From 62d7fa9555924ab8e152b546711d27add640b102 Mon Sep 17 00:00:00 2001 From: notaz Date: Thu, 19 Jan 2012 01:33:12 +0200 Subject: refactor gpu plugins and Makefiles name common gpu code gpulib, reduce amount of copy-paste in plugin Makefiles --- plugins/dfxvideo/draw_fb.c | 2 +- plugins/dfxvideo/gpulib_if.c | 398 ++++++++++++++++++++++++++ plugins/gpu_neon/Makefile | 59 +--- plugins/gpu_neon/Makefile.test | 46 --- plugins/gpu_neon/gpu.c | 636 ----------------------------------------- plugins/gpu_neon/gpu.h | 131 --------- plugins/gpu_neon/peops_if.c | 398 -------------------------- plugins/gpu_neon/psx_gpu_if.c | 2 +- plugins/gpu_neon/test.c | 109 ------- plugins/gpu_neon/unai_if.cpp | 291 ------------------- plugins/gpu_neon/vout_fb.c | 121 -------- plugins/gpu_neon/vout_sdl.c | 95 ------ plugins/gpu_unai/Makefile | 41 +-- plugins/gpu_unai/gpu.cpp | 2 +- plugins/gpu_unai/gpulib_if.cpp | 291 +++++++++++++++++++ plugins/gpulib/Makefile | 38 +++ plugins/gpulib/Makefile.test | 45 +++ plugins/gpulib/cspace.c | 21 ++ plugins/gpulib/cspace.h | 12 + plugins/gpulib/cspace_neon.s | 132 +++++++++ plugins/gpulib/gpu.c | 636 +++++++++++++++++++++++++++++++++++++++++ plugins/gpulib/gpu.h | 131 +++++++++ plugins/gpulib/gpulib.mak | 54 ++++ plugins/gpulib/test.c | 109 +++++++ plugins/gpulib/vout_fb.c | 121 ++++++++ plugins/gpulib/vout_sdl.c | 95 ++++++ 26 files changed, 2112 insertions(+), 1904 deletions(-) create mode 100644 plugins/dfxvideo/gpulib_if.c delete mode 100644 plugins/gpu_neon/Makefile.test delete mode 100644 plugins/gpu_neon/gpu.c delete mode 100644 plugins/gpu_neon/gpu.h delete mode 100644 plugins/gpu_neon/peops_if.c delete mode 100644 plugins/gpu_neon/test.c delete mode 100644 plugins/gpu_neon/unai_if.cpp delete mode 100644 plugins/gpu_neon/vout_fb.c delete mode 100644 plugins/gpu_neon/vout_sdl.c create mode 100644 plugins/gpu_unai/gpulib_if.cpp create mode 100644 plugins/gpulib/Makefile create mode 100644 plugins/gpulib/Makefile.test create mode 100644 plugins/gpulib/cspace.c create mode 100644 plugins/gpulib/cspace.h create mode 100644 plugins/gpulib/cspace_neon.s create mode 100644 plugins/gpulib/gpu.c create mode 100644 plugins/gpulib/gpu.h create mode 100644 plugins/gpulib/gpulib.mak create mode 100644 plugins/gpulib/test.c create mode 100644 plugins/gpulib/vout_fb.c create mode 100644 plugins/gpulib/vout_sdl.c (limited to 'plugins') diff --git a/plugins/dfxvideo/draw_fb.c b/plugins/dfxvideo/draw_fb.c index 8814da8..22351a4 100644 --- a/plugins/dfxvideo/draw_fb.c +++ b/plugins/dfxvideo/draw_fb.c @@ -9,8 +9,8 @@ #include "gpu.h" +#include "../gpulib/cspace.h" #include "../../frontend/plugin_lib.h" -#include "../../frontend/cspace.h" #include "../../frontend/pcnt.h" // misc globals diff --git a/plugins/dfxvideo/gpulib_if.c b/plugins/dfxvideo/gpulib_if.c new file mode 100644 index 0000000..82bc38d --- /dev/null +++ b/plugins/dfxvideo/gpulib_if.c @@ -0,0 +1,398 @@ +/*************************************************************************** + copyright : (C) 2001 by Pete Bernert, 2011 notaz + + ***************************************************************************/ +/*************************************************************************** + * * + * 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. See also the license.txt file for * + * additional informations. * + * * + ***************************************************************************/ + +#include +#include +#include +#include "../gpulib/gpu.h" + +#define u32 uint32_t + +#define INFO_TW 0 +#define INFO_DRAWSTART 1 +#define INFO_DRAWEND 2 +#define INFO_DRAWOFF 3 + +#define SHADETEXBIT(x) ((x>>24) & 0x1) +#define SEMITRANSBIT(x) ((x>>25) & 0x1) +#define PSXRGB(r,g,b) ((g<<10)|(b<<5)|r) + +#define DATAREGISTERMODES unsigned short + +#define DR_NORMAL 0 +#define DR_VRAMTRANSFER 1 + +#define GPUSTATUS_READYFORVRAM 0x08000000 + +// byteswappings + +#define SWAP16(x) ({ uint16_t y=(x); (((y)>>8 & 0xff) | ((y)<<8 & 0xff00)); }) +#define SWAP32(x) ({ uint32_t y=(x); (((y)>>24 & 0xfful) | ((y)>>8 & 0xff00ul) | ((y)<<8 & 0xff0000ul) | ((y)<<24 & 0xff000000ul)); }) + +#ifdef __BIG_ENDIAN__ + +// big endian config +#define HOST2LE32(x) SWAP32(x) +#define HOST2BE32(x) (x) +#define LE2HOST32(x) SWAP32(x) +#define BE2HOST32(x) (x) + +#define HOST2LE16(x) SWAP16(x) +#define HOST2BE16(x) (x) +#define LE2HOST16(x) SWAP16(x) +#define BE2HOST16(x) (x) + +#else + +// little endian config +#define HOST2LE32(x) (x) +#define HOST2BE32(x) SWAP32(x) +#define LE2HOST32(x) (x) +#define BE2HOST32(x) SWAP32(x) + +#define HOST2LE16(x) (x) +#define HOST2BE16(x) SWAP16(x) +#define LE2HOST16(x) (x) +#define BE2HOST16(x) SWAP16(x) + +#endif + +#define GETLEs16(X) ((int16_t)GETLE16((uint16_t *)X)) +#define GETLEs32(X) ((int16_t)GETLE32((uint16_t *)X)) + +#define GETLE16(X) LE2HOST16(*(uint16_t *)X) +#define GETLE32_(X) LE2HOST32(*(uint32_t *)X) +#define GETLE16D(X) ({uint32_t val = GETLE32(X); (val<<16 | val >> 16);}) +#define PUTLE16(X, Y) do{*((uint16_t *)X)=HOST2LE16((uint16_t)Y);}while(0) +#define PUTLE32_(X, Y) do{*((uint32_t *)X)=HOST2LE16((uint32_t)Y);}while(0) +#ifdef __arm__ +#define GETLE32(X) (*(uint16_t *)(X)|(((uint16_t *)(X))[1]<<16)) +#define PUTLE32(X, Y) do{uint16_t *p_=(uint16_t *)(X);uint32_t y_=Y;p_[0]=y_;p_[1]=y_>>16;}while(0) +#else +#define GETLE32 GETLE32_ +#define PUTLE32 PUTLE32_ +#endif + +///////////////////////////////////////////////////////////////////////////// + +typedef struct VRAMLOADTTAG +{ + short x; + short y; + short Width; + short Height; + short RowsRemaining; + short ColsRemaining; + unsigned short *ImagePtr; +} VRAMLoad_t; + +///////////////////////////////////////////////////////////////////////////// + +typedef struct PSXPOINTTAG +{ + int32_t x; + int32_t y; +} PSXPoint_t; + +typedef struct PSXSPOINTTAG +{ + short x; + short y; +} PSXSPoint_t; + +typedef struct PSXRECTTAG +{ + short x0; + short x1; + short y0; + short y1; +} PSXRect_t; + +// linux defines for some windows stuff + +#define FALSE 0 +#define TRUE 1 +#define BOOL unsigned short +#define LOWORD(l) ((unsigned short)(l)) +#define HIWORD(l) ((unsigned short)(((uint32_t)(l) >> 16) & 0xFFFF)) +#define max(a,b) (((a) > (b)) ? (a) : (b)) +#define min(a,b) (((a) < (b)) ? (a) : (b)) +#define DWORD uint32_t +#ifndef __int64 +#define __int64 long long int +#endif + +typedef struct RECTTAG +{ + int left; + int top; + int right; + int bottom; +}RECT; + +///////////////////////////////////////////////////////////////////////////// + +typedef struct TWINTAG +{ + PSXRect_t Position; + int xmask, ymask; +} TWin_t; + +///////////////////////////////////////////////////////////////////////////// + +typedef struct PSXDISPLAYTAG +{ + PSXPoint_t DisplayModeNew; + PSXPoint_t DisplayMode; + PSXPoint_t DisplayPosition; + PSXPoint_t DisplayEnd; + + int32_t Double; + int32_t Height; + int32_t PAL; + int32_t InterlacedNew; + int32_t Interlaced; + int32_t RGB24New; + int32_t RGB24; + PSXSPoint_t DrawOffset; + int32_t Disabled; + PSXRect_t Range; + +} PSXDisplay_t; + +///////////////////////////////////////////////////////////////////////////// + +// draw.c + +extern int32_t GlobalTextAddrX,GlobalTextAddrY,GlobalTextTP; +extern int32_t GlobalTextABR,GlobalTextPAGE; +extern short ly0,lx0,ly1,lx1,ly2,lx2,ly3,lx3; +extern long lLowerpart; +extern BOOL bCheckMask; +extern unsigned short sSetMask; +extern unsigned long lSetMask; +extern short g_m1; +extern short g_m2; +extern short g_m3; +extern short DrawSemiTrans; + +// prim.c + +extern BOOL bUsingTWin; +extern TWin_t TWin; +extern void (*primTableJ[256])(unsigned char *); +extern void (*primTableSkip[256])(unsigned char *); +extern unsigned short usMirror; +extern int iDither; +extern uint32_t dwCfgFixes; +extern uint32_t dwActFixes; +extern int iUseFixes; +extern int iUseDither; +extern BOOL bDoVSyncUpdate; +extern int32_t drawX; +extern int32_t drawY; +extern int32_t drawW; +extern int32_t drawH; + +// gpu.h + +#define OPAQUEON 10 +#define OPAQUEOFF 11 + +#define KEY_RESETTEXSTORE 1 +#define KEY_SHOWFPS 2 +#define KEY_RESETOPAQUE 4 +#define KEY_RESETDITHER 8 +#define KEY_RESETFILTER 16 +#define KEY_RESETADVBLEND 32 +#define KEY_BADTEXTURES 128 +#define KEY_CHECKTHISOUT 256 + +#if !defined(__BIG_ENDIAN__) || defined(__x86_64__) || defined(__i386__) +#ifndef __LITTLE_ENDIAN__ +#define __LITTLE_ENDIAN__ +#endif +#endif + +#ifdef __LITTLE_ENDIAN__ +#define RED(x) (x & 0xff) +#define BLUE(x) ((x>>16) & 0xff) +#define GREEN(x) ((x>>8) & 0xff) +#define COLOR(x) (x & 0xffffff) +#elif defined __BIG_ENDIAN__ +#define RED(x) ((x>>24) & 0xff) +#define BLUE(x) ((x>>8) & 0xff) +#define GREEN(x) ((x>>16) & 0xff) +#define COLOR(x) SWAP32(x & 0xffffff) +#endif + +PSXDisplay_t PSXDisplay; +unsigned char *psxVub; +signed char *psxVsb; +unsigned short *psxVuw; +unsigned short *psxVuw_eom; +signed short *psxVsw; +uint32_t *psxVul; +int32_t *psxVsl; + +long lGPUstatusRet; +uint32_t lGPUInfoVals[16]; + +VRAMLoad_t VRAMWrite; +VRAMLoad_t VRAMRead; + +DATAREGISTERMODES DataWriteMode; +DATAREGISTERMODES DataReadMode; + +BOOL bCheckMask = FALSE; +unsigned short sSetMask = 0; +unsigned long lSetMask = 0; +long lLowerpart; + +#include "soft.c" +#include "prim.c" + +///////////////////////////////////////////////////////////////////////////// + +int renderer_init(void) +{ + psxVub=(void *)gpu.vram; + + psxVsb=(signed char *)psxVub; // different ways of accessing PSX VRAM + psxVsw=(signed short *)psxVub; + psxVsl=(int32_t *)psxVub; + psxVuw=(unsigned short *)psxVub; + psxVul=(uint32_t *)psxVub; + + psxVuw_eom=psxVuw+1024*512; // pre-calc of end of vram + + PSXDisplay.RGB24 = FALSE; // init some stuff + PSXDisplay.Interlaced = FALSE; + PSXDisplay.DrawOffset.x = 0; + PSXDisplay.DrawOffset.y = 0; + PSXDisplay.DisplayMode.x= 320; + PSXDisplay.DisplayMode.y= 240; + PSXDisplay.Disabled = FALSE; + PSXDisplay.Range.x0=0; + PSXDisplay.Range.x1=0; + PSXDisplay.Double = 1; + + DataWriteMode = DR_NORMAL; + lGPUstatusRet = 0x14802000; + + return 0; +} + +extern const unsigned char cmd_lengths[256]; + +void do_cmd_list(unsigned int *list, int list_len) +{ + unsigned int cmd, len; + + unsigned int *list_end = list + list_len; + + for (; list < list_end; list += 1 + len) + { + cmd = *list >> 24; + len = cmd_lengths[cmd]; + + primTableJ[cmd]((void *)list); + + switch(cmd) + { + case 0x48 ... 0x4F: + { + u32 num_vertexes = 1; + u32 *list_position = &(list[2]); + + while(1) + { + if((*list_position & 0xf000f000) == 0x50005000 || list_position >= list_end) + break; + + list_position++; + num_vertexes++; + } + + if(num_vertexes > 2) + len += (num_vertexes - 2); + + break; + } + + case 0x58 ... 0x5F: + { + u32 num_vertexes = 1; + u32 *list_position = &(list[2]); + + while(1) + { + if((*list_position & 0xf000f000) == 0x50005000 || list_position >= list_end) + break; + + list_position += 2; + num_vertexes++; + } + + if(num_vertexes > 2) + len += (num_vertexes - 2) * 2; + + break; + } + +#ifdef TEST + case 0xA0: // sys -> vid + { + short *slist = (void *)list; + u32 load_width = slist[4]; + u32 load_height = slist[5]; + u32 load_size = load_width * load_height; + + len += load_size / 2; + break; + } +#endif + } + } +} + +void renderer_sync_ecmds(uint32_t *ecmds) +{ + cmdTexturePage((unsigned char *)&ecmds[1]); + cmdTextureWindow((unsigned char *)&ecmds[2]); + cmdDrawAreaStart((unsigned char *)&ecmds[3]); + cmdDrawAreaEnd((unsigned char *)&ecmds[4]); + cmdDrawOffset((unsigned char *)&ecmds[5]); + cmdSTP((unsigned char *)&ecmds[6]); +} + +void renderer_update_caches(int x, int y, int w, int h) +{ +} + +void renderer_flush_queues(void) +{ +} + +void renderer_set_interlace(int enable, int is_odd) +{ +} + +#include "../../frontend/plugin_lib.h" + +void renderer_set_config(const struct rearmed_cbs *cbs) +{ + iUseDither = cbs->gpu_peops.iUseDither; + dwActFixes = cbs->gpu_peops.dwActFixes; +} diff --git a/plugins/gpu_neon/Makefile b/plugins/gpu_neon/Makefile index 2d942f8..93839d2 100644 --- a/plugins/gpu_neon/Makefile +++ b/plugins/gpu_neon/Makefile @@ -1,59 +1,28 @@ CC = $(CROSS_COMPILE)gcc -CXX = $(CROSS_COMPILE)g++ -CC_ = $(CC) ARCH = $(shell $(CC) -v 2>&1 | grep -i 'target:' | awk '{print $$2}' | awk -F '-' '{print $$1}') -CFLAGS += -ggdb -Wall -fPIC -O2 -CFLAGS += -fno-strict-aliasing -ifdef MAEMO -CFLAGS += -DMAEMO -endif +HAVE_NEON = $(shell $(CC) -E -dD $(CFLAGS_) ../gpulib/gpu.h | grep -q '__ARM_NEON__ 1' && echo 1) -SRC += gpu.c +# FIXME: rework NEON detection to get rid of this ifeq "$(ARCH)" "arm" -ARM_CORTEXA8 ?