From 73d95387f5bb98bdb67b8f69fb5b8979c682be2e Mon Sep 17 00:00:00 2001 From: Nigel Tao Date: Sun, 24 Jan 2021 22:04:53 +1100 Subject: [PATCH] Change some more s/sse128/sse42/ --- internal/cgen/base/fundamental-public.h | 22 ++--- internal/cgen/base/pixconv-submodule.c | 48 +++++------ internal/cgen/data/data.go | 52 ++++++------ internal/cgen/statement.go | 2 +- release/c/wuffs-unsupported-snapshot.c | 102 ++++++++++++------------ std/png/decode_filter_sse128.wuffs | 16 ++-- std/png/decode_png.wuffs | 8 +- 7 files changed, 125 insertions(+), 125 deletions(-) diff --git a/internal/cgen/base/fundamental-public.h b/internal/cgen/base/fundamental-public.h index 4655d231..bc47e823 100644 --- a/internal/cgen/base/fundamental-public.h +++ b/internal/cgen/base/fundamental-public.h @@ -70,12 +70,12 @@ // WUFFS_BASE__CPU_ARCH__X86_64__ETC are bits returned by // wuffs_base__cpu_arch__x86_64__capabilities. -// - "SSE128" means all of SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2 and POPCNT. -// - "AVX256" means all of AVX and AVX2. +// - "SSE42" means all of SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2 and POPCNT. +// - "AVX2" means all of AVX and AVX2. // - "AVX512ETC" is reserved, pending need. Note that AVX-512 consists of // multiple extensions that may be implemented independently. -#define WUFFS_BASE__CPU_ARCH__X86_64__SSE128 0x01 -#define WUFFS_BASE__CPU_ARCH__X86_64__AVX256 0x02 +#define WUFFS_BASE__CPU_ARCH__X86_64__SSE42 0x01 +#define WUFFS_BASE__CPU_ARCH__X86_64__AVX2 0x02 static inline uint32_t // wuffs_base__cpu_arch__x86_64__capabilities() { @@ -87,9 +87,9 @@ wuffs_base__cpu_arch__x86_64__capabilities() { unsigned int ecx1 = 0; unsigned int edx1 = 0; if (__get_cpuid(1, &eax1, &ebx1, &ecx1, &edx1)) { - const unsigned int sse128_ecx1 = bit_SSE4_2 | bit_POPCNT; - if ((ecx1 & sse128_ecx1) == sse128_ecx1) { - ret |= WUFFS_BASE__CPU_ARCH__X86_64__SSE128; + const unsigned int sse42_ecx1 = bit_SSE4_2 | bit_POPCNT; + if ((ecx1 & sse42_ecx1) == sse42_ecx1) { + ret |= WUFFS_BASE__CPU_ARCH__X86_64__SSE42; } } @@ -100,7 +100,7 @@ wuffs_base__cpu_arch__x86_64__capabilities() { if (__get_cpuid_count(7, 0, &eax7, &ebx7, &ecx7, &edx7)) { const unsigned int avx256_ebx7 = bit_AVX2; if ((ebx7 & avx256_ebx7) == avx256_ebx7) { - ret |= WUFFS_BASE__CPU_ARCH__X86_64__AVX256; + ret |= WUFFS_BASE__CPU_ARCH__X86_64__AVX2; } } @@ -111,15 +111,15 @@ wuffs_base__cpu_arch__x86_64__capabilities() { } static inline bool // -wuffs_base__cpu_arch__have_sse128() { +wuffs_base__cpu_arch__have_sse42() { #if defined(WUFFS_BASE__CPU_ARCH__X86_64) unsigned int eax1 = 0; unsigned int ebx1 = 0; unsigned int ecx1 = 0; unsigned int edx1 = 0; if (__get_cpuid(1, &eax1, &ebx1, &ecx1, &edx1)) { - const unsigned int sse128_ecx1 = bit_SSE4_2 | bit_POPCNT; - return (ecx1 & sse128_ecx1) == sse128_ecx1; + const unsigned int sse42_ecx1 = bit_SSE4_2 | bit_POPCNT; + return (ecx1 & sse42_ecx1) == sse42_ecx1; } #endif // defined(WUFFS_BASE__CPU_ARCH__X86_64) return false; diff --git a/internal/cgen/base/pixconv-submodule.c b/internal/cgen/base/pixconv-submodule.c index 0af33930..a0167abe 100644 --- a/internal/cgen/base/pixconv-submodule.c +++ b/internal/cgen/base/pixconv-submodule.c @@ -535,12 +535,12 @@ wuffs_base__pixel_swizzler__swap_rgb_bgr(uint8_t* dst_ptr, __attribute__((target("sse4.2"))) #endif static uint64_t // -wuffs_base__pixel_swizzler__swap_rgbx_bgrx__sse128(uint8_t* dst_ptr, - size_t dst_len, - uint8_t* dst_palette_ptr, - size_t dst_palette_len, - const uint8_t* src_ptr, - size_t src_len) { +wuffs_base__pixel_swizzler__swap_rgbx_bgrx__sse42(uint8_t* dst_ptr, + size_t dst_len, + uint8_t* dst_palette_ptr, + size_t dst_palette_len, + const uint8_t* src_ptr, + size_t src_len) { size_t len = (dst_len < src_len ? dst_len : src_len) / 4; uint8_t* d = dst_ptr; const uint8_t* s = src_ptr; @@ -1744,12 +1744,12 @@ wuffs_base__pixel_swizzler__bgrw__bgrx(uint8_t* dst_ptr, __attribute__((target("sse4.2"))) #endif static uint64_t // -wuffs_base__pixel_swizzler__bgrw__rgb__sse128(uint8_t* dst_ptr, - size_t dst_len, - uint8_t* dst_palette_ptr, - size_t dst_palette_len, - const uint8_t* src_ptr, - size_t src_len) { +wuffs_base__pixel_swizzler__bgrw__rgb__sse42(uint8_t* dst_ptr, + size_t dst_len, + uint8_t* dst_palette_ptr, + size_t dst_palette_len, + const uint8_t* src_ptr, + size_t src_len) { size_t dst_len4 = dst_len / 4; size_t src_len3 = src_len / 3; size_t len = (dst_len4 < src_len3) ? dst_len4 : src_len3; @@ -2122,12 +2122,12 @@ wuffs_base__pixel_swizzler__xxxx__index_binary_alpha__src_over( __attribute__((target("sse4.2"))) #endif static uint64_t // -wuffs_base__pixel_swizzler__xxxx__y__sse128(uint8_t* dst_ptr, - size_t dst_len, - uint8_t* dst_palette_ptr, - size_t dst_palette_len, - const uint8_t* src_ptr, - size_t src_len) { +wuffs_base__pixel_swizzler__xxxx__y__sse42(uint8_t* dst_ptr, + size_t dst_len, + uint8_t* dst_palette_ptr, + size_t dst_palette_len, + const uint8_t* src_ptr, + size_t src_len) { size_t dst_len4 = dst_len / 4; size_t len = (dst_len4 < src_len) ? dst_len4 : src_len; uint8_t* d = dst_ptr; @@ -2254,8 +2254,8 @@ wuffs_base__pixel_swizzler__prepare__y(wuffs_base__pixel_swizzler* p, case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY: case WUFFS_BASE__PIXEL_FORMAT__RGBX: #if defined(WUFFS_BASE__CPU_ARCH__X86_64) - if (wuffs_base__cpu_arch__have_sse128()) { - return wuffs_base__pixel_swizzler__xxxx__y__sse128; + if (wuffs_base__cpu_arch__have_sse42()) { + return wuffs_base__pixel_swizzler__xxxx__y__sse42; } #endif return wuffs_base__pixel_swizzler__xxxx__y; @@ -2559,8 +2559,8 @@ wuffs_base__pixel_swizzler__prepare__rgb(wuffs_base__pixel_swizzler* p, case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY: case WUFFS_BASE__PIXEL_FORMAT__BGRX: #if defined(WUFFS_BASE__CPU_ARCH__X86_64) - if (wuffs_base__cpu_arch__have_sse128()) { - return wuffs_base__pixel_swizzler__bgrw__rgb__sse128; + if (wuffs_base__cpu_arch__have_sse42()) { + return wuffs_base__pixel_swizzler__bgrw__rgb__sse42; } #endif return wuffs_base__pixel_swizzler__bgrw__rgb; @@ -2606,8 +2606,8 @@ wuffs_base__pixel_swizzler__prepare__rgba_nonpremul( switch (blend) { case WUFFS_BASE__PIXEL_BLEND__SRC: #if defined(WUFFS_BASE__CPU_ARCH__X86_64) - if (wuffs_base__cpu_arch__have_sse128()) { - return wuffs_base__pixel_swizzler__swap_rgbx_bgrx__sse128; + if (wuffs_base__cpu_arch__have_sse42()) { + return wuffs_base__pixel_swizzler__swap_rgbx_bgrx__sse42; } #endif return wuffs_base__pixel_swizzler__swap_rgbx_bgrx; diff --git a/internal/cgen/data/data.go b/internal/cgen/data/data.go index 0c872ab4..8357d536 100644 --- a/internal/cgen/data/data.go +++ b/internal/cgen/data/data.go @@ -58,8 +58,8 @@ const BaseFundamentalPublicH = "" + "" + "// --------\n\n// Define WUFFS_CONFIG__STATIC_FUNCTIONS to make all of Wuffs' functions have\n// static storage. The motivation is discussed in the \"ALLOW STATIC\n// IMPLEMENTATION\" section of\n// https://raw.githubusercontent.com/nothings/stb/master/docs/stb_howto.txt\n#if defined(WUFFS_CONFIG__STATIC_FUNCTIONS)\n#define WUFFS_BASE__MAYBE_STATIC static\n#else\n#define WUFFS_BASE__MAYBE_STATIC\n#endif // defined(WUFFS_CONFIG__STATIC_FUNCTIONS)\n\n" + "" + - "// ---------------- CPU Architecture\n\n// WUFFS_BASE__CPU_ARCH__X86_64__ETC are bits returned by\n// wuffs_base__cpu_arch__x86_64__capabilities.\n// - \"SSE128\" means all of SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2 and POPCNT.\n// - \"AVX256\" means all of AVX and AVX2.\n// - \"AVX512ETC\" is reserved, pending need. Note that AVX-512 consists of\n// multiple extensions that may be implemented independently.