diff --git a/src_c/alphablit.c b/src_c/alphablit.c
index f428dd1bac..7174713154 100644
--- a/src_c/alphablit.c
+++ b/src_c/alphablit.c
@@ -24,7 +24,6 @@
#define NO_PYGAME_C_API
#include "_surface.h"
-#include "simd_shared.h"
#include "simd_blitters.h"
static void
@@ -57,6 +56,12 @@ blit_blend_rgba_max(SDL_BlitInfo *info);
static void
blit_blend_premultiplied(SDL_BlitInfo *info);
+void
+premul_surf_color_by_alpha_non_simd(SDL_Surface *src,
+ PG_PixelFormat *src_format,
+ SDL_Palette *src_palette, SDL_Surface *dst,
+ PG_PixelFormat *dst_format,
+ SDL_Palette *dst_palette);
static int
SoftBlitPyGame(SDL_Surface *src, SDL_Rect *srcrect, SDL_Surface *dst,
@@ -164,8 +169,7 @@ SoftBlitPyGame(SDL_Surface *src, SDL_Rect *srcrect, SDL_Surface *dst,
case 0: {
if (info.src_blend != SDL_BLENDMODE_NONE &&
info.src->Amask) {
-#if !defined(__EMSCRIPTEN__)
-#if SDL_BYTEORDER == SDL_LIL_ENDIAN
+#ifndef PG_SIMD_DISABLED
if (PG_SURF_BytesPerPixel(src) == 4 &&
PG_SURF_BytesPerPixel(dst) == 4 &&
info.src->Rmask == info.dst->Rmask &&
@@ -174,7 +178,8 @@ SoftBlitPyGame(SDL_Surface *src, SDL_Rect *srcrect, SDL_Surface *dst,
/* If our source and destination are the same ARGB
32bit format we can use SSE2/NEON/AVX2 to speed
up the blend */
- if (pg_has_avx2() && (src != dst)) {
+#ifdef PG_HAS_AVX2
+ if (PG_HAS_AVX2() && (src != dst)) {
if (info.src_blanket_alpha != 255) {
alphablit_alpha_avx2_argb_surf_alpha(
&info);
@@ -192,8 +197,9 @@ SoftBlitPyGame(SDL_Surface *src, SDL_Rect *srcrect, SDL_Surface *dst,
}
break;
}
-#if PG_ENABLE_SSE_NEON
- if ((pg_HasSSE_NEON()) && (src != dst)) {
+#endif /* PG_HAS_AVX2 */
+#ifdef PG_HAS_SSE2_OR_NEON
+ if ((PG_HAS_SSE2_OR_NEON()) && (src != dst)) {
if (info.src_blanket_alpha != 255) {
alphablit_alpha_sse2_argb_surf_alpha(
&info);
@@ -211,10 +217,9 @@ SoftBlitPyGame(SDL_Surface *src, SDL_Rect *srcrect, SDL_Surface *dst,
}
break;
}
-#endif /* PG_ENABLE_SSE_NEON */
+#endif /* PG_HAS_SSE2_OR_NEON */
}
-#endif /* SDL_BYTEORDER == SDL_LIL_ENDIAN */
-#endif /* __EMSCRIPTEN__ */
+#endif /* PG_SIMD_DISABLED */
alphablit_alpha(&info);
}
else if (info.src_has_colorkey) {
@@ -226,8 +231,8 @@ SoftBlitPyGame(SDL_Surface *src, SDL_Rect *srcrect, SDL_Surface *dst,
break;
}
case PYGAME_BLEND_ADD: {
-#if !defined(__EMSCRIPTEN__)
-#if SDL_BYTEORDER == SDL_LIL_ENDIAN
+#ifndef PG_SIMD_DISABLED
+#ifdef PG_HAS_AVX2
if (PG_SURF_BytesPerPixel(src) == 4 &&
PG_SURF_BytesPerPixel(dst) == 4 &&
info.src->Rmask == info.dst->Rmask &&
@@ -235,11 +240,12 @@ SoftBlitPyGame(SDL_Surface *src, SDL_Rect *srcrect, SDL_Surface *dst,
info.src->Bmask == info.dst->Bmask &&
!(info.src->Amask != 0 && info.dst->Amask != 0 &&
info.src->Amask != info.dst->Amask) &&
- pg_has_avx2() && (src != dst)) {
+ PG_HAS_AVX2() && (src != dst)) {
blit_blend_rgb_add_avx2(&info);
break;
}
-#if PG_ENABLE_SSE_NEON
+#endif /* PG_HAS_AVX2 */
+#ifdef PG_HAS_SSE2_OR_NEON
if (PG_SURF_BytesPerPixel(src) == 4 &&
PG_SURF_BytesPerPixel(dst) == 4 &&
info.src->Rmask == info.dst->Rmask &&
@@ -247,19 +253,18 @@ SoftBlitPyGame(SDL_Surface *src, SDL_Rect *srcrect, SDL_Surface *dst,
info.src->Bmask == info.dst->Bmask &&
!(info.src->Amask != 0 && info.dst->Amask != 0 &&
info.src->Amask != info.dst->Amask) &&
- pg_HasSSE_NEON() && (src != dst)) {
+ PG_HAS_SSE2_OR_NEON() && (src != dst)) {
blit_blend_rgb_add_sse2(&info);
break;
}
-#endif /* PG_ENABLE_SSE_NEON */
-#endif /* SDL_BYTEORDER == SDL_LIL_ENDIAN */
-#endif /* __EMSCRIPTEN__ */
+#endif /* PG_HAS_SSE2_OR_NEON */
+#endif /* PG_SIMD_DISABLED */
blit_blend_add(&info);
break;
}
case PYGAME_BLEND_SUB: {
-#if !defined(__EMSCRIPTEN__)
-#if SDL_BYTEORDER == SDL_LIL_ENDIAN
+#ifndef PG_SIMD_DISABLED
+#ifdef PG_HAS_AVX2
if (PG_SURF_BytesPerPixel(src) == 4 &&
PG_SURF_BytesPerPixel(dst) == 4 &&
info.src->Rmask == info.dst->Rmask &&
@@ -267,11 +272,12 @@ SoftBlitPyGame(SDL_Surface *src, SDL_Rect *srcrect, SDL_Surface *dst,
info.src->Bmask == info.dst->Bmask &&
!(info.src->Amask != 0 && info.dst->Amask != 0 &&
info.src->Amask != info.dst->Amask) &&
- pg_has_avx2() && (src != dst)) {
+ PG_HAS_AVX2() && (src != dst)) {
blit_blend_rgb_sub_avx2(&info);
break;
}
-#if PG_ENABLE_SSE_NEON
+#endif /* PG_HAS_AVX2 */
+#ifdef PG_HAS_SSE2_OR_NEON
if (PG_SURF_BytesPerPixel(src) == 4 &&
PG_SURF_BytesPerPixel(dst) == 4 &&
info.src->Rmask == info.dst->Rmask &&
@@ -279,19 +285,18 @@ SoftBlitPyGame(SDL_Surface *src, SDL_Rect *srcrect, SDL_Surface *dst,
info.src->Bmask == info.dst->Bmask &&
!(info.src->Amask != 0 && info.dst->Amask != 0 &&
info.src->Amask != info.dst->Amask) &&
- pg_HasSSE_NEON() && (src != dst)) {
+ PG_HAS_SSE2_OR_NEON() && (src != dst)) {
blit_blend_rgb_sub_sse2(&info);
break;
}
-#endif /* PG_ENABLE_SSE_NEON */
-#endif /* SDL_BYTEORDER == SDL_LIL_ENDIAN */
-#endif /* __EMSCRIPTEN__ */
+#endif /* PG_HAS_SSE2_OR_NEON */
+#endif /* PG_SIMD_DISABLED */
blit_blend_sub(&info);
break;
}
case PYGAME_BLEND_MULT: {
-#if !defined(__EMSCRIPTEN__)
-#if SDL_BYTEORDER == SDL_LIL_ENDIAN
+#ifndef PG_SIMD_DISABLED
+#ifdef PG_HAS_AVX2
if (PG_SURF_BytesPerPixel(src) == 4 &&
PG_SURF_BytesPerPixel(dst) == 4 &&
info.src->Rmask == info.dst->Rmask &&
@@ -299,11 +304,12 @@ SoftBlitPyGame(SDL_Surface *src, SDL_Rect *srcrect, SDL_Surface *dst,
info.src->Bmask == info.dst->Bmask &&
!(info.src->Amask != 0 && info.dst->Amask != 0 &&
info.src->Amask != info.dst->Amask) &&
- pg_has_avx2() && (src != dst)) {
+ PG_HAS_AVX2() && (src != dst)) {
blit_blend_rgb_mul_avx2(&info);
break;
}
-#if PG_ENABLE_SSE_NEON
+#endif /* PG_HAS_AVX2 */
+#ifdef PG_HAS_SSE2_OR_NEON
if (PG_SURF_BytesPerPixel(src) == 4 &&
PG_SURF_BytesPerPixel(dst) == 4 &&
info.src->Rmask == info.dst->Rmask &&
@@ -311,19 +317,18 @@ SoftBlitPyGame(SDL_Surface *src, SDL_Rect *srcrect, SDL_Surface *dst,
info.src->Bmask == info.dst->Bmask &&
!(info.src->Amask != 0 && info.dst->Amask != 0 &&
info.src->Amask != info.dst->Amask) &&
- pg_HasSSE_NEON() && (src != dst)) {
+ PG_HAS_SSE2_OR_NEON() && (src != dst)) {
blit_blend_rgb_mul_sse2(&info);
break;
}
-#endif /* PG_ENABLE_SSE_NEON */