= 1 -ifeq "$(ARM_CORTEXA8)" "1" -CFLAGS += -mcpu=cortex-a8 -mtune=cortex-a8 -mfpu=neon -mfloat-abi=softfp -SRC += ../../frontend/cspace_neon.s -else -CFLAGS += -mcpu=arm926ej-s -mtune=arm926ej-s -SRC += ../../frontend/cspace.c -endif -SRC += vout_fb.c -EXT = so -endif -ifeq "$(ARCH)" "x86_64" -CFLAGS += `sdl-config --cflags` -m32 -LDFLAGS += `sdl-config --libs` -SRC += vout_sdl.c -EXT = so.x86 + ARM_CORTEXA8 ?= 1 + ifeq "$(ARM_CORTEXA8)" "1" + CFLAGS_ += -mcpu=cortex-a8 -mtune=cortex-a8 -mfpu=neon -mfloat-abi=softfp + endif endif -HAVE_NEON = $(shell $(CC_) -E -dD $(CFLAGS) gpu.h | grep -q '__ARM_NEON__ 1' && echo 1) +CFLAGS += -ggdb -Wall -fPIC -O2 -TARGETS = gpu_neon.$(EXT) gpu_peops.$(EXT) gpu_unai.$(EXT) +SRC += psx_gpu_if.c -gpu_neon.$(EXT): SRC += psx_gpu_if.c -gpu_neon.$(EXT): CFLAGS += -DTEXTURE_CACHE_4BPP -DTEXTURE_CACHE_8BPP +CFLAGS += -DTEXTURE_CACHE_4BPP -DTEXTURE_CACHE_8BPP ifeq "$(HAVE_NEON)" "1" -gpu_neon.$(EXT): SRC += psx_gpu/psx_gpu_arm_neon.S -gpu_neon.$(EXT): CFLAGS += -DNEON_BUILD +SRC += psx_gpu/psx_gpu_arm_neon.S +CFLAGS += -DNEON_BUILD else -gpu_neon.$(EXT): CFLAGS += -fno-strict-aliasing -endif -gpu_peops.$(EXT): SRC += peops_if.c -gpu_peops.$(EXT): CFLAGS += -fno-strict-aliasing -gpu_unai.$(EXT): SRC += unai_if.cpp -gpu_unai.$(EXT): CC_ = $(CXX) -ifeq "$(ARCH)" "arm" -gpu_unai.$(EXT): SRC += ../gpu_unai/gpu_arm.s +CFLAGS += -fno-strict-aliasing endif -all: $(TARGETS) - -$(TARGETS): $(SRC) - $(CC_) -o $@ $(SRC) $(CFLAGS) $(LDFLAGS) -shared -Wl,-soname,$@ - -clean: - $(RM) $(TARGETS) +BIN_GPULIB = gpu_neon.so +include ../gpulib/gpulib.mak diff --git a/plugins/gpu_neon/Makefile.test b/plugins/gpu_neon/Makefile.test deleted file mode 100644 index 114f86e..0000000 --- a/plugins/gpu_neon/Makefile.test +++ /dev/null @@ -1,46 +0,0 @@ -CC = $(CROSS_COMPILE)gcc -CXX = $(CROSS_COMPILE)g++ -CC_ = $(CROSS_COMPILE)gcc - -ARCH = $(shell $(CC) -v 2>&1 | grep -i 'target:' | awk '{print $$2}' | awk -F '-' '{print $$1}') -HAVE_NEON = $(shell $(CC_) -E -dD $(CFLAGS) gpu.h | grep -q '__ARM_NEON__ 1' && echo 1) - -CFLAGS += -ggdb -Wall -DTEST -ifndef DEBUG -CFLAGS += -O2 -endif -ifeq "$(ARCH)" "arm" -CFLAGS += -mcpu=cortex-a8 -mtune=cortex-a8 -mfpu=neon -mfloat-abi=softfp -endif -ifeq "$(ARCH)" "x86_64" -CFLAGS += -m32 -endif - -TARGETS = test_neon test_peops test_unai - -SRC += test.c - -all: $(TARGETS) - -test_neon: SRC += psx_gpu_if.c -test_neon: psx_gpu_if.c psx_gpu/*.c -test_neon: CFLAGS += -DTEXTURE_CACHE_4BPP -DTEXTURE_CACHE_8BPP -ifeq "$(HAVE_NEON)" "1" -test_neon: SRC += psx_gpu/psx_gpu_arm_neon.S -test_neon: CFLAGS += -DNEON_BUILD -else -test_neon: CFLAGS += -fno-strict-aliasing -endif -test_peops: SRC += peops_if.c -test_peops: CFLAGS += -fno-strict-aliasing -test_unai: SRC += unai_if.cpp -test_unai: CC_ = $(CXX) -ifeq "$(ARCH)" "arm" -test_unai: SRC += ../gpu_unai/gpu_arm.s -endif - -$(TARGETS): $(SRC) - $(CC_) -o $@ $(SRC) $(CFLAGS) $(LDFLAGS) - -clean: - $(RM) $(TARGETS) diff --git a/plugins/gpu_neon/gpu.c b/plugins/gpu_neon/gpu.c deleted file mode 100644 index 5b16cd4..0000000 --- a/plugins/gpu_neon/gpu.c +++ /dev/null @@ -1,636 +0,0 @@ -/* - * (C) Gražvydas "notaz" Ignotas, 2011-2012 - * - * This work is licensed under the terms of any of these licenses - * (at your option): - * - GNU GPL, version 2 or later. - * - GNU LGPL, version 2.1 or later. - * See the COPYING file in the top-level directory. - */ - -#include -#include -#include "gpu.h" - -#define ARRAY_SIZE(x) (sizeof(x) / sizeof((x)[0])) -#define unlikely(x) __builtin_expect((x), 0) -#define noinline __attribute__((noinline)) - -#define gpu_log(fmt, ...) \ - printf("%d:%03d: " fmt, *gpu.state.frame_count, *gpu.state.hcnt, ##__VA_ARGS__) - -//#define log_io gpu_log -#define log_io(...) -//#define log_anomaly gpu_log -#define log_anomaly(...) - -struct psx_gpu gpu __attribute__((aligned(2048))); - -static noinline int do_cmd_buffer(uint32_t *data, int count); -static void finish_vram_transfer(int is_read); - -static noinline void do_cmd_reset(void) -{ - if (unlikely(gpu.cmd_len > 0)) - do_cmd_buffer(gpu.cmd_buffer, gpu.cmd_len); - gpu.cmd_len = 0; - - if (unlikely(gpu.dma.h > 0)) - finish_vram_transfer(gpu.dma_start.is_read); - gpu.dma.h = 0; -} - -static noinline void do_reset(void) -{ - unsigned int i; - - do_cmd_reset(); - - memset(gpu.regs, 0, sizeof(gpu.regs)); - for (i = 0; i < sizeof(gpu.ex_regs) / sizeof(gpu.ex_regs[0]); i++) - gpu.ex_regs[i] = (0xe0 + i) << 24; - gpu.status.reg = 0x14802000; - gpu.gp0 = 0; - gpu.regs[3] = 1; - gpu.screen.hres = gpu.screen.w = 256; - gpu.screen.vres = gpu.screen.h = 240; -} - -static noinline void update_width(void) -{ - int sw = gpu.screen.x2 - gpu.screen.x1; - if (sw <= 0 || sw >= 2560) - // full width - gpu.screen.w = gpu.screen.hres; - else - gpu.screen.w = sw * gpu.screen.hres / 2560; -} - -static noinline void update_height(void) -{ - int sh = gpu.screen.y2 - gpu.screen.y1; - if (gpu.status.dheight) - sh *= 2; - if (sh <= 0) - sh = gpu.screen.vres; - - gpu.screen.h = sh; -} - -static noinline void decide_frameskip(void) -{ - if (gpu.frameskip.active) - gpu.frameskip.cnt++; - else { - gpu.frameskip.cnt = 0; - gpu.frameskip.frame_ready = 1; - } - - if (!gpu.frameskip.active && *gpu.frameskip.advice) - gpu.frameskip.active = 1; - else if (gpu.frameskip.set > 0 && gpu.frameskip.cnt < gpu.frameskip.set) - gpu.frameskip.active = 1; - else - gpu.frameskip.active = 0; -} - -static noinline void decide_frameskip_allow(uint32_t cmd_e3) -{ - // no frameskip if it decides to draw to display area, - // but not for interlace since it'll most likely always do that - uint32_t x = cmd_e3 & 0x3ff; - uint32_t y = (cmd_e3 >> 10) & 0x3ff; - gpu.frameskip.allow = gpu.status.interlace || - (uint32_t)(x - gpu.screen.x) >= (uint32_t)gpu.screen.w || - (uint32_t)(y - gpu.screen.y) >= (uint32_t)gpu.screen.h; -} - -static noinline void get_gpu_info(uint32_t data) -{ - switch (data & 0x0f) { - case 0x02: - case 0x03: - case 0x04: - case 0x05: - gpu.gp0 = gpu.ex_regs[data & 7] & 0xfffff; - break; - case 0x06: - gpu.gp0 = gpu.ex_regs[5] & 0xfffff; - break; - case 0x07: - gpu.gp0 = 2; - break; - default: - gpu.gp0 = 0; - break; - } -} - -long GPUinit(void) -{ - int ret; - ret = vout_init(); - ret |= renderer_init(); - - gpu.state.frame_count = &gpu.zero; - gpu.state.hcnt = &gpu.zero; - gpu.frameskip.active = 0; - gpu.cmd_len = 0; - do_reset(); - - return ret; -} - -long GPUshutdown(void) -{ - return vout_finish(); -} - -void GPUwriteStatus(uint32_t data) -{ - static const short hres[8] = { 256, 368, 320, 384, 512, 512, 640, 640 }; - static const short vres[4] = { 240, 480, 256, 480 }; - uint32_t cmd = data >> 24; - - if (cmd < ARRAY_SIZE(gpu.regs)) { - if (cmd > 1 && cmd != 5 && gpu.regs[cmd] == data) - return; - gpu.regs[cmd] = data; - } - - gpu.state.fb_dirty = 1; - - switch (cmd) { - case 0x00: - do_reset(); - break; - case 0x01: - do_cmd_reset(); - break; - case 0x03: - gpu.status.blanking = data & 1; - break; - case 0x04: - gpu.status.dma = data & 3; - break; - case 0x05: - gpu.screen.x = data & 0x3ff; - gpu.screen.y = (data >> 10) & 0x3ff; - if (gpu.frameskip.set) { - decide_frameskip_allow(gpu.ex_regs[3]); - if (gpu.frameskip.last_flip_frame != *gpu.state.frame_count) { - decide_frameskip(); - gpu.frameskip.last_flip_frame = *gpu.state.frame_count; - } - } - break; - case 0x06: - gpu.screen.x1 = data & 0xfff; - gpu.screen.x2 = (data >> 12) & 0xfff; - update_width(); - break; - case 0x07: - gpu.screen.y1 = data & 0x3ff; - gpu.screen.y2 = (data >> 10) & 0x3ff; - update_height(); - break; - case 0x08: - gpu.status.reg = (gpu.status.reg & ~0x7f0000) | ((data & 0x3F) << 17) | ((data & 0x40) << 10); - gpu.screen.hres = hres[(gpu.status.reg >> 16) & 7]; - gpu.screen.vres = vres[(gpu.status.reg >> 19) & 3]; - update_width(); - update_height(); - break; - default: - if ((cmd & 0xf0) == 0x10) - get_gpu_info(data); - break; - } -} - -const unsigned char cmd_lengths[256] = -{ - 0, 0, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, - 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, - 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, - 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, - 2, 2, 2, 2, 3, 3, 3, 3, 1, 1, 1, 1, 0, 0, 0, 0, // 60 - 1, 1, 1, 1, 2, 2, 2, 2, 1, 1, 1, 1, 2, 2, 2, 2, - 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, - 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, - 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, - 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 -}; - -#define VRAM_MEM_XY(x, y) &gpu.vram[(y) * 1024 + (x)] - -static inline void do_vram_line(int x, int y, uint16_t *mem, int l, int is_read) -{ - uint16_t *vram = VRAM_MEM_XY(x, y); - if (is_read) - memcpy(mem, vram, l * 2); - else - memcpy(vram, mem, l * 2); -} - -static int do_vram_io(uint32_t *data, int count, int is_read) -{ - int count_initial = count; - uint16_t *sdata = (uint16_t *)data; - int x = gpu.dma.x, y = gpu.dma.y; - int w = gpu.dma.w, h = gpu.dma.h; - int o = gpu.dma.offset; - int l; - count *= 2; // operate in 16bpp pixels - - if (gpu.dma.offset) { - l = w - gpu.dma.offset; - if (count < l) - l = count; - - do_vram_line(x + o, y, sdata, l, is_read); - - if (o + l < w) - o += l; - else { - o = 0; - y++; - h--; - } - sdata += l; - count -= l; - } - - for (; h > 0 && count >= w; sdata += w, count -= w, y++, h--) { - y &= 511; - do_vram_line(x, y, sdata, w, is_read); - } - - if (h > 0) { - if (count > 0) { - y &= 511; - do_vram_line(x, y, sdata, count, is_read); - o = count; - count = 0; - } - } - else - finish_vram_transfer(is_read); - gpu.dma.y = y; - gpu.dma.h = h; - gpu.dma.offset = o; - - return count_initial - count / 2; -} - -static void start_vram_transfer(uint32_t pos_word, uint32_t size_word, int is_read) -{ - if (gpu.dma.h) - log_anomaly("start_vram_transfer while old unfinished\n"); - - gpu.dma.x = pos_word & 0x3ff; - gpu.dma.y = (pos_word >> 16) & 0x1ff; - gpu.dma.w = ((size_word - 1) & 0x3ff) + 1; - gpu.dma.h = (((size_word >> 16) - 1) & 0x1ff) + 1; - gpu.dma.offset = 0; - gpu.dma.is_read = is_read; - gpu.dma_start = gpu.dma; - - renderer_flush_queues(); - if (is_read) { - gpu.status.img = 1; - // XXX: wrong for width 1 - memcpy(&gpu.gp0, VRAM_MEM_XY(gpu.dma.x, gpu.dma.y), 4); - gpu.state.last_vram_read_frame = *gpu.state.frame_count; - } - - log_io("start_vram_transfer %c (%d, %d) %dx%d\n", is_read ? 