\n#define WUFFS_BASE__CPU_ARCH__X86_64__SSE128 0x01\n#define WUFFS_BASE__CPU_ARCH__X86_64__AVX256 0x02\n\nstatic inline uint32_t //\nwuffs_base__cpu_arch__x86_64__capabilities() {\n#if defined(WUFFS_BASE__CPU_ARCH__X86_64)\n uint32_t ret = 0;\n\n unsigned int eax1 = 0;\n unsigned int ebx1 = 0;\n unsigned int ecx1 = 0;\n unsigned int edx1 = 0;\n if (__get_cpuid(1, &eax1, &ebx1, &ecx1, &edx1)) {\n const unsigned int sse128_ecx1 = bit_SSE4_2 | bit_POPCNT;\n if ((ecx1 & sse128_ecx1) == sse128_ecx1) {\n ret |= WUFFS_BASE__CPU_ARCH__X86_64__SSE128;\n }\n }\n\n unsigned int eax7 = 0;\n unsigned int ebx7 = 0;\n unsigned in" + - "t ecx7 = 0;\n unsigned int edx7 = 0;\n if (__get_cpuid_count(7, 0, &eax7, &ebx7, &ecx7, &edx7)) {\n const unsigned int avx256_ebx7 = bit_AVX2;\n if ((ebx7 & avx256_ebx7) == avx256_ebx7) {\n ret |= WUFFS_BASE__CPU_ARCH__X86_64__AVX256;\n }\n }\n\n return ret;\n#else\n return 0;\n#endif // defined(WUFFS_BASE__CPU_ARCH__X86_64)\n}\n\nstatic inline bool //\nwuffs_base__cpu_arch__have_sse128() {\n#if defined(WUFFS_BASE__CPU_ARCH__X86_64)\n unsigned int eax1 = 0;\n unsigned int ebx1 = 0;\n unsigned int ecx1 = 0;\n unsigned int edx1 = 0;\n if (__get_cpuid(1, &eax1, &ebx1, &ecx1, &edx1)) {\n const unsigned int sse128_ecx1 = bit_SSE4_2 | bit_POPCNT;\n return (ecx1 & sse128_ecx1) == sse128_ecx1;\n }\n#endif // defined(WUFFS_BASE__CPU_ARCH__X86_64)\n return false;\n}\n\n" + + "// ---------------- CPU Architecture\n\n// WUFFS_BASE__CPU_ARCH__X86_64__ETC are bits returned by\n// wuffs_base__cpu_arch__x86_64__capabilities.\n// - \"SSE42\" means all of SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2 and POPCNT.\n// - \"AVX2\" means all of AVX and AVX2.\n// - \"AVX512ETC\" is reserved, pending need. Note that AVX-512 consists of\n// multiple extensions that may be implemented independently.\n#define WUFFS_BASE__CPU_ARCH__X86_64__SSE42 0x01\n#define WUFFS_BASE__CPU_ARCH__X86_64__AVX2 0x02\n\nstatic inline uint32_t //\nwuffs_base__cpu_arch__x86_64__capabilities() {\n#if defined(WUFFS_BASE__CPU_ARCH__X86_64)\n uint32_t ret = 0;\n\n unsigned int eax1 = 0;\n unsigned int ebx1 = 0;\n unsigned int ecx1 = 0;\n unsigned int edx1 = 0;\n if (__get_cpuid(1, &eax1, &ebx1, &ecx1, &edx1)) {\n const unsigned int sse42_ecx1 = bit_SSE4_2 | bit_POPCNT;\n if ((ecx1 & sse42_ecx1) == sse42_ecx1) {\n ret |= WUFFS_BASE__CPU_ARCH__X86_64__SSE42;\n }\n }\n\n unsigned int eax7 = 0;\n unsigned int ebx7 = 0;\n unsigned int ecx7 = " + + "0;\n unsigned int edx7 = 0;\n if (__get_cpuid_count(7, 0, &eax7, &ebx7, &ecx7, &edx7)) {\n const unsigned int avx256_ebx7 = bit_AVX2;\n if ((ebx7 & avx256_ebx7) == avx256_ebx7) {\n ret |= WUFFS_BASE__CPU_ARCH__X86_64__AVX2;\n }\n }\n\n return ret;\n#else\n return 0;\n#endif // defined(WUFFS_BASE__CPU_ARCH__X86_64)\n}\n\nstatic inline bool //\nwuffs_base__cpu_arch__have_sse42() {\n#if defined(WUFFS_BASE__CPU_ARCH__X86_64)\n unsigned int eax1 = 0;\n unsigned int ebx1 = 0;\n unsigned int ecx1 = 0;\n unsigned int edx1 = 0;\n if (__get_cpuid(1, &eax1, &ebx1, &ecx1, &edx1)) {\n const unsigned int sse42_ecx1 = bit_SSE4_2 | bit_POPCNT;\n return (ecx1 & sse42_ecx1) == sse42_ecx1;\n }\n#endif // defined(WUFFS_BASE__CPU_ARCH__X86_64)\n return false;\n}\n\n" + "" + "// ---------------- Fundamentals\n\n// Wuffs assumes that:\n// - converting a uint32_t to a size_t will never overflow.\n// - converting a size_t to a uint64_t will never overflow.\n#ifdef __WORDSIZE\n#if (__WORDSIZE != 32) && (__WORDSIZE != 64)\n#error \"Wuffs requires a word size of either 32 or 64 bits\"\n#endif\n#endif\n\n#if defined(__clang__)\n#define WUFFS_BASE__POTENTIALLY_UNUSED_FIELD __attribute__((unused))\n#else\n#define WUFFS_BASE__POTENTIALLY_UNUSED_FIELD\n#endif\n\n// Clang also defines \"__GNUC__\".