-#endif /* SDL_BYTEORDER == SDL_LIL_ENDIAN */
-#endif /* __EMSCRIPTEN__ */
+#endif /* PG_HAS_SSE2_OR_NEON */
+#endif /* PG_SIMD_DISABLED */
blit_blend_mul(&info);
break;
}
case PYGAME_BLEND_MIN: {
-#if !defined(__EMSCRIPTEN__)
-#if SDL_BYTEORDER == SDL_LIL_ENDIAN
+#ifndef PG_SIMD_DISABLED
+#ifdef PG_HAS_AVX2
if (PG_SURF_BytesPerPixel(src) == 4 &&
PG_SURF_BytesPerPixel(dst) == 4 &&
info.src->Rmask == info.dst->Rmask &&
@@ -331,11 +336,12 @@ SoftBlitPyGame(SDL_Surface *src, SDL_Rect *srcrect, SDL_Surface *dst,
info.src->Bmask == info.dst->Bmask &&
!(info.src->Amask != 0 && info.dst->Amask != 0 &&
info.src->Amask != info.dst->Amask) &&
- pg_has_avx2() && (src != dst)) {
+ PG_HAS_AVX2() && (src != dst)) {
blit_blend_rgb_min_avx2(&info);
break;
}
-#if PG_ENABLE_SSE_NEON
+#endif /* PG_HAS_AVX2 */
+#ifdef PG_HAS_SSE2_OR_NEON
if (PG_SURF_BytesPerPixel(src) == 4 &&
PG_SURF_BytesPerPixel(dst) == 4 &&
info.src->Rmask == info.dst->Rmask &&
@@ -343,19 +349,18 @@ SoftBlitPyGame(SDL_Surface *src, SDL_Rect *srcrect, SDL_Surface *dst,
info.src->Bmask == info.dst->Bmask &&
!(info.src->Amask != 0 && info.dst->Amask != 0 &&
info.src->Amask != info.dst->Amask) &&
- pg_HasSSE_NEON() && (src != dst)) {
+ PG_HAS_SSE2_OR_NEON() && (src != dst)) {
blit_blend_rgb_min_sse2(&info);
break;
}
-#endif /* PG_ENABLE_SSE_NEON */
-#endif /* SDL_BYTEORDER == SDL_LIL_ENDIAN */
-#endif /* __EMSCRIPTEN__ */
+#endif /* PG_HAS_SSE2_OR_NEON */
+#endif /* PG_SIMD_DISABLED */
blit_blend_min(&info);
break;
}
case PYGAME_BLEND_MAX: {
-#if !defined(__EMSCRIPTEN__)
-#if SDL_BYTEORDER == SDL_LIL_ENDIAN
+#ifndef PG_SIMD_DISABLED
+#ifdef PG_HAS_AVX2
if (PG_SURF_BytesPerPixel(src) == 4 &&
PG_SURF_BytesPerPixel(dst) == 4 &&
info.src->Rmask == info.dst->Rmask &&
@@ -363,11 +368,12 @@ SoftBlitPyGame(SDL_Surface *src, SDL_Rect *srcrect, SDL_Surface *dst,
info.src->Bmask == info.dst->Bmask &&
!(info.src->Amask != 0 && info.dst->Amask != 0 &&
info.src->Amask != info.dst->Amask) &&
- pg_has_avx2() && (src != dst)) {
+ PG_HAS_AVX2() && (src != dst)) {
blit_blend_rgb_max_avx2(&info);
break;
}
-#if PG_ENABLE_SSE_NEON
+#endif /* PG_HAS_AVX2 */
+#ifdef PG_HAS_SSE2_OR_NEON
if (PG_SURF_BytesPerPixel(src) == 4 &&
PG_SURF_BytesPerPixel(dst) == 4 &&
info.src->Rmask == info.dst->Rmask &&
@@ -375,181 +381,181 @@ SoftBlitPyGame(SDL_Surface *src, SDL_Rect *srcrect, SDL_Surface *dst,
info.src->Bmask == info.dst->Bmask &&
!(info.src->Amask != 0 && info.dst->Amask != 0 &&
info.src->Amask != info.dst->Amask) &&
- pg_HasSSE_NEON() && (src != dst)) {
+ PG_HAS_SSE2_OR_NEON() && (src != dst)) {
blit_blend_rgb_max_sse2(&info);
break;
}
-#endif /* PG_ENABLE_SSE_NEON */
-#endif /* SDL_BYTEORDER == SDL_LIL_ENDIAN */
-#endif /* __EMSCRIPTEN__ */
+#endif /* PG_HAS_SSE2_OR_NEON */
+#endif /* PG_SIMD_DISABLED */
blit_blend_max(&info);
break;
}
case PYGAME_BLEND_RGBA_ADD: {
-#if !defined(__EMSCRIPTEN__)
-#if SDL_BYTEORDER == SDL_LIL_ENDIAN
+#ifndef PG_SIMD_DISABLED
+#ifdef PG_HAS_AVX2
if (PG_SURF_BytesPerPixel(src) == 4 &&
PG_SURF_BytesPerPixel(dst) == 4 &&
info.src->Rmask == info.dst->Rmask &&
info.src->Gmask == info.dst->Gmask &&
info.src->Bmask == info.dst->Bmask &&
info.src_blend != SDL_BLENDMODE_NONE &&
- pg_has_avx2() && (src != dst)) {
+ PG_HAS_AVX2() && (src != dst)) {
blit_blend_rgba_add_avx2(&info);
break;
}
-#if PG_ENABLE_SSE_NEON
+#endif /* PG_HAS_AVX2 */
+#ifdef PG_HAS_SSE2_OR_NEON
if (PG_SURF_BytesPerPixel(src) == 4 &&
PG_SURF_BytesPerPixel(dst) == 4 &&
info.src->Rmask == info.dst->Rmask &&
info.src->Gmask == info.dst->Gmask &&
info.src->Bmask == info.dst->Bmask &&
info.src_blend != SDL_BLENDMODE_NONE &&
- pg_HasSSE_NEON() && (src != dst)) {
+ PG_HAS_SSE2_OR_NEON() && (src != dst)) {
blit_blend_rgba_add_sse2(&info);
break;
}
-#endif /* PG_ENABLE_SSE_NEON */
-#endif /* SDL_BYTEORDER == SDL_LIL_ENDIAN */
-#endif /* __EMSCRIPTEN__ */
+#endif /* PG_HAS_SSE2_OR_NEON */
+#endif /* PG_SIMD_DISABLED */
blit_blend_rgba_add(&info);
break;
}
case PYGAME_BLEND_RGBA_SUB: {
-#if !defined(__EMSCRIPTEN__)
-#if SDL_BYTEORDER == SDL_LIL_ENDIAN
+#ifndef PG_SIMD_DISABLED
+#ifdef PG_HAS_AVX2
if (PG_SURF_BytesPerPixel(src) == 4 &&
PG_SURF_BytesPerPixel(dst) == 4 &&
info.src->Rmask == info.dst->Rmask &&
info.src->Gmask == info.dst->Gmask &&
info.src->Bmask == info.dst->Bmask &&
info.src_blend != SDL_BLENDMODE_NONE &&
- pg_has_avx2() && (src != dst)) {
+ PG_HAS_AVX2() && (src != dst)) {
blit_blend_rgba_sub_avx2(&info);
break;
}
-#if PG_ENABLE_SSE_NEON
+#endif /* PG_HAS_AVX2 */
+#ifdef PG_HAS_SSE2_OR_NEON
if (PG_SURF_BytesPerPixel(src) == 4 &&
PG_SURF_BytesPerPixel(dst) == 4 &&
info.src->Rmask == info.dst->Rmask &&
info.src->Gmask == info.dst->Gmask &&
info.src->Bmask == info.dst->Bmask &&
info.src_blend != SDL_BLENDMODE_NONE &&
- pg_HasSSE_NEON() && (src != dst)) {
+ PG_HAS_SSE2_OR_NEON() && (src != dst)) {
blit_blend_rgba_sub_sse2(&info);
break;
}
-#endif /* PG_ENABLE_SSE_NEON */
-#endif /* SDL_BYTEORDER == SDL_LIL_ENDIAN */
-#endif /* __EMSCRIPTEN__ */
+#endif /* PG_HAS_SSE2_OR_NEON */
+#endif /* PG_SIMD_DISABLED */
blit_blend_rgba_sub(&info);
break;
}
case PYGAME_BLEND_RGBA_MULT: {
-#if !defined(__EMSCRIPTEN__)
-#if SDL_BYTEORDER == SDL_LIL_ENDIAN
+#ifndef PG_SIMD_DISABLED
+#ifdef PG_HAS_AVX2
if (PG_SURF_BytesPerPixel(src) == 4 &&
PG_SURF_BytesPerPixel(dst) == 4 &&
info.src->Rmask == info.dst->Rmask &&
info.src->Gmask == info.dst->Gmask &&
info.src->Bmask == info.dst->Bmask &&
info.src_blend != SDL_BLENDMODE_NONE &&
- pg_has_avx2() && (src != dst)) {
+ PG_HAS_AVX2() && (src != dst)) {
blit_blend_rgba_mul_avx2(&info);
break;
}
-#if PG_ENABLE_SSE_NEON
+#endif /* PG_HAS_AVX2 */
+#ifdef PG_HAS_SSE2_OR_NEON
if (PG_SURF_BytesPerPixel(src) == 4 &&
PG_SURF_BytesPerPixel(dst) == 4 &&
info.src->Rmask == info.dst->Rmask &&
info.src->Gmask == info.dst->Gmask &&
info.src->Bmask == info.dst->Bmask &&
info.src_blend != SDL_BLENDMODE_NONE &&
- pg_HasSSE_NEON() && (src != dst)) {
+ PG_HAS_SSE2_OR_NEON() && (src != dst)) {
blit_blend_rgba_mul_sse2(&info);
break;
}
-#endif /* PG_ENABLE_SSE_NEON */