'r' : 'w', - gpu.dma.x, gpu.dma.y, gpu.dma.w, gpu.dma.h); -} - -static void finish_vram_transfer(int is_read) -{ - if (is_read) - gpu.status.img = 0; - else - renderer_update_caches(gpu.dma_start.x, gpu.dma_start.y, - gpu.dma_start.w, gpu.dma_start.h); -} - -static noinline int do_cmd_buffer(uint32_t *data, int count) -{ - int len, cmd, start, pos; - int vram_dirty = 0; - - // process buffer - for (start = pos = 0; pos < count; ) - { - cmd = -1; - len = 0; - - if (gpu.dma.h) { - pos += do_vram_io(data + pos, count - pos, 0); - if (pos == count) - break; - start = pos; - } - - // do look-ahead pass to detect SR changes and VRAM i/o - while (pos < count) { - uint32_t *list = data + pos; - cmd = list[0] >> 24; - len = 1 + cmd_lengths[cmd]; - - //printf(" %3d: %02x %d\n", pos, cmd, len); - if ((cmd & 0xf4) == 0x24) { - // flat textured prim - gpu.ex_regs[1] &= ~0x1ff; - gpu.ex_regs[1] |= list[4] & 0x1ff; - } - else if ((cmd & 0xf4) == 0x34) { - // shaded textured prim - gpu.ex_regs[1] &= ~0x1ff; - gpu.ex_regs[1] |= list[5] & 0x1ff; - } - else if (cmd == 0xe3) - decide_frameskip_allow(list[0]); - - if (2 <= cmd && cmd < 0xc0) - vram_dirty = 1; - else if ((cmd & 0xf8) == 0xe0) - gpu.ex_regs[cmd & 7] = list[0]; - - if (pos + len > count) { - cmd = -1; - break; // incomplete cmd - } - if (cmd == 0xa0 || cmd == 0xc0) - break; // image i/o - pos += len; - } - - if (pos - start > 0) { - if (!gpu.frameskip.active || !gpu.frameskip.allow) - do_cmd_list(data + start, pos - start); - start = pos; - } - - if (cmd == 0xa0 || cmd == 0xc0) { - // consume vram write/read cmd - start_vram_transfer(data[pos + 1], data[pos + 2], cmd == 0xc0); - pos += len; - } - else if (cmd == -1) - break; - } - - gpu.status.reg &= ~0x1fff; - gpu.status.reg |= gpu.ex_regs[1] & 0x7ff; - gpu.status.reg |= (gpu.ex_regs[6] & 3) << 11; - - if (gpu.frameskip.active) - renderer_sync_ecmds(gpu.ex_regs); - gpu.state.fb_dirty |= vram_dirty; - - return count - pos; -} - -static void flush_cmd_buffer(void) -{ - int left = do_cmd_buffer(gpu.cmd_buffer, gpu.cmd_len); - if (left > 0) - memmove(gpu.cmd_buffer, gpu.cmd_buffer + gpu.cmd_len - left, left * 4); - gpu.cmd_len = left; -} - -void GPUwriteDataMem(uint32_t *mem, int count) -{ - int left; - - log_io("gpu_dma_write %p %d\n", mem, count); - - if (unlikely(gpu.cmd_len > 0)) - flush_cmd_buffer(); - - left = do_cmd_buffer(mem, count); - if (left) - log_anomaly("GPUwriteDataMem: discarded %d/%d words\n", left, count); -} - -void GPUwriteData(uint32_t data) -{ - log_io("gpu_write %08x\n", data); - gpu.cmd_buffer[gpu.cmd_len++] = data; - if (gpu.cmd_len >= CMD_BUFFER_LEN) - flush_cmd_buffer(); -} - -long GPUdmaChain(uint32_t *rambase, uint32_t start_addr) -{ - uint32_t addr, *list; - uint32_t *llist_entry = NULL; - int len, left, count; - long cpu_cycles = 0; - - if (unlikely(gpu.cmd_len > 0)) - flush_cmd_buffer(); - - // ff7 sends it's main list twice, detect this - if (*gpu.state.frame_count == gpu.state.last_list.frame && - *gpu.state.hcnt - gpu.state.last_list.hcnt <= 1 && - gpu.state.last_list.cycles > 2048) - { - llist_entry = rambase + (gpu.state.last_list.addr & 0x1fffff) / 4; - *llist_entry |= 0x800000; - } - - log_io("gpu_dma_chain\n"); - addr = start_addr & 0xffffff; - for (count = 0; addr != 0xffffff; count++) - { - list = rambase + (addr & 0x1fffff) / 4; - len = list[0] >> 24; - addr = list[0] & 0xffffff; - cpu_cycles += 10; - if (len > 0) - cpu_cycles += 5 + len; - - log_io(".chain %08x #%d\n", (list - rambase) * 4, len); - - // loop detection marker - // (bit23 set causes DMA error on real machine, so - // unlikely to be ever set by the game) - list[0] |= 0x800000; - - if (len) { - left = do_cmd_buffer(list + 1, len); - if (left) - log_anomaly("GPUdmaChain: discarded %d/%d words\n", left, len); - } - - if (addr & 0x800000) - break; - } - - // remove loop detection markers - addr = start_addr & 0x1fffff; - while (count-- > 0) { - list = rambase + addr / 4; - addr = list[0] & 0x1fffff; - list[0] &= ~0x800000; - } - if (llist_entry) - *llist_entry &= ~0x800000; - - gpu.state.last_list.frame = *gpu.state.frame_count; - gpu.state.last_list.hcnt = *gpu.state.hcnt; - gpu.state.last_list.cycles = cpu_cycles; - gpu.state.last_list.addr = start_addr; - - return cpu_cycles; -} - -void GPUreadDataMem(uint32_t *mem, int count) -{ - log_io("gpu_dma_read %p %d\n", mem, count); - - if (unlikely(gpu.cmd_len > 0)) - flush_cmd_buffer(); - - if (gpu.dma.h) - do_vram_io(mem, count, 1); -} - -uint32_t GPUreadData(void) -{ - uint32_t ret; - - if (unlikely(gpu.cmd_len > 0)) - flush_cmd_buffer(); - - ret = gpu.gp0; - if (gpu.dma.h) - do_vram_io(&ret, 1, 1); - - log_io("gpu_read %08x\n", ret); - return ret; -} - -uint32_t GPUreadStatus(void) -{ - uint32_t ret; - - if (unlikely(gpu.cmd_len > 0)) - flush_cmd_buffer(); - - ret = gpu.status.reg; - log_io("gpu_read_status %08x\n", ret); - return ret; -} - -struct GPUFreeze -{ - uint32_t ulFreezeVersion; // should be always 1 for now (set by main emu) - uint32_t ulStatus; // current gpu status - uint32_t ulControl[256]; // latest control register values - unsigned char psxVRam[1024*1024*2]; // current VRam image (full 2 MB for ZN) -}; - -long GPUfreeze(uint32_t type, struct GPUFreeze *freeze) -{ - int i; - - switch (type) { - case 1: // save - if (gpu.cmd_len > 0) - flush_cmd_buffer(); - memcpy(freeze->psxVRam, gpu.vram, sizeof(gpu.vram)); - memcpy(freeze->ulControl, gpu.regs, sizeof(gpu.regs)); - memcpy(freeze->ulControl + 0xe0, gpu.ex_regs, sizeof(gpu.ex_regs)); - freeze->ulStatus = gpu.status.reg; - break; - case 0: // load - memcpy(gpu.vram, freeze->psxVRam, sizeof(gpu.vram)); - memcpy(gpu.regs, freeze->ulControl, sizeof(gpu.regs)); - memcpy(gpu.ex_regs, freeze->ulControl + 0xe0, sizeof(gpu.ex_regs)); - gpu.status.reg = freeze->ulStatus; - for (i = 8; i > 0; i--) { - gpu.regs[i] ^= 1; // avoid reg change detection - GPUwriteStatus((i << 24) | (gpu.regs[i] ^ 1)); - } - renderer_sync_ecmds(gpu.ex_regs); - renderer_update_caches(0, 0, 1024, 512); - break; - } - - return 1; -} - -void GPUupdateLace(void) -{ - if (gpu.cmd_len > 0) - flush_cmd_buffer(); - renderer_flush_queues(); - - if (gpu.status.blanking || !gpu.state.fb_dirty) - return; - - if (gpu.frameskip.set) { - if (!gpu.frameskip.frame_ready) { - if (*gpu.state.frame_count - gpu.frameskip.last_flip_frame < 9) - return; - gpu.frameskip.active = 0; - } - gpu.frameskip.frame_ready = 0; - } - - vout_update(); - gpu.state.fb_dirty = 0; -} - -void GPUvBlank(int is_vblank, int lcf) -{ - int interlace = gpu.state.allow_interlace - && gpu.status.interlace && gpu.status.dheight; - // interlace doesn't look nice on progressive displays, - // so we have this "auto" mode here for games that don't read vram - if (gpu.state.allow_interlace == 2 - && *gpu.state.frame_count - gpu.state.last_vram_read_frame > 1) - { - interlace = 0; - } - if (interlace || interlace != gpu.state.old_interlace) { - gpu.state.old_interlace = interlace; - - if (gpu.cmd_len > 0) - flush_cmd_buffer(); - renderer_flush_queues(); - renderer_set_interlace(interlace, !lcf); - } -} - -#include "../../frontend/plugin_lib.h" - -void GPUrearmedCallbacks(const struct rearmed_cbs *cbs) -{ - gpu.frameskip.set = cbs->frameskip; - gpu.frameskip.advice = &cbs->fskip_advice; - gpu.frameskip.active = 0; - gpu.frameskip.frame_ready = 1; - gpu.state.hcnt = cbs->gpu_hcnt; - gpu.state.frame_count = cbs->gpu_frame_count; - gpu.state.allow_interlace = cbs->gpu_neon.allow_interlace; - - if (cbs->pl_vout_set_raw_vram) - cbs->pl_vout_set_raw_vram(gpu.vram); - renderer_set_config(cbs); - vout_set_config(cbs); -} - -// vim:shiftwidth=2:expandtab diff --git a/plugins/gpu_neon/gpu.h b/plugins/gpu_neon/gpu.h deleted file mode 100644 index d9ad416..0000000 --- a/plugins/gpu_neon/gpu.h +++ /dev/null @@ -1,131 +0,0 @@ -/* - * (C) Gražvydas "notaz" Ignotas, 2011 - * - * This work is licensed under the terms of any of these licenses - * (at your option): - * - GNU GPL, version 2 or later. - * - GNU LGPL, version 2.1 or later. - * See the COPYING file in the top-level directory. - */ - -#include - -#ifdef __cplusplus -extern "C" { -#endif - -#define CMD_BUFFER_LEN 1024 - -struct psx_gpu { - uint16_t vram[1024 * 512]; - uint16_t guard[1024 * 512]; // overdraw guard - uint32_t cmd_buffer[CMD_BUFFER_LEN]; - uint32_t regs[16]; - union { - uint32_t reg; - struct { - uint32_t tx:4; // 0 texture page - uint32_t ty:1; - uint32_t abr:2; - uint32_t tp:2; // 7 t.p. mode (4,8,15bpp) - uint32_t dtd:1; // 9 dither - uint32_t dfe:1; - uint32_t md:1; // 11 set mask bit when drawing - uint32_t me:1; // 12 no draw on mask - uint32_t unkn:3; - uint32_t width1:1; // 16 - uint32_t width0:2; - uint32_t dheight:1; // 19 double height - uint32_t video:1; // 20 NTSC,PAL - uint32_t rgb24:1; - uint32_t interlace:1; // 22 interlace on - uint32_t blanking:1; // 23 display not enabled - uint32_t unkn2:2; - uint32_t busy:1; // 26 !busy drawing - uint32_t img:1; // 27 ready to DMA image data - uint32_t com:1; // 28 ready for commands - uint32_t dma:2; // 29 off, ?, to vram, from vram - uint32_t lcf:1; // 31 - }; - } status; - uint32_t gp0; - uint32_t ex_regs[8]; - struct { - int hres, vres; - int x, y, w, h; - int x1, x2; - int y1, y2; - } screen; - struct { - int x, y, w, h; - short int offset, is_read; - } dma, dma_start; - int cmd_len; - uint32_t zero; - struct { - uint32_t fb_dirty:1; - uint32_t old_interlace:1; - uint32_t allow_interlace:2; - uint32_t *frame_count; - uint32_t *hcnt; /* hsync count */ - struct { - uint32_t addr; - uint32_t cycles; - uint32_t frame; - uint32_t hcnt; - } last_list; - uint32_t last_vram_read_frame; - } state; - struct { - int32_t set:3; /* -1 auto, 0 off, 1-3 fixed */ - int32_t cnt:3; /* amount skipped in a row */ - uint32_t active:1; - uint32_t allow:1; - uint32_t frame_ready:1; - const int *advice; - uint32_t last_flip_frame; - } frameskip; -}; - -extern struct psx_gpu gpu; - -extern const unsigned char cmd_lengths[256]; - -void do_cmd_list(uint32_t *list, int count); - -struct rearmed_cbs; - -int renderer_init(void); -void renderer_sync_ecmds(uint32_t * ecmds); -void renderer_update_caches(int x, int y, int w, int h); -void renderer_flush_queues(void); -void renderer_set_interlace(int enable, int is_odd); -void renderer_set_config(const struct rearmed_cbs *config); - -int vout_init(void); -int vout_finish(void); -void vout_update(void); -void vout_set_config(const struct rearmed_cbs *config); - -/* listing these here for correct linkage if rasterizer uses c++ */ -struct GPUFreeze; - -long GPUinit(void); -long GPUshutdown(void); -void GPUwriteDataMem(uint32_t *mem, int count); -long GPUdmaChain(uint32_t *rambase, uint32_t addr); -void GPUwriteData(uint32_t data); -void GPUreadDataMem(uint32_t *mem, int count); -uint32_t GPUreadData(void); -uint32_t GPUreadStatus(void); -void GPUwriteStatus(uint32_t data); -long GPUfreeze(uint32_t type, struct GPUFreeze *freeze); -void GPUupdateLace(void); -long GPUopen(void **dpy); -long GPUclose(void); -void GPUvBlank(int is_vblank, int lcf); -void GPUrearmedCallbacks(const struct rearmed_cbs *cbs_); - -#ifdef __cplusplus -} -#endif diff --git a/plugins/gpu_neon/peops_if.c b/plugins/gpu_neon/peops_if.c deleted file mode 100644 index 191fce1..0000000 --- a/plugins/gpu_neon/peops_if.c +++ /dev/null @@ -1,398 +0,0 @@ -/*************************************************************************** - copyright : (C) 2001 by Pete Bernert, 2011 notaz - - ***************************************************************************/ -/*************************************************************************** - * * - * 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. See also the license.txt file for * - * additional informations. * - * * - ***************************************************************************/ - -#include -#include -#include -#include "gpu.h" - -#define u32 uint32_t - -#define INFO_TW 0 -#define INFO_DRAWSTART 1 -#define INFO_DRAWEND 2 -#define INFO_DRAWOFF 3 - -#define SHADETEXBIT(x) ((x>>24) & 0x1) -#define SEMITRANSBIT(x) ((x>>25) & 0x1) -#define PSXRGB(r,g,b) ((g<<10)|(b<<5)|r) - -#define DATAREGISTERMODES unsigned short - -#define DR_NORMAL 0 -#define DR_VRAMTRANSFER 1 - -#define GPUSTATUS_READYFORVRAM 0x08000000 - -// byteswappings - -#define SWAP16(x) ({ uint16_t y=(x); (((y)>>8 & 0xff) | ((y)<<8 & 0xff00)); }) -#define SWAP32(x) ({ uint32_t y=(x); (((y)>>24 & 0xfful) | ((y)>>8 & 0xff00ul) | ((y)<<8 & 0xff0000ul) | ((y)<<24 & 0xff000000ul)); }) - -#ifdef __BIG_ENDIAN__ - -// big endian config -#define HOST2LE32(x) SWAP32(x) -#define HOST2BE32(x) (x) -#define LE2HOST32(x) SWAP32(x) -#define BE2HOST32(x) (x) - -#define HOST2LE16(x) SWAP16(x) -#define HOST2BE16(x) (x) -#define LE2HOST16(x) SWAP16(x) -#define BE2HOST16(x) (x) - -#else - -// little endian config -#define HOST2LE32(x) (x) -#define HOST2BE32(x) SWAP32(x) -#define LE2HOST32(x) (x) -#define BE2HOST32(x) SWAP32(x) - -#define HOST2LE16(x) (x) -#define HOST2BE16(x) SWAP16(x) -#define LE2HOST16(x) (x) -#define BE2HOST16(x) SWAP16(x) - -#endif - -#define GETLEs16(X) ((int16_t)GETLE16((uint16_t *)X)) -#define GETLEs32(X) ((int16_t)GETLE32((uint16_t *)X)) - -#define GETLE16(X) LE2HOST16(*(uint16_t *)X) -#define GETLE32_(X) LE2HOST32(*(uint32_t *)X) -#define GETLE16D(X) ({uint32_t val = GETLE32(X); (val<<16 | val >> 16);}) -#define PUTLE16(X, Y) do{*((uint16_t *)X)=HOST2LE16((uint16_t)Y);}while(0) -#define PUTLE32_(X, Y) do{*((uint32_t *)X)=HOST2LE16((uint32_t)Y);}while(0) -#ifdef __arm__ -#define GETLE32(X) (*(uint16_t *)(X)|(((uint16_t *)(X))[1]<<16)) -#define PUTLE32(X, Y) do{uint16_t *p_=(uint16_t *)(X);uint32_t y_=Y;p_[0]=y_;p_[1]=y_>>16;}while(0) -#else -#define GETLE32 GETLE32_ -#define PUTLE32 PUTLE32_ -#endif - -///////////////////////////////////////////////////////////////////////////// - -typedef struct VRAMLOADTTAG -{ - short x; - short y; - short Width; - short Height; - short RowsRemaining; - short ColsRemaining; - unsigned short *ImagePtr; -} VRAMLoad_t; - -///////////////////////////////////////////////////////////////////////////// - -typedef struct PSXPOINTTAG -{ - int32_t x; - int32_t y; -} PSXPoint_t; - -typedef struct PSXSPOINTTAG -{ - short x; - short y; -} PSXSPoint_t; - -typedef struct PSXRECTTAG -{ - short x0; - short x1; - short y0; - short y1; -} PSXRect_t; - -// linux defines for some windows stuff - -#define FALSE 0 -#define TRUE 1 -#define BOOL unsigned short -#define LOWORD(l) ((unsigned short)(l)) -#define HIWORD(l) ((unsigned short)(((uint32_t)(l) >> 16) & 0xFFFF)) -#define max(a,b) (((a) > (b)) ? (a) : (b)) -#define min(a,b) (((a) < (b)) ? (a) : (b)) -#define DWORD uint32_t -#ifndef __int64 -#define __int64 long long int -#endif - -typedef struct RECTTAG -{ - int left; - int top; - int right; - int bottom; -}RECT; - -///////////////////////////////////////////////////////////////////////////// - -typedef struct TWINTAG -{ - PSXRect_t Position; - int xmask, ymask; -} TWin_t; - -///////////////////////////////////////////////////////////////////////////// - -typedef struct PSXDISPLAYTAG -{ - PSXPoint_t DisplayModeNew; - PSXPoint_t DisplayMode; - PSXPoint_t DisplayPosition; - PSXPoint_t DisplayEnd; - - int32_t Double; - int32_t Height; - int32_t PAL; - int32_t InterlacedNew; - int32_t Interlaced; - int32_t RGB24New; - int32_t RGB24; - PSXSPoint_t DrawOffset; - int32_t Disabled; - PSXRect_t Range; - -} PSXDisplay_t; - -///////////////////////////////////////////////////////////////////////////// - -// draw.c - -extern int32_t GlobalTextAddrX,GlobalTextAddrY,GlobalTextTP; -extern int32_t GlobalTextABR,GlobalTextPAGE; -extern short ly0,lx0,ly1,lx1,ly2,lx2,ly3,lx3; -extern long lLowerpart; -extern BOOL bCheckMask; -extern unsigned short sSetMask; -extern unsigned long lSetMask; -extern short g_m1; -extern short g_m2; -extern short g_m3; -extern short DrawSemiTrans; - -// prim.c - -extern BOOL bUsingTWin; -extern TWin_t TWin; -extern void (*primTableJ[256])(unsigned char *); -extern void (*primTableSkip[256])(unsigned char *); -extern unsigned short usMirror; -extern int iDither; -extern uint32_t dwCfgFixes; -extern uint32_t dwActFixes; -extern int iUseFixes; -extern int iUseDither; -extern BOOL bDoVSyncUpdate; -extern int32_t drawX; -extern int32_t drawY; -extern int32_t drawW; -extern int32_t drawH; - -// gpu.h - -#define OPAQUEON 10 -#define OPAQUEOFF 11 - -#define KEY_RESETTEXSTORE 1 -#define KEY_SHOWFPS 2 -#define KEY_RESETOPAQUE 4 -#define KEY_RESETDITHER 8 -#define KEY_RESETFILTER 16 -#define KEY_RESETADVBLEND 32 -#define KEY_BADTEXTURES 128 -#define KEY_CHECKTHISOUT 256 - -#if !defined(__BIG_ENDIAN__) || defined(__x86_64__) || defined(__i386__) -#ifndef __LITTLE_ENDIAN__ -#define __LITTLE_ENDIAN__ -#endif -#endif - -#ifdef __LITTLE_ENDIAN__ -#define RED(x) (x & 0xff) -#define BLUE(x) ((x>>16) & 0xff) -#define GREEN(x) ((x>>8) & 0xff) -#define COLOR(x) (x & 0xffffff) -#elif defined __BIG_ENDIAN__ -#define RED(x) ((x>>24) & 0xff) -#define BLUE(x) ((x>>8) & 0xff) -#define GREEN(x) ((x>>16) & 0xff) -#define COLOR(x) SWAP32(x & 0xffffff) -#endif - -PSXDisplay_t PSXDisplay; -unsigned char *psxVub; -signed char *psxVsb; -unsigned short *psxVuw; -unsigned short *psxVuw_eom; -signed short *psxVsw; -uint32_t *psxVul; -int32_t *psxVsl; - -long lGPUstatusRet; -uint32_t lGPUInfoVals[16]; - -VRAMLoad_t VRAMWrite; -VRAMLoad_t VRAMRead; - -DATAREGISTERMODES DataWriteMode; -DATAREGISTERMODES DataReadMode; - -BOOL bCheckMask = FALSE; -unsigned short sSetMask = 0; -unsigned long lSetMask = 0; -long lLowerpart; - -#include "../dfxvideo/soft.c" -#include "../dfxvideo/prim.c" - -///////////////////////////////////////////////////////////////////////////// - -int renderer_init(void) -{ - psxVub=(void *)gpu.vram; - - psxVsb=(signed char *)psxVub; // different ways of accessing PSX VRAM - psxVsw=(signed short *)psxVub; - psxVsl=(int32_t *)psxVub; - psxVuw=(unsigned short *)psxVub; - psxVul=(uint32_t *)psxVub; - - psxVuw_eom=psxVuw+1024*512; // pre-calc of end of vram - - PSXDisplay.RGB24 = FALSE; // init some stuff - PSXDisplay.Interlaced = FALSE; - PSXDisplay.DrawOffset.x = 0; - PSXDisplay.DrawOffset.y = 0; - PSXDisplay.DisplayMode.x= 320; - PSXDisplay.DisplayMode.y= 240; - PSXDisplay.Disabled = FALSE; - PSXDisplay.Range.x0=0; - PSXDisplay.Range.x1=0; - PSXDisplay.Double = 1; - - DataWriteMode = DR_NORMAL; - lGPUstatusRet = 0x14802000; - - return 0; -} - -extern const unsigned char cmd_lengths[256]; - -void do_cmd_list(unsigned int *list, int list_len) -{ - unsigned int cmd, len; - - unsigned int *list_end = list + list_len; - - for (; list < list_end; list += 1 + len) - { - cmd = *list >> 24; - len = cmd_lengths[cmd]; - - primTableJ[cmd]((void *)list); - - switch(cmd) - { - case 0x48 ... 0x4F: - { - u32 num_vertexes = 1; - u32 *list_position = &(list[2]); - - while(1) - { - if((*list_position & 0xf000f000) == 0x50005000 || list_position >= list_end) - break; - - list_position++; - num_vertexes++; - } - - if(num_vertexes > 2) - len += (num_vertexes - 2); - - break; - } - - case 0x58 ... 0x5F: - { - u32 num_vertexes = 1; - u32 *list_position = &(list[2]); - - while(1) - { - if((*list_position & 0xf000f000) == 0x50005000 || list_position >= list_end) - break; - - list_position += 2; - num_vertexes++; - } - - if(num_vertexes > 2) - len += (num_vertexes - 2) * 2; - - break; - } - -#ifdef TEST - case 0xA0: // sys -> vid - { - short *slist = (void *)list; - u32 load_width = slist[4]; - u32 load_height = slist[5]; - u32 load_size = load_width * load_height; - - len += load_size / 2; - break; - } -#endif - } - } -} - -void renderer_sync_ecmds(uint32_t *ecmds) -{ - cmdTexturePage((unsigned char *)&ecmds[1]); - cmdTextureWindow((unsigned char *)&ecmds[2]); - cmdDrawAreaStart((unsigned char *)&ecmds[3]); - cmdDrawAreaEnd((unsigned char *)&ecmds[4]); - cmdDrawOffset((unsigned char *)&ecmds[5]); - cmdSTP((unsigned char *)&ecmds[6]); -} - -void renderer_update_caches(int x, int y, int w, int h) -{ -} - -void renderer_flush_queues(void) -{ -} - -void renderer_set_interlace(int enable, int is_odd) -{ -} - -#include "../../frontend/plugin_lib.h" - -void renderer_set_config(const struct rearmed_cbs *cbs) -{ - iUseDither = cbs->gpu_peops.iUseDither; - dwActFixes = cbs->gpu_peops.dwActFixes; -} diff --git a/plugins/gpu_neon/psx_gpu_if.c b/plugins/gpu_neon/psx_gpu_if.c index 1c462e1..3ff6e48 100644 --- a/plugins/gpu_neon/psx_gpu_if.c +++ b/plugins/gpu_neon/psx_gpu_if.c @@ -15,7 +15,7 @@ extern const unsigned char cmd_lengths[256]; #include "psx_gpu/psx_gpu.c" #include "psx_gpu/psx_gpu_parse.c" -#include "gpu.h" +#include "../gpulib/gpu.h" static psx_gpu_struct egpu __attribute__((aligned(256))); diff --git a/plugins/gpu_neon/test.c b/plugins/gpu_neon/test.c deleted file mode 100644 index e523e20..0000000 --- a/plugins/gpu_neon/test.c +++ /dev/null @@ -1,109 +0,0 @@ -#include -#include -#include -#include "gpu.h" - -static inline unsigned int pcnt_get(void) -{ - unsigned int val; -#ifdef __ARM_ARCH_7A__ - asm volatile("mrc p15, 0, %0, c9, c13, 0" : "=r"(val)); -#else - val = 0; -#endif - return val; -} - -static inline void pcnt_init(void) -{ -#ifdef __ARM_ARCH_7A__ - int v; - asm volatile("mrc p15, 0, %0, c9, c12, 0" : "=r"(v)); - v |= 5; // master enable, ccnt reset - v &= ~8; // ccnt divider 0 - asm volatile("mcr p15, 0, %0, c9, c12, 0" :: "r"(v)); - // enable cycle counter - asm volatile("mcr p15, 0, %0, c9, c12, 1" :: "r"(1<<31)); -#endif -} - -const unsigned char cmd_lengths[256] = -{ - 0, 0, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, - 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, - 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, - 2, 2, 2, 2, 0, 0, 0, 0, 3, 3, 3, 3, 3, 3, 3, 3, // 40 - 3, 3, 3, 3, 0, 0, 0, 0, 4, 4, 4, 4, 4, 4, 4, 4, - 2, 2, 2, 2, 3, 3, 3, 3, 1, 1, 1, 1, 2, 2, 2, 2, // 60 - 1, 1, 1, 1, 2, 2, 2, 2, 1, 1, 1, 1, 2, 2, 2, 2, - 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, - 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, - 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, - 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 -}; - -struct psx_gpu gpu __attribute__((aligned(64))); - -typedef struct -{ - uint16_t vram[1024 * 512]; - uint32_t gpu_register[15]; - uint32_t status; -} gpu_dump_struct; - -static gpu_dump_struct state; - -int main(int argc, char *argv[]) -{ - unsigned int start_cycles; - uint32_t *list; - int size; - FILE *state_file; - FILE *list_file; - FILE *out_file; - - if (argc != 3 && argc != 4) - { - printf("usage:\n%s [vram_out]\n", argv[0]); - return 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); - - list = (uint32_t *)malloc(size); - fread(list, 1, size, list_file); - fclose(list_file); - - pcnt_init(); - renderer_init(); - memcpy(gpu.vram, state.vram, sizeof(gpu.vram)); - if ((state.gpu_register[8] & 0x24) == 0x24) - renderer_set_interlace(1, !(state.status >> 31)); - - start_cycles = pcnt_get(); - - do_cmd_list(list, size / 4); - renderer_flush_queues(); - - printf("%u\n", pcnt_get() - start_cycles); - - if (argc >= 4) { - out_file = fopen(argv[3], "wb"); - fwrite(gpu.vram, 1, sizeof(gpu.vram), out_file); - fclose(out_file); - } - - return 0; -} diff --git a/plugins/gpu_neon/unai_if.cpp b/plugins/gpu_neon/unai_if.cpp deleted file mode 100644 index 8917f2a..0000000 --- a/plugins/gpu_neon/unai_if.cpp +++ /dev/null @@ -1,291 +0,0 @@ -/*************************************************************************** -* Copyright (C) 2010 PCSX4ALL Team * -* Copyright (C) 2010 Unai * -* Copyright (C) 2011 notaz * -* * -* 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. * -* * -* You should have received a copy of the GNU General Public License * -* along with this program; if not, write to the * -* Free Software Foundation, Inc., * -* 51 Franklin Street, Fifth Floor, Boston, MA 02111-1307 USA. * -***************************************************************************/ - -#include -#include -#include -#include "gpu.h" - -#define u8 uint8_t -#define s8 int8_t -#define u16 uint16_t -#define s16 int16_t -#define u32 uint32_t -#define s32 int32_t -#define s64 int64_t - -#define INLINE static - -#define FRAME_BUFFER_SIZE (1024*512*2) -#define FRAME_WIDTH 1024 -#define FRAME_HEIGHT 512 -#define FRAME_OFFSET(x,y) (((y)<<10)+(x)) - -#define isSkip 0 /* skip frame (info coming from GPU) */ -#define alt_fps 0 -static int linesInterlace; /* internal lines interlace */ -static int force_interlace; - -static bool light = true; /* lighting */ -static bool blend = true; /* blending */ -static bool FrameToRead = false; /* load image in progress */ -static bool FrameToWrite = false; /* store image in progress */ - -static bool enableAbbeyHack = false; /* Abe's Odyssey hack */ - -static u8 BLEND_MODE; -static u8 TEXT_MODE; -static u8 Masking; - -static u16 PixelMSB; -static u16 PixelData; - -/////////////////////////////////////////////////////////////////////////////// -// GPU Global data -/////////////////////////////////////////////////////////////////////////////// - -// Dma Transfers info -static s32 px,py; -static s32 x_end,y_end; -static u16* pvram; - -static s32 PacketCount; -static s32 PacketIndex; - -// Rasterizer status -static u32 TextureWindow [4]; -static u32 DrawingArea [4]; -static u32 DrawingOffset [2]; - -static u16* TBA; -static u16* CBA; - -// Inner Loops -static s32 u4, du4; -static s32 v4, dv4; -static s32 r4, dr4; -static s32 g4, dg4; -static s32 b4, db4; -static u32 lInc; -static u32 tInc, tMsk; - -union GPUPacket -{ - u32 U4[16]; - s32 S4[16]; - u16 U2[32]; - s16 S2[32]; - u8 U1[64]; - s8 S1[64]; -}; - -static GPUPacket PacketBuffer; -static u16 *GPU_FrameBuffer; -static u32 GPU_GP1; - -/////////////////////////////////////////////////////////////////////////////// - -#include "../gpu_unai/gpu_fixedpoint.h" - -// Inner loop driver instanciation file -#include "../gpu_unai/gpu_inner.h" - -// GPU Raster Macros -#define GPU_RGB16(rgb) ((((rgb)&0xF80000)>>9)|(((rgb)&0xF800)>>6)|(((rgb)&0xF8)>>3)) - -#define GPU_EXPANDSIGN_POLY(x) (((s32)(x)<<20)>>20) -//#define GPU_EXPANDSIGN_POLY(x) (((s32)(x)<<21)>>21) -#define GPU_EXPANDSIGN_SPRT(x) (((s32)(x)<<21)>>21) - -//#define GPU_TESTRANGE(x) { if((u32)(x+1024) > 2047) return; } -#define GPU_TESTRANGE(x) { if ((x<-1023) || (x>1023)) return; } - -#define GPU_SWAP(a,b,t) {(t)=(a);(a)=(b);(b)=(t);} - -// GPU internal image drawing functions -#include "../gpu_unai/gpu_raster_image.h" - -// GPU internal line drawing functions -#include "../gpu_unai/gpu_raster_line.h" - -// GPU internal polygon drawing functions -#include "../gpu_unai/gpu_raster_polygon.h" - -// GPU internal sprite drawing functions -#include "../gpu_unai/gpu_raster_sprite.h" - -// GPU command buffer execution/store -#include "../gpu_unai/gpu_command.h" - -#define unai_do_prim(cmd, list, len) \ - memcpy(PacketBuffer.U4, list, (len) * 4); \ - gpuSendPacketFunction(cmd) - -///////////////////////////////////////////////////////////////////////////// - -int renderer_init(void) -{ - GPU_FrameBuffer = (u16 *)gpu.vram; - - // s_invTable - for(int i=1;i<=(1<>1); - #else - v *= double(0x80000000); - #endif - s_invTable[i-1]=s32(v); - } - - return 0; -} - -extern const unsigned char cmd_lengths[256]; - -void do_cmd_list(unsigned int *list, int list_len) -{ - unsigned int cmd, len; - unsigned int *list_end = list + list_len; - - linesInterlace = force_interlace; -#ifndef __ARM_ARCH_7A__ /* XXX */ - linesInterlace |= gpu.status.interlace; -#endif - - for (; list < list_end; list += 1 + len) - { - cmd = *list >> 24; - len = cmd_lengths[cmd]; - - switch(cmd) - { - case 0x48 ... 