(VITA) Try to use NEON

Makefile.griffin

@@ -211,7 +211,7 @@ else ifeq ($(platform), vita)
   WANT_ZLIB						:= 1
   INCLUDE += -Ideps/libvita2d/include
   INCLUDE += -Ideps/zlib
-  PLATCFLAGS := -mthumb -mfloat-abi=hard -fsingle-precision-constant \
+  PLATCFLAGS := -marm -mfloat-abi=hard -fsingle-precision-constant \
        -mword-relocations -fno-unwind-tables -fno-asynchronous-unwind-tables -ftree-vectorize -fno-optimize-sibling-calls
   LIBS += -lSceKernel_stub -lSceDisplay_stub -lSceGxm_stub -lSceNet_stub -lSceNetCtl_stub\
 			 -lSceSysmodule_stub -lSceCtrl_stub -lSceTouch_stub -lSceAudio_stub -lSceFiber_stub\
@@ -226,8 +226,10 @@ else ifeq ($(platform), vita)
 				  deps/libvita2d/shader/compiled/texture_f_gxp.o \
 				  deps/libvita2d/shader/compiled/texture_tint_f_gxp.o
   PLATOBJS +=			  libretro-common/conversion/s16_to_float_neon.o \
-				  audio/drivers_resampler/sinc_resampler_neon.o \
-              audio/drivers_resampler/cc_resampler_neon.o
+                    libretro-common/conversion/float_to_s16_neon.o \
+                    memory/neon/memcpy-neon.o \
+				            audio/drivers_resampler/sinc_resampler_neon.o \
+                    audio/drivers_resampler/cc_resampler_neon.o
 
   LIBDIRS += -L.
   LDFLAGS +=  -Wl,-q
@@ -404,7 +406,7 @@ all: $(EXT_TARGET)
 
 %.velf: %.elf
 ifneq ($(DEBUG), 1)
-	arm-vita-eabi-strip -g $<
+  arm-vita-eabi-strip -g $<
 endif
 	vita-elf-create $< $@
 	vita-make-fself -s $@ eboot.bin

audio/drivers_resampler/sinc_resampler.c

@@ -334,7 +334,7 @@ static void process_sinc(rarch_sinc_resampler_t *resamp, float *out_buffer)
    /* movehl { X, R, X, L } == { X, R, X, R } */
    _mm_store_ss(out_buffer + 1, _mm_movehl_ps(sum, sum));
 }
-#elif defined(__ARM_NEON__) && !defined(VITA)
+#elif defined(__ARM_NEON__)
 
 #if SINC_COEFF_LERP
 #error "NEON asm does not support SINC lerp."
@@ -437,7 +437,7 @@ static void *resampler_sinc_new(const struct resampler_config *config,
    }
 
    /* Be SIMD-friendly. */
-#if (defined(__AVX__) && ENABLE_AVX) || (defined(__ARM_NEON__)&& !defined(VITA))
+#if (defined(__AVX__) && ENABLE_AVX) || (defined(__ARM_NEON__))
    re->taps     = (re->taps + 7) & ~7;
 #else
    re->taps     = (re->taps + 3) & ~3;
@@ -460,7 +460,7 @@ static void *resampler_sinc_new(const struct resampler_config *config,
    init_sinc_table(re, cutoff, re->phase_table,
          1 << PHASE_BITS, re->taps, SINC_COEFF_LERP);
 
-#if defined(__ARM_NEON__) && !defined(VITA)
+#if defined(__ARM_NEON__) 
    process_sinc_func = mask & RESAMPLER_SIMD_NEON 
       ? process_sinc_neon : process_sinc_C;
 #endif

deps/libvita2d/source/vita2d_texture.c

@@ -66,7 +66,7 @@ vita2d_texture *vita2d_create_empty_texture_format(unsigned int w, unsigned int
 	}
 
 	/* Clear the texture */
-	memset(texture_data, 0, tex_size);
+	//memset(texture_data, 0, tex_size);
 
 	/* Create the gxm texture */
 	sceGxmTextureInitLinear(
@@ -94,7 +94,7 @@ vita2d_texture *vita2d_create_empty_texture_format(unsigned int w, unsigned int
 			return NULL;
 		}
 
-		memset(texture_palette, 0, pal_size);
+		//memset(texture_palette, 0, pal_size);
 
 		sceGxmTextureSetPalette(&texture->gxm_tex, texture_palette);
 	} else {

gfx/drivers/vita2d_gfx.c

@@ -33,6 +33,9 @@
 #include "../../runloop.h"
 #include "../video_coord_array.h"
 
+extern void *memcpy_neon(void *dst, const void *src, size_t n);
+
+
 static void vita2d_gfx_set_viewport(void *data, unsigned viewport_width,
       unsigned viewport_height, bool force_full, bool allow_rotate);
       