\n#if defined(__GNUC__)\n#define WUFFS_BASE__POTENTIALLY_UNUSED __attribute__((unused))\n#define WUFFS_BASE__WARN_UNUSED_RESULT __attribute__((warn_unused_result))\n#else\n#define WUFFS_BASE__POTENTIALLY_UNUSED\n#define WUFFS_BASE__WARN_UNUSED_RESULT\n#endif\n\n" + "" + @@ -557,9 +557,9 @@ const BasePixConvSubmoduleC = "" + " >>= 8;\n dg >>= 8;\n db >>= 8;\n\n // Combine components.\n return (db << 0) | (dg << 8) | (dr << 16) | (da << 24);\n}\n\n" + "" + "// --------\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__squash_align4_bgr_565_888(uint8_t* dst_ptr,\n size_t dst_len,\n uint8_t* dst_palette_ptr,\n size_t dst_palette_len,\n const uint8_t* src_ptr,\n size_t src_len) {\n size_t len = (dst_len < src_len ? dst_len : src_len) / 4;\n uint8_t* d = dst_ptr;\n const uint8_t* s = src_ptr;\n\n size_t n = len;\n while (n--) {\n uint32_t argb = wuffs_base__peek_u32le__no_bounds_check(s);\n uint32_t b5 = 0x1F & (argb >> (8 - 5));\n uint32_t g6 = 0x3F & (argb >> (16 - 6));\n uint32_t r5 = 0x1F & (argb >> (24 - 5));\n uint32_t alpha = argb & 0xFF000000;\n wuffs_base__poke_u32le__no_bounds_check(\n d, alpha | (r5 << 11) | (g6 << 5) | (b5 << 0));\n s += 4;\n d += 4;\n }\n return len;\n}\n\nstatic u" + - "int64_t //\nwuffs_base__pixel_swizzler__swap_rgb_bgr(uint8_t* dst_ptr,\n size_t dst_len,\n uint8_t* dst_palette_ptr,\n size_t dst_palette_len,\n const uint8_t* src_ptr,\n size_t src_len) {\n size_t len = (dst_len < src_len ? dst_len : src_len) / 3;\n uint8_t* d = dst_ptr;\n const uint8_t* s = src_ptr;\n\n size_t n = len;\n while (n--) {\n uint8_t b0 = s[0];\n uint8_t b1 = s[1];\n uint8_t b2 = s[2];\n d[0] = b2;\n d[1] = b1;\n d[2] = b0;\n s += 3;\n d += 3;\n }\n return len;\n}\n\n#if defined(WUFFS_BASE__CPU_ARCH__X86_64)\n#if defined(__GNUC__)\n__attribute__((target(\"sse4.2\")))\n#endif\nstatic uint64_t //\nwuffs_base__pixel_swizzler__swap_rgbx_bgrx__sse128(uint8_t* dst_ptr,\n size_t dst_len,\n uint8_t* dst_palette" + - "_ptr,\n size_t dst_palette_len,\n const uint8_t* src_ptr,\n size_t src_len) {\n size_t len = (dst_len < src_len ? dst_len : src_len) / 4;\n uint8_t* d = dst_ptr;\n const uint8_t* s = src_ptr;\n size_t n = len;\n\n __m128i shuffle = _mm_set_epi8(+0x0F, +0x0C, +0x0D, +0x0E, //\n +0x0B, +0x08, +0x09, +0x0A, //\n +0x07, +0x04, +0x05, +0x06, //\n +0x03, +0x00, +0x01, +0x02);\n\n while (n >= 4) {\n __m128i x;\n x = _mm_lddqu_si128((const __m128i*)(const void*)s);\n x = _mm_shuffle_epi8(x, shuffle);\n _mm_storeu_si128((__m128i*)(void*)d, x);\n\n s += 4 * 4;\n d += 4 * 4;\n n -= 4;\n }\n\n while (n--) {\n uint8_t b0 = s[0];\n uint8_t b1 = s[1];\n uint8_t b2 = s[2];\n uint8_t b3 = s[3];\n d[0] = b2;\n d[1] = b1;\n d[2] = b0;\n d[3] = b3;\n s += 4;\n " + - "d += 4;\n }\n return len;\n}\n#endif // defined(WUFFS_BASE__CPU_ARCH__X86_64)\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__swap_rgbx_bgrx(uint8_t* dst_ptr,\n size_t dst_len,\n uint8_t* dst_palette_ptr,\n size_t dst_palette_len,\n const uint8_t* src_ptr,\n size_t src_len) {\n size_t len = (dst_len < src_len ? dst_len : src_len) / 4;\n uint8_t* d = dst_ptr;\n const uint8_t* s = src_ptr;\n\n size_t n = len;\n while (n--) {\n uint8_t b0 = s[0];\n uint8_t b1 = s[1];\n uint8_t b2 = s[2];\n uint8_t b3 = s[3];\n d[0] = b2;\n d[1] = b1;\n d[2] = b0;\n d[3] = b3;\n s += 4;\n d += 4;\n }\n return len;\n}\n\n" + + "int64_t //\nwuffs_base__pixel_swizzler__swap_rgb_bgr(uint8_t* dst_ptr,\n size_t dst_len,\n uint8_t* dst_palette_ptr,\n size_t dst_palette_len,\n const uint8_t* src_ptr,\n size_t src_len) {\n size_t len = (dst_len < src_len ? dst_len : src_len) / 3;\n uint8_t* d = dst_ptr;\n const uint8_t* s = src_ptr;\n\n size_t n = len;\n while (n--) {\n uint8_t b0 = s[0];\n uint8_t b1 = s[1];\n uint8_t b2 = s[2];\n d[0] = b2;\n d[1] = b1;\n d[2] = b0;\n s += 3;\n d += 3;\n }\n return len;\n}\n\n#if defined(WUFFS_BASE__CPU_ARCH__X86_64)\n#if defined(__GNUC__)\n__attribute__((target(\"sse4.2\")))\n#endif\nstatic uint64_t //\nwuffs_base__pixel_swizzler__swap_rgbx_bgrx__sse42(uint8_t* dst_ptr,\n size_t dst_len,\n