0x4F: - { - u32 num_vertexes = 1; - u32 *list_position = &(list[2]); - u32 PRIM = cmd; - - memcpy(&PacketBuffer.U4[0], list, 4 * 3); - gpuDrawLF(gpuPixelDrivers [ (Blending_Mode | Masking | Blending | (PixelMSB>>3)) >> 1]); - - while(1) - { - if((*list_position & 0xf000f000) == 0x50005000 || list_position >= list_end) - break; - - PacketBuffer.U4[1] = PacketBuffer.U4[2]; - PacketBuffer.U4[2] = *list_position++; - gpuDrawLF(gpuPixelDrivers [ (Blending_Mode | Masking | Blending | (PixelMSB>>3)) >> 1]); - - num_vertexes++; - } - - if(num_vertexes > 2) - len += (num_vertexes - 2); - - break; - } - - case 0x58 ... 0x5F: - { - u32 num_vertexes = 1; - u32 *list_position = &(list[2]); - u32 PRIM = cmd; - - memcpy(&PacketBuffer.U4[0], list, 4 * 4); - gpuDrawLG(gpuPixelDrivers [ (Blending_Mode | Masking | Blending | (PixelMSB>>3)) >> 1]); - - while(1) - { - if((*list_position & 0xf000f000) == 0x50005000 || list_position >= list_end) - break; - - PacketBuffer.U4[0] = PacketBuffer.U4[2]; - PacketBuffer.U4[1] = PacketBuffer.U4[3]; - PacketBuffer.U4[2] = *list_position++; - PacketBuffer.U4[3] = *list_position++; - gpuDrawLG(gpuPixelDrivers [ (Blending_Mode | Masking | Blending | (PixelMSB>>3)) >> 1]); - - num_vertexes++; - } - - if(num_vertexes > 2) - len += (num_vertexes - 2) * 2; - - break; - } - -#ifdef TEST - case 0xA0: // sys -> vid - { - u32 load_width = list[2] & 0xffff; - u32 load_height = list[2] >> 16; - u32 load_size = load_width * load_height; - - len += load_size / 2; - break; - } -#endif - - default: - unai_do_prim(cmd, list, len + 1); - break; - } - } -} - -void renderer_sync_ecmds(uint32_t *ecmds) -{ - unai_do_prim(0xe1, &ecmds[1], 1); - unai_do_prim(0xe2, &ecmds[2], 1); - unai_do_prim(0xe3, &ecmds[3], 1); - unai_do_prim(0xe4, &ecmds[4], 1); - unai_do_prim(0xe5, &ecmds[5], 1); - unai_do_prim(0xe6, &ecmds[6], 1); -} - -void renderer_update_caches(int x, int y, int w, int h) -{ -} - -void renderer_flush_queues(void) -{ -} - -void renderer_set_interlace(int enable, int is_odd) -{ -} - -#ifndef TEST - -#include "../../frontend/plugin_lib.h" - -void renderer_set_config(const struct rearmed_cbs *cbs) -{ - force_interlace = cbs->gpu_unai.lineskip; - enableAbbeyHack = cbs->gpu_unai.abe_hack; - light = !cbs->gpu_unai.no_light; - blend = !cbs->gpu_unai.no_blend; -} - -#endif diff --git a/plugins/gpu_neon/vout_fb.c b/plugins/gpu_neon/vout_fb.c deleted file mode 100644 index 20c8ff3..0000000 --- a/plugins/gpu_neon/vout_fb.c +++ /dev/null @@ -1,121 +0,0 @@ -/* - * (C) Gražvydas "notaz" Ignotas, 2011 - * - * This work is licensed under the terms of any of these licenses - * (at your option): - * - GNU GPL, version 2 or later. - * - GNU LGPL, version 2.1 or later. - * See the COPYING file in the top-level directory. - */ - -#include "gpu.h" -#include "../../frontend/plugin_lib.h" -#include "../../frontend/cspace.h" - -static const struct rearmed_cbs *cbs; -static void *screen_buf; - -int vout_init(void) -{ - return 0; -} - -int vout_finish(void) -{ - return 0; -} - -static void check_mode_change(void) -{ - static uint32_t old_status; - static int old_h; - - // width|rgb24 change? - if ((gpu.status.reg ^ old_status) & ((7<<16)|(1<<21)) || gpu.screen.h != old_h) - { - old_status = gpu.status.reg; - old_h = gpu.screen.h; - screen_buf = cbs->pl_vout_set_mode(gpu.screen.hres, - gpu.screen.h, gpu.status.rgb24 ? 24 : 16); - } -} - -static void blit(void) -{ - int x = gpu.screen.x & ~1; // alignment needed by blitter - int y = gpu.screen.y; - int w = gpu.screen.w; - int h = gpu.screen.h; - uint16_t *vram = gpu.vram; - int stride = gpu.screen.hres; - int fb_offs, doffs; - uint8_t *dest; - - fb_offs = y * 1024 + x; - dest = (uint8_t *)screen_buf; - - // only do centering, at least for now - doffs = (stride - w) / 2 & ~1; - - if (gpu.status.rgb24) - { -#ifndef MAEMO - dest += (doffs / 8) * 24; - for (; h-- > 0; dest += stride * 3, fb_offs += 1024) - { - fb_offs &= 1024*512-1; - bgr888_to_rgb888(dest, vram + fb_offs, w * 3); - } -#else - dest += doffs * 2; - for (; h-- > 0; dest += stride * 2, fb_offs += 1024) - { - fb_offs &= 1024*512-1; - bgr888_to_rgb565(dest, vram + fb_offs, w * 3); - } -#endif - } - else - { - dest += doffs * 2; - for (; h-- > 0; dest += stride * 2, fb_offs += 1024) - { - fb_offs &= 1024*512-1; - bgr555_to_rgb565(dest, vram + fb_offs, w * 2); - } - } - - screen_buf = cbs->pl_vout_flip(); -} - -void vout_update(void) -{ - check_mode_change(); - if (cbs->pl_vout_raw_flip) - cbs->pl_vout_raw_flip(gpu.screen.x, gpu.screen.y); - else - blit(); -} - -long GPUopen(void **unused) -{ - gpu.frameskip.active = 0; - gpu.frameskip.frame_ready = 1; - - cbs->pl_vout_open(); - screen_buf = cbs->pl_vout_flip(); - return 0; -} - -long GPUclose(void) -{ - cbs->pl_vout_close(); - return 0; -} - -void vout_set_config(const struct rearmed_cbs *cbs_) -{ - cbs = cbs_; -} - -// vim:shiftwidth=2:expandtab diff --git a/plugins/gpu_neon/vout_sdl.c b/plugins/gpu_neon/vout_sdl.c deleted file mode 100644 index db1ae96..0000000 --- a/plugins/gpu_neon/vout_sdl.c +++ /dev/null @@ -1,95 +0,0 @@ -/* - * (C) Gražvydas "notaz" Ignotas, 2011 - * - * This work is licensed under the terms of any of these licenses - * (at your option): - * - GNU GPL, version 2 or later. - * - GNU LGPL, version 2.1 or later. - * See the COPYING file in the top-level directory. - */ - -#include -#include -#include -#include "gpu.h" - -static SDL_Surface *screen; -static Display *x11_display; - -int vout_init(void) -{ - SDL_SysWMinfo wminfo; - int ret; - - ret = SDL_Init(SDL_INIT_VIDEO); - if (ret != 0) { - fprintf(stderr, "SDL_Init failed: %s\n", SDL_GetError()); - return ret; - } - - screen = SDL_SetVideoMode(1024, 512, 32, 0); - if (screen == NULL) { - fprintf(stderr, "SDL_SetVideoMode failed: %s\n", SDL_GetError()); - SDL_Quit(); - return -1; - } - - SDL_VERSION(&wminfo.version); - ret = SDL_GetWMInfo(&wminfo); - if (ret == 1) - x11_display = wminfo.info.x11.display; - - return 0; -} - -int vout_finish(void) -{ - SDL_Quit(); - return 0; -} - -void vout_update(void) -{ - uint32_t *d; - int i; - - SDL_LockSurface(screen); - if (gpu.status.rgb24) - { - uint8_t *s; - int y; - for (y = 0; y < 512; y++) { - s = (uint8_t *)gpu.vram + y * 2*1024; - d = (uint32_t *)screen->pixels + y * 1024; - for (i = 0; i < 1024 * 2 / 3; i++, s += 3) - d[i] = (s[0] << 16) | (s[1] << 8) | s[2]; - } - } - else - { - uint16_t *s = gpu.vram; - d = (uint32_t *)screen->pixels; - for (i = 0; i < 1024 * 512; i++) - d[i] = (((uint32_t)s[i] << 19) & 0xf80000) | ((s[i] << 6) & 0xf800) | - ((s[i] >> 7) & 0xf8); - } - SDL_UnlockSurface(screen); - SDL_UpdateRect(screen, 0, 0, 1024, 512); -} - -long GPUopen(void **dpy) -{ - *dpy = x11_display; - return 0; -} - -long GPUclose(void) -{ - return 0; -} - -void vout_set_config(const struct rearmed_cbs *cbs) -{ -} - -// vim:shiftwidth=2:expandtab diff --git a/plugins/gpu_unai/Makefile b/plugins/gpu_unai/Makefile index e4b6e6b..dd9e4ca 100644 --- a/plugins/gpu_unai/Makefile +++ b/plugins/gpu_unai/Makefile @@ -1,34 +1,17 @@ -CC = $(CROSS_COMPILE)gcc +CC = $(CROSS_COMPILE)g++ -CFLAGS += -ggdb -fPIC -Wall -DREARMED -ifndef DEBUG -CFLAGS += -O2 -ffast-math -fomit-frame-pointer -endif -ifdef MAEMO -CFLAGS += -DMAEMO -endif -ARM_CORTEXA8 ?= 1 -ifeq "$(ARM_CORTEXA8)" "1" -CFLAGS += -mcpu=cortex-a8 -mtune=cortex-a8 -mfpu=neon -mfloat-abi=softfp -SRC += ../../frontend/cspace_neon.s -else -CFLAGS += -mcpu=arm926ej-s -mtune=arm926ej-s -SRC += ../../frontend/cspace.c -endif - -SRC += gpu.cpp gpu_arm.s +ARCH = $(shell $(CC) -v 2>&1 | grep -i 'target:' | awk '{print $$2}' | awk -F '-' '{print $$1}') -TARGET = gpuPCSX4ALL.so -LDFLAGS += -shared -Wl,-soname,$(TARGET) +CFLAGS += -DREARMED +CFLAGS += -ggdb -Wall -fPIC -O2 -ffast-math --include Makefile.local +SRC_STANDALONE += gpu.cpp +SRC_GPULIB += gpulib_if.cpp -all: $(TARGET) - -$(TARGET): $(SRC) - $(CC) -o $@ $^ $(CFLAGS) $(LDFLAGS) - -# $(TARGET): *.h +ifeq "$(ARCH)" "arm" +SRC += gpu_arm.s +endif -clean: - $(RM) $(TARGET) $(OBJ) +BIN_STANDLALONE = gpuPCSX4ALL.so +BIN_GPULIB = gpu_unai.so +include ../gpulib/gpulib.mak diff --git a/plugins/gpu_unai/gpu.cpp b/plugins/gpu_unai/gpu.cpp index f0539c0..5a0ad2c 100644 --- a/plugins/gpu_unai/gpu.cpp +++ b/plugins/gpu_unai/gpu.cpp @@ -821,7 +821,7 @@ void GPU_updateLace(void) #else #include "../../frontend/plugin_lib.h" -#include "../../frontend/cspace.h" +#include "../gpulib/cspace.h" extern "C" { diff --git a/plugins/gpu_unai/gpulib_if.cpp b/plugins/gpu_unai/gpulib_if.cpp new file mode 100644 index 0000000..22fe31e --- /dev/null +++ b/plugins/gpu_unai/gpulib_if.cpp @@ -0,0 +1,291 @@ +/*************************************************************************** +* Copyright (C) 2010 PCSX4ALL Team * +* Copyright (C) 2010 Unai * +* Copyright (C) 2011 notaz * +* * +* 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. * +* * +* You should have received a copy of the GNU General Public License * +* along with this program; if not, write to the * +* Free Software Foundation, Inc., * +* 51 Franklin Street, Fifth Floor, Boston, MA 02111-1307 USA. * +***************************************************************************/ + +#include +#include +#include +#include "../gpulib/gpu.h" + +#define u8 uint8_t +#define s8 int8_t +#define u16 uint16_t +#define s16 int16_t +#define u32 uint32_t +#define s32 int32_t +#define s64 int64_t + +#define INLINE static + +#define FRAME_BUFFER_SIZE (1024*512*2) +#define FRAME_WIDTH 1024 +#define FRAME_HEIGHT 512 +#define FRAME_OFFSET(x,y) (((y)<<10)+(x)) + +#define isSkip 0 /* skip frame (info coming from GPU) */ +#define alt_fps 0 +static int linesInterlace; /* internal lines interlace */ +static int force_interlace; + +static bool light = true; /* lighting */ +static bool blend = true; /* blending */ +static bool FrameToRead = false; /* load image in progress */ +static bool FrameToWrite = false; /* store image in progress */ + +static bool enableAbbeyHack = false; /* Abe's Odyssey hack */ + +static u8 BLEND_MODE; +static u8 TEXT_MODE; +static u8 Masking; + +static u16 PixelMSB; +static u16 PixelData; + +/////////////////////////////////////////////////////////////////////////////// +// GPU Global data +/////////////////////////////////////////////////////////////////////////////// + +// Dma Transfers info +static s32 px,py; +static s32 x_end,y_end; +static u16* pvram; + +static s32 PacketCount; +static s32 PacketIndex; + +// Rasterizer status +static u32 TextureWindow [4]; +static u32 DrawingArea [4]; +static u32 DrawingOffset [2]; + +static u16* TBA; +static u16* CBA; + +// Inner Loops +static s32 u4, du4; +static s32 v4, dv4; +static s32 r4, dr4; +static s32 g4, dg4; +static s32 b4, db4; +static u32 lInc; +static u32 tInc, tMsk; + +union GPUPacket +{ + u32 U4[16]; + s32 S4[16]; + u16 U2[32]; + s16 S2[32]; + u8 U1[64]; + s8 S1[64]; +}; + +static GPUPacket PacketBuffer; +static u16 *GPU_FrameBuffer; +static u32 GPU_GP1; + +/////////////////////////////////////////////////////////////////////////////// + +#include "../gpu_unai/gpu_fixedpoint.h" + +// Inner loop driver instanciation file +#include "../gpu_unai/gpu_inner.h" + +// GPU Raster Macros +#define GPU_RGB16(rgb) ((((rgb)&0xF80000)>>9)|(((rgb)&0xF800)>>6)|(((rgb)&0xF8)>>3)) + +#define GPU_EXPANDSIGN_POLY(x) (((s32)(x)<<20)>>20) +//#define GPU_EXPANDSIGN_POLY(x) (((s32)(x)<<21)>>21) +#define GPU_EXPANDSIGN_SPRT(x) (((s32)(x)<<21)>>21) + +//#define GPU_TESTRANGE(x) { if((u32)(x+1024) > 2047) return; } +#define GPU_TESTRANGE(x) { if ((x<-1023) || (x>1023)) return; } + +#define GPU_SWAP(a,b,t) {(t)=(a);(a)=(b);(b)=(t);} + +// GPU internal image drawing functions +#include "../gpu_unai/gpu_raster_image.h" + +// GPU internal line drawing functions +#include "../gpu_unai/gpu_raster_line.h" + +// GPU internal polygon drawing functions +#include "../gpu_unai/gpu_raster_polygon.h" + +// GPU internal sprite drawing functions +#include "../gpu_unai/gpu_raster_sprite.h" + +// GPU command buffer execution/store +#include "../gpu_unai/gpu_command.h" + +#define unai_do_prim(cmd, list, len) \ + memcpy(PacketBuffer.U4, list, (len) * 4); \ + gpuSendPacketFunction(cmd) + +///////////////////////////////////////////////////////////////////////////// + +int renderer_init(void) +{ + GPU_FrameBuffer = (u16 *)gpu.vram; + + // s_invTable + for(int i=1;i<=(1<>1); + #else + v *= double(0x80000000); + #endif + s_invTable[i-1]=s32(v); + } + + return 0; +} + +extern const unsigned char cmd_lengths[256]; + +void do_cmd_list(unsigned int *list, int list_len) +{ + unsigned int cmd, len; + unsigned int *list_end = list + list_len; + + linesInterlace = force_interlace; +#ifndef __ARM_ARCH_7A__ /* XXX */ + linesInterlace |= gpu.status.interlace; +#endif + + for (; list < list_end; list += 1 + len) + { + cmd = *list >> 24; + len = cmd_lengths[cmd]; + + switch(cmd) + { + case 0x48 ... 0x4F: + { + u32 num_vertexes = 1; + u32 *list_position = &(list[2]); + u32 PRIM = cmd; + + memcpy(&PacketBuffer.U4[0], list, 4 * 3); + gpuDrawLF(gpuPixelDrivers [ (Blending_Mode | Masking | Blending | (PixelMSB>>3)) >> 1]); + + while(1) + { + if((*list_position & 0xf000f000) == 0x50005000 || list_position >= list_end) + break; + + PacketBuffer.U4[1] = PacketBuffer.U4[2]; + PacketBuffer.U4[2] = *list_position++; + gpuDrawLF(gpuPixelDrivers [ (Blending_Mode | Masking | Blending | (PixelMSB>>3)) >> 1]); + + num_vertexes++; + } + + if(num_vertexes > 2) + len += (num_vertexes - 2); + + break; + } + + case 0x58 ... 0x5F: + { + u32 num_vertexes = 1; + u32 *list_position = &(list[2]); + u32 PRIM = cmd; + + memcpy(&PacketBuffer.U4[0], list, 4 * 4); + gpuDrawLG(gpuPixelDrivers [ (Blending_Mode | Masking | Blending | (PixelMSB>>3)) >> 1]); + + while(1) + { + if((*list_position & 0xf000f000) == 0x50005000 || list_position >= list_end) + break; + + PacketBuffer.U4[0] = PacketBuffer.U4[2]; + PacketBuffer.U4[1] = PacketBuffer.U4[3]; + PacketBuffer.U4[2] = *list_position++; + PacketBuffer.U4[3] = *list_position++; + gpuDrawLG(gpuPixelDrivers [ (Blending_Mode | Masking | Blending | (PixelMSB>>3)) >> 1]); + + num_vertexes++; + } + + if(num_vertexes > 2) + len += (num_vertexes - 2) * 2; + + break; + } + +#ifdef TEST + case 0xA0: // sys -> vid + { + u32 load_width = list[2] & 0xffff; + u32 load_height = list[2] >> 16; + u32 load_size = load_width * load_height; + + len += load_size / 2; + break; + } +#endif + + default: + unai_do_prim(cmd, list, len + 1); + break; + } + } +} + +void renderer_sync_ecmds(uint32_t *ecmds) +{ + unai_do_prim(0xe1, &ecmds[1], 1); + unai_do_prim(0xe2, &ecmds[2], 1); + unai_do_prim(0xe3, &ecmds[3], 1); + unai_do_prim(0xe4, &ecmds[4], 1); + unai_do_prim(0xe5, &ecmds[5], 1); + unai_do_prim(0xe6, &ecmds[6], 1); +} + +void renderer_update_caches(int x, int y, int w, int h) +{ +} + +void renderer_flush_queues(void) +{ +} + +void renderer_set_interlace(int enable, int is_odd) +{ +} + +#ifndef TEST + +#include "../../frontend/plugin_lib.h" + +void renderer_set_config(const struct rearmed_cbs *cbs) +{ + force_interlace = cbs->gpu_unai.lineskip; + enableAbbeyHack = cbs->gpu_unai.abe_hack; + light = !cbs->gpu_unai.no_light; + blend = !cbs->gpu_unai.no_blend; +} + +#endif diff --git a/plugins/gpulib/Makefile b/plugins/gpulib/Makefile new file mode 100644 index 0000000..a244524 --- /dev/null +++ b/plugins/gpulib/Makefile @@ -0,0 +1,38 @@ +CC = $(CROSS_COMPILE)gcc +AS = $(CROSS_COMPILE)as +AR = $(CROSS_COMPILE)ar + +ARCH = $(shell $(CC) -v 2>&1 | grep -i 'target:' | awk '{print $$2}' | awk -F '-' '{print $$1}') + +CFLAGS += -ggdb -Wall -fPIC -O2 + +OBJS += gpu.o + +ifeq "$(ARCH)" "arm" +ARM_CORTEXA8 ?= 1 +ifeq "$(ARM_CORTEXA8)" "1" +CFLAGS += -mcpu=cortex-a8 -mtune=cortex-a8 -mfpu=neon -mfloat-abi=softfp +ASFLAGS += -mcpu=cortex-a8 -mfpu=neon +OBJS += cspace_neon.o +else +CFLAGS += -mcpu=arm926ej-s -mtune=arm926ej-s +ASFLAGS += -mcpu=arm926ej-s -mfloat-abi=softfp +OBJS += cspace.o +endif +OBJS += vout_fb.o +EXT = a +else +CFLAGS += `sdl-config --cflags` -m32 +OBJS += vout_sdl.o +EXT = x86.a +endif + +TARGET = gpulib.$(EXT) + +all: $(TARGET) + +$(TARGET): $(OBJS) + $(AR) crs $@ $^ + +clean: + $(RM) $(TARGET) $(OBJS) diff --git a/plugins/gpulib/Makefile.test b/plugins/gpulib/Makefile.test new file mode 100644 index 0000000..196b1f9 --- /dev/null +++ b/plugins/gpulib/Makefile.test @@ -0,0 +1,45 @@ +CC = $(CROSS_COMPILE)gcc +CXX = $(CROSS_COMPILE)g++ +CC_ = $(CROSS_COMPILE)gcc + +ARCH = $(shell $(CC) -v 2>&1 | grep -i 'target:' | awk '{print $$2}' | awk -F '-' '{print $$1}') +HAVE_NEON = $(shell $(CC_) -E -dD $(CFLAGS) gpu.h | grep -q '__ARM_NEON__ 1' && echo 1) + +CFLAGS += -ggdb -Wall -DTEST +ifndef DEBUG +CFLAGS += -O2 +endif +ifeq "$(ARCH)" "arm" +CFLAGS += -mcpu=cortex-a8 -mtune=cortex-a8 -mfpu=neon -mfloat-abi=softfp +endif +ifeq "$(ARCH)" "x86_64" +CFLAGS += -m32 +endif + +TARGETS = test_neon test_peops test_unai + +SRC += test.c + +all: $(TARGETS) + +test_neon: SRC += ../gpu_neon/psx_gpu_if.c +test_neon: CFLAGS += -DTEXTURE_CACHE_4BPP -DTEXTURE_CACHE_8BPP +ifeq "$(HAVE_NEON)" "1" +test_neon: SRC += ../gpu_neon/psx_gpu/psx_gpu_arm_neon.S +test_neon: CFLAGS += -DNEON_BUILD +else +test_neon: CFLAGS += -fno-strict-aliasing +endif +test_peops: SRC += ../dfxvideo/gpulib_if.c +test_peops: CFLAGS += -fno-strict-aliasing +test_unai: SRC += ../gpu_unai/gpulib_if.cpp +test_unai: CC_ = $(CXX) +ifeq "$(ARCH)" "arm" +test_unai: SRC += ../gpu_unai/gpu_arm.s +endif + +$(TARGETS): $(SRC) + $(CC_) -o $@ $(SRC) $(CFLAGS) $(LDFLAGS) + +clean: + $(RM) $(TARGETS) diff --git a/plugins/gpulib/cspace.c b/plugins/gpulib/cspace.c new file mode 100644 index 0000000..eee56ce --- /dev/null +++ b/plugins/gpulib/cspace.c @@ -0,0 +1,21 @@ +#include "cspace.h" + +void bgr555_to_rgb565(void *dst_, const void *src_, int bytes) +{ + unsigned int *src = (unsigned int *)src_; + unsigned int *dst = (unsigned int *)dst_; + unsigned int p; + int x; + + for (x = 0; x < bytes / 4; x++) { + p = src[x]; + p = ((p & 0x7c007c00) >> 10) | ((p & 0x03e003e0) << 1) + | ((p & 0x001f001f) << 11); + dst[x] = p; + } +} + +// TODO? +void bgr888_to_rgb888(void *dst, const void *src, int bytes) {} +void bgr888_to_rgb565(void *dst, const void *src, int bytes) {} + diff --git a/plugins/gpulib/cspace.h b/plugins/gpulib/cspace.h new file mode 100644 index 0000000..644143b --- /dev/null +++ b/plugins/gpulib/cspace.h @@ -0,0 +1,12 @@ +#ifdef __cplusplus +extern "C" +{ +#endif + +void bgr555_to_rgb565(void *dst, const void *src, int bytes); +void bgr888_to_rgb888(void *dst, const void *src, int bytes); +void bgr888_to_rgb565(void *dst, const void *src, int bytes); + +#ifdef __cplusplus +} +#endif diff --git a/plugins/gpulib/cspace_neon.s b/plugins/gpulib/cspace_neon.s new file mode 100644 index 0000000..abc0381 --- /dev/null +++ b/plugins/gpulib/cspace_neon.s @@ -0,0 +1,132 @@ +/* + * (C) Gražvydas "notaz" Ignotas, 2010 + * + * This work is licensed under the terms of any of these licenses + * (at your option): + * - GNU GPL, version 2 or later. + * - GNU LGPL, version 2.1 or later. + * See the COPYING file in the top-level directory. + */ + +.text +.align 2 + +.global bgr555_to_rgb565 +bgr555_to_rgb565: + pld [r1] + mov r3, #0x07c0 + vdup.16 q15, r3 + subs r2, r2, #64 + blt btr16_end64 +0: + pld [r1, #64*2] + vldmia r1!, {q0-q3} + vshl.u16 q4, q0, #11 + vshl.u16 q5, q1, #11 + vshl.u16 q6, q2, #11 + vshl.u16 q7, q3, #11 + vsri.u16 q4, q0, #10 + vsri.u16 q5, q1, #10 + vsri.u16 q6, q2, #10 + vsri.u16 q7, q3, #10 + vshl.u16 q0, q0, #1 + vshl.u16 q1, q1, #1 + vshl.u16 q2, q2, #1 + vshl.u16 q3, q3, #1 + vbit q4, q0, q15 + vbit q5, q1, q15 + vbit q6, q2, q15 + vbit q7, q3, q15 + vstmia r0!, {q4-q7} + subs r2, r2, #64 + bge 0b + +btr16_end64: + adds r2, r2, #64 + bxeq lr + subs r2, r2, #16 + blt btr16_end16 + + @ handle the remainder (reasonably rare) +0: + vld1.16 {q0}, [r1]! + vshl.u16 q1, q0, #11 + vshl.u16 q2, q0, #1 + vsri.u16 q1, q0, #10 + vbit q1, q2, q15 + subs r2, r2, #16 + vst1.16 {q1}, [r0]! + bge 0b + +btr16_end16: + adds r2, r2, #16 + bxeq lr + subs r2, r2, #8 + bxlt lr + + @ very rare + vld1.16 d0, [r1]! + vshl.u16 d1, d0, #11 + vshl.u16 d2, d0, #1 + vsri.u16 d1, d0, #10 + vbit d1, d2, d30 + vst1.16 d1, [r0]! + bx lr + + +.global bgr888_to_rgb888 +bgr888_to_rgb888: + pld [r1] + @ r2 /= 48 + mov r2, r2, lsr #4 + movw r3, #0x5556 + movt r3, #0x5555 + umull r12,r2, r3, r2 +0: + pld [r1, #48*3] + vld3.8 {d0-d2}, [r1, :64]! + vld3.8 {d3-d5}, [r1, :64]! + vswp d0, d2 + vswp d3, d5 + vst3.8 {d0-d2}, [r0, :64]! + vst3.8 {d3-d5}, [r0, :64]! + subs r2, r2, #1 + bne 0b + + bx lr + + +.global bgr888_to_rgb565 +bgr888_to_rgb565: + pld [r1] + @ r2 /= 48 + mov r2, r2, lsr #4 + movw r3, #0x5556 + movt r3, #0x5555 + umull r12,r2, r3, r2 + + mov r3, #0x07e0 + vdup.16 q15, r3 +0: + pld [r1, #48*3] + vld3.8 {d1-d3}, [r1, :64]! + vld3.8 {d5-d7}, [r1, :64]! + + vshll.u8 q8, d2, #3 @ g + vshll.u8 q9, d6, #3 + vshr.u8 d0, d3, #3 @ b + vshr.u8 d4, d7, #3 + vzip.8 d0, d1 @ rb + vzip.8 d4, d5 + vbit q0, q8, q15 + vbit q2, q9, q15 + + vstmia r0!, {d0,d1} + vstmia r0!, {d4,d5} + subs r2, r2, #1 + bne 0b + + bx lr + + +@ vim:filetype=armasm diff --git a/plugins/gpulib/gpu.c b/plugins/gpulib/gpu.c new file mode 100644 index 0000000..5b16cd4 --- /dev/null +++ b/plugins/gpulib/gpu.c @@ -0,0 +1,636 @@ +/* + * (C) Gražvydas "notaz" Ignotas, 2011-2012 + * + * This work is licensed under the terms of any of these licenses + * (at your option): + * - GNU GPL, version 2 or later. + * - GNU LGPL, version 2.1 or later. + * See the COPYING file in the top-level directory. + */ + +#include +#include +#include "gpu.h" + +#define ARRAY_SIZE(x) (sizeof(x) / sizeof((x)[0])) +#define unlikely(x) __builtin_expect((x), 0) +#define noinline __attribute__((noinline)) + +#define gpu_log(fmt, ...) \ + printf("%d:%03d: " fmt, *gpu.state.frame_count, *gpu.state.hcnt, ##__VA_ARGS__) + +//#define log_io gpu_log +#define log_io(...) +//#define log_anomaly gpu_log +#define log_anomaly(...) + +struct psx_gpu gpu __attribute__((aligned(2048))); + +static noinline int do_cmd_buffer(uint32_t *data, int count); +static void finish_vram_transfer(int is_read); + +static noinline void do_cmd_reset(void) +{ + if (unlikely(gpu.cmd_len > 0)) + do_cmd_buffer(gpu.cmd_buffer, gpu.cmd_len); + gpu.cmd_len = 0; + + if (unlikely(gpu.dma.h > 0)) + finish_vram_transfer(gpu.dma_start.is_read); + gpu.dma.h = 0; +} + +static noinline void do_reset(void) +{ + unsigned int i; + + do_cmd_reset(); + + memset(gpu.regs, 0, sizeof(gpu.regs)); + for (i = 0; i < sizeof(gpu.ex_regs) / sizeof(gpu.ex_regs[0]); i++) + gpu.ex_regs[i] = (0xe0 + i) << 24; + gpu.status.reg = 0x14802000; + gpu.gp0 = 0; + gpu.regs[3] = 1; + gpu.screen.hres = gpu.screen.w = 256; + gpu.screen.vres = gpu.screen.h = 240; +} + +static noinline void update_width(void) +{ + int sw = gpu.screen.x2 - gpu.screen.x1; + if (sw <= 0 || sw >= 2560) + // full width + gpu.screen.w = gpu.screen.hres; + else + gpu.screen.w = sw * gpu.screen.hres / 2560; +} + +static noinline void update_height(void) +{ + int sh = gpu.screen.y2 - gpu.screen.y1; + if (gpu.status.dheight) + sh *= 2; + if (sh <= 0) + sh = gpu.screen.vres; + + gpu.screen.h = sh; +} + +static noinline void decide_frameskip(void) +{ + if (gpu.frameskip.active) + gpu.frameskip.cnt++; + else { + gpu.frameskip.cnt = 