(result < 0)                                                             \
+      result = 0;                                                              \
+                                                                               \
+    (dest).e[_i] = result;                                                     \
+  })                                                                           \
+
+#define sub_8x16b(dest, source_a, source_b)                                    \
+  foreach_element(8, (dest).e[_i] = (source_a).e[_i] - (source_b).e[_i])       \
+
+#define sub_16x8b(dest, source_a, source_b)                                    \
+  foreach_element(16, (dest).e[_i] = (source_a).e[_i] - (source_b).e[_i])      \
+
+#define orn_8x16b(dest, source_a, source_b)                                    \
+  foreach_element(8, (dest).e[_i] = (source_a).e[_i] | ~((source_b).e[_i]))    \
+
+#define and_4x16b(dest, source_a, source_b)                                    \
+  foreach_element(4, (dest).e[_i] = (source_a).e[_i] & (source_b).e[_i])       \
+
+#define and_8x16b(dest, source_a, source_b)                                    \
+  foreach_element(8, (dest).e[_i] = (source_a).e[_i] & (source_b).e[_i])       \
+
+#define and_4x32b(dest, source_a, source_b)                                    \
+  foreach_element(4, (dest).e[_i] = (source_a).e[_i] & (source_b).e[_i])       \
+
+#define and_16x8b(dest, source_a, source_b)                                    \
+  foreach_element(16, (dest).e[_i] = (source_a).e[_i] & (source_b).e[_i])      \
+
+#define and_8x8b(dest, source_a, source_b)                                     \
+  foreach_element(8, (dest).e[_i] = (source_a).e[_i] & (source_b).e[_i])       \
+
+#define and_2x32b(dest, source_a, source_b)                                    \
+  foreach_element(2, (dest).e[_i] = (source_a).e[_i] & (source_b).e[_i])       \
+
+#define bic_8x8b(dest, source_a, source_b)                                     \
+  foreach_element(8, (dest).e[_i] = (source_a).e[_i] & ~((source_b).e[_i]))    \
+
+#define bic_8x16b(dest, source_a, source_b)                                    \
+  foreach_element(8, (dest).e[_i] = (source_a).e[_i] & ~((source_b).e[_i]))    \
+
+#define bic_immediate_4x16b(dest, value)                                       \
+  foreach_element(4, (dest).e[_i] = (dest).e[_i] & ~(value))                   \
+
+#define bic_immediate_8x16b(dest, value)                                       \
+  foreach_element(8, (dest).e[_i] = (dest).e[_i] & ~(value))                   \
+
+#define or_8x16b(dest, source_a, source_b)                                     \
+  foreach_element(8, (dest).e[_i] = (source_a).e[_i] | (source_b).e[_i])       \
+
+#define or_immediate_8x16b(dest, source_a, value)                              \
+  foreach_element(8, (dest).e[_i] = (source_a).e[_i] | (value))                \
+
+#define eor_8x16b(dest, source_a, source_b)                                    \
+  foreach_element(8, (dest).e[_i] = (source_a).e[_i] ^ (source_b).e[_i])       \
+
+#define eor_4x32b(dest, source_a, source_b)                                    \
+  foreach_element(4, (dest).e[_i] = (source_a).e[_i] ^ (source_b).e[_i])       \
+
+#define eor_2x32b(dest, source_a, source_b)                                    \
+  foreach_element(2, (dest).e[_i] = (source_a).e[_i] ^ (source_b).e[_i])       \
+
+#define zip_8x16b(dest, source_a, source_b)                                    \
+  foreach_element(8, (dest).e[_i] =                                            \
+   (u8)(source_a).e[_i] | ((u8)(source_b).e[_i] << 8))                         \
+
+#define zip_2x64b(dest, source_a, source_b)                                    \
+  foreach_element(2, (dest).e[_i] =                                            \
+   (u64)(source_a).e[_i] | ((u64)(source_b).e[_i] << 32))                      \
+
+#define unzip_8x8b(dest_a, dest_b, source)                                     \
+  foreach_element(8,                                                           \
+  {                                                                            \
+    (dest_a).e[_i] = (source).e[_i];                                           \
+    (dest_b).e[_i] = ((source).e[_i]) >> 8;                                    \
+  })                                                                           \
+
+#define unzip_16x8b(dest_a, dest_b, source_a, source_b)                        \
+  foreach_element(8,                                                           \
+  {                                                                            \
+    (dest_a).e[_i] = (source_a).e[_i];                                         \
+    (dest_b).e[_i] = (source_a).e[_i] >> 8;                                    \
+  });                                                                          \
+  foreach_element(8,                                                           \
+  {                                                                            \
+    (dest_a).e[_i + 8] = (source_b).e[_i];                                     \
+    (dest_b).e[_i + 8] = (source_b).e[_i] >> 8;                                \
+  })                                                                           \
+
+#define tbl_16(dest, indexes, table)                                           \
+  foreach_element(8,                                                           \
+  {                                                                            \