@@ -139,7 +142,7 @@ static bool vita2d_gfx_frame(void *data, const void *frame,
    if (frame)
    {
       if(!(vita->texture&&vita2d_texture_get_datap(vita->texture)==frame)){
-        unsigned i, j;
+        unsigned i;
         unsigned int stride;
 
         if ((width != vita->width || height != vita->height) && vita->texture)
@@ -167,8 +170,7 @@ static bool vita2d_gfx_frame(void *data, const void *frame,
            const uint32_t     *frame32 = frame;
 
            for (i = 0; i < height; i++)
-              for (j = 0; j < width; j++)
-                 tex32[j + i*stride] = frame32[j + i*pitch];
+              memcpy_neon(&tex32[i*stride],&frame32[i*pitch],pitch*sizeof(uint32_t));
         }
         else
         {
@@ -178,8 +180,7 @@ static bool vita2d_gfx_frame(void *data, const void *frame,
            const uint16_t *frame16 = frame;
 
            for (i = 0; i < height; i++)
-              for (j = 0; j < width; j++)
-                 tex16[j + i*stride] = frame16[j + i*pitch];
+              memcpy_neon(&tex16[i*stride],&frame16[i*pitch],width*sizeof(uint16_t));
         }
       }
    }
@@ -700,7 +701,7 @@ static void vita_set_texture_enable(void *data, bool state, bool full_screen)
 static uintptr_t vita_load_texture(void *video_data, void *data,
       bool threaded, enum texture_filter_type filter_type)
 {
-   unsigned int stride, pitch, j, k;
+   unsigned int stride, pitch, j;
    struct texture_image *image = (struct texture_image*)data;
    struct vita2d_texture *texture = vita2d_create_empty_texture_format(image->width, 
      image->height,SCE_GXM_TEXTURE_FORMAT_U8U8U8U8_ARGB);
@@ -717,8 +718,7 @@ static uintptr_t vita_load_texture(void *video_data, void *data,
    const uint32_t     *frame32 = image->pixels;
    pitch = image->width;
    for (j = 0; j < image->height; j++)
-      for (k = 0; k < image->width; k++)
-         tex32[k + j*stride] = frame32[k + j*pitch];
+         memcpy_neon(&tex32[j*stride],&frame32[j*pitch],pitch*sizeof(uint32_t));
 
    return (uintptr_t)texture;
 }

libretro-common/conversion/float_to_s16.c

@@ -122,7 +122,7 @@ void convert_float_to_s16_altivec(int16_t *out,
    }
    convert_float_to_s16_C(out, in, samples_in);
 }
-#elif defined(__ARM_NEON__) && !defined(VITA)
+#elif defined(__ARM_NEON__)
 void convert_float_s16_asm(int16_t *out, const float *in, size_t samples);
 /**
  * convert_float_to_s16_neon:
@@ -210,7 +210,7 @@ void convert_float_to_s16_init_simd(void)
    unsigned cpu = cpu_features_get();
 
    (void)cpu;
-#if defined(__ARM_NEON__) && !defined(VITA)
+#if defined(__ARM_NEON__)
    convert_float_to_s16_arm = (cpu & RETRO_SIMD_NEON) ?
       convert_float_to_s16_neon : convert_float_to_s16_C;
 #endif

libretro-common/conversion/s16_to_float.c

@@ -122,7 +122,7 @@ void convert_s16_to_float_altivec(float *out,
    convert_s16_to_float_C(out, in, samples_in, gain);
 }
 
-#elif defined(__ARM_NEON__) && !defined(VITA)
+#elif defined(__ARM_NEON__)
 /* Avoid potential hard-float/soft-float ABI issues. */
 void convert_s16_float_asm(float *out, const int16_t *in,
       size_t samples, const float *gain);
@@ -235,7 +235,7 @@ void convert_s16_to_float_init_simd(void)
    unsigned cpu = cpu_features_get();
 
    (void)cpu;
-#if defined(__ARM_NEON__) && !defined(VITA)
+#if defined(__ARM_NEON__)
    convert_s16_to_float_arm = (cpu & RETRO_SIMD_NEON) ?
       convert_s16_to_float_neon : convert_s16_to_float_C;
 #endif

libretro-common/include/conversion/float_to_s16.h

@@ -74,7 +74,7 @@ void convert_float_to_s16_SSE2(int16_t *out,
  **/
 void convert_float_to_s16_altivec(int16_t *out,
       const float *in, size_t samples);
-#elif defined(__ARM_NEON__) && !defined(VITA)
+#elif defined(__ARM_NEON__)
 #define convert_float_to_s16 convert_float_to_s16_arm
 
 void (*convert_float_to_s16_arm)(int16_t *out,

libretro-common/include/conversion/s16_to_float.h

@@ -65,7 +65,7 @@ void convert_s16_to_float_SSE2(float *out,
 void convert_s16_to_float_altivec(float *out,
       const int16_t *in, size_t samples, float gain);
 
-#elif defined(__ARM_NEON__) && !defined(VITA)
+#elif defined(__ARM_NEON__)
 #define convert_s16_to_float convert_s16_to_float_arm
 
 void (*convert_s16_to_float_arm)(float *out,