Merge branch 'enhancement'

Conflicts:
	frontend/libretro.c

14 files changed, 390 insertions, 75 deletions

diff --git a/frontend/common/plat.h b/frontend/common/plat.h
index 0a9fc0b..1fb8767 100644
--- a/frontend/common/plat.h
+++ b/frontend/common/plat.h
@@ -45,6 +45,10 @@ int  plat_is_dir(const char *path);
 int  plat_wait_event(int *fds_hnds, int count, int timeout_ms);
 void plat_sleep_ms(int ms);
 
+void *plat_mmap(unsigned long addr, size_t size, int need_exec, int is_fixed);
+void *plat_mremap(void *ptr, size_t oldsize, size_t newsize);
+void  plat_munmap(void *ptr, size_t size);
+
 /* timers, to be used for time diff and must refer to the same clock */
 unsigned int plat_get_ticks_ms(void);
 unsigned int plat_get_ticks_us(void);
diff --git a/frontend/libpicofe b/frontend/libpicofe
new file mode 160000
+Subproject 6ce097ba2f3cd1c269bacd032b775b6d296433f
diff --git a/frontend/libretro.c b/frontend/libretro.c
index 4305aa7..1eb2147 100644
--- a/frontend/libretro.c
+++ b/frontend/libretro.c
@@ -13,6 +13,7 @@
 #include "../libpcsxcore/psxcounters.h"
 #include "../libpcsxcore/new_dynarec/new_dynarec.h"
 #include "../plugins/dfsound/out.h"
+#include "../plugins/gpulib/cspace.h"
 #include "main.h"
 #include "plugin.h"
 #include "plugin_lib.h"
@@ -26,7 +27,6 @@ static retro_environment_t environ_cb;
 static retro_audio_sample_batch_t audio_batch_cb;
 
 static void *vout_buf;
-static int vout_width, vout_height;
 static int samples_sent, samples_to_send;
 static int plugins_opened;
 static int native_rgb565;
@@ -42,14 +42,10 @@ static int vout_open(void)
 	return 0;
 }
 
-static void *vout_set_mode(int w, int h, int bpp)
+static void vout_set_mode(int w, int h, int bpp)
 {
-	vout_width = w;
-	vout_height = h;
-	return vout_buf;
 }
 
-/* FIXME: either teach PCSX to blit to RGB1555 or RetroArch to support RGB565 */
 static void convert(void *buf, size_t bytes)
 {
 	unsigned int i, v, *p = buf;
@@ -60,14 +56,39 @@ static void convert(void *buf, size_t bytes)
 	}
 }
 
-static void *vout_flip(void)
+static void vout_flip(const void *vram, int stride, int bgr24, int w, int h)
 {
-	pl_rearmed_cbs.flip_cnt++;
-	if (!native_rgb565)
-		convert(vout_buf,  vout_width * vout_height * 2);
-	video_cb(vout_buf, vout_width, vout_height, vout_width * 2);
+	unsigned short *dest = vout_buf;
+	const unsigned short *src = vram;
+	int dstride = w, h1 = h;
+
+	if (vram == NULL) {
+		// blanking
+		memset(pl_vout_buf, 0, dstride * h * 2);
+		goto out;
+	}
 
-	return vout_buf;
+	if (bgr24)
+	{
+		// XXX: could we switch to RETRO_PIXEL_FORMAT_XRGB8888 here?
+		for (; h1-- > 0; dest += dstride, src += stride)
+		{
+			bgr888_to_rgb565(dest, src, w * 3);
+		}
+	}
+	else
+	{
+		for (; h1-- > 0; dest += dstride, src += stride)
+		{
+			bgr555_to_rgb565(dest, src, w * 2);
+		}
+	}
+
+out:
+	if (!native_rgb565)
+		convert(vout_buf, w * h * 2);
+	video_cb(vout_buf, w, h, w * 2);
+	pl_rearmed_cbs.flip_cnt++;
 }
 
 static void vout_close(void)
diff --git a/frontend/linux/plat.c b/frontend/linux/plat.c
index b7152b5..4ed1e65 100644
--- a/frontend/linux/plat.c
+++ b/frontend/linux/plat.c
@@ -17,9 +17,17 @@
 #include <time.h>
 #include <unistd.h>
 #include <sys/mman.h>
+#include <errno.h>
 
 #include "../common/plat.h"
 
+/* XXX: maybe unhardcode pagesize? */
+#define HUGETLB_PAGESIZE (2 * 1024 * 1024)
+#define HUGETLB_THRESHOLD (HUGETLB_PAGESIZE / 2)
+#ifndef MAP_HUGETLB
+#define MAP_HUGETLB 0x40000 /* arch specific */
+#endif
+
 
 int plat_is_dir(const char *path)
 {
@@ -126,16 +134,36 @@ int plat_wait_event(int *fds_hnds, int count, int timeout_ms)
 	return ret;
 }
 