+    u32 index = indexes.e[_i];                                                 \
+    if(index < 16)                                                             \
+      (dest).e[_i] = table.e[index];                                           \
+    else                                                                       \
+      (dest).e[_i] = 0;                                                        \
+  })                                                                           \
+
+#define cmpeqz_8x16b(dest, source)                                             \
+  foreach_element(8, (dest).e[_i] = ~(((source).e[_i] == 0) - 1))              \
+
+#define cmpltz_8x16b(dest, source)                                             \
+  foreach_element(8, (dest).e[_i] = ((s16)(source).e[_i] >> 15))               \
+
+#define cmpltz_4x32b(dest, source)                                             \
+  foreach_element(4, (dest).e[_i] = ((s32)(source).e[_i] >> 31))               \
+
+#define cmpltz_2x32b(dest, source)                                             \
+  foreach_element(2, (dest).e[_i] = ((s32)(source).e[_i] >> 31))               \
+
+#define cmplte_4x16b(dest, source_a, source_b)                                 \
+  foreach_element(4, (dest).e[_i] = ~((source_a.e[_i] <= source_b.e[_i]) - 1)) \
+
+#define cmplt_4x16b(dest, source_a, source_b)                                  \
+  foreach_element(4, (dest).e[_i] = ~((source_a.e[_i] < source_b.e[_i]) - 1))  \
+
+#define cmpgt_4x16b(dest, source_a, source_b)                                  \
+  foreach_element(4, (dest).e[_i] = ~((source_a.e[_i] > source_b.e[_i]) - 1))  \
+
+#define tst_8x16b(dest, source_a, source_b)                                    \
+  foreach_element(8,                                                           \
+   (dest).e[_i] = ~(((source_a.e[_i] & source_b.e[_i]) != 0) - 1))             \
+
+#define andi_8x8b(dest, source_a, value)                                       \
+  foreach_element(8, (dest).e[_i] = (source_a).e[_i] & value)                  \
+
+#define average_8x16b(dest, source_a, source_b)                                \
+  foreach_element(8,                                                           \
+   (dest).e[_i] = ((source_a).e[_i] + (source_b).e[_i]) >> 1)                  \
+
+
+#define mul_8x8b(dest, source_a, source_b)                                     \
+  foreach_element(8, (dest).e[_i] = (source_a).e[_i] * (source_b).e[_i])       \
+
+#define mul_8x16b(dest, source_a, source_b)                                    \
+  foreach_element(8, (dest).e[_i] = (source_a).e[_i] * (source_b).e[_i])       \
+
+#define mul_2x32b(dest, source_a, source_b)                                    \
+  foreach_element(2, (dest).e[_i] = (source_a).e[_i] * (source_b).e[_i])       \
+
+#define mul_4x32b(dest, source_a, source_b)                                    \
+  foreach_element(4, (dest).e[_i] = (source_a).e[_i] * (source_b).e[_i])       \
+
+#define mul_long_8x8b(dest, source_a, source_b)                                \
+  foreach_element(8, (dest).e[_i] = (source_a).e[_i] * (source_b).e[_i])       \
+
+#define mul_long_4x16b(dest, source_a, source_b)                               \
+  foreach_element(4, (dest).e[_i] = (source_a).e[_i] * (source_b).e[_i])       \
+
+#define mul_long_2x32b(dest, source_a, source_b)                               \
+  foreach_element(2,                                                           \
+   (dest).e[_i] = (source_a).e[_i] * (s64)((source_b).e[_i]))                  \
+
+#define mul_scalar_2x32b(dest, source, value)                                  \
+  foreach_element(2, (dest).e[_i] = (source).e[_i] * value)                    \
+
+#define mul_scalar_long_8x16b(dest, source, value)                             \
+  foreach_element(8, (dest).e[_i] = (source).e[_i] * value)                    \
+
+#define mul_scalar_long_2x32b(dest, source, value)                             \
+  foreach_element(2, (dest).e[_i] = (source).e[_i] * value)                    \
+
+#define mla_2x32b(dest, source_a, source_b)                                    \
+  foreach_element(2, (dest).e[_i] += (source_a).e[_i] * (source_b).e[_i])      \
+
+#define mla_4x32b(dest, source_a, source_b)                                    \
+  foreach_element(4, (dest).e[_i] += (source_a).e[_i] * (source_b).e[_i])      \
+
+#define mla_scalar_long_2x32b(dest, source, value)                             \
+  foreach_element(2, (dest).e[_i] += (source).e[_i] * value)                   \
+
+#define mla_long_8x8b(dest, source_a, source_b)                                \
+  foreach_element(8, (dest).e[_i] += (source_a).e[_i] * (source_b).e[_i])      \
+
+#define mla_long_2x32b(dest, source_a, source_b)                               \
+  foreach_element(2, (dest).e[_i] += (source_a).e[_i] * (s64)(source_b).e[_i]) \
+
+#define mla_scalar_4x32b(dest, source, value)                                  \
+  foreach_element(4, (dest).e[_i] += (source).e[_i] * value)                   \
+
+#define mla_scalar_2x32b(dest, source, value)                                  \
+  foreach_element(2, (dest).e[_i] += (source).e[_i] * value)                   \
+
+#define mls_scalar_4x32b(dest, source, value)                                  \
+  foreach_element(4, (dest).e[_i] -= (source).e[_i] * value)                   \
+
+#define mls_scalar_2x32b(dest, source, value)                                  \