uint8_t* dst_palette_pt" + + "r,\n size_t dst_palette_len,\n const uint8_t* src_ptr,\n size_t src_len) {\n size_t len = (dst_len < src_len ? dst_len : src_len) / 4;\n uint8_t* d = dst_ptr;\n const uint8_t* s = src_ptr;\n size_t n = len;\n\n __m128i shuffle = _mm_set_epi8(+0x0F, +0x0C, +0x0D, +0x0E, //\n +0x0B, +0x08, +0x09, +0x0A, //\n +0x07, +0x04, +0x05, +0x06, //\n +0x03, +0x00, +0x01, +0x02);\n\n while (n >= 4) {\n __m128i x;\n x = _mm_lddqu_si128((const __m128i*)(const void*)s);\n x = _mm_shuffle_epi8(x, shuffle);\n _mm_storeu_si128((__m128i*)(void*)d, x);\n\n s += 4 * 4;\n d += 4 * 4;\n n -= 4;\n }\n\n while (n--) {\n uint8_t b0 = s[0];\n uint8_t b1 = s[1];\n uint8_t b2 = s[2];\n uint8_t b3 = s[3];\n d[0] = b2;\n d[1] = b1;\n d[2] = b0;\n d[3] = b3;\n s += 4;\n d += 4" + + ";\n }\n return len;\n}\n#endif // defined(WUFFS_BASE__CPU_ARCH__X86_64)\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__swap_rgbx_bgrx(uint8_t* dst_ptr,\n size_t dst_len,\n uint8_t* dst_palette_ptr,\n size_t dst_palette_len,\n const uint8_t* src_ptr,\n size_t src_len) {\n size_t len = (dst_len < src_len ? dst_len : src_len) / 4;\n uint8_t* d = dst_ptr;\n const uint8_t* s = src_ptr;\n\n size_t n = len;\n while (n--) {\n uint8_t b0 = s[0];\n uint8_t b1 = s[1];\n uint8_t b2 = s[2];\n uint8_t b3 = s[3];\n d[0] = b2;\n d[1] = b1;\n d[2] = b0;\n d[3] = b3;\n s += 4;\n d += 4;\n }\n return len;\n}\n\n" + "" + "// --------\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__squash_tight_4x8_4x16le(uint8_t* dst_ptr,\n size_t dst_len,\n uint8_t* dst_palette_ptr,\n size_t dst_palette_len,\n const uint8_t* src_ptr,\n size_t src_len) {\n size_t dst_len4 = dst_len / 4;\n size_t src_len8 = src_len / 8;\n size_t len = (dst_len4 < src_len8) ? dst_len4 : src_len8;\n uint8_t* d = dst_ptr;\n const uint8_t* s = src_ptr;\n\n size_t n = len;\n while (n >= 1) {\n wuffs_base__poke_u32le__no_bounds_check(\n d + (0 * 4), wuffs_base__color_u64__as__color_u32(\n wuffs_base__peek_u64le__no_bounds_check(s + (0 * 8))));\n\n s += 1 * 8;\n d += 1 * 4;\n n -= 1;\n }\n return len;\n}\n\n" + "" + @@ -600,10 +600,10 @@ const BasePixConvSubmoduleC = "" + "ler__bgra_premul__rgba_nonpremul__src_over(\n uint8_t* dst_ptr,\n size_t dst_len,\n uint8_t* dst_palette_ptr,\n size_t dst_palette_len,\n const uint8_t* src_ptr,\n size_t src_len) {\n size_t dst_len4 = dst_len / 4;\n size_t src_len4 = src_len / 4;\n size_t len = (dst_len4 < src_len4) ? dst_len4 : src_len4;\n uint8_t* d = dst_ptr;\n const uint8_t* s = src_ptr;\n size_t n = len;\n\n // TODO: unroll.\n\n while (n >= 1) {\n uint32_t d0 = wuffs_base__peek_u32le__no_bounds_check(d + (0 * 4));\n uint32_t s0 = wuffs_base__swap_u32_argb_abgr(\n wuffs_base__peek_u32le__no_bounds_check(s + (0 * 4)));\n wuffs_base__poke_u32le__no_bounds_check(\n d + (0 * 4), wuffs_base__composite_premul_nonpremul_u32_axxx(d0, s0));\n\n s += 1 * 4;\n d += 1 * 4;\n n -= 1;\n }\n\n return len;\n}\n\n" + "" + "// --------\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__bgrw__bgr(uint8_t* dst_ptr,\n size_t dst_len,\n uint8_t* dst_palette_ptr,\n size_t dst_palette_len,\n const uint8_t* src_ptr,\n size_t src_len) {\n size_t dst_len4 = dst_len / 4;\n size_t src_len3 = src_len / 3;\n size_t len = (dst_len4 < src_len3) ? dst_len4 : src_len3;\n uint8_t* d = dst_ptr;\n const uint8_t* s = src_ptr;\n size_t n = len;\n\n // TODO: unroll.\n\n while (n >= 1) {\n wuffs_base__poke_u32le__no_bounds_check(\n d + (0 * 4),\n 0xFF000000 | wuffs_base__peek_u24le__no_bounds_check(s + (0 * 3)));\n\n s += 1 * 3;\n d += 1 * 4;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__bgrw__bgrx(uint8_t* dst_ptr,\n size_t dst_len,\n uint8_t* dst_palett" + - "e_ptr,\n size_t dst_palette_len,\n const uint8_t* src_ptr,\n size_t src_len) {\n size_t dst_len4 = dst_len / 4;\n size_t src_len4 = src_len / 4;\n size_t len = (dst_len4 < src_len4) ? dst_len4 : src_len4;\n uint8_t* d = dst_ptr;\n const uint8_t* s = src_ptr;\n size_t n = len;\n\n // TODO: unroll.