-#endif /* SDL_BYTEORDER == SDL_LIL_ENDIAN */
-#endif /* __EMSCRIPTEN__ */
+#endif /* PG_HAS_SSE2_OR_NEON */
+#endif /* PG_SIMD_DISABLED */
blit_blend_rgba_mul(&info);
break;
}
case PYGAME_BLEND_RGBA_MIN: {
-#if !defined(__EMSCRIPTEN__)
-#if SDL_BYTEORDER == SDL_LIL_ENDIAN
+#ifndef PG_SIMD_DISABLED
+#ifdef PG_HAS_AVX2
if (PG_SURF_BytesPerPixel(src) == 4 &&
PG_SURF_BytesPerPixel(dst) == 4 &&
info.src->Rmask == info.dst->Rmask &&
info.src->Gmask == info.dst->Gmask &&
info.src->Bmask == info.dst->Bmask &&
info.src_blend != SDL_BLENDMODE_NONE &&
- pg_has_avx2() && (src != dst)) {
+ PG_HAS_AVX2() && (src != dst)) {
blit_blend_rgba_min_avx2(&info);
break;
}
-#if PG_ENABLE_SSE_NEON
+#endif /* PG_HAS_AVX2 */
+#ifdef PG_HAS_SSE2_OR_NEON
if (PG_SURF_BytesPerPixel(src) == 4 &&
PG_SURF_BytesPerPixel(dst) == 4 &&
info.src->Rmask == info.dst->Rmask &&
info.src->Gmask == info.dst->Gmask &&
info.src->Bmask == info.dst->Bmask &&
info.src_blend != SDL_BLENDMODE_NONE &&
- pg_HasSSE_NEON() && (src != dst)) {
+ PG_HAS_SSE2_OR_NEON() && (src != dst)) {
blit_blend_rgba_min_sse2(&info);
break;
}
-#endif /* PG_ENABLE_SSE_NEON */
-#endif /* SDL_BYTEORDER == SDL_LIL_ENDIAN */
-#endif /* __EMSCRIPTEN__ */
+#endif /* PG_HAS_SSE2_OR_NEON */
+#endif /* PG_SIMD_DISABLED */
blit_blend_rgba_min(&info);
break;
}
case PYGAME_BLEND_RGBA_MAX: {
-#if !defined(__EMSCRIPTEN__)
-#if SDL_BYTEORDER == SDL_LIL_ENDIAN
+#ifndef PG_SIMD_DISABLED
+#ifdef PG_HAS_AVX2
if (PG_SURF_BytesPerPixel(src) == 4 &&
PG_SURF_BytesPerPixel(dst) == 4 &&
info.src->Rmask == info.dst->Rmask &&
info.src->Gmask == info.dst->Gmask &&
info.src->Bmask == info.dst->Bmask &&
info.src_blend != SDL_BLENDMODE_NONE &&
- pg_has_avx2() && (src != dst)) {
+ PG_HAS_AVX2() && (src != dst)) {
blit_blend_rgba_max_avx2(&info);
break;
}
-#if PG_ENABLE_SSE_NEON
+#endif /* PG_HAS_AVX2 */
+#ifdef PG_HAS_SSE2_OR_NEON
if (PG_SURF_BytesPerPixel(src) == 4 &&
PG_SURF_BytesPerPixel(dst) == 4 &&
info.src->Rmask == info.dst->Rmask &&
info.src->Gmask == info.dst->Gmask &&
info.src->Bmask == info.dst->Bmask &&
info.src_blend != SDL_BLENDMODE_NONE &&
- pg_HasSSE_NEON() && (src != dst)) {
+ PG_HAS_SSE2_OR_NEON() && (src != dst)) {
blit_blend_rgba_max_sse2(&info);
break;
}
-#endif /* PG_ENABLE_SSE_NEON */
-#endif /* SDL_BYTEORDER == SDL_LIL_ENDIAN */
-#endif /* __EMSCRIPTEN__ */
+#endif /* PG_HAS_SSE2_OR_NEON */
+#endif /* PG_SIMD_DISABLED */
blit_blend_rgba_max(&info);
break;
}
case PYGAME_BLEND_PREMULTIPLIED: {
-#if !defined(__EMSCRIPTEN__)
-#if SDL_BYTEORDER == SDL_LIL_ENDIAN
+#ifndef PG_SIMD_DISABLED
+#ifdef PG_HAS_AVX2
if (PG_SURF_BytesPerPixel(src) == 4 &&
PG_SURF_BytesPerPixel(dst) == 4 &&
info.src->Rmask == info.dst->Rmask &&
info.src->Gmask == info.dst->Gmask &&
info.src->Bmask == info.dst->Bmask &&
info.src_blend != SDL_BLENDMODE_NONE &&
- pg_has_avx2() && (src != dst)) {
+ PG_HAS_AVX2() && (src != dst)) {
blit_blend_premultiplied_avx2(&info);
break;
}
-#if PG_ENABLE_SSE_NEON
+#endif /* PG_HAS_AVX2 */
+#ifdef PG_HAS_SSE2_OR_NEON
if (PG_SURF_BytesPerPixel(src) == 4 &&
PG_SURF_BytesPerPixel(dst) == 4 &&
info.src->Rmask == info.dst->Rmask &&
@@ -557,13 +563,12 @@ SoftBlitPyGame(SDL_Surface *src, SDL_Rect *srcrect, SDL_Surface *dst,
info.src->Bmask == info.dst->Bmask &&
info.src->Amask == 0xFF000000 &&
info.src_blend != SDL_BLENDMODE_NONE &&
- pg_HasSSE_NEON() && (src != dst)) {
+ PG_HAS_SSE2_OR_NEON() && (src != dst)) {
blit_blend_premultiplied_sse2(&info);
break;
}
-#endif /* PG_ENABLE_SSE_NEON */
-#endif /* SDL_BYTEORDER == SDL_LIL_ENDIAN */
-#endif /* __EMSCRIPTEN__ */
+#endif /* PG_HAS_SSE2_OR_NEON */
+#endif /* PG_SIMD_DISABLED */
blit_blend_premultiplied(&info);
break;
@@ -2958,26 +2963,20 @@ premul_surf_color_by_alpha(SDL_Surface *src, SDL_Surface *dst)
if (src_blend == SDL_BLENDMODE_NONE && !(src_format->Amask != 0))
return -1;
// since we know dst is a copy of src we can simplify the normal checks
-#if !defined(__EMSCRIPTEN__)
-#if SDL_BYTEORDER == SDL_LIL_ENDIAN
- if ((PG_SURF_BytesPerPixel(src) == 4) && pg_has_avx2()) {
+#ifndef PG_SIMD_DISABLED
+#if PG_HAS_AVX2
+ if ((PG_SURF_BytesPerPixel(src) == 4) && PG_HAS_AVX2()) {
premul_surf_color_by_alpha_avx2(src, dst);
return 0;
}
-#if defined(__SSE2__)
- if ((PG_SURF_BytesPerPixel(src) == 4) && SDL_HasSSE2()) {
- premul_surf_color_by_alpha_sse2(src, dst);
- return 0;
- }
-#endif /* __SSE2__*/
-#if PG_ENABLE_ARM_NEON
- if ((PG_SURF_BytesPerPixel(src) == 4) && SDL_HasNEON()) {
+#endif /* PG_HAS_AVX2 */
+#ifdef PG_HAS_SSE2_OR_NEON
+ if ((PG_SURF_BytesPerPixel(src) == 4) && PG_HAS_SSE2_OR_NEON()) {
premul_surf_color_by_alpha_sse2(src, dst);
return 0;
}
-#endif /* PG_ENABLE_ARM_NEON */
-#endif /* SDL_BYTEORDER == SDL_LIL_ENDIAN */
-#endif /* __EMSCRIPTEN__ */
+#endif /* PG_HAS_SSE2_OR_NEON */
+#endif /* PG_SIMD_DISABLED */
premul_surf_color_by_alpha_non_simd(src, src_format, src_palette, dst,
dst_format, dst_palette);
return 0;
diff --git a/src_c/simd_blitters.h b/src_c/simd_blitters.h
index 0c02207c26..0996271885 100644
--- a/src_c/simd_blitters.h
+++ b/src_c/simd_blitters.h
@@ -1,14 +1,9 @@
#define NO_PYGAME_C_API
#include "_surface.h"
#include "_blit_info.h"
+#include "simd_shared.h"
-#if !defined(PG_ENABLE_ARM_NEON) && defined(__aarch64__)
-// arm64 has neon optimisations enabled by default, even when fpu=neon is not
-// passed
-#define PG_ENABLE_ARM_NEON 1
-#endif
-
-#if (defined(__SSE2__) || defined(PG_ENABLE_ARM_NEON))
+#ifdef PG_HAS_SSE2_OR_NEON
void
alphablit_alpha_sse2_argb_surf_alpha(SDL_BlitInfo *info);
void
@@ -37,26 +32,11 @@ void
blit_blend_rgb_min_sse2(SDL_BlitInfo *info);
void
blit_blend_premultiplied_sse2(SDL_BlitInfo *info);
-#endif /* (defined(__SSE2__) || defined(PG_ENABLE_ARM_NEON)) */
-
-/* Deliberately putting these outside of the preprocessor guards as I want to
- move to a system of trusting the runtime checks to head to the right
- function and having a fallback function there if pygame is not compiled
- with the right stuff (this is the strategy used for AVX2 right now.