+  foreach_element(2, (dest).e[_i] -= (source).e[_i] * value)                   \
+
+#define mls_scalar_long_2x32b(dest, source, value)                             \
+  foreach_element(2, (dest).e[_i] -= (source).e[_i] * value)                   \
+
+#define rev_2x32b(dest, source)                                                \
+{                                                                              \
+  u32 tmp = source.e[1];                                                       \
+  (dest).e[1] = source.e[0];                                                   \
+  (dest).e[0] = tmp;                                                           \
+}                                                                              \
+
+#define abs_4x32b(dest, source)                                                \
+  foreach_element(4, (dest).e[_i] = abs(source.e[_i]))                         \
+
+#define abs_2x32b(dest, source)                                                \
+  foreach_element(2, (dest).e[_i] = abs(source.e[_i]))                         \
+
+#define neg_2x32b(dest, source)                                                \
+  foreach_element(2, (dest).e[_i] = -((source).e[_i]))                         \
+
+
+#define shrq_narrow_8x16b(dest, source, shift)                                 \
+  foreach_element(8,                                                           \
+  {                                                                            \
+    u32 result = ((source).e[_i]) >> shift;                                    \
+    if(result > 0xFF)                                                          \
+      result = 0xFF;                                                           \
+    (dest).e[_i] = result;                                                     \
+  })                                                                           \
+
+#define min_8x16b(dest, source_a, source_b)                                    \
+  foreach_element(8,                                                           \
+  {                                                                            \
+    s32 result = (source_a).e[_i];                                             \
+    if((source_b).e[_i] < result)                                              \
+      result = (source_b).e[_i];                                               \
+    (dest).e[_i] = result;                                                     \
+  })                                                                           \
+
+#define min_8x8b(dest, source_a, source_b)                                     \
+  foreach_element(8,                                                           \
+  {                                                                            \
+    u32 result = (source_a).e[_i];                                             \
+    if((source_b).e[_i] < result)                                              \
+      result = (source_b).e[_i];                                               \
+    (dest).e[_i] = result;                                                     \
+  })                                                                           \
+
+#define min_16x8b(dest, source_a, source_b)                                    \
+  foreach_element(16,                                                          \
+  {                                                                            \
+    u32 result = (source_a).e[_i];                                             \
+    if((source_b).e[_i] < result)                                              \
+      result = (source_b).e[_i];                                               \
+    (dest).e[_i] = result;                                                     \
+  })                                                                           \
+
+#define max_8x16b(dest, source_a, source_b)                                    \
+  foreach_element(8,                                                           \
+  {                                                                            \
+    s32 result = (source_a).e[_i];                                             \
+    if((source_b).e[_i] > result)                                              \
+      result = (source_b).e[_i];                                               \
+    (dest).e[_i] = result;                                                     \
+  })                                                                           \
+
+#define bsl_8x16b(dest_mask, source_a, source_b)                               \
+  foreach_element(8, dest_mask.e[_i] = ((source_a).e[_i] & dest_mask.e[_i]) |  \
+   ((source_b).e[_i] & ~(dest_mask.e[_i])))                                    \
+
+#define bif_8x16b(dest, source, mask)                                          \
+  foreach_element(8, dest.e[_i] = ((source).e[_i] & ~(mask.e[_i])) |           \
+   ((dest).e[_i] & mask.e[_i]))                                                \
+
+#define bsl_4x32b(dest_mask, source_a, source_b)                               \
+  foreach_element(4, dest_mask.e[_i] = ((source_a).e[_i] & dest_mask.e[_i]) |  \
+   ((source_b).e[_i] & ~(dest_mask.e[_i])))                                    \
+
+#define bit_4x16b(dest, source, mask)                                          \
+  foreach_element(4, dest.e[_i] = ((source).e[_i] & mask.e[_i]) |              \
+   ((dest).e[_i] & ~(mask.e[_i])))                                             \
+
+#endif