\n\n while (n >= 1) {\n wuffs_base__poke_u32le__no_bounds_check(\n d + (0 * 4),\n 0xFF000000 | wuffs_base__peek_u32le__no_bounds_check(s + (0 * 4)));\n\n s += 1 * 4;\n d += 1 * 4;\n n -= 1;\n }\n\n return len;\n}\n\n#if defined(WUFFS_BASE__CPU_ARCH__X86_64)\n#if defined(__GNUC__)\n__attribute__((target(\"sse4.2\")))\n#endif\nstatic uint64_t //\nwuffs_base__pixel_swizzler__bgrw__rgb__sse128(uint8_t* dst_ptr,\n size_t dst_len,\n uint8_t* dst_palette_ptr,\n size_t dst_palette" + - "_len,\n const uint8_t* src_ptr,\n size_t src_len) {\n size_t dst_len4 = dst_len / 4;\n size_t src_len3 = src_len / 3;\n size_t len = (dst_len4 < src_len3) ? dst_len4 : src_len3;\n uint8_t* d = dst_ptr;\n const uint8_t* s = src_ptr;\n size_t n = len;\n\n __m128i shuffle = _mm_set_epi8(+0x00, +0x09, +0x0A, +0x0B, //\n +0x00, +0x06, +0x07, +0x08, //\n +0x00, +0x03, +0x04, +0x05, //\n +0x00, +0x00, +0x01, +0x02);\n __m128i or_ff = _mm_set_epi8(-0x01, +0x00, +0x00, +0x00, //\n -0x01, +0x00, +0x00, +0x00, //\n -0x01, +0x00, +0x00, +0x00, //\n -0x01, +0x00, +0x00, +0x00);\n\n while (n >= 6) {\n __m128i x;\n x = _mm_lddqu_si128((const __m128i*)(const void*)s);\n x = _mm_shuffle_epi8(x, shuffle);\n x = _mm_or_si128(x, or_ff);\n _mm_storeu_si" + - "128((__m128i*)(void*)d, x);\n\n s += 4 * 3;\n d += 4 * 4;\n n -= 4;\n }\n\n while (n >= 1) {\n uint8_t b0 = s[0];\n uint8_t b1 = s[1];\n uint8_t b2 = s[2];\n d[0] = b2;\n d[1] = b1;\n d[2] = b0;\n d[3] = 0xFF;\n\n s += 1 * 3;\n d += 1 * 4;\n n -= 1;\n }\n\n return len;\n}\n#endif // defined(WUFFS_BASE__CPU_ARCH__X86_64)\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__bgrw__rgb(uint8_t* dst_ptr,\n size_t dst_len,\n uint8_t* dst_palette_ptr,\n size_t dst_palette_len,\n const uint8_t* src_ptr,\n size_t src_len) {\n size_t dst_len4 = dst_len / 4;\n size_t src_len3 = src_len / 3;\n size_t len = (dst_len4 < src_len3) ? dst_len4 : src_len3;\n uint8_t* d = dst_ptr;\n const uint8_t* s = src_ptr;\n size_t n = len;\n\n while (n >= 1) {\n uint8_t b0 = s[0];\n uint8_t b1 = s[1];\n uint8_t b2 = s[2];\n d[0] = b2;\n " + - "d[1] = b1;\n d[2] = b0;\n d[3] = 0xFF;\n\n s += 1 * 3;\n d += 1 * 4;\n n -= 1;\n }\n\n return len;\n}\n\n" + + "e_ptr,\n size_t dst_palette_len,\n const uint8_t* src_ptr,\n size_t src_len) {\n size_t dst_len4 = dst_len / 4;\n size_t src_len4 = src_len / 4;\n size_t len = (dst_len4 < src_len4) ? dst_len4 : src_len4;\n uint8_t* d = dst_ptr;\n const uint8_t* s = src_ptr;\n size_t n = len;\n\n // TODO: unroll.\n\n while (n >= 1) {\n wuffs_base__poke_u32le__no_bounds_check(\n d + (0 * 4),\n 0xFF000000 | wuffs_base__peek_u32le__no_bounds_check(s + (0 * 4)));\n\n s += 1 * 4;\n d += 1 * 4;\n n -= 1;\n }\n\n return len;\n}\n\n#if defined(WUFFS_BASE__CPU_ARCH__X86_64)\n#if defined(__GNUC__)\n__attribute__((target(\"sse4.2\")))\n#endif\nstatic uint64_t //\nwuffs_base__pixel_swizzler__bgrw__rgb__sse42(uint8_t* dst_ptr,\n size_t dst_len,\n uint8_t* dst_palette_ptr,\n size_t dst_palette_len" + + ",\n const uint8_t* src_ptr,\n size_t src_len) {\n size_t dst_len4 = dst_len / 4;\n size_t src_len3 = src_len / 3;\n size_t len = (dst_len4 < src_len3) ? dst_len4 : src_len3;\n uint8_t* d = dst_ptr;\n const uint8_t* s = src_ptr;\n size_t n = len;\n\n __m128i shuffle = _mm_set_epi8(+0x00, +0x09, +0x0A, +0x0B, //\n +0x00, +0x06, +0x07, +0x08, //\n +0x00, +0x03, +0x04, +0x05, //\n +0x00, +0x00, +0x01, +0x02);\n __m128i or_ff = _mm_set_epi8(-0x01, +0x00, +0x00, +0x00, //\n -0x01, +0x00, +0x00, +0x00, //\n -0x01, +0x00, +0x00, +0x00, //\n -0x01, +0x00, +0x00, +0x00);\n\n while (n >= 6) {\n __m128i x;\n x = _mm_lddqu_si128((const __m128i*)(const void*)s);\n x = _mm_shuffle_epi8(x, shuffle);\n x = _mm_or_si128(x, or_ff);\n _mm_storeu_si128((_" + + "_m128i*)(void*)d, x);\n\n