- Potentially I might want to shift both these into a slightly different
- file as they are not exactly blits (though v. similar) - or I could rename
- the SIMD trilogy of files to replace the word blit with something more
- generic like surface_ops*/
-
-void
-premul_surf_color_by_alpha_non_simd(SDL_Surface *src,
- PG_PixelFormat *src_format,
- SDL_Palette *src_palette, SDL_Surface *dst,
- PG_PixelFormat *dst_format,
- SDL_Palette *dst_palette);
void
premul_surf_color_by_alpha_sse2(SDL_Surface *src, SDL_Surface *dst);
+#endif /* PG_HAS_SSE2_OR_NEON */
+#ifdef PG_HAS_AVX2
void
alphablit_alpha_avx2_argb_no_surf_alpha_opaque_dst(SDL_BlitInfo *info);
void
@@ -87,3 +67,4 @@ void
blit_blend_premultiplied_avx2(SDL_BlitInfo *info);
void
premul_surf_color_by_alpha_avx2(SDL_Surface *src, SDL_Surface *dst);
+#endif
diff --git a/src_c/simd_blitters_avx2.c b/src_c/simd_blitters_avx2.c
index d113b2b87e..c2e4bf9340 100644
--- a/src_c/simd_blitters_avx2.c
+++ b/src_c/simd_blitters_avx2.c
@@ -1,44 +1,16 @@
#include "simd_blitters.h"
-#if defined(HAVE_IMMINTRIN_H) && !defined(SDL_DISABLE_IMMINTRIN_H)
-#include <immintrin.h>
-#endif /* defined(HAVE_IMMINTRIN_H) && !defined(SDL_DISABLE_IMMINTRIN_H) */
-
-#define BAD_AVX2_FUNCTION_CALL \
- printf( \
- "Fatal Error: Attempted calling an AVX2 function when both compile " \
- "time and runtime support is missing. If you are seeing this " \
- "message, you have stumbled across a pygame bug, please report it " \
- "to the devs!"); \
- PG_EXIT(1)
-
-/* helper function that does a runtime check for AVX2. It has the added
- * functionality of also returning 0 if compile time support is missing */
-int
-pg_has_avx2()
-{
-#if defined(__AVX2__) && defined(HAVE_IMMINTRIN_H) && \
- !defined(SDL_DISABLE_IMMINTRIN_H)
- return SDL_HasAVX2();
-#else
- return 0;
-#endif /* defined(__AVX2__) && defined(HAVE_IMMINTRIN_H) && \
- !defined(SDL_DISABLE_IMMINTRIN_H) */
-}
-
/* This returns 1 when avx2 is available at runtime but support for it isn't
* compiled in, 0 in all other cases */
int
pg_avx2_at_runtime_but_uncompiled()
{
if (SDL_HasAVX2()) {
-#if defined(__AVX2__) && defined(HAVE_IMMINTRIN_H) && \
- !defined(SDL_DISABLE_IMMINTRIN_H)
+#ifdef PG_HAS_AVX2
return 0;
#else
return 1;
-#endif /* defined(__AVX2__) && defined(HAVE_IMMINTRIN_H) && \
- !defined(SDL_DISABLE_IMMINTRIN_H) */
+#endif /* PG_HAS_AVX2 */
}
return 0;
}
@@ -190,8 +162,7 @@ pg_avx2_at_runtime_but_uncompiled()
_mm256_srli_epi16( \
_mm256_mulhi_epu16(MM256I, _mm256_set1_epi16((short)0x8081)), 7);
-#if defined(__AVX2__) && defined(HAVE_IMMINTRIN_H) && \
- !defined(SDL_DISABLE_IMMINTRIN_H)
+#ifdef PG_HAS_AVX2
void
alphablit_alpha_avx2_argb_no_surf_alpha_opaque_dst(SDL_BlitInfo *info)
{
@@ -258,17 +229,7 @@ alphablit_alpha_avx2_argb_no_surf_alpha_opaque_dst(SDL_BlitInfo *info)
* surfaces. */
pixels_dst = _mm256_and_si256(pixels_dst, mask_out_alpha);)
}
-#else
-void
-alphablit_alpha_avx2_argb_no_surf_alpha_opaque_dst(SDL_BlitInfo *info)
-{
- BAD_AVX2_FUNCTION_CALL;
-}
-#endif /* defined(__AVX2__) && defined(HAVE_IMMINTRIN_H) && \
- !defined(SDL_DISABLE_IMMINTRIN_H) */
-#if defined(__AVX2__) && defined(HAVE_IMMINTRIN_H) && \
- !defined(SDL_DISABLE_IMMINTRIN_H)
void
alphablit_alpha_avx2_argb_no_surf_alpha(SDL_BlitInfo *info)
{
@@ -324,17 +285,7 @@ alphablit_alpha_avx2_argb_no_surf_alpha(SDL_BlitInfo *info)
shuff_dst =
_mm256_blendv_epi8(shuff_dst, new_dst_alpha, combine_rgba_mask);))
}
-#else
-void
-alphablit_alpha_avx2_argb_no_surf_alpha(SDL_BlitInfo *info)
-{
- BAD_AVX2_FUNCTION_CALL;
-}
-#endif /* defined(__AVX2__) && defined(HAVE_IMMINTRIN_H) && \
- !defined(SDL_DISABLE_IMMINTRIN_H) */
-#if defined(__AVX2__) && defined(HAVE_IMMINTRIN_H) && \
- !defined(SDL_DISABLE_IMMINTRIN_H)
void
alphablit_alpha_avx2_argb_surf_alpha(SDL_BlitInfo *info)
{
@@ -406,17 +357,6 @@ alphablit_alpha_avx2_argb_surf_alpha(SDL_BlitInfo *info)
shuff_dst =
_mm256_blendv_epi8(shuff_dst, new_dst_alpha, combine_rgba_mask);))
}
-#else
-void
-alphablit_alpha_avx2_argb_surf_alpha(SDL_BlitInfo *info)
-{
- BAD_AVX2_FUNCTION_CALL;
-}
-#endif /* defined(__AVX2__) && defined(HAVE_IMMINTRIN_H) && \
- !defined(SDL_DISABLE_IMMINTRIN_H) */
-
-#if defined(__AVX2__) && defined(HAVE_IMMINTRIN_H) && \
- !defined(SDL_DISABLE_IMMINTRIN_H)
void
blit_blend_rgba_mul_avx2(SDL_BlitInfo *info)
{
@@ -524,17 +464,6 @@ blit_blend_rgba_mul_avx2(SDL_BlitInfo *info)
dstp = (Uint32 *)dstp256 + dstskip;
}
}
-#else
-void
-blit_blend_rgba_mul_avx2(SDL_BlitInfo *info)
-{
- BAD_AVX2_FUNCTION_CALL;
-}
-#endif /* defined(__AVX2__) && defined(HAVE_IMMINTRIN_H) && \
- !defined(SDL_DISABLE_IMMINTRIN_H) */
-
-#if defined(__AVX2__) && defined(HAVE_IMMINTRIN_H) && \
- !defined(SDL_DISABLE_IMMINTRIN_H)
void
blit_blend_rgb_mul_avx2(SDL_BlitInfo *info)
{
@@ -653,17 +582,6 @@ blit_blend_rgb_mul_avx2(SDL_BlitInfo *info)
dstp = (Uint32 *)dstp256 + dstskip;
}
}
-#else
-void
-blit_blend_rgb_mul_avx2(SDL_BlitInfo *info)
-{
- BAD_AVX2_FUNCTION_CALL;
-}
-#endif /* defined(__AVX2__) && defined(HAVE_IMMINTRIN_H) && \
- !defined(SDL_DISABLE_IMMINTRIN_H) */
-
-#if defined(__AVX2__) && defined(HAVE_IMMINTRIN_H) && \
- !defined(SDL_DISABLE_IMMINTRIN_H)
void
blit_blend_rgba_add_avx2(SDL_BlitInfo *info)
{
@@ -725,17 +643,6 @@ blit_blend_rgba_add_avx2(SDL_BlitInfo *info)
dstp = (Uint32 *)dstp256 + dstskip;
}
}
-#else
-void
-blit_blend_rgba_add_avx2(SDL_BlitInfo *info)
-{
- BAD_AVX2_FUNCTION_CALL;
-}
-#endif /* defined(__AVX2__) && defined(HAVE_IMMINTRIN_H) && \
- !defined(SDL_DISABLE_IMMINTRIN_H) */
-
-#if defined(__AVX2__) && defined(HAVE_IMMINTRIN_H) && \
- !defined(SDL_DISABLE_IMMINTRIN_H)
void
blit_blend_rgb_add_avx2(SDL_BlitInfo *info)
{
@@ -805,17 +712,6 @@ blit_blend_rgb_add_avx2(SDL_BlitInfo *info)
dstp = (Uint32 *)dstp256 + dstskip;
}
}
-#else
-void
-blit_blend_rgb_add_avx2(SDL_BlitInfo *info)
-{
- BAD_AVX2_FUNCTION_CALL;
-}
-#endif /* defined(__AVX2__) && defined(HAVE_IMMINTRIN_H) && \
- !defined(SDL_DISABLE_IMMINTRIN_H) */
-
-#if defined(__AVX2__) && defined(HAVE_IMMINTRIN_H) && \
- !defined(SDL_DISABLE_IMMINTRIN_H)
void
blit_blend_rgba_sub_avx2(SDL_BlitInfo *info)
{
@@ -877,17 +773,6 @@ blit_blend_rgba_sub_avx2(SDL_BlitInfo *info)
dstp = (Uint32 *)dstp256 + dstskip;
}
}
-#else
-void
-blit_blend_rgba_sub_avx2(SDL_BlitInfo *info)
-{
- BAD_AVX2_FUNCTION_CALL;
-}
-#endif /* defined(__AVX2__) && defined(HAVE_IMMINTRIN_H) && \
- !defined(SDL_DISABLE_IMMINTRIN_H) */
-
-#if defined(__AVX2__) && defined(HAVE_IMMINTRIN_H) && \
- !defined(SDL_DISABLE_IMMINTRIN_H)
void
blit_blend_rgb_sub_avx2(SDL_BlitInfo *info)
{
@@ -957,17 +842,6 @@ blit_blend_rgb_sub_avx2(SDL_BlitInfo *info)
dstp = (Uint32 *)dstp256 + dstskip;
}
}
-#else
-void
-blit_blend_rgb_sub_avx2(SDL_BlitInfo *info)
-{
- BAD_AVX2_FUNCTION_CALL;
-}
-#endif /* defined(__AVX2__) && defined(HAVE_IMMINTRIN_H) && \
- !defined(SDL_DISABLE_IMMINTRIN_H) */
-
-#if defined(__AVX2__) && defined(HAVE_IMMINTRIN_H) && \
- !defined(SDL_DISABLE_IMMINTRIN_H)
void
blit_blend_rgba_max_avx2(SDL_BlitInfo *info)
{
@@ -1029,17 +903,6 @@ blit_blend_rgba_max_avx2(SDL_BlitInfo *info)
dstp = (Uint32 *)dstp256 + dstskip;
}
}
-#else
-void
-blit_blend_rgba_max_avx2(SDL_BlitInfo *info)
-{
- BAD_AVX2_FUNCTION_CALL;
-}
-#endif /* defined(__AVX2__) && defined(HAVE_IMMINTRIN_H) && \
- !defined(SDL_DISABLE_IMMINTRIN_H) */
-
-#if defined(__AVX2__) && defined(HAVE_IMMINTRIN_H) && \
- !defined(SDL_DISABLE_IMMINTRIN_H)
void
blit_blend_rgb_max_avx2(SDL_BlitInfo *info)
{
@@ -1109,17 +972,6 @@ blit_blend_rgb_max_avx2(SDL_BlitInfo *info)
dstp = (Uint32 *)dstp256 + dstskip;
}
}
-#else
-void
-blit_blend_rgb_max_avx2(SDL_BlitInfo *info)
-{
- BAD_AVX2_FUNCTION_CALL;
-}
-#endif /* defined(__AVX2__) && defined(HAVE_IMMINTRIN_H) && \
- !defined(SDL_DISABLE_IMMINTRIN_H) */
-
-#if defined(__AVX2__) && defined(HAVE_IMMINTRIN_H) && \
- !defined(SDL_DISABLE_IMMINTRIN_H)
void
blit_blend_rgba_min_avx2(SDL_BlitInfo *info)
{
@@ -1181,17 +1033,6 @@ blit_blend_rgba_min_avx2(SDL_BlitInfo *info)
dstp = (Uint32 *)dstp256 + dstskip;
}
}
-#else
-void
-blit_blend_rgba_min_avx2(SDL_BlitInfo *info)
-{
- BAD_AVX2_FUNCTION_CALL;
-}
-#endif /* defined(__AVX2__) && defined(HAVE_IMMINTRIN_H) && \
- !defined(SDL_DISABLE_IMMINTRIN_H) */
-
-#if defined(__AVX2__) && defined(HAVE_IMMINTRIN_H) && \
- !defined(SDL_DISABLE_IMMINTRIN_H)
void
blit_blend_rgb_min_avx2(SDL_BlitInfo *info)
{
@@ -1261,17 +1102,6 @@ blit_blend_rgb_min_avx2(SDL_BlitInfo *info)
dstp = (Uint32 *)dstp256 + dstskip;
}
}
-#else
-void
-blit_blend_rgb_min_avx2(SDL_BlitInfo *info)
-{
- BAD_AVX2_FUNCTION_CALL;
-}
-#endif /* defined(__AVX2__) && defined(HAVE_IMMINTRIN_H) && \
- !defined(SDL_DISABLE_IMMINTRIN_H) */
-
-#if defined(__AVX2__) && defined(HAVE_IMMINTRIN_H) && \
- !defined(SDL_DISABLE_IMMINTRIN_H)
void
blit_blend_premultiplied_avx2(SDL_BlitInfo *info)
{
@@ -1521,14 +1351,6 @@ blit_blend_premultiplied_avx2(SDL_BlitInfo *info)
dstp = (Uint32 *)dstp256 + dstskip;
}
}
-#else
-void
-blit_blend_premultiplied_avx2(SDL_BlitInfo *info)
-{
- BAD_AVX2_FUNCTION_CALL;
-}
-#endif /* defined(__AVX2__) && defined(HAVE_IMMINTRIN_H) && \
- !defined(SDL_DISABLE_IMMINTRIN_H) */
#define PREMUL_ALPHA_CODE \
/* extract the alpha */ \
@@ -1558,8 +1380,6 @@ blit_blend_premultiplied_avx2(SDL_BlitInfo *info)
/*add the original alpha back in*/ \
mm_dst = _mm256_or_si256(mm_dst, mm_alpha_in);
-#if defined(__AVX2__) && defined(HAVE_IMMINTRIN_H) && \
- !defined(SDL_DISABLE_IMMINTRIN_H)
void
premul_surf_color_by_alpha_avx2(SDL_Surface *src, SDL_Surface *dst)
{
@@ -1635,11 +1455,4 @@ premul_surf_color_by_alpha_avx2(SDL_Surface *src, SDL_Surface *dst)
dstp += dst_skip;
}
}
-#else
-void
-premul_surf_color_by_alpha_avx2(SDL_Surface *src, SDL_Surface *dst)
-{
- BAD_AVX2_FUNCTION_CALL;
-}
-#endif /* defined(__AVX2__) && defined(HAVE_IMMINTRIN_H) && \
-!defined(SDL_DISABLE_IMMINTRIN_H) */
+#endif /* PG_HAS_AVX2 */
diff --git a/src_c/simd_blitters_sse2.c b/src_c/simd_blitters_sse2.c
index 65bb926721..ae5e2bbb95 100644
--- a/src_c/simd_blitters_sse2.c
+++ b/src_c/simd_blitters_sse2.c
@@ -1,24 +1,5 @@
#include "simd_blitters.h"
-#if PG_ENABLE_ARM_NEON
-// sse2neon.h is from here: https://github.com/DLTcollab/sse2neon
-#include "include/sse2neon.h"
-#endif /* PG_ENABLE_ARM_NEON */
-
-/* See if we are compiled 64 bit on GCC or MSVC */
-#if _WIN32 || _WIN64
-#if _WIN64
-#define ENV64BIT
-#endif
-#endif
-
-// Check GCC
-#if __GNUC__
-#if __x86_64__ || __ppc64__ || __aarch64__
-#define ENV64BIT
-#endif
-#endif
-
/* This returns 1 when sse2 is available at runtime but support for it isn't
* compiled in, 0 in all other cases */
int
@@ -139,7 +120,7 @@ pg_neon_at_runtime_but_uncompiled()
dstp = (Uint32 *)dstp128 + dstskip; \
}
-#if defined(__SSE2__) || defined(PG_ENABLE_ARM_NEON)
+#ifdef PG_HAS_SSE2_OR_NEON
void
alphablit_alpha_sse2_argb_surf_alpha(SDL_BlitInfo *info)
{
@@ -959,4 +940,4 @@ blit_blend_rgba_max_sse2(SDL_BlitInfo *info)
SETUP_SSE2_BLITTER
RUN_SSE2_BLITTER({ mm128_dst = _mm_max_epu8(mm128_dst, mm128_src); })
}
-#endif /* __SSE2__ || PG_ENABLE_ARM_NEON*/
+#endif /* PG_HAS_SSE2_OR_NEON */
diff --git a/src_c/simd_fill.h b/src_c/simd_fill.h
index db80008c1d..35b605efdd 100644
--- a/src_c/simd_fill.h
+++ b/src_c/simd_fill.h
@@ -1,44 +1,9 @@
#define NO_PYGAME_C_API
#include "_surface.h"
-
-#if !defined(PG_ENABLE_ARM_NEON) && defined(__aarch64__)
-// arm64 has neon optimisations enabled by default, even when fpu=neon is not
-// passed
-#define PG_ENABLE_ARM_NEON 1
-#endif
-
-/* See if we are compiled 64 bit on GCC or MSVC */
-#if _WIN32 || _WIN64
-#if _WIN64
-#define ENV64BIT
-#endif
-#endif
-
-// Check GCC
-#if __GNUC__
-#if __x86_64__ || __ppc64__ || __aarch64__
-#define ENV64BIT
-#endif
-#endif
-
-#if defined(__SSE2__)
-#define PG_ENABLE_SSE_NEON 1
-#elif PG_ENABLE_ARM_NEON
-#define PG_ENABLE_SSE_NEON 1
-#else
-#define PG_ENABLE_SSE_NEON 0
-#endif
-
-int
-_pg_has_avx2();
-
-/* This returns True if either SSE2 or NEON is present at runtime.
- * Relevant because they use the same codepaths. Only the relevant runtime
- * SDL cpu feature check is compiled in.*/
-int
-_pg_HasSSE_NEON();
+#include "simd_shared.h"
// AVX2 functions
+#ifdef PG_HAS_AVX2
int
surface_fill_blend_add_avx2(SDL_Surface *surface, SDL_Rect *rect,
Uint32 color);
@@ -70,7 +35,10 @@ surface_fill_blend_max_avx2(SDL_Surface *surface, SDL_Rect *rect,
int
surface_fill_blend_rgba_max_avx2(SDL_Surface *surface, SDL_Rect *rect,
Uint32 color);
+#endif
+
// SSE2 functions
+#ifdef PG_HAS_SSE2_OR_NEON
int
surface_fill_blend_add_sse2(SDL_Surface *surface, SDL_Rect *rect,
Uint32 color);
@@ -101,3 +69,4 @@ surface_fill_blend_max_sse2(SDL_Surface *surface, SDL_Rect *rect,
int
surface_fill_blend_rgba_max_sse2(SDL_Surface *surface, SDL_Rect *rect,
Uint32 color);
+#endif
diff --git a/src_c/simd_shared.h b/src_c/simd_shared.h
index a697fffb0c..ca7e9a76ab 100644
--- a/src_c/simd_shared.h
+++ b/src_c/simd_shared.h
@@ -10,8 +10,6 @@ int
pg_neon_at_runtime_but_uncompiled();
int
pg_avx2_at_runtime_but_uncompiled();
-int
-pg_has_avx2();
#if !defined(PG_ENABLE_ARM_NEON) && defined(__aarch64__)
// arm64 has neon optimisations enabled by default, even when fpu=neon is not
@@ -33,31 +31,27 @@ pg_has_avx2();
#endif
#endif
-/* This defines PG_ENABLE_SSE_NEON as True if either SSE or NEON is available
- * at compile time. Since we do compile time translation of SSE2->NEON, they
- * have the same code paths, so this reduces code duplication of those paths.
- */
-#if defined(__SSE2__)
-#define PG_ENABLE_SSE_NEON 1
-#elif PG_ENABLE_ARM_NEON
-#define PG_ENABLE_SSE_NEON 1
-#else
-#define PG_ENABLE_SSE_NEON 0
-#endif
+#if defined(__EMSCRIPTEN__) || SDL_BYTEORDER != SDL_LIL_ENDIAN
+#define PG_SIMD_DISABLED 1
+#endif /* PG_SIMD_DISABLED */
-/* This returns True if either SSE2 or NEON is present at runtime.