0; + gpu.frameskip.frame_ready = 1; + } + + if (!gpu.frameskip.active && *gpu.frameskip.advice) + gpu.frameskip.active = 1; + else if (gpu.frameskip.set > 0 && gpu.frameskip.cnt < gpu.frameskip.set) + gpu.frameskip.active = 1; + else + gpu.frameskip.active = 0; +} + +static noinline void decide_frameskip_allow(uint32_t cmd_e3) +{ + // no frameskip if it decides to draw to display area, + // but not for interlace since it'll most likely always do that + uint32_t x = cmd_e3 & 0x3ff; + uint32_t y = (cmd_e3 >> 10) & 0x3ff; + gpu.frameskip.allow = gpu.status.interlace || + (uint32_t)(x - gpu.screen.x) >= (uint32_t)gpu.screen.w || + (uint32_t)(y - gpu.screen.y) >= (uint32_t)gpu.screen.h; +} + +static noinline void get_gpu_info(uint32_t data) +{ + switch (data & 0x0f) { + case 0x02: + case 0x03: + case 0x04: + case 0x05: + gpu.gp0 = gpu.ex_regs[data & 7] & 0xfffff; + break; + case 0x06: + gpu.gp0 = gpu.ex_regs[5] & 0xfffff; + break; + case 0x07: + gpu.gp0 = 2; + break; + default: + gpu.gp0 = 0; + break; + } +} + +long GPUinit(void) +{ + int ret; + ret = vout_init(); + ret |= renderer_init(); + + gpu.state.frame_count = &gpu.zero; + gpu.state.hcnt = &gpu.zero; + gpu.frameskip.active = 0; + gpu.cmd_len = 0; + do_reset(); + + return ret; +} + +long GPUshutdown(void) +{ + return vout_finish(); +} + +void GPUwriteStatus(uint32_t data) +{ + static const short hres[8] = { 256, 368, 320, 384, 512, 512, 640, 640 }; + static const short vres[4] = { 240, 480, 256, 480 }; + uint32_t cmd = data >> 24; + + if (cmd < ARRAY_SIZE(gpu.regs)) { + if (cmd > 1 && cmd != 5 && gpu.regs[cmd] == data) + return; + gpu.regs[cmd] = data; + } + + gpu.state.fb_dirty = 1; + + switch (cmd) { + case 0x00: + do_reset(); + break; + case 0x01: + do_cmd_reset(); + break; + case 0x03: + gpu.status.blanking = data & 1; + break; + case 0x04: + gpu.status.dma = data & 3; + break; + case 0x05: + gpu.screen.x = data & 0x3ff; + gpu.screen.y = (data >> 10) & 0x3ff; + if (gpu.frameskip.set) { + decide_frameskip_allow(gpu.ex_regs[3]); + if (gpu.frameskip.last_flip_frame != *gpu.state.frame_count) { + decide_frameskip(); + gpu.frameskip.last_flip_frame = *gpu.state.frame_count; + } + } + break; + case 0x06: + gpu.screen.x1 = data & 0xfff; + gpu.screen.x2 = (data >> 12) & 0xfff; + update_width(); + break; + case 0x07: + gpu.screen.y1 = data & 0x3ff; + gpu.screen.y2 = (data >> 10) & 0x3ff; + update_height(); + break; + case 0x08: + gpu.status.reg = (gpu.status.reg & ~0x7f0000) | ((data & 0x3F) << 17) | ((data & 0x40) << 10); + gpu.screen.hres = hres[(gpu.status.reg >> 16) & 7]; + gpu.screen.vres = vres[(gpu.status.reg >> 19) & 3]; + update_width(); + update_height(); + break; + default: + if ((cmd & 0xf0) == 0x10) + get_gpu_info(data); + break; + } +} + +const unsigned char cmd_lengths[256] = +{ + 0, 0, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, + 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, + 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, + 2, 2, 2, 2, 3, 3, 3, 3, 1, 1, 1, 1, 0, 0, 0, 0, // 60 + 1, 1, 1, 1, 2, 2, 2, 2, 1, 1, 1, 1, 2, 2, 2, 2, + 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, + 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, + 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, + 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 +}; + +#define VRAM_MEM_XY(x, y) &gpu.vram[(y) * 1024 + (x)] + +static inline void do_vram_line(int x, int y, uint16_t *mem, int l, int is_read) +{ + uint16_t *vram = VRAM_MEM_XY(x, y); + if (is_read) + memcpy(mem, vram, l * 2); + else + memcpy(vram, mem, l * 2); +} + +static int do_vram_io(uint32_t *data, int count, int is_read) +{ + int count_initial = count; + uint16_t *sdata = (uint16_t *)data; + int x = gpu.dma.x, y = gpu.dma.y; + int w = gpu.dma.w, h = gpu.dma.h; + int o = gpu.dma.offset; + int l; + count *= 2; // operate in 16bpp pixels + + if (gpu.dma.offset) { + l = w - gpu.dma.offset; + if (count < l) + l = count; + + do_vram_line(x + o, y, sdata, l, is_read); + + if (o + l < w) + o += l; + else { + o = 0; + y++; + h--; + } + sdata += l; + count -= l; + } + + for (; h > 0 && count >= w; sdata += w, count -= w, y++, h--) { + y &= 511; + do_vram_line(x, y, sdata, w, is_read); + } + + if (h > 0) { + if (count > 0) { + y &= 511; + do_vram_line(x, y, sdata, count, is_read); + o = count; + count = 0; + } + } + else + finish_vram_transfer(is_read); + gpu.dma.y = y; + gpu.dma.h = h; + gpu.dma.offset = o; + + return count_initial - count / 2; +} + +static void start_vram_transfer(uint32_t pos_word, uint32_t size_word, int is_read) +{ + if (gpu.dma.h) + log_anomaly("start_vram_transfer while old unfinished\n"); + + gpu.dma.x = pos_word & 0x3ff; + gpu.dma.y = (pos_word >> 16) & 0x1ff; + gpu.dma.w = ((size_word - 1) & 0x3ff) + 1; + gpu.dma.h = (((size_word >> 16) - 1) & 0x1ff) + 1; + gpu.dma.offset = 0; + gpu.dma.is_read = is_read; + gpu.dma_start = gpu.dma; + + renderer_flush_queues(); + if (is_read) { + gpu.status.img = 1; + // XXX: wrong for width 1 + memcpy(&gpu.gp0, VRAM_MEM_XY(gpu.dma.x, gpu.dma.y), 4); + gpu.state.last_vram_read_frame = *gpu.state.frame_count; + } + + log_io("start_vram_transfer %c (%d, %d) %dx%d\n", is_read ? 'r' : 'w', + gpu.dma.x, gpu.dma.y, gpu.dma.w, gpu.dma.h); +} + +static void finish_vram_transfer(int is_read) +{ + if (is_read) + gpu.status.img = 0; + else + renderer_update_caches(gpu.dma_start.x, gpu.dma_start.y, + gpu.dma_start.w, gpu.dma_start.h); +} + +static noinline int do_cmd_buffer(uint32_t *data, int count) +{ + int len, cmd, start, pos; + int vram_dirty = 0; + + // process buffer + for (start = pos = 0; pos < count; ) + { + cmd = -1; + len = 0; + + if (gpu.dma.h) { + pos += do_vram_io(data + pos, count - pos, 0); + if (pos == count) + break; + start = pos; + } + + // do look-ahead pass to detect SR changes and VRAM i/o + while (pos < count) { + uint32_t *list = data + pos; + cmd = list[0] >> 24; + len = 1 + cmd_lengths[cmd]; + + //printf(" %3d: %02x %d\n", pos, cmd, len); + if ((cmd & 0xf4) == 0x24) { + // flat textured prim + gpu.ex_regs[1] &= ~0x1ff; + gpu.ex_regs[1] |= list[4] & 0x1ff; + } + else if ((cmd & 0xf4) == 0x34) { + // shaded textured prim + gpu.ex_regs[1] &= ~0x1ff; + gpu.ex_regs[1] |= list[5] & 0x1ff; + } + else if (cmd == 0xe3) + decide_frameskip_allow(list[0]); + + if (2 <= cmd && cmd < 0xc0) + vram_dirty = 1; + else if ((cmd & 0xf8) == 0xe0) + gpu.ex_regs[cmd & 7] = list[0]; + + if (pos + len > count) { + cmd = -1; + break; // incomplete cmd + } + if (cmd == 0xa0 || cmd == 0xc0) + break; // image i/o + pos += len; + } + + if (pos - start > 0) { + if (!gpu.frameskip.active || !gpu.frameskip.allow) + do_cmd_list(data + start, pos - start); + start = pos; + } + + if (cmd == 0xa0 || cmd == 0xc0) { + // consume vram write/read cmd + start_vram_transfer(data[pos + 1], data[pos + 2], cmd == 0xc0); + pos += len; + } + else if (cmd == -1) + break; + } + + gpu.status.reg &= ~0x1fff; + gpu.status.reg |= gpu.ex_regs[1] & 0x7ff; + gpu.status.reg |= (gpu.ex_regs[6] & 3) << 11; + + if (gpu.frameskip.active) + renderer_sync_ecmds(gpu.ex_regs); + gpu.state.fb_dirty |= vram_dirty; + + return count - pos; +} + +static void flush_cmd_buffer(void) +{ + int left = do_cmd_buffer(gpu.cmd_buffer, gpu.cmd_len); + if (left > 0) + memmove(gpu.cmd_buffer, gpu.cmd_buffer + gpu.cmd_len - left, left * 4); + gpu.cmd_len = left; +} + +void GPUwriteDataMem(uint32_t *mem, int count) +{ + int left; + + log_io("gpu_dma_write %p %d\n", mem, count); + + if (unlikely(gpu.cmd_len > 0)) + flush_cmd_buffer(); + + left = do_cmd_buffer(mem, count); + if (left) + log_anomaly("GPUwriteDataMem: discarded %d/%d words\n", left, count); +} + +void GPUwriteData(uint32_t data) +{ + log_io("gpu_write %08x\n", data); + gpu.cmd_buffer[gpu.cmd_len++] = data; + if (gpu.cmd_len >= CMD_BUFFER_LEN) + flush_cmd_buffer(); +} + +long GPUdmaChain(uint32_t *rambase, uint32_t start_addr) +{ + uint32_t addr, *list; + uint32_t *llist_entry = NULL; + int len, left, count; + long cpu_cycles = 0; + + if (unlikely(gpu.cmd_len > 0)) + flush_cmd_buffer(); + + // ff7 sends it's main list twice, detect this + if (*gpu.state.frame_count == gpu.state.last_list.frame && + *gpu.state.hcnt - gpu.state.last_list.hcnt <= 1 && + gpu.state.last_list.cycles > 2048) + { + llist_entry = rambase + (gpu.state.last_list.addr & 0x1fffff) / 4; + *llist_entry |= 0x800000; + } + + log_io("gpu_dma_chain\n"); + addr = start_addr & 0xffffff; + for (count = 0; addr != 0xffffff; count++) + { + list = rambase + (addr & 0x1fffff) / 4; + len = list[0] >> 24; + addr = list[0] & 0xffffff; + cpu_cycles += 10; + if (len > 0) + cpu_cycles += 5 + len; + + log_io(".chain %08x #%d\n", (list - rambase) * 4, len); + + // loop detection marker + // (bit23 set causes DMA error on real machine, so + // unlikely to be ever set by the game) + list[0] |= 0x800000; + + if (len) { + left = do_cmd_buffer(list + 1, len); + if (left) + log_anomaly("GPUdmaChain: discarded %d/%d words\n", left, len); + } + + if (addr & 0x800000) + break; + } + + // remove loop detection markers + addr = start_addr & 0x1fffff; + while (count-- > 0) { + list = rambase + addr / 4; + addr = list[0] & 0x1fffff; + list[0] &= ~0x800000; + } + if (llist_entry) + *llist_entry &= ~0x800000; + + gpu.state.last_list.frame = *gpu.state.frame_count; + gpu.state.last_list.hcnt = *gpu.state.hcnt; + gpu.state.last_list.cycles = cpu_cycles; + gpu.state.last_list.addr = start_addr; + + return cpu_cycles; +} + +void GPUreadDataMem(uint32_t *mem, int count) +{ + log_io("gpu_dma_read %p %d\n", mem, count); + + if (unlikely(gpu.cmd_len > 0)) + flush_cmd_buffer(); + + if (gpu.dma.h) + do_vram_io(mem, count, 1); +} + +uint32_t GPUreadData(void) +{ + uint32_t ret; + + if (unlikely(gpu.cmd_len > 0)) + flush_cmd_buffer(); + + ret = gpu.gp0; + if (gpu.dma.h) + do_vram_io(&ret, 1, 1); + + log_io("gpu_read %08x\n", ret); + return ret; +} + +uint32_t GPUreadStatus(void) +{ + uint32_t ret; + + if (unlikely(gpu.cmd_len > 0)) + flush_cmd_buffer(); + + ret = gpu.status.reg; + log_io("gpu_read_status %08x\n", ret); + return ret; +} + +struct GPUFreeze +{ + uint32_t ulFreezeVersion; // should be always 1 for now (set by main emu) + uint32_t ulStatus; // current gpu status + uint32_t ulControl[256]; // latest control register values + unsigned char psxVRam[1024*1024*2]; // current VRam image (full 2 MB for ZN) +}; + +long GPUfreeze(uint32_t type, struct GPUFreeze *freeze) +{ + int i; + + switch (type) { + case 1: // save + if (gpu.cmd_len > 0) + flush_cmd_buffer(); + memcpy(freeze->psxVRam, gpu.vram, sizeof(gpu.vram)); + memcpy(freeze->ulControl, gpu.regs, sizeof(gpu.regs)); + memcpy(freeze->ulControl + 0xe0, gpu.ex_regs, sizeof(gpu.ex_regs)); + freeze->ulStatus = gpu.status.reg; + break; + case 0: // load + memcpy(gpu.vram, freeze->psxVRam, sizeof(gpu.vram)); + memcpy(gpu.regs, freeze->ulControl, sizeof(gpu.regs)); + memcpy(gpu.ex_regs, freeze->ulControl + 0xe0, sizeof(gpu.ex_regs)); + gpu.status.reg = freeze->ulStatus; + for (i = 8; i > 0; i--) { + gpu.regs[i] ^= 1; // avoid reg change detection + GPUwriteStatus((i << 24) | (gpu.regs[i] ^ 1)); + } + renderer_sync_ecmds(gpu.ex_regs); + renderer_update_caches(0, 0, 1024, 512); + break; + } + + return 1; +} + +void GPUupdateLace(void) +{ + if (gpu.cmd_len > 0) + flush_cmd_buffer(); + renderer_flush_queues(); + + if (gpu.status.blanking || !gpu.state.fb_dirty) + return; + + if (gpu.frameskip.set) { + if (!gpu.frameskip.frame_ready) { + if (*gpu.state.frame_count - gpu.frameskip.last_flip_frame < 9) + return; + gpu.frameskip.active = 0; + } + gpu.frameskip.frame_ready = 0; + } + + vout_update(); + gpu.state.fb_dirty = 0; +} + +void GPUvBlank(int is_vblank, int lcf) +{ + int interlace = gpu.state.allow_interlace + && gpu.status.interlace && gpu.status.dheight; + // interlace doesn't look nice on progressive displays, + // so we have this "auto" mode here for games that don't read vram + if (gpu.state.allow_interlace == 2 + && *gpu.state.frame_count - gpu.state.last_vram_read_frame > 1) + { + interlace = 0; + } + if (interlace || interlace != gpu.state.old_interlace) { + gpu.state.old_interlace = interlace; + + if (gpu.cmd_len > 0) + flush_cmd_buffer(); + renderer_flush_queues(); + renderer_set_interlace(interlace, !lcf); + } +} + +#include "../../frontend/plugin_lib.h" + +void