s += 4 * 3;\n d += 4 * 4;\n n -= 4;\n }\n\n while (n >= 1) {\n uint8_t b0 = s[0];\n uint8_t b1 = s[1];\n uint8_t b2 = s[2];\n d[0] = b2;\n d[1] = b1;\n d[2] = b0;\n d[3] = 0xFF;\n\n s += 1 * 3;\n d += 1 * 4;\n n -= 1;\n }\n\n return len;\n}\n#endif // defined(WUFFS_BASE__CPU_ARCH__X86_64)\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__bgrw__rgb(uint8_t* dst_ptr,\n size_t dst_len,\n uint8_t* dst_palette_ptr,\n size_t dst_palette_len,\n const uint8_t* src_ptr,\n size_t src_len) {\n size_t dst_len4 = dst_len / 4;\n size_t src_len3 = src_len / 3;\n size_t len = (dst_len4 < src_len3) ? dst_len4 : src_len3;\n uint8_t* d = dst_ptr;\n const uint8_t* s = src_ptr;\n size_t n = len;\n\n while (n >= 1) {\n uint8_t b0 = s[0];\n uint8_t b1 = s[1];\n uint8_t b2 = s[2];\n d[0] = b2;\n d[1] =" + + " b1;\n d[2] = b0;\n d[3] = 0xFF;\n\n s += 1 * 3;\n d += 1 * 4;\n n -= 1;\n }\n\n return len;\n}\n\n" + "" + "// --------\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__xxx__index__src(uint8_t* dst_ptr,\n size_t dst_len,\n uint8_t* dst_palette_ptr,\n size_t dst_palette_len,\n const uint8_t* src_ptr,\n size_t src_len) {\n if (dst_palette_len != 1024) {\n return 0;\n }\n size_t dst_len3 = dst_len / 3;\n size_t len = (dst_len3 < src_len) ? dst_len3 : src_len;\n uint8_t* d = dst_ptr;\n const uint8_t* s = src_ptr;\n size_t n = len;\n\n const size_t loop_unroll_count = 4;\n\n // The comparison in the while condition is \">\", not \">=\", because with\n // \">=\", the last 4-byte store could write past the end of the dst slice.\n //\n // Each 4-byte store writes one too many bytes, but a subsequent store\n // will overwrite that with the correct byte. There is always another\n // store, whether a 4-byte store in this loop" + " or a 1-byte store in the\n // next loop.\n while (n > loop_unroll_count) {\n wuffs_base__poke_u32le__no_bounds_check(\n d + (0 * 3), wuffs_base__peek_u32le__no_bounds_check(\n dst_palette_ptr + ((size_t)s[0] * 4)));\n wuffs_base__poke_u32le__no_bounds_check(\n d + (1 * 3), wuffs_base__peek_u32le__no_bounds_check(\n dst_palette_ptr + ((size_t)s[1] * 4)));\n wuffs_base__poke_u32le__no_bounds_check(\n d + (2 * 3), wuffs_base__peek_u32le__no_bounds_check(\n dst_palette_ptr + ((size_t)s[2] * 4)));\n wuffs_base__poke_u32le__no_bounds_check(\n d + (3 * 3), wuffs_base__peek_u32le__no_bounds_check(\n dst_palette_ptr + ((size_t)s[3] * 4)));\n\n s += loop_unroll_count * 1;\n d += loop_unroll_count * 3;\n n -= loop_unroll_count;\n }\n\n while (n >= 1) {\n uint32_t s0 = wuffs_base__peek_u32le__no_bounds_check(dst_palette_ptr +\n ((size_t)" + @@ -615,27 +615,27 @@ const BasePixConvSubmoduleC = "" + "// --------\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__xxxx__index__src(uint8_t* dst_ptr,\n size_t dst_len,\n uint8_t* dst_palette_ptr,\n size_t dst_palette_len,\n const uint8_t* src_ptr,\n size_t src_len) {\n if (dst_palette_len != 1024) {\n return 0;\n }\n size_t dst_len4 = dst_len / 4;\n size_t len = (dst_len4 < src_len) ? dst_len4 : src_len;\n uint8_t* d = dst_ptr;\n const uint8_t* s = src_ptr;\n size_t n = len;\n\n const size_t loop_unroll_count = 4;\n\n while (n >= loop_unroll_count) {\n wuffs_base__poke_u32le__no_bounds_check(\n d + (0 * 4), wuffs_base__peek_u32le__no_bounds_check(\n dst_palette_ptr + ((size_t)s[0] * 4)));\n wuffs_base__poke_u32le__no_bounds_check(\n d + (1 * 4), wuffs_base__peek_u32le__no_bounds_check(\n ds" + "t_palette_ptr + ((size_t)s[1] * 4)));\n wuffs_base__poke_u32le__no_bounds_check(\n d + (2 * 4), wuffs_base__peek_u32le__no_bounds_check(\n dst_palette_ptr + ((size_t)s[2] * 4)));\n wuffs_base__poke_u32le__no_bounds_check(\n d + (3 * 4), wuffs_base__peek_u32le__no_bounds_check(\n dst_palette_ptr + ((size_t)s[3] * 4)));\n\n s += loop_unroll_count * 1;\n d += loop_unroll_count * 4;\n n -= loop_unroll_count;\n }\n\n while (n >= 