- * Relevant because they use the same codepaths. Only the relevant runtime
- * SDL cpu feature check is compiled in.*/
-int
-pg_HasSSE_NEON()
-{
-#if defined(__SSE2__)
- return SDL_HasSSE2();
+#ifndef PG_SIMD_DISABLED
+
+#ifdef __SSE2__
+#define PG_HAS_SSE2_OR_NEON SDL_HasSSE2
#elif PG_ENABLE_ARM_NEON
- return SDL_HasNEON();
-#else
- return 0;
-#endif
-}
+// sse2neon.h is from here: https://github.com/DLTcollab/sse2neon
+#include "include/sse2neon.h"
+#define PG_HAS_SSE2_OR_NEON SDL_HasNEON
+#endif /* PG_ENABLE_ARM_NEON */
+
+#if defined(__AVX2__) && defined(HAVE_IMMINTRIN_H) && \
+ !defined(SDL_DISABLE_IMMINTRIN_H)
+#include <immintrin.h>
+#define PG_HAS_AVX2 SDL_HasAVX2
+#endif /* defined(__AVX2__) && defined(HAVE_IMMINTRIN_H) && \
+ * !defined(SDL_DISABLE_IMMINTRIN_H) */
+
+#endif /* PG_SIMD_DISABLED */
#endif // SIMD_SHARED_H
diff --git a/src_c/simd_surface_fill_avx2.c b/src_c/simd_surface_fill_avx2.c
index 0c0d27ef21..0426b8f801 100644
--- a/src_c/simd_surface_fill_avx2.c
+++ b/src_c/simd_surface_fill_avx2.c
@@ -1,31 +1,5 @@
#include "simd_fill.h"
-#if defined(HAVE_IMMINTRIN_H) && !defined(SDL_DISABLE_IMMINTRIN_H)
-#include <immintrin.h>
-#endif /* defined(HAVE_IMMINTRIN_H) && !defined(SDL_DISABLE_IMMINTRIN_H) */
-
-#define BAD_AVX2_FUNCTION_CALL \
- printf( \
- "Fatal Error: Attempted calling an AVX2 function when both compile " \
- "time and runtime support is missing. If you are seeing this " \
- "message, you have stumbled across a pygame bug, please report it " \
- "to the devs!"); \
- PG_EXIT(1)
-
-/* helper function that does a runtime check for AVX2. It has the added
- * functionality of also returning 0 if compile time support is missing */
-int
-_pg_has_avx2()
-{
-#if defined(__AVX2__) && defined(HAVE_IMMINTRIN_H) && \
- !defined(SDL_DISABLE_IMMINTRIN_H)
- return SDL_HasAVX2();
-#else
- return 0;
-#endif /* defined(__AVX2__) && defined(HAVE_IMMINTRIN_H) && \
- !defined(SDL_DISABLE_IMMINTRIN_H) */
-}
-
#define SETUP_AVX2_FILLER(COLOR_PROCESS_CODE) \
/* initialize surface data */ \
int width = rect->w, height = rect->h; \
@@ -139,20 +113,6 @@ _pg_has_avx2()
return 0; \
}
-#define INVALID_DEFS(NAME) \
- int surface_fill_blend_##NAME##_avx2(SDL_Surface *surface, \
- SDL_Rect *rect, Uint32 color) \
- { \
- BAD_AVX2_FUNCTION_CALL; \
- return -1; \
- } \
- int surface_fill_blend_rgba_##NAME##_avx2(SDL_Surface *surface, \
- SDL_Rect *rect, Uint32 color) \
- { \
- BAD_AVX2_FUNCTION_CALL; \
- return -1; \
- }
-
#define ADD_CODE mm256_dst = _mm256_adds_epu8(mm256_dst, mm256_color);
#define SUB_CODE mm256_dst = _mm256_subs_epu8(mm256_dst, mm256_color);
#define MIN_CODE mm256_dst = _mm256_min_epu8(mm256_dst, mm256_color);
@@ -164,18 +124,10 @@ _pg_has_avx2()
shuff_dst = _mm256_srli_epi16(shuff_dst, 8); \
}
-#if defined(__AVX2__) && defined(HAVE_IMMINTRIN_H) && \
- !defined(SDL_DISABLE_IMMINTRIN_H)
+#ifdef PG_HAS_AVX2
FILLERS(add, color &= ~amask;, ADD_CODE)
FILLERS(sub, color &= ~amask;, SUB_CODE)
FILLERS(min, color |= amask;, MIN_CODE)
FILLERS(max, color &= ~amask;, MAX_CODE)
FILLERS_SHUFF(mult, color |= amask;, MULT_CODE)
-#else
-INVALID_DEFS(add)
-INVALID_DEFS(sub)
-INVALID_DEFS(min)
-INVALID_DEFS(max)
-INVALID_DEFS(mult)
-#endif /* defined(__AVX2__) && defined(HAVE_IMMINTRIN_H) && \
- !defined(SDL_DISABLE_IMMINTRIN_H) */
+#endif /* PG_HAS_AVX2 */
diff --git a/src_c/simd_surface_fill_sse2.c b/src_c/simd_surface_fill_sse2.c
index 7e4a80b030..9f0f23ad55 100644
--- a/src_c/simd_surface_fill_sse2.c
+++ b/src_c/simd_surface_fill_sse2.c
@@ -1,30 +1,5 @@
#include "simd_fill.h"
-#if PG_ENABLE_ARM_NEON
-// sse2neon.h is from here: https://github.com/DLTcollab/sse2neon
-#include "include/sse2neon.h"
-#endif /* PG_ENABLE_ARM_NEON */
-
-#define BAD_SSE2_FUNCTION_CALL \
- printf( \
- "Fatal Error: Attempted calling an SSE2 function when both compile " \
- "time and runtime support is missing. If you are seeing this " \
- "message, you have stumbled across a pygame bug, please report it " \
- "to the devs!"); \
- PG_EXIT(1)
-
-int
-_pg_HasSSE_NEON()
-{
-#if defined(__SSE2__)
- return SDL_HasSSE2();
-#elif PG_ENABLE_ARM_NEON
- return SDL_HasNEON();
-#else
- return 0;
-#endif
-}
-
#define SETUP_SSE2_FILLER(COLOR_PROCESS_CODE) \
/* initialize surface data */ \
int width = rect->w, height = rect->h; \
@@ -130,20 +105,6 @@ _pg_HasSSE_NEON()
return 0; \
}
-#define INVALID_DEFS(NAME) \
- int surface_fill_blend_##NAME##_sse2(SDL_Surface *surface, \
- SDL_Rect *rect, Uint32 color) \
- { \
- BAD_SSE2_FUNCTION_CALL; \
- return -1; \
- } \
- int surface_fill_blend_rgba_##NAME##_sse2(SDL_Surface *surface, \
- SDL_Rect *rect, Uint32 color) \
- { \
- BAD_SSE2_FUNCTION_CALL; \
- return -1; \
- }
-
#define ADD_CODE mm128_dst = _mm_adds_epu8(mm128_dst, mm128_color);
#define SUB_CODE mm128_dst = _mm_subs_epu8(mm128_dst, mm128_color);
#define MIN_CODE mm128_dst = _mm_min_epu8(mm128_dst, mm128_color);
@@ -155,16 +116,10 @@ _pg_HasSSE_NEON()
shuff_dst = _mm_srli_epi16(shuff_dst, 8); \
}
-#if defined(__SSE2__) || defined(PG_ENABLE_ARM_NEON)
+#ifdef PG_HAS_SSE2_OR_NEON
FILLERS(add, color &= ~amask;, ADD_CODE)
FILLERS(sub, color &= ~amask;, SUB_CODE)
FILLERS(min, color |= amask;, MIN_CODE)
FILLERS(max, color &= ~amask;, MAX_CODE)
FILLERS_SHUFF(mult, color |= amask;, MULT_CODE)
-#else
-INVALID_DEFS(add)
-INVALID_DEFS(sub)
-INVALID_DEFS(min)
-INVALID_DEFS(max)
-INVALID_DEFS(mult)
-#endif /* defined(__SSE2__) || defined(PG_ENABLE_ARM_NEON) */
+#endif /* PG_HAS_SSE2_OR_NEON */
diff --git a/src_c/simd_transform.h b/src_c/simd_transform.h
index 45b85f651f..b9882540c6 100644
--- a/src_c/simd_transform.h
+++ b/src_c/simd_transform.h
@@ -1,6 +1,8 @@
#define NO_PYGAME_C_API
#include "_surface.h"
+#include "simd_shared.h"
+
/**
* MACRO borrowed from SSE2NEON - useful for making the shuffling family of
* intrinsics easier to understand by indicating clearly what will go where.
@@ -16,14 +18,8 @@
#define _PG_SIMD_SHUFFLE(fp3, fp2, fp1, fp0) \
(((fp3) << 6) | ((fp2) << 4) | ((fp1) << 2) | ((fp0)))
-#if !defined(PG_ENABLE_ARM_NEON) && defined(__aarch64__)
-// arm64 has neon optimisations enabled by default, even when fpu=neon is not
-// passed
-#define PG_ENABLE_ARM_NEON 1
-#endif
-
// SSE2 functions
-#if defined(__SSE2__) || defined(PG_ENABLE_ARM_NEON)
+#ifdef PG_HAS_SSE2_OR_NEON
void
grayscale_sse2(SDL_Surface *src, SDL_Surface *newsurf);
@@ -43,10 +39,12 @@ filter_expand_Y_SSE2(Uint8 *srcpix, Uint8 *dstpix, int width, int srcpitch,
void
invert_sse2(SDL_Surface *src, SDL_Surface *newsurf);
-#endif /* (defined(__SSE2__) || defined(PG_ENABLE_ARM_NEON)) */
+#endif /* PG_HAS_SSE2_OR_NEON */
// AVX2 functions
+#ifdef PG_HAS_AVX2
void
grayscale_avx2(SDL_Surface *src, SDL_Surface *newsurf);
void
invert_avx2(SDL_Surface *src, SDL_Surface *newsurf);
+#endif /* PG_HAS_AVX2 */
diff --git a/src_c/simd_transform_avx2.c b/src_c/simd_transform_avx2.c
index 18c27ac355..ac458fc917 100644
--- a/src_c/simd_transform_avx2.c
+++ b/src_c/simd_transform_avx2.c
@@ -1,50 +1,6 @@
#include "simd_transform.h"
-#if defined(HAVE_IMMINTRIN_H) && !defined(SDL_DISABLE_IMMINTRIN_H)
-#include <immintrin.h>
-#endif /* defined(HAVE_IMMINTRIN_H) && !defined(SDL_DISABLE_IMMINTRIN_H) */
-
-#define BAD_AVX2_FUNCTION_CALL \
- printf( \
- "Fatal Error: Attempted calling an AVX2 function when both compile " \
- "time and runtime support is missing. If you are seeing this " \
- "message, you have stumbled across a pygame bug, please report it " \
- "to the devs!"); \
- PG_EXIT(1)
-
-/* helper function that does a runtime check for AVX2. It has the added
- * functionality of also returning 0 if compile time support is missing */
-int
-pg_has_avx2()
-{
-#if defined(__AVX2__) && defined(HAVE_IMMINTRIN_H) && \
- !defined(SDL_DISABLE_IMMINTRIN_H)
- return SDL_HasAVX2();
-#else
- return 0;
-#endif /* defined(__AVX2__) && defined(HAVE_IMMINTRIN_H) && \
- !defined(SDL_DISABLE_IMMINTRIN_H) */
-}
-
-/* This returns 1 when avx2 is available at runtime but support for it isn't
- * compiled in, 0 in all other cases */
-int
-pg_avx2_at_runtime_but_uncompiled()
-{
- if (SDL_HasAVX2()) {
-#if defined(__AVX2__) && defined(HAVE_IMMINTRIN_H) && \
- !defined(SDL_DISABLE_IMMINTRIN_H)
- return 0;
-#else
- return 1;
-#endif /* defined(__AVX2__) && defined(HAVE_IMMINTRIN_H) && \
- !defined(SDL_DISABLE_IMMINTRIN_H) */
- }
- return 0;
-}
-
-#if defined(__AVX2__) && defined(HAVE_IMMINTRIN_H) && \
- !defined(SDL_DISABLE_IMMINTRIN_H)
+#ifdef PG_HAS_AVX2
void
grayscale_avx2(SDL_Surface *src, SDL_Surface *newsurf)
{
@@ -298,16 +254,4 @@ invert_avx2(SDL_Surface *src, SDL_Surface *newsurf)
srcp256 = (__m256i *)srcp;
}
}
-#else
-void
-grayscale_avx2(SDL_Surface *src, SDL_Surface *newsurf)
-{
- BAD_AVX2_FUNCTION_CALL;
-}
-void
-invert_avx2(SDL_Surface *src, SDL_Surface *newsurf)
-{
- BAD_AVX2_FUNCTION_CALL;
-}
-#endif /* defined(__AVX2__) && defined(HAVE_IMMINTRIN_H) && \
- !defined(SDL_DISABLE_IMMINTRIN_H) */
+#endif /* PG_HAS_AVX2 */
diff --git a/src_c/simd_transform_sse2.c b/src_c/simd_transform_sse2.c
index 174ee09378..bf916549a5 100644
--- a/src_c/simd_transform_sse2.c
+++ b/src_c/simd_transform_sse2.c
@@ -1,41 +1,6 @@
#include "simd_transform.h"
-#if PG_ENABLE_ARM_NEON
-// sse2neon.h is from here: https://github.com/DLTcollab/sse2neon
-#include "include/sse2neon.h"
-#endif /* PG_ENABLE_ARM_NEON */
-
-/* This returns 1 when sse2 is available at runtime but support for it isn't
- * compiled in, 0 in all other cases */
-int
-pg_sse2_at_runtime_but_uncompiled()
-{
- if (SDL_HasSSE2()) {
-#ifdef __SSE2__
- return 0;
-#else
- return 1;
-#endif /* __SSE2__ */
- }
- return 0;
-}
-
-/* This returns 1 when neon is available at runtime but support for it isn't
- * compiled in, 0 in all other cases */
-int
-pg_neon_at_runtime_but_uncompiled()
-{
- if (SDL_HasNEON()) {
-#if PG_ENABLE_ARM_NEON
- return 0;
-#else
- return 1;
-#endif /* PG_ENABLE_ARM_NEON */
- }
- return 0;
-}
-
-#if (defined(__SSE2__) || defined(PG_ENABLE_ARM_NEON))
+#ifdef PG_HAS_SSE2_OR_NEON
// For some reason this is not defined on some non Windows compilers
#define _pg_loadu_si32(p) _mm_cvtsi32_si128(*(unsigned int const *)(p))
@@ -686,4 +651,4 @@ invert_sse2(SDL_Surface *src, SDL_Surface *newsurf)
}
}
-#endif /* __SSE2__ || PG_ENABLE_ARM_NEON*/
+#endif /* PG_HAS_SSE2_OR_NEON */
diff --git a/src_c/surface_fill.c b/src_c/surface_fill.c
index 6fa3f984dd..a536be09e3 100644
--- a/src_c/surface_fill.c
+++ b/src_c/surface_fill.c
@@ -922,215 +922,215 @@ surface_fill_blend(SDL_Surface *surface, SDL_Rect *rect, Uint32 color,
switch (blendargs) {
case PYGAME_BLEND_ADD: {
-#if !defined(__EMSCRIPTEN__)
-#if SDL_BYTEORDER == SDL_LIL_ENDIAN
+#ifndef PG_SIMD_DISABLED
if (PG_SURF_BytesPerPixel(surface) == 4) {
- if (_pg_has_avx2()) {
+#ifdef PG_HAS_AVX2
+ if (PG_HAS_AVX2()) {
result = surface_fill_blend_add_avx2(surface, rect, color);
break;
}
-#if PG_ENABLE_SSE_NEON
- if (_pg_HasSSE_NEON()) {
+#endif /* PG_HAS_AVX2 */
+#ifdef PG_HAS_SSE2_OR_NEON
+ if (PG_HAS_SSE2_OR_NEON()) {
result = surface_fill_blend_add_sse2(surface, rect, color);
break;
}
-#endif /* PG_ENABLE_SSE_NEON */
+#endif /* PG_HAS_SSE2_OR_NEON */
}
-#endif /* SDL_BYTEORDER == SDL_LIL_ENDIAN */
-#endif /* __EMSCRIPTEN__ */
+#endif /* PG_SIMD_DISABLED */
result = surface_fill_blend_add(surface, rect, color);
break;
}
case PYGAME_BLEND_SUB: {
-#if !defined(__EMSCRIPTEN__)
-#if SDL_BYTEORDER == SDL_LIL_ENDIAN
+#ifndef PG_SIMD_DISABLED
if (PG_SURF_BytesPerPixel(surface) == 4) {
- if (_pg_has_avx2()) {
+#ifdef PG_HAS_AVX2
+ if (PG_HAS_AVX2()) {
result = surface_fill_blend_sub_avx2(surface, rect, color);
break;
}
-#if PG_ENABLE_SSE_NEON
- if (_pg_HasSSE_NEON()) {
+#endif /* PG_HAS_AVX2 */
+#ifdef PG_HAS_SSE2_OR_NEON
+ if (PG_HAS_SSE2_OR_NEON()) {
result = surface_fill_blend_sub_sse2(surface, rect, color);
break;
}
-#endif /* PG_ENABLE_SSE_NEON */
+#endif /* PG_HAS_SSE2_OR_NEON */
}
-#endif /* SDL_BYTEORDER == SDL_LIL_ENDIAN */
-#endif /* __EMSCRIPTEN__ */
+#endif /* PG_SIMD_DISABLED */
result = surface_fill_blend_sub(surface, rect, color);
break;
}
case PYGAME_BLEND_MULT: {
-#if !defined(__EMSCRIPTEN__)
-#if SDL_BYTEORDER == SDL_LIL_ENDIAN
+#ifndef PG_SIMD_DISABLED
if (PG_SURF_BytesPerPixel(surface) == 4) {
- if (_pg_has_avx2()) {
+#ifdef PG_HAS_AVX2
+ if (PG_HAS_AVX2()) {
result =
surface_fill_blend_mult_avx2(surface, rect, color);
break;
}
-#if PG_ENABLE_SSE_NEON
- if (_pg_HasSSE_NEON()) {
+#endif /* PG_HAS_AVX2 */
+#ifdef PG_HAS_SSE2_OR_NEON
+ if (PG_HAS_SSE2_OR_NEON()) {
result =
surface_fill_blend_mult_sse2(surface, rect, color);
break;
}
-#endif /* PG_ENABLE_SSE_NEON */
+#endif /* PG_HAS_SSE2_OR_NEON */
}
-#endif /* SDL_BYTEORDER == SDL_LIL_ENDIAN */
-#endif /* __EMSCRIPTEN__ */
+#endif /* PG_SIMD_DISABLED */
result = surface_fill_blend_mult(surface, rect, color);
break;
}
case PYGAME_BLEND_MIN: {
-#if !defined(__EMSCRIPTEN__)
-#if SDL_BYTEORDER == SDL_LIL_ENDIAN
+#ifndef PG_SIMD_DISABLED
if (PG_SURF_BytesPerPixel(surface) == 4) {
- if (_pg_has_avx2()) {
+#ifdef PG_HAS_AVX2
+ if (PG_HAS_AVX2()) {
result = surface_fill_blend_min_avx2(surface, rect, color);
break;
}
-#if PG_ENABLE_SSE_NEON
- if (_pg_HasSSE_NEON()) {
+#endif /* PG_HAS_AVX2 */
+#ifdef PG_HAS_SSE2_OR_NEON
+ if (PG_HAS_SSE2_OR_NEON()) {
result = surface_fill_blend_min_sse2(surface, rect, color);
break;
}
-#endif /* PG_ENABLE_SSE_NEON */
+#endif /* PG_HAS_SSE2_OR_NEON */
}
-#endif /* SDL_BYTEORDER == SDL_LIL_ENDIAN */
-#endif /* __EMSCRIPTEN__ */
+#endif /* PG_SIMD_DISABLED */
result = surface_fill_blend_min(surface, rect, color);
break;
}
case PYGAME_BLEND_MAX: {
-#if !defined(__EMSCRIPTEN__)
-#if SDL_BYTEORDER == SDL_LIL_ENDIAN
+#ifndef PG_SIMD_DISABLED
if (PG_SURF_BytesPerPixel(surface) == 4) {
- if (_pg_has_avx2()) {
+#ifdef PG_HAS_AVX2
+ if (PG_HAS_AVX2()) {
result = surface_fill_blend_max_avx2(surface, rect, color);
break;
}
-#if PG_ENABLE_SSE_NEON
- if (_pg_HasSSE_NEON()) {
+#endif /* PG_HAS_AVX2 */
+#ifdef PG_HAS_SSE2_OR_NEON
+ if (PG_HAS_SSE2_OR_NEON()) {
result = surface_fill_blend_max_sse2(surface, rect, color);
break;
}
-#endif /* PG_ENABLE_SSE_NEON */
+#endif /* PG_HAS_SSE2_OR_NEON */
}
-#endif /* SDL_BYTEORDER == SDL_LIL_ENDIAN */
-#endif /* __EMSCRIPTEN__ */
+#endif /* PG_SIMD_DISABLED */
result = surface_fill_blend_max(surface, rect, color);
break;
}
case PYGAME_BLEND_RGBA_ADD: {
-#if !defined(__EMSCRIPTEN__)
-#if SDL_BYTEORDER == SDL_LIL_ENDIAN
+#ifndef PG_SIMD_DISABLED
if (PG_SURF_BytesPerPixel(surface) == 4) {
- if (_pg_has_avx2()) {
+#ifdef PG_HAS_AVX2
+ if (PG_HAS_AVX2()) {
result =
surface_fill_blend_rgba_add_avx2(surface, rect, color);
break;
}
-#if PG_ENABLE_SSE_NEON
- if (_pg_HasSSE_NEON()) {
+#endif /* PG_HAS_AVX2 */
+#ifdef PG_HAS_SSE2_OR_NEON
+ if (PG_HAS_SSE2_OR_NEON()) {
result =
surface_fill_blend_rgba_add_sse2(surface, rect, color);
break;
}
-#endif /* PG_ENABLE_SSE_NEON */
+#endif /* PG_HAS_SSE2_OR_NEON */
}
-#endif /* SDL_BYTEORDER == SDL_LIL_ENDIAN */
-#endif /* __EMSCRIPTEN__ */
+#endif /* PG_SIMD_DISABLED */