GPUrearmedCallbacks(const struct rearmed_cbs *cbs) +{ + gpu.frameskip.set = cbs->frameskip; + gpu.frameskip.advice = &cbs->fskip_advice; + gpu.frameskip.active = 0; + gpu.frameskip.frame_ready = 1; + gpu.state.hcnt = cbs->gpu_hcnt; + gpu.state.frame_count = cbs->gpu_frame_count; + gpu.state.allow_interlace = cbs->gpu_neon.allow_interlace; + + if (cbs->pl_vout_set_raw_vram) + cbs->pl_vout_set_raw_vram(gpu.vram); + renderer_set_config(cbs); + vout_set_config(cbs); +} + +// vim:shiftwidth=2:expandtab diff --git a/plugins/gpulib/gpu.h b/plugins/gpulib/gpu.h new file mode 100644 index 0000000..d9ad416 --- /dev/null +++ b/plugins/gpulib/gpu.h @@ -0,0 +1,131 @@ +/* + * (C) Gražvydas "notaz" Ignotas, 2011 + * + * This work is licensed under the terms of any of these licenses + * (at your option): + * - GNU GPL, version 2 or later. + * - GNU LGPL, version 2.1 or later. + * See the COPYING file in the top-level directory. + */ + +#include + +#ifdef __cplusplus +extern "C" { +#endif + +#define CMD_BUFFER_LEN 1024 + +struct psx_gpu { + uint16_t vram[1024 * 512]; + uint16_t guard[1024 * 512]; // overdraw guard + uint32_t cmd_buffer[CMD_BUFFER_LEN]; + uint32_t regs[16]; + union { + uint32_t reg; + struct { + uint32_t tx:4; // 0 texture page + uint32_t ty:1; + uint32_t abr:2; + uint32_t tp:2; // 7 t.p. mode (4,8,15bpp) + uint32_t dtd:1; // 9 dither + uint32_t dfe:1; + uint32_t md:1; // 11 set mask bit when drawing + uint32_t me:1; // 12 no draw on mask + uint32_t unkn:3; + uint32_t width1:1; // 16 + uint32_t width0:2; + uint32_t dheight:1; // 19 double height + uint32_t video:1; // 20 NTSC,PAL + uint32_t rgb24:1; + uint32_t interlace:1; // 22 interlace on + uint32_t blanking:1; // 23 display not enabled + uint32_t unkn2:2; + uint32_t busy:1; // 26 !busy drawing + uint32_t img:1; // 27 ready to DMA image data + uint32_t com:1; // 28 ready for commands + uint32_t dma:2; // 29 off, ?, to vram, from vram + uint32_t lcf:1; // 31 + }; + } status; + uint32_t gp0; + uint32_t ex_regs[8]; + struct { + int hres, vres; + int x, y, w, h; + int x1, x2; + int y1, y2; + } screen; + struct { + int x, y, w, h; + short int offset, is_read; + } dma, dma_start; + int cmd_len; + uint32_t zero; + struct { + uint32_t fb_dirty:1; + uint32_t old_interlace:1; + uint32_t allow_interlace:2; + uint32_t *frame_count; + uint32_t *hcnt; /* hsync count */ + struct { + uint32_t addr; + uint32_t cycles; + uint32_t frame; + uint32_t hcnt; + } last_list; + uint32_t last_vram_read_frame; + } state; + struct { + int32_t set:3; /* -1 auto, 0 off, 1-3 fixed */ + int32_t cnt:3; /* amount skipped in a row */ + uint32_t active:1; + uint32_t allow:1; + uint32_t frame_ready:1; + const int *advice; + uint32_t last_flip_frame; + } frameskip; +}; + +extern struct psx_gpu gpu; + +extern const unsigned char cmd_lengths[256]; + +void do_cmd_list(uint32_t *list, int count); + +struct rearmed_cbs; + +int renderer_init(void); +void renderer_sync_ecmds(uint32_t * ecmds); +void renderer_update_caches(int x, int y, int w, int h); +void renderer_flush_queues(void); +void renderer_set_interlace(int enable, int is_odd); +void renderer_set_config(const struct rearmed_cbs *config); + +int vout_init(void); +int vout_finish(void); +void vout_update(void); +void vout_set_config(const struct rearmed_cbs *config); + +/* listing these here for correct linkage if rasterizer uses c++ */ +struct GPUFreeze; + +long GPUinit(void); +long GPUshutdown(void); +void GPUwriteDataMem(uint32_t *mem, int count); +long GPUdmaChain(uint32_t *rambase, uint32_t addr); +void GPUwriteData(uint32_t data); +void GPUreadDataMem(uint32_t *mem, int count); +uint32_t GPUreadData(void); +uint32_t GPUreadStatus(void); +void GPUwriteStatus(uint32_t data); +long GPUfreeze(uint32_t type, struct GPUFreeze *freeze); +void GPUupdateLace(void); +long GPUopen(void **dpy); +long GPUclose(void); +void GPUvBlank(int is_vblank, int lcf); +void GPUrearmedCallbacks(const struct rearmed_cbs *cbs_); + +#ifdef __cplusplus +} +#endif diff --git a/plugins/gpulib/gpulib.mak b/plugins/gpulib/gpulib.mak new file mode 100644 index 0000000..cec6f5a --- /dev/null +++ b/plugins/gpulib/gpulib.mak @@ -0,0 +1,54 @@ +# depends on ARCH definition +# always adding gpulib to LDLIBS in case cspace is needed + +LDFLAGS += -shared +ifeq "$(ARCH)" "arm" + ARM_CORTEXA8 ?= 1 + ifeq "$(ARM_CORTEXA8)" "1" + CFLAGS += -mcpu=cortex-a8 -mtune=cortex-a8 -mfpu=neon -mfloat-abi=softfp + ASFLAGS += -mcpu=cortex-a8 -mfpu=neon + else + CFLAGS += -mcpu=arm926ej-s -mtune=arm926ej-s + ASFLAGS += -mcpu=arm926ej-s -mfloat-abi=softfp + endif + EXT = +else + CFLAGS += -m32 + LDFLAGS += -m32 + LDLIBS_GPULIB += `sdl-config --libs` + EXT = .x86 +endif +ifdef MAEMO + CFLAGS += -DMAEMO +endif +ifdef DEBUG + CFLAGS += -O0 +endif + +GPULIB_A = ../gpulib/gpulib$(EXT).a +LDLIBS += $(GPULIB_A) + +ifdef BIN_STANDLALONE +TARGETS += $(BIN_STANDLALONE)$(EXT) +endif +ifdef BIN_GPULIB +TARGETS += $(BIN_GPULIB)$(EXT) +endif + +all: $(GPULIB_A) $(TARGETS) + +ifdef BIN_STANDLALONE +$(BIN_STANDLALONE)$(EXT): $(SRC) $(SRC_STANDALONE) + $(CC) -o $@ $^ $(CFLAGS) $(LDFLAGS) $(LDLIBS) $(LDLIBS_STANDALONE) +endif + +ifdef BIN_GPULIB +$(BIN_GPULIB)$(EXT): $(SRC) $(SRC_GPULIB) + $(CC) -o $@ $^ $(CFLAGS) $(LDFLAGS) $(LDLIBS) $(LDLIBS_GPULIB) +endif + +$(GPULIB_A): + make -C ../gpulib/ all + +clean: + $(RM) $(TARGETS) diff --git a/plugins/gpulib/test.c b/plugins/gpulib/test.c new file mode 100644 index 0000000..e523e20 --- /dev/null +++ b/plugins/gpulib/test.c @@ -0,0 +1,109 @@ +#include +#include +#include +#include "gpu.h" + +static inline unsigned int pcnt_get(void) +{ + unsigned int val; +#ifdef __ARM_ARCH_7A__ + asm volatile("mrc p15, 0, %0, c9, c13, 0" : "=r"(val)); +#else + val = 0; +#endif + return val; +} + +static inline void pcnt_init(void) +{ +#ifdef __ARM_ARCH_7A__ + int v; + asm volatile("mrc p15, 0, %0, c9, c12, 0" : "=r"(v)); + v |= 5; // master enable, ccnt reset + v &= ~8; // ccnt divider 0 + asm volatile("mcr p15, 0, %0, c9, c12, 0" :: "r"(v)); + // enable cycle counter + asm volatile("mcr p15, 0, %0, c9, c12, 1" :: "r"(1<<31)); +#endif +} + +const unsigned char cmd_lengths[256] = +{ + 0, 0, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, + 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, + 2, 2, 2, 2, 0, 0, 0, 0, 3, 3, 3, 3, 3, 3, 3, 3, // 40 + 3, 3, 3, 3, 0, 0, 0, 0, 4, 4, 4, 4, 4, 4, 4, 4, + 2, 2, 2, 2, 3, 3, 3, 3, 1, 1, 1, 1, 2, 2, 2, 2, // 60 + 1, 1, 1, 1, 2, 2, 2, 2, 1, 1, 1, 1, 2, 2, 2, 2, + 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, + 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, + 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, + 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 +}; + +struct psx_gpu gpu __attribute__((aligned(64))); + +typedef struct +{ + uint16_t vram[1024 * 512]; + uint32_t gpu_register[15]; + uint32_t status; +} gpu_dump_struct; + +static gpu_dump_struct state; + +int main(int argc, char *argv[]) +{ + unsigned int start_cycles; + uint32_t *list; + int size; + FILE *state_file; + FILE *list_file; + FILE *out_file; + + if (argc != 3 && argc != 4) + { + printf("usage:\n%s [vram_out]\n", argv[0]); + return 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); + + list = (uint32_t *)malloc(size); + fread(list, 1, size, list_file); + fclose(list_file); + + pcnt_init(); + renderer_init(); + memcpy(gpu.vram, state.vram, sizeof(gpu.vram)); + if ((state.gpu_register[8] & 0x24) == 0x24) + renderer_set_interlace(1, !(state.status >> 31)); + + start_cycles = pcnt_get(); + + do_cmd_list(list, size / 4); + renderer_flush_queues(); + + printf("%u\n", pcnt_get() - start_cycles); + + if (argc >= 4) { + out_file = fopen(argv[3], "wb"); + fwrite(gpu.vram, 1, sizeof(gpu.vram), out_file); + fclose(out_file); + } + + return 0; +} diff --git a/plugins/gpulib/vout_fb.c b/plugins/gpulib/vout_fb.c new file mode 100644 index 0000000..021efa7 --- /dev/null +++ b/plugins/gpulib/vout_fb.c @@ -0,0 +1,121 @@ +/* + * (C) Gražvydas "notaz" Ignotas, 2011 + * + * This work is licensed under the terms of any of these licenses + * (at your option): + * - GNU GPL, version 2 or later. + * - GNU LGPL, version 2.1 or later. + * See the COPYING file in the top-level directory. + */ + +#include "gpu.h" +#include "cspace.h" +#include "../../frontend/plugin_lib.h" + +static const struct rearmed_cbs *cbs; +static void *screen_buf; + +int vout_init(void) +{ + return 0; +} + +int vout_finish(void) +{ + return 0; +} + +static void check_mode_change(void) +{ + static uint32_t old_status; + static int old_h; + + // width|rgb24 change? + if ((gpu.status.reg ^ old_status) & ((7<<16)|(1<<21)) || gpu.screen.h != old_h) + { + old_status = gpu.status.reg; + old_h = gpu.screen.h; + screen_buf = cbs->pl_vout_set_mode(gpu.screen.hres, + gpu.screen.h, gpu.status.rgb24 ? 24 : 16); + } +} + +static void blit(void) +{ + int x = gpu.screen.x & ~1; // alignment needed by blitter + int y = gpu.screen.y; + int w = gpu.screen.w; + int h = gpu.screen.h; + uint16_t *vram = gpu.vram; + int stride = gpu.screen.hres; + int fb_offs, doffs; + uint8_t *dest; + + fb_offs = y * 1024 + x; + dest = (uint8_t *)screen_buf; + + // only do centering, at least for now + doffs = (stride - w) / 2 & ~1; + + if (gpu.status.rgb24) + { +#ifndef MAEMO + dest += (doffs / 8) * 24; + for (; h-- > 0; dest += stride * 3, fb_offs += 1024) + { + fb_offs &= 1024*512-1; + bgr888_to_rgb888(dest, vram + fb_offs, w * 3); + } +#else + dest += doffs * 2; + for (; h-- > 0; dest += stride * 2, fb_offs += 1024) + { + fb_offs &= 1024*512-1; + bgr888_to_rgb565(dest, vram + fb_offs, w * 3); + } +#endif + } + else + { + dest += doffs * 2; + for (; h-- > 0; dest += stride * 2, fb_offs += 1024) + { + fb_offs &= 1024*512-1; + bgr555_to_rgb565(dest, vram + fb_offs, w * 2); + } + } + + screen_buf = cbs->pl_vout_flip(); +} + +void vout_update(void) +{ + check_mode_change(); + if (cbs->pl_vout_raw_flip) + cbs->pl_vout_raw_flip(gpu.screen.x, gpu.screen.y); + else + blit(); +} + +long GPUopen(void **unused) +{ + gpu.frameskip.active = 0; + gpu.frameskip.frame_ready = 1; + + cbs->pl_vout_open(); + screen_buf = cbs->pl_vout_flip(); + return 0; +} + +long GPUclose(void) +{ + cbs->pl_vout_close(); + return 0; +} + +void vout_set_config(const struct rearmed_cbs *cbs_) +{ + cbs = cbs_; +} + +// vim:shiftwidth=2:expandtab diff --git a/plugins/gpulib/vout_sdl.c b/plugins/gpulib/vout_sdl.c new file mode 100644 index 0000000..db1ae96 --- /dev/null +++ b/plugins/gpulib/vout_sdl.c @@ -0,0 +1,95 @@ +/* + * (C) Gražvydas "notaz" Ignotas, 2011 + * + * This work is licensed under the terms of any of these licenses + * (at your option): + * - GNU GPL, version 2 or later. + * - GNU LGPL, version 2.1 or later. + * See the COPYING file in the top-level directory. + */ + +#include +#include +#include +#include "gpu.h" + +static SDL_Surface *screen; +static Display *x11_display; + +int vout_init(void) +{ + SDL_SysWMinfo wminfo; + int ret; + + ret = SDL_Init(SDL_INIT_VIDEO); + if (ret != 0) { + fprintf(stderr, "SDL_Init failed: %s\n", SDL_GetError()); + return ret; + } + + screen = SDL_SetVideoMode(1024, 512, 32, 0); + if (screen == NULL) { + fprintf(stderr, "SDL_SetVideoMode failed: %s\n", SDL_GetError()); + SDL_Quit(); + return -1; + } + + SDL_VERSION(&wminfo.version); + ret = SDL_GetWMInfo(&wminfo); + if (ret == 1) + x11_display = wminfo.info.x11.display; + + return 0; +} + +int vout_finish(void) +{ + SDL_Quit(); + return 0; +} + +void vout_update(void) +{ + uint32_t *d; + int i; + + SDL_LockSurface(screen); + if (gpu.status.rgb24) + { + uint8_t *s; + int y; + for (y = 0; y < 512; y++) { + s = (uint8_t *)gpu.vram + y * 2*1024; + d = (uint32_t *)screen->pixels + y * 1024; + for (i = 0; i < 1024 * 2 / 3; i++, s += 3) + d[i] = (s[0] << 16) | (s[1] << 8) | s[2]; + } + } + else + { + uint16_t *s = gpu.vram; + d = (uint32_t *)screen->pixels; + for (i = 0; i < 1024 * 512; i++) + d[i] = (((uint32_t)s[i] << 19) & 0xf80000) | ((s[i] << 6) & 0xf800) | + ((s[i] >> 7) & 0xf8); + } + SDL_UnlockSurface(screen); + SDL_UpdateRect(screen, 0, 0, 1024, 512); +} + +long GPUopen(void **dpy) +{ + *dpy = x11_display; + return 0; +} + +long GPUclose(void) +{ + return 0; +} + +void vout_set_config(const struct rearmed_cbs *cbs) +{ +} + +// vim:shiftwidth=2:expandtab -- cgit v1.2.3