1) {\n wuffs_base__poke_u32le__no_bounds_check(\n d + (0 * 4), wuffs_base__peek_u32le__no_bounds_check(\n dst_palette_ptr + ((size_t)s[0] * 4)));\n\n s += 1 * 1;\n d += 1 * 4;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__xxxx__index_binary_alpha__src_over(\n uint8_t* dst_ptr,\n size_t dst_len,\n uint8_t* dst_palette_ptr,\n size_t dst_palette_len,\n const uint8_t* src_ptr,\n size_t src_len) {\n if (dst_palette_len != 1024) {\n return 0;\n }\n " + " size_t dst_len4 = dst_len / 4;\n size_t len = (dst_len4 < src_len) ? dst_len4 : src_len;\n uint8_t* d = dst_ptr;\n const uint8_t* s = src_ptr;\n size_t n = len;\n\n const size_t loop_unroll_count = 4;\n\n while (n >= loop_unroll_count) {\n uint32_t s0 = wuffs_base__peek_u32le__no_bounds_check(dst_palette_ptr +\n ((size_t)s[0] * 4));\n if (s0) {\n wuffs_base__poke_u32le__no_bounds_check(d + (0 * 4), s0);\n }\n uint32_t s1 = wuffs_base__peek_u32le__no_bounds_check(dst_palette_ptr +\n ((size_t)s[1] * 4));\n if (s1) {\n wuffs_base__poke_u32le__no_bounds_check(d + (1 * 4), s1);\n }\n uint32_t s2 = wuffs_base__peek_u32le__no_bounds_check(dst_palette_ptr +\n ((size_t)s[2] * 4));\n if (s2) {\n wuffs_base__poke_u32le__no_bounds_check(d + (2 * 4), s2);\n }\n uint32_t s3 = wuffs_base__peek_u32le__no_bounds_check(dst_palette_" + - "ptr +\n ((size_t)s[3] * 4));\n if (s3) {\n wuffs_base__poke_u32le__no_bounds_check(d + (3 * 4), s3);\n }\n\n s += loop_unroll_count * 1;\n d += loop_unroll_count * 4;\n n -= loop_unroll_count;\n }\n\n while (n >= 1) {\n uint32_t s0 = wuffs_base__peek_u32le__no_bounds_check(dst_palette_ptr +\n ((size_t)s[0] * 4));\n if (s0) {\n wuffs_base__poke_u32le__no_bounds_check(d + (0 * 4), s0);\n }\n\n s += 1 * 1;\n d += 1 * 4;\n n -= 1;\n }\n\n return len;\n}\n\n#if defined(WUFFS_BASE__CPU_ARCH__X86_64)\n#if defined(__GNUC__)\n__attribute__((target(\"sse4.2\")))\n#endif\nstatic uint64_t //\nwuffs_base__pixel_swizzler__xxxx__y__sse128(uint8_t* dst_ptr,\n size_t dst_len,\n uint8_t* dst_palette_ptr,\n size_t dst_palette_len,\n con" + - "st uint8_t* src_ptr,\n size_t src_len) {\n size_t dst_len4 = dst_len / 4;\n size_t len = (dst_len4 < src_len) ? dst_len4 : src_len;\n uint8_t* d = dst_ptr;\n const uint8_t* s = src_ptr;\n size_t n = len;\n\n __m128i shuffle = _mm_set_epi8(+0x03, +0x03, +0x03, +0x03, //\n +0x02, +0x02, +0x02, +0x02, //\n +0x01, +0x01, +0x01, +0x01, //\n +0x00, +0x00, +0x00, +0x00);\n __m128i or_ff = _mm_set_epi8(-0x01, +0x00, +0x00, +0x00, //\n -0x01, +0x00, +0x00, +0x00, //\n -0x01, +0x00, +0x00, +0x00, //\n -0x01, +0x00, +0x00, +0x00);\n\n while (n >= 4) {\n __m128i x;\n x = _mm_cvtsi32_si128((int)(wuffs_base__peek_u32le__no_bounds_check(s)));\n x = _mm_shuffle_epi8(x, shuffle);\n x = _mm_or_si128(x, or_ff);\n _mm_storeu_si128((__m128i*)(void*)d, x);\n\n s += 4 * 1;\n d += 4 * 4;\n n -= 4" + - ";\n }\n\n while (n >= 1) {\n wuffs_base__poke_u32le__no_bounds_check(\n d + (0 * 4), 0xFF000000 | (0x010101 * (uint32_t)s[0]));\n\n s += 1 * 1;\n d += 1 * 4;\n n -= 1;\n }\n\n return len;\n}\n#endif // defined(WUFFS_BASE__CPU_ARCH__X86_64)\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__xxxx__y(uint8_t* dst_ptr,\n size_t dst_len,\n uint8_t* dst_palette_ptr,\n size_t dst_palette_len,\n const uint8_t* src_ptr,\n size_t src_len) {\n size_t dst_len4 = dst_len / 4;\n size_t len = (dst_len4 < src_len) ? dst_len4 : src_len;\n uint8_t* d = dst_ptr;\n const uint8_t* s = src_ptr;\n size_t n = len;\n\n while (n >= 1) {\n wuffs_base__poke_u32le__no_bounds_check(\n d + (0 * 4), 0xFF000000 | (0x010101 * (uint32_t)s[0]));\n\n s += 1 * 1;\n d += 1 * 4;\n n -= 1;\n }\n\n return len;\n}\n\n" + + "ptr +\n ((size_t)s[3] * 4));\n if (s3) {\n wuffs_base__poke_u32le__no_bounds_check(d + (3 * 4), s3);\n }\n\n s += loop_unroll_count * 1;\n d += loop_unroll_count * 4;\n n -= loop_unroll_count;\n }\n\n while (n >= 1) {\n uint32_t s0 