result = surface_fill_blend_rgba_add(surface, rect, color);
break;
}
case PYGAME_BLEND_RGBA_SUB: {
-#if !defined(__EMSCRIPTEN__)
-#if SDL_BYTEORDER == SDL_LIL_ENDIAN
+#ifndef PG_SIMD_DISABLED
if (PG_SURF_BytesPerPixel(surface) == 4) {
- if (_pg_has_avx2()) {
+#ifdef PG_HAS_AVX2
+ if (PG_HAS_AVX2()) {
result =
surface_fill_blend_rgba_sub_avx2(surface, rect, color);
break;
}
-#if PG_ENABLE_SSE_NEON
- if (_pg_HasSSE_NEON()) {
+#endif /* PG_HAS_AVX2 */
+#ifdef PG_HAS_SSE2_OR_NEON
+ if (PG_HAS_SSE2_OR_NEON()) {
result =
surface_fill_blend_rgba_sub_sse2(surface, rect, color);
break;
}
-#endif /* PG_ENABLE_SSE_NEON */
+#endif /* PG_HAS_SSE2_OR_NEON */
}
-#endif /* SDL_BYTEORDER == SDL_LIL_ENDIAN */
-#endif /* __EMSCRIPTEN__ */
+#endif /* PG_SIMD_DISABLED */
result = surface_fill_blend_rgba_sub(surface, rect, color);
break;
}
case PYGAME_BLEND_RGBA_MULT: {
-#if !defined(__EMSCRIPTEN__)
-#if SDL_BYTEORDER == SDL_LIL_ENDIAN
+#ifndef PG_SIMD_DISABLED
if (PG_SURF_BytesPerPixel(surface) == 4) {
- if (_pg_has_avx2()) {
+#ifdef PG_HAS_AVX2
+ if (PG_HAS_AVX2()) {
result = surface_fill_blend_rgba_mult_avx2(surface, rect,
color);
break;
}
-#if PG_ENABLE_SSE_NEON
- if (_pg_HasSSE_NEON()) {
+#endif /* PG_HAS_AVX2 */
+#ifdef PG_HAS_SSE2_OR_NEON
+ if (PG_HAS_SSE2_OR_NEON()) {
result = surface_fill_blend_rgba_mult_sse2(surface, rect,
color);
break;
}
-#endif /* PG_ENABLE_SSE_NEON */
+#endif /* PG_HAS_SSE2_OR_NEON */
}
-#endif /* SDL_BYTEORDER == SDL_LIL_ENDIAN */
-#endif /* __EMSCRIPTEN__ */
+#endif /* PG_SIMD_DISABLED */
result = surface_fill_blend_rgba_mult(surface, rect, color);
break;
}
case PYGAME_BLEND_RGBA_MIN: {
-#if !defined(__EMSCRIPTEN__)
-#if SDL_BYTEORDER == SDL_LIL_ENDIAN
+#ifndef PG_SIMD_DISABLED
if (PG_SURF_BytesPerPixel(surface) == 4) {
- if (_pg_has_avx2()) {
+#ifdef PG_HAS_AVX2
+ if (PG_HAS_AVX2()) {
result =
surface_fill_blend_rgba_min_avx2(surface, rect, color);
break;
}
-#if PG_ENABLE_SSE_NEON
- if (_pg_HasSSE_NEON()) {
+#endif /* PG_HAS_AVX2 */
+#ifdef PG_HAS_SSE2_OR_NEON
+ if (PG_HAS_SSE2_OR_NEON()) {
result =
surface_fill_blend_rgba_min_sse2(surface, rect, color);
break;
}
-#endif /* PG_ENABLE_SSE_NEON */
+#endif /* PG_HAS_SSE2_OR_NEON */
}
-#endif /* SDL_BYTEORDER == SDL_LIL_ENDIAN */
-#endif /* __EMSCRIPTEN__ */
+#endif /* PG_SIMD_DISABLED */
result = surface_fill_blend_rgba_min(surface, rect, color);
break;
}
case PYGAME_BLEND_RGBA_MAX: {
-#if !defined(__EMSCRIPTEN__)
-#if SDL_BYTEORDER == SDL_LIL_ENDIAN
+#ifndef PG_SIMD_DISABLED
if (PG_SURF_BytesPerPixel(surface) == 4) {
- if (_pg_has_avx2()) {
+#ifdef PG_HAS_AVX2
+ if (PG_HAS_AVX2()) {
result =
surface_fill_blend_rgba_max_avx2(surface, rect, color);
break;
}
-#if PG_ENABLE_SSE_NEON
- if (_pg_HasSSE_NEON()) {
+#endif /* PG_HAS_AVX2 */
+#ifdef PG_HAS_SSE2_OR_NEON
+ if (PG_HAS_SSE2_OR_NEON()) {
result =
surface_fill_blend_rgba_max_sse2(surface, rect, color);
break;
}
-#endif /* PG_ENABLE_SSE_NEON */
+#endif /* PG_HAS_SSE2_OR_NEON */
}
-#endif /* SDL_BYTEORDER == SDL_LIL_ENDIAN */
-#endif /* __EMSCRIPTEN__ */
+#endif /* PG_SIMD_DISABLED */
result = surface_fill_blend_rgba_max(surface, rect, color);
break;
}
diff --git a/src_c/transform.c b/src_c/transform.c
index ab69290b2d..41ec1e26d4 100644
--- a/src_c/transform.c
+++ b/src_c/transform.c
@@ -33,7 +33,6 @@
#include <math.h>
#include <string.h>
-#include "simd_shared.h"
#include "simd_transform.h"
#include "scale.h"
@@ -1255,8 +1254,8 @@ smoothscale_init(struct _module_state *st)
return;
}
-#if !defined(__EMSCRIPTEN__)
-#if PG_ENABLE_SSE_NEON
+#ifndef PG_SIMD_DISABLED
+#ifdef PG_HAS_SSE2_OR_NEON
if (SDL_HasSSE2()) {
st->filter_type = "SSE2";
st->filter_shrink_X = filter_shrink_X_SSE2;
@@ -1273,8 +1272,8 @@ smoothscale_init(struct _module_state *st)
st->filter_expand_Y = filter_expand_Y_SSE2;
return;
}
-#endif /* PG_ENABLE_SSE_NEON */
-#endif /* !__EMSCRIPTEN__ */
+#endif /* PG_HAS_SSE2_OR_NEON */
+#endif /* PG_SIMD_DISABLED */
#ifdef SCALE_MMX_SUPPORT
if (SDL_HasSSE()) {
st->filter_type = "SSE";
@@ -1655,8 +1654,8 @@ surf_set_smoothscale_backend(PyObject *self, PyObject *args, PyObject *kwargs)
"%s not supported on this machine", type);
}
#endif /* ~defined(SCALE_MMX_SUPPORT) */
-#if !defined(__EMSCRIPTEN__)
-#if PG_ENABLE_SSE_NEON
+#ifndef PG_SIMD_DISABLED
+#ifdef PG_HAS_SSE2_OR_NEON
else if (strcmp(type, "SSE2") == 0) {
if (!SDL_HasSSE2()) {
return RAISE(PyExc_ValueError,
@@ -1680,8 +1679,8 @@ surf_set_smoothscale_backend(PyObject *self, PyObject *args, PyObject *kwargs)
st->filter_expand_X = filter_expand_X_SSE2;
st->filter_expand_Y = filter_expand_Y_SSE2;
}
-#endif /* PG_ENABLE_SSE_NEON */
-#endif /* !__EMSCRIPTEN__ */
+#endif /* PG_HAS_SSE2_OR_NEON */
+#endif /* PG_SIMD_DISABLED */
else {
return PyErr_Format(PyExc_ValueError, "Unknown backend type %s", type);
}
@@ -2196,31 +2195,30 @@ grayscale(pgSurfaceObject *srcobj, pgSurfaceObject *dstobj)
return (SDL_Surface *)(RAISE(pgExc_SDLError, SDL_GetError()));
}
-#if defined(__EMSCRIPTEN__)
- grayscale_non_simd(src, src_format, newsurf, newsurf_format);
-#else // !defined(__EMSCRIPTEN__)
+#ifndef PG_DISABLE_SIMD
if (PG_FORMAT_BytesPerPixel(src_format) == 4 &&
src_format->Rmask == newsurf_format->Rmask &&
src_format->Gmask == newsurf_format->Gmask &&
src_format->Bmask == newsurf_format->Bmask && (src->pitch % 4 == 0) &&
(newsurf->pitch == (newsurf->w * 4))) {
- if (pg_has_avx2()) {
+#ifdef PG_HAS_AVX2
+ if (PG_HAS_AVX2()) {
grayscale_avx2(src, newsurf);
+ goto end;
}
-#if defined(__SSE2__) || defined(PG_ENABLE_ARM_NEON)
- else if (pg_HasSSE_NEON()) {
+#endif /* PG_HAS_AVX2 */
+#ifdef PG_HAS_SSE2_OR_NEON
+ if (PG_HAS_SSE2_OR_NEON()) {
grayscale_sse2(src, newsurf);
+ goto end;
}
-#endif // defined(__SSE2__) || defined(PG_ENABLE_ARM_NEON)
- else {
- grayscale_non_simd(src, src_format, newsurf, newsurf_format);
- }
+#endif /* PG_HAS_SSE2_OR_NEON */
}
- else {
- grayscale_non_simd(src, src_format, newsurf, newsurf_format);
- }
-#endif // !defined(__EMSCRIPTEN__)
+#endif /* PG_DISABLE_SIMD */
+ grayscale_non_simd(src, src_format, newsurf, newsurf_format);
+ goto end; /* Silence warning that the end label is not used. */
+end:
SDL_UnlockSurface(newsurf);
return newsurf;
@@ -4009,31 +4007,30 @@ invert(pgSurfaceObject *srcobj, pgSurfaceObject *dstobj)
return (SDL_Surface *)(RAISE(pgExc_SDLError, SDL_GetError()));
}
-#if defined(__EMSCRIPTEN__)
- invert_non_simd(src, src_format, newsurf, newsurf_format);
-#else // !defined(__EMSCRIPTEN__)
+#ifndef PG_DISABLE_SIMD
if (PG_FORMAT_BytesPerPixel(src_format) == 4 &&
src_format->Rmask == newsurf_format->Rmask &&
src_format->Gmask == newsurf_format->Gmask &&
src_format->Bmask == newsurf_format->Bmask && (src->pitch % 4 == 0) &&
(newsurf->pitch == (newsurf->w * 4))) {
- if (pg_has_avx2()) {
+#ifdef PG_HAS_AVX2
+ if (PG_HAS_AVX2()) {
invert_avx2(src, newsurf);
+ goto end;
}
-#if defined(__SSE2__) || defined(PG_ENABLE_ARM_NEON)
- else if (pg_HasSSE_NEON()) {
+#endif /* PG_HAS_AVX2 */
+#ifdef PG_HAS_SSE2_OR_NEON
+ if (PG_HAS_SSE2_OR_NEON()) {
invert_sse2(src, newsurf);
+ goto end;
}
-#endif // defined(__SSE2__) || defined(PG_ENABLE_ARM_NEON)
- else {
- invert_non_simd(src, src_format, newsurf, newsurf_format);
- }
+#endif /* PG_HAS_SSE2_OR_NEON */
}
- else {
- invert_non_simd(src, src_format, newsurf, newsurf_format);
- }
-#endif // !defined(__EMSCRIPTEN__)
+#endif // PG_DISABLE_SIMD
+ invert_non_simd(src, src_format, newsurf, newsurf_format);
+ goto end; /* Silence warning that the end label is not used. */
+end:
SDL_UnlockSurface(newsurf);
return newsurf;