-void *plat_mmap(unsigned long addr, size_t size)
+void *plat_mmap(unsigned long addr, size_t size, int need_exec, int is_fixed)
 {
+	static int hugetlb_disabled;
+	int prot = PROT_READ | PROT_WRITE;
+	int flags = MAP_PRIVATE | MAP_ANONYMOUS;
 	void *req, *ret;
 
 	req = (void *)addr;
-	ret = mmap(req, size, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0);
+	if (need_exec)
+		prot |= PROT_EXEC;
+	if (is_fixed)
+		flags |= MAP_FIXED;
+	if (size >= HUGETLB_THRESHOLD && !hugetlb_disabled)
+		flags |= MAP_HUGETLB;
+
+	ret = mmap(req, size, prot, flags, -1, 0);
+	if (ret == MAP_FAILED && (flags & MAP_HUGETLB)) {
+		fprintf(stderr,
+			"warning: failed to do hugetlb mmap (%p, %zu): %d\n",
+			req, size, errno);
+		hugetlb_disabled = 1;
+		flags &= ~MAP_HUGETLB;
+		ret = mmap(req, size, prot, flags, -1, 0);
+	}
 	if (ret == MAP_FAILED)
 		return NULL;
-	if (ret != req)
-		printf("warning: mmaped to %p, requested %p\n", ret, req);
+
+	if (req != NULL && ret != req)
+		fprintf(stderr,
+			"warning: mmaped to %p, requested %p\n", ret, req);
 
 	return ret;
 }
@@ -155,7 +183,18 @@ void *plat_mremap(void *ptr, size_t oldsize, size_t newsize)
 
 void plat_munmap(void *ptr, size_t size)
 {
-	munmap(ptr, size);
+	int ret;
+
+	ret = munmap(ptr, size);
+	if (ret != 0 && (size & (HUGETLB_PAGESIZE - 1))) {
+		// prehaps an autorounded hugetlb mapping?
+		size = (size + HUGETLB_PAGESIZE - 1) & ~(HUGETLB_PAGESIZE - 1);
+		ret = munmap(ptr, size);
+	}
+	if (ret != 0) {
+		fprintf(stderr,
+			"munmap(%p, %zu) failed: %d\n", ptr, size, errno);
+	}
 }
 
 /* lprintf */
diff --git a/frontend/main.c b/frontend/main.c
index 19e8319..56b5cb7 100644
--- a/frontend/main.c
+++ b/frontend/main.c
@@ -143,6 +143,8 @@ void emu_set_default_config(void)
 	Config.PsxAuto = 1;
 
 	pl_rearmed_cbs.gpu_neon.allow_interlace = 2; // auto
+	pl_rearmed_cbs.gpu_neon.enhancement_enable =
+	pl_rearmed_cbs.gpu_neon.enhancement_no_main = 0;
 	pl_rearmed_cbs.gpu_peops.iUseDither = 0;
 	pl_rearmed_cbs.gpu_peops.dwActFixes = 1<<7;
 	pl_rearmed_cbs.gpu_unai.abe_hack =
@@ -230,6 +232,14 @@ do_state_slot:
 			pl_rearmed_cbs.frameskip == 0 ? "OFF" : "1" );
 		plugin_call_rearmed_cbs();
 		break;
+	case SACTION_SWITCH_DISPMODE:
+		pl_switch_dispmode();
+		plugin_call_rearmed_cbs();
+		if (GPU_open != NULL && GPU_close != NULL) {
+			GPU_close();
+			GPU_open(&gpuDisp, "PCSX", NULL);
+		}
+		break;
 	case SACTION_SCREENSHOT:
 		{
 			char buf[MAXPATHLEN];
diff --git a/frontend/main.h b/frontend/main.h
index bdb4870..a03db8b 100644
--- a/frontend/main.h
+++ b/frontend/main.h
@@ -65,6 +65,7 @@ enum sched_action {
 	SACTION_NEXT_SSLOT,
 	SACTION_PREV_SSLOT,
 	SACTION_TOGGLE_FSKIP,
+	SACTION_SWITCH_DISPMODE,
 	SACTION_SCREENSHOT,
 	SACTION_VOLUME_UP,
 	SACTION_VOLUME_DOWN,
diff --git a/frontend/menu.c b/frontend/menu.c
index 42a53e1..d3ce06c 100644
--- a/frontend/menu.c
+++ b/frontend/menu.c
@@ -75,6 +75,7 @@ typedef enum
 	MA_OPT_SCALER,
 	MA_OPT_SCALER2,
 	MA_OPT_FILTERING,
+	MA_OPT_FILTERING2,
 	MA_OPT_SCALER_C,
 } menu_id;
 
@@ -87,7 +88,7 @@ static int psx_clock;
 static int memcard1_sel, memcard2_sel;
 int g_opts, g_scaler;
 int soft_scaling, analog_deadzone; // for Caanoo
-int filter;
+int filter, soft_filter;
 
 #ifdef __ARM_ARCH_7A__
 #define DEFAULT_PSX_CLOCK 57
@@ -213,6 +214,7 @@ static void menu_set_defconfig(void)
 	frameskip = 0;
 	analog_deadzone = 50;
 	soft_scaling = 1;
+	soft_filter = 0;
 	psx_clock = DEFAULT_PSX_CLOCK;
 
 	region = 0;
@@ -274,6 +276,7 @@ static const struct {
 	CE_INTVAL(g_layer_w),
 	CE_INTVAL(g_layer_h),
 	CE_INTVAL(filter),
+	CE_INTVAL(soft_filter),
 	CE_INTVAL(state_slot),
 	CE_INTVAL(cpu_clock),
 	CE_INTVAL(g_opts),
@@ -288,6 +291,8 @@ static const struct {
 	CE_INTVAL_P(gpu_unai.no_light),
 	CE_INTVAL_P(gpu_unai.no_blend),
 	CE_INTVAL_P(gpu_neon.allow_interlace),
+	CE_INTVAL_P(gpu_neon.enhancement_enable),
+	CE_INTVAL_P(gpu_neon.enhancement_no_main),
 	CE_INTVAL_P(gpu_peopsgl.bDrawDither),
 	CE_INTVAL_P(gpu_peopsgl.iFilterType),
 	CE_INTVAL_P(gpu_peopsgl.iFrameTexType),
@@ -661,6 +666,7 @@ me_bind_action emuctrl_actions[] =
 	{ "Next Save Slot   ", 1 << SACTION_NEXT_SSLOT },
 	{ "Toggle Frameskip ", 1 << SACTION_TOGGLE_FSKIP },
 	{ "Take Screenshot  ", 1 << SACTION_SCREENSHOT },
+	{ "Switch Renderer  ", 1 << SACTION_SWITCH_DISPMODE },
 	{ "Enter Menu       ", 1 << SACTION_ENTER_MENU },
 #ifdef __ARM_ARCH_7A__ /* XXX */
 	{ "Minimize         ", 1 << SACTION_MINIMIZE },
@@ -1029,9 +1035,15 @@ static int menu_loop_keyconfig(int id, int keys)
 // ------------ gfx options menu ------------
 
 static const char *men_scaler[] = { "1x1", "scaled 4:3", "integer scaled 4:3", "fullscreen", "custom", NULL };
+static const char *men_soft_filter[] = { "None",
+#ifdef __ARM_NEON__
+	"scale2x", "eagle2x",
+#endif
+	NULL };
+static const char *men_dummy[] = { NULL };
 static const char h_cscaler[]   = "Displays the scaler layer, you can resize it\n"
 				  "using d-pad or move it using R+d-pad";
-static const char *men_dummy[] = { NULL };
+static const char h_soft_filter[] = "Works only if game uses low resolution modes";
 
 static int menu_loop_cscaler(int id, int keys)
 {
@@ -1090,6 +1102,7 @@ static menu_entry e_menu_gfx_options[] =
 	mee_enum      ("Scaler",                   MA_OPT_SCALER, g_scaler, men_scaler),
 	mee_onoff     ("Software Scaling",         MA_OPT_SCALER2, soft_scaling, 1),
 	mee_enum      ("Filter",                   MA_OPT_FILTERING, filter, men_dummy),
+	mee_enum_h    ("Software Filter",          MA_OPT_FILTERING2, soft_filter, men_soft_filter, h_soft_filter),
 //	mee_onoff     ("Vsync",                    0, vsync, 1),
 	mee_cust_h    ("Setup custom scaler",      MA_OPT_SCALER_C, menu_loop_cscaler, NULL, h_cscaler),
 	mee_end,
@@ -1118,18 +1131,26 @@ void menu_set_filter_list(void *filters)
 
 #ifdef __ARM_NEON__
 
-static const char h_gpu_neon[] = "Configure built-in NEON GPU plugin";
+static const char h_gpu_neon[] =
+	"Configure built-in NEON GPU plugin";
+static const char h_gpu_neon_enhanced[] =
+	"Renders in double resolution at the cost of lower performance\n"
+	"(not available for high resolution games)";
+static const char h_gpu_neon_enhanced_hack[] =
+	"Speed hack for above option (glitches some games)";
 static const char *men_gpu_interlace[] = { "Off", "On", "Auto", NULL };
 
 static menu_entry e_menu_plugin_gpu_neon[] =
 {
 	mee_enum      ("Enable interlace mode",      0, pl_rearmed_cbs.gpu_neon.allow_interlace, men_gpu_interlace),
+	mee_onoff_h   ("Enhanced resolution (slow)", 0, pl_rearmed_cbs.gpu_neon.enhancement_enable, 1, h_gpu_neon_enhanced),
+	mee_onoff_h   ("Enhanced res. speed hack",   0, pl_rearmed_cbs.gpu_neon.enhancement_no_main, 1, h_gpu_neon_enhanced_hack),
 	mee_end,
 };
 
 static int menu_loop_plugin_gpu_neon(int id, int keys)
 {
-	int sel = 0;
+	static int sel = 0;
 	me_loop(e_menu_plugin_gpu_neon, &sel);
 	return 0;
 }
@@ -2247,6 +2268,7 @@ void menu_init(void)
 #ifndef __ARM_ARCH_7A__ /* XXX */
 	me_enable(e_menu_gfx_options, MA_OPT_SCALER, 0);
 	me_enable(e_menu_gfx_options, MA_OPT_FILTERING, 0);
+	me_enable(e_menu_gfx_options, MA_OPT_FILTERING2, 0);
 	me_enable(e_menu_gfx_options, MA_OPT_SCALER_C, 0);
 	me_enable(e_menu_keyconfig, MA_CTRL_NUBS_BTNS, 0);
 #else
diff --git a/frontend/menu.h b/frontend/menu.h
index 2062acd..221be15 100644
--- a/frontend/menu.h
+++ b/frontend/menu.h
@@ -22,9 +22,15 @@ enum g_scaler_opts {
 	SCALE_CUSTOM,
 };
 
+enum g_soft_filter_opts {
+	SOFT_FILTER_NONE,
+	SOFT_FILTER_SCALE2X,
+	SOFT_FILTER_EAGLE2X,
+};
+
 extern int g_opts, g_scaler;
 extern int soft_scaling, analog_deadzone;
-extern int filter;
+extern int filter, soft_filter;
 
 extern int g_menuscreen_w;
 extern int g_menuscreen_h;
diff --git a/frontend/pandora/pcsx.sh b/frontend/pandora/pcsx.sh
index 0957b94..bc1d6c5 100755
--- a/frontend/pandora/pcsx.sh
+++ b/frontend/pandora/pcsx.sh
@@ -5,10 +5,15 @@ nub0mode=`cat /proc/pandora/nub0/mode`
 nub1mode=`cat /proc/pandora/nub1/mode`
 /usr/pandora/scripts/op_nubchange.sh absolute absolute
 
+# 4MB for RAM (2+align) + 2MB for vram (1+overdraw) + 10MB for gpu_neon (8+overdraw)
+# no big deal if this fails, only performance loss
+sudo -n /usr/pandora/scripts/op_hugetlb.sh 16
+
 ./pcsx "$@"
 
 # restore stuff if pcsx crashes
 ./picorestore
 sudo -n /usr/pandora/scripts/op_lcdrate.sh 60
+sudo -n /usr/pandora/scripts/op_hugetlb.sh 0
 
 /usr/pandora/scripts/op_nubchange.sh $nub0mode $nub1mode
diff --git a/frontend/plat_omap.c b/frontend/plat_omap.c
index b01c634..e5b6c04 100644
--- a/frontend/plat_omap.c
+++ b/frontend/plat_omap.c
@@ -52,8 +52,9 @@ static int omap_setup_layer_(int fd, int enabled, int x, int y, int w, int h)
 			perror("SETUP_PLANE");
 	}
 
-	if (mi.size < 640*512*3*3) {
-		mi.size = 640*512*3*3;
+	// upto 1024x512 (2x resolution enhancement)
+	if (mi.size < 1024*512*2 * 3) {
+		mi.size = 1024*512*2 * 3;
 		ret = ioctl(fd, OMAPFB_SETUP_MEM, &mi);
 		if (ret != 0) {
 			perror("SETUP_MEM");
diff --git a/frontend/plat_pandora.c b/frontend/plat_pandora.c
index 9ec747d..b82450c 100644
--- a/frontend/plat_pandora.c
+++ b/frontend/plat_pandora.c
@@ -65,6 +65,7 @@ static const struct in_default_bind in_evdev_defbinds[] = {
 	{ KEY_4,        IN_BINDTYPE_EMU, SACTION_NEXT_SSLOT },
 	{ KEY_5,        IN_BINDTYPE_EMU, SACTION_TOGGLE_FSKIP },
 	{ KEY_6,        IN_BINDTYPE_EMU, SACTION_SCREENSHOT },
+	{ KEY_7,        IN_BINDTYPE_EMU, SACTION_SWITCH_DISPMODE },
 	{ 0, 0, 0 }
 };
 
diff --git a/frontend/plat_pollux.c b/frontend/plat_pollux.c
index 1dafb7c..52a09b1 100644
--- a/frontend/plat_pollux.c
+++ b/frontend/plat_pollux.c
@@ -305,12 +305,13 @@ static void spend_cycles(int loops)
 #define DMA_REG(x) memregl[(DMA_BASE6 + x) >> 2]
 
 /* this takes ~1.5ms, while ldm/stm ~1.95ms */
-static void raw_flip_dma(int x, int y)
+static void raw_flip_dma(const void *vram, int stride, int bgr24, int w, int h)
 {
+	unsigned int pixel_offset = psx_vram - (unsigned short *)vram;
 	unsigned int dst = fb_paddrs[fb_work_buf] +
 			(fb_offset_y * 320 + fb_offset_x) * psx_bpp / 8;
-	int spsx_line = y + psx_offset_y;
-	int spsx_offset = (x + psx_offset_x) & 0x3f8;
+	int spsx_line = pixel_offset / 1024 + psx_offset_y;
+	int spsx_offset = (pixel_offset + psx_offset_x) & 0x3f8;
 	int dst_stride = 320 * psx_bpp / 8;
 	int len = psx_src_width * psx_bpp / 8;
 	int i;
@@ -344,7 +345,7 @@ static void raw_flip_dma(int x, int y)
 
 	if (psx_bpp == 16) {
 		pl_vout_buf = g_menuscreen_ptr;
-		pl_print_hud(fb_offset_x);
+		pl_print_hud(w, h, fb_offset_x);
 	}
 
 	g_menuscreen_ptr = fb_flip();
@@ -354,26 +355,24 @@ static void raw_flip_dma(int x, int y)
 }
 
 #define make_flip_func(name, blitfunc)                                                  \
-static void name(int x, int y)                                                          \
+static void name(const void *vram_, int stride, int bgr24, int w, int h)                \
 {                                                                                       \
-        unsigned short *vram = psx_vram;                                                \
+        const unsigned short *vram = vram_;                                             \
         unsigned char *dst = (unsigned char *)g_menuscreen_ptr +                        \
                         (fb_offset_y * 320 + fb_offset_x) * psx_bpp / 8;                \
-        unsigned int src = (y + psx_offset_y) * 1024 + x + psx_offset_x;                \
         int dst_stride = 320 * psx_bpp / 8;                                             \
         int len = psx_src_width * psx_bpp / 8;                                          \
         int i;                                                                          \
                                                                                         \
         pcnt_start(PCNT_BLIT);                                                          \
                                                                                         \
-        for (i = psx_src_height; i > 0; i--, src += psx_step * 1024, dst += dst_stride) { \
-                src &= 1024*512-1;                                                      \
-                blitfunc(dst, vram + src, len);                                         \
-        }                                                                               \
+        vram += psx_offset_y * 1024 + psx_offset_x;                                     \
+        for (i = psx_src_height; i > 0; i--, vram += psx_step * 1024, dst += dst_stride)\
+                blitfunc(dst, vram, len);                                               \
                                                                                         \
         if (psx_bpp == 16) {                                                            \
                 pl_vout_buf = g_menuscreen_ptr;                                         \
-                pl_print_hud(fb_offset_x);                                              \
+                pl_print_hud(w, h, fb_offset_x);                                        \
         }                                                                               \
                                                                                         \
         g_menuscreen_ptr = fb_flip();                                                   \
@@ -402,20 +401,20 @@ void *plat_gvideo_set_mode(int *w_, int *h_, int *bpp_)
 
 	switch (w + (bpp != 16) + !soft_scaling) {
 	case 640:
-		pl_rearmed_cbs.pl_vout_raw_flip = raw_flip_soft_640;
+		pl_rearmed_cbs.pl_vout_flip = raw_flip_soft_640;
 		w_max = 640;
 		break;
 	case 512:
-		pl_rearmed_cbs.pl_vout_raw_flip = raw_flip_soft_512;
+		pl_rearmed_cbs.pl_vout_flip = raw_flip_soft_512;
 		w_max = 512;
 		break;
 	case 384:
 	case 368:
-		pl_rearmed_cbs.pl_vout_raw_flip = raw_flip_soft_368;
+		pl_rearmed_cbs.pl_vout_flip = raw_flip_soft_368;
 		w_max = 368;
 		break;
 	default:
-		pl_rearmed_cbs.pl_vout_raw_flip = have_warm ? raw_flip_dma : raw_flip_soft;
+		pl_rearmed_cbs.pl_vout_flip = have_warm ? raw_flip_dma : raw_flip_soft;
 		w_max = 320;
 		break;
 	}
@@ -621,7 +620,7 @@ void plat_init(void)
 	if (mixerdev == -1)
 		perror("open(/dev/mixer)");
 
-	pl_rearmed_cbs.pl_vout_raw_flip = have_warm ? raw_flip_dma : raw_flip_soft;
+	pl_rearmed_cbs.pl_vout_flip = have_warm ? raw_flip_dma : raw_flip_soft;
 	pl_rearmed_cbs.pl_vout_set_raw_vram = pl_vout_set_raw_vram;
 
 	psx_src_width = 320;
diff --git a/frontend/plugin_lib.c b/frontend/plugin_lib.c
index 4dbb9a7..3ee5947 100644
--- a/frontend/plugin_lib.c
+++ b/frontend/plugin_lib.c
@@ -21,13 +21,17 @@
 #include "linux/fbdev.h"
 #include "common/fonts.h"
 #include "common/input.h"
+#include "common/plat.h"
 #include "menu.h"
 #include "main.h"
 #include "plat.h"
 #include "pcnt.h"
 #include "pl_gun_ts.h"
+#include "libpicofe/arm/neon_scale2x.h"
+#include "libpicofe/arm/neon_eagle2x.h"
 #include "../libpcsxcore/new_dynarec/new_dynarec.h"
 #include "../libpcsxcore/psemu_plugin_defs.h"
+#include "../plugins/gpulib/cspace.h"
 
 int in_type1, in_type2;
 int in_a1[2] = { 127, 127 }, in_a2[2] = { 127, 127 };
@@ -38,6 +42,7 @@ void *tsdev;
 void *pl_vout_buf;
 int g_layer_x, g_layer_y, g_layer_w, g_layer_h;
 static int pl_vout_w, pl_vout_h, pl_vout_bpp; /* output display/layer */
+static int pl_vout_scale;
 static int psx_w, psx_h, psx_bpp;
 static int vsync_cnt;
 static int is_pal, frame_interval, frame_interval1024;
@@ -113,10 +118,8 @@ static __attribute__((noinline)) void draw_active_chans(int vout_w, int vout_h)
 	}
 }
 
-void pl_print_hud(int xborder)
+void pl_print_hud(int w, int h, int xborder)
 {
-	int w = pl_vout_w, h = pl_vout_h;
-
 	if (h < 16)
 		return;
 
@@ -184,55 +187,142 @@ static void update_layer_size(int w, int h)
 	if (g_layer_h > g_menuscreen_h) g_layer_h = g_menuscreen_h;
 }
 
-static void *pl_vout_set_mode(int w, int h, int bpp)
+// XXX: this is platform specific really
+static int resolution_ok(int w, int h)
 {
+	return w <= 1024 && h <= 512;
+}
+
+static void pl_vout_set_mode(int w, int h, int bpp)
+{
+	int vout_w, vout_h, vout_bpp;
+
 	// special h handling, Wipeout likes to change it by 1-6
 	static int vsync_cnt_ms_prev;
 	if ((unsigned int)(vsync_cnt - vsync_cnt_ms_prev) < 5*60)
 		h = (h + 7) & ~7;
 	vsync_cnt_ms_prev = vsync_cnt;
 
-	if (w == psx_w && h == psx_h && bpp == psx_bpp)
-		return pl_vout_buf;
+	vout_w = psx_w = w;
+	vout_h = psx_h = h;
+	vout_bpp = psx_bpp = bpp;
+
+	pl_vout_scale = 1;
+#ifdef __ARM_NEON__
+	if (soft_filter) {
+		if (resolution_ok(w * 2, h * 2) && bpp == 16) {
+			vout_w *= 2;
+			vout_h *= 2;
+			pl_vout_scale = 2;
+		}
+		else {
+			// filter unavailable
+			hud_msg[0] = 0;
+		}
+	}
+#endif
 
-	pl_vout_w = psx_w = w;
-	pl_vout_h = psx_h = h;
-	pl_vout_bpp = psx_bpp = bpp;
+	if (pl_vout_buf != NULL && vout_w == pl_vout_w && vout_h == pl_vout_h
+	    && vout_bpp == pl_vout_bpp)
+		return;
 
-	update_layer_size(pl_vout_w, pl_vout_h);
+	update_layer_size(vout_w, vout_h);
 
-	pl_vout_buf = plat_gvideo_set_mode(&pl_vout_w, &pl_vout_h, &pl_vout_bpp);
-	if (pl_vout_buf == NULL && pl_rearmed_cbs.pl_vout_raw_flip == NULL)
+	pl_vout_buf = plat_gvideo_set_mode(&vout_w, &vout_h, &vout_bpp);
+	if (pl_vout_buf == NULL)
 		fprintf(stderr, "failed to set mode %dx%d@%d\n",
 			psx_w, psx_h, psx_bpp);
+	else {
+		pl_vout_w = vout_w;
+		pl_vout_h = vout_h;
+		pl_vout_bpp = vout_bpp;
+	}
 
 	menu_notify_mode_change(pl_vout_w, pl_vout_h, pl_vout_bpp);
-
-	return pl_vout_buf;
 }
 
-// only used if raw flip is not defined
-static void *pl_vout_flip(void)
+static void pl_vout_flip(const void *vram, int stride, int bgr24, int w, int h)
 {
-	pl_rearmed_cbs.flip_cnt++;
+	static int doffs_old, clear_counter;
+	unsigned char *dest = pl_vout_buf;
+	const unsigned short *src = vram;
+	int dstride = pl_vout_w, h1 = h;
+	int doffs;
+
+	if (dest == NULL)
+		goto out;
+
+	if (vram == NULL) {
+		// blanking
+		memset(pl_vout_buf, 0, dstride * pl_vout_h * pl_vout_bpp / 8);
+		goto out;
+	}
+
+	// borders
+	doffs = (dstride - w * pl_vout_scale) / 2 & ~1;
+	dest += doffs * 2;
+
+	if (doffs > doffs_old)
+		clear_counter = 2;
+	doffs_old = doffs;
+
+	if (clear_counter > 0) {
+		memset(pl_vout_buf, 0, dstride * pl_vout_h * pl_vout_bpp / 8);
+		clear_counter--;
+	}
 
-	if (pl_vout_buf != NULL)
-		pl_print_hud(0);
+	if (bgr24)
+	{
+		if (pl_rearmed_cbs.only_16bpp) {
+			for (; h1-- > 0; dest += dstride * 2, src += stride)
+			{
+				bgr888_to_rgb565(dest, src, w * 3);
+			}
+		}
+		else {
+			dest -= doffs * 2;
+			dest += (doffs / 8) * 24;
 
+			for (; h1-- > 0; dest += dstride * 3, src += stride)
+			{
+				bgr888_to_rgb888(dest, src, w * 3);
+			}
+		}
+	}
+#ifdef __ARM_NEON__
+	else if (soft_filter == SOFT_FILTER_SCALE2X && pl_vout_scale == 2)
+	{
+		neon_scale2x_16_16(src, (void *)dest, w,
+			stride * 2, dstride * 2, h1);
+	}
+	else if (soft_filter == SOFT_FILTER_EAGLE2X && pl_vout_scale == 2)
+	{
+		neon_eagle2x_16_16(src, (void *)dest, w,
+			stride * 2, dstride * 2, h1);
+	}
+#endif
+	else
+	{
+		for (; h1-- > 0; dest += dstride * 2, src += stride)
+		{
+			bgr555_to_rgb565(dest, src, w * 2);
+		}
+	}
+
+	pl_print_hud(w * pl_vout_scale, h * pl_vout_scale, 0);
+
+out:
 	// let's flip now
 	pl_vout_buf = plat_gvideo_flip();
-	return pl_vout_buf;
+	pl_rearmed_cbs.flip_cnt++;
 }
 
 static int pl_vout_open(void)
 {
 	struct timeval now;
-	int h;
 
-	// force mode update
-	h = psx_h;
-	psx_h--;
-	pl_vout_buf = pl_vout_set_mode(psx_w, h, psx_bpp);
+	// force mode update on pl_vout_set_mode() call from gpulib/vout_pl
+	pl_vout_buf = NULL;
 
 	plat_gvideo_open(is_pal);
 
@@ -249,6 +339,11 @@ static void pl_vout_close(void)
 	plat_gvideo_close();
 }
 
+static void pl_set_gpu_caps(int caps)
+{
+	pl_rearmed_cbs.gpu_caps = caps;
+}
+
 void *pl_prepare_screenshot(int *w, int *h, int *bpp)
 {
 	void *ret = plat_prepare_screenshot(w, h, bpp);
@@ -262,6 +357,75 @@ void *pl_prepare_screenshot(int *w, int *h, int *bpp)
 	return pl_vout_buf;
 }
 
+/* display/redering mode switcher */
+static int dispmode_default(void)
+{
+	pl_rearmed_cbs.gpu_neon.enhancement_enable = 0;
+	soft_filter = SOFT_FILTER_NONE;
+	snprintf(hud_msg, sizeof(hud_msg), "default mode");
+	return 1;
+}
+
+int dispmode_doubleres(void)
+{
+	if (!(pl_rearmed_cbs.gpu_caps & GPU_CAP_SUPPORTS_2X)
+	    || !resolution_ok(psx_w * 2, psx_h * 2) || psx_bpp != 16)
+		return 0;
+
+	dispmode_default();
+	pl_rearmed_cbs.gpu_neon.enhancement_enable = 1;
+	snprintf(hud_msg, sizeof(hud_msg), "double resolution");
+	return 1;
+}
+
+int dispmode_scale2x(void)
+{
+	if (psx_bpp != 16)
+		return 0;
+
+	dispmode_default();
+	soft_filter = SOFT_FILTER_SCALE2X;
+	snprintf(hud_msg, sizeof(hud_msg), "scale2x");
+	return 1;
+}
+
+int dispmode_eagle2x(void)
+{
+	if (psx_bpp != 16)
+		return 0;
+
+	dispmode_default();
+	soft_filter = SOFT_FILTER_EAGLE2X;
+	snprintf(hud_msg, sizeof(hud_msg), "eagle2x");
+	return 1;
+}
+
+static int (*dispmode_switchers[])(void) = {
+	dispmode_default,
+#ifdef __ARM_NEON__
+	dispmode_doubleres,
+	dispmode_scale2x,
+	dispmode_eagle2x,
+#endif
+};
+
+static int dispmode_current;
+
+void pl_switch_dispmode(void)
+{
+	if (pl_rearmed_cbs.gpu_caps & GPU_CAP_OWNS_DISPLAY)
+		return;
+
+	while (1) {
+		dispmode_current++;
+		if (dispmode_current >=
+		    sizeof(dispmode_switchers) / sizeof(dispmode_switchers[0]))
+			dispmode_current = 0;
+		if (dispmode_switchers[dispmode_current]())
+			break;
+	}
+}
+
 #ifndef MAEMO
 static void update_analogs(void)
 {
@@ -442,16 +606,31 @@ void pl_timing_prepare(int is_pal_)
 
 static void pl_text_out16_(int x, int y, const char *text)
 {
-	int i, l, len = strlen(text), w = pl_vout_w;
-	unsigned short *screen = (unsigned short *)pl_vout_buf + x + y * w;
+	int i, l, w = pl_vout_w;
+	unsigned short *screen;
 	unsigned short val = 0xffff;
 
-	for (i = 0; i < len; i++, screen += 8)
+	x &= ~1;
+	screen = (unsigned short *)pl_vout_buf + x + y * w;
+	for (i = 0; ; i++, screen += 8)
 	{
+		char c = text[i];
+		if (c == 0)
+			break;
+		if (c == ' ')
+			continue;
+
 		for (l = 0; l < 8; l++)
 		{
-			unsigned char fd = fontdata8x8[text[i] * 8 + l];
+			unsigned char fd = fontdata8x8[c * 8 + l];
 			unsigned short *s = screen + l * w;
+			unsigned int *s32 = (void *)s;
+
+			s32[0] = (s32[0] >> 1) & 0x7bef7bef;
+			s32[1] = (s32[1] >> 1) & 0x7bef7bef;
+			s32[2] = (s32[2] >> 1) & 0x7bef7bef;
+			s32[3] = (s32[3] >> 1) & 0x7bef7bef;
+
 			if (fd&0x80) s[0] = val;
 			if (fd&0x40) s[1] = val;
 			if (fd&0x20) s[2] = val;
@@ -484,12 +663,26 @@ static void pl_get_layer_pos(int *x, int *y, int *w, int *h)
 	*h = g_layer_h;
 }
 
+static void *pl_mmap(unsigned int size)
+{
+	return plat_mmap(0, size, 0, 0);
+}
+
+static void pl_munmap(void *ptr, unsigned int size)
+{
+	plat_munmap(ptr, size);
+}
+
 struct rearmed_cbs pl_rearmed_cbs = {
 	pl_get_layer_pos,
 	pl_vout_open,
 	pl_vout_set_mode,
 	pl_vout_flip,
 	pl_vout_close,
+
+	.mmap = pl_mmap,
+	.munmap = pl_munmap,
+	.pl_set_gpu_caps = pl_set_gpu_caps,
 };
 
 /* watchdog */
diff --git a/frontend/plugin_lib.h b/frontend/plugin_lib.h
index bcf74ac..332fbc2 100644
--- a/frontend/plugin_lib.h
+++ b/frontend/plugin_lib.h
@@ -31,7 +31,8 @@ void  pl_text_out16(int x, int y, const char *texto, ...);
 void  pl_start_watchdog(void);
 void *pl_prepare_screenshot(int *w, int *h, int *bpp);
 void  pl_init(void);
-void  pl_print_hud(int xborder);
+void  pl_print_hud(int width, int height, int xborder);
+void  pl_switch_dispmode(void);
 
 void  pl_timing_prepare(int is_pal);
 void  pl_frame_limit(void);
@@ -41,12 +42,15 @@ void  pl_update_gun(int *xn, int *xres, int *y, int *in);
 struct rearmed_cbs {
 	void  (*pl_get_layer_pos)(int *x, int *y, int *w, int *h);
 	int   (*pl_vout_open)(void);
-	void *(*pl_vout_set_mode)(int w, int h, int bpp);
-	void *(*pl_vout_flip)(void);
+	void  (*pl_vout_set_mode)(int w, int h, int bpp);
+	void  (*pl_vout_flip)(const void *vram, int stride, int bgr24,
+			      int w, int h);
 	void  (*pl_vout_close)(void);
-	// these are only used by some frontends
-	void  (*pl_vout_raw_flip)(int x, int y);
+	void *(*mmap)(unsigned int size);
+	void  (*munmap)(void *ptr, unsigned int size);
+	// only used by some frontends
 	void  (*pl_vout_set_raw_vram)(void *vram);
+	void  (*pl_set_gpu_caps)(int caps);
 	// some stats, for display by some plugins
 	int flips_per_sec, cpu_usage;
 	float vsps_cur; // currect vsync/s
@@ -60,6 +64,8 @@ struct rearmed_cbs {
 	unsigned int only_16bpp; // platform is 16bpp-only
 	struct {
 		int   allow_interlace; // 0 off, 1 on, 2 guess
+		int   enhancement_enable;
+		int   enhancement_no_main;
 	} gpu_neon;
 	struct {
 		int   iUseDither;
@@ -78,10 +84,17 @@ struct rearmed_cbs {
 		int   iUseMask, bOpaquePass, bAdvancedBlend, bUseFastMdec;
 		int   iVRamSize, iTexGarbageCollection;
 	} gpu_peopsgl;
+	// misc
+	int gpu_caps;
 };
 
 extern struct rearmed_cbs pl_rearmed_cbs;
 
+enum gpu_plugin_caps {
+	GPU_CAP_OWNS_DISPLAY = (1 << 0),
+	GPU_CAP_SUPPORTS_2X = (1 << 1),
+};
+
 #ifndef ARRAY_SIZE
 #define ARRAY_SIZE(x) (sizeof(x) / sizeof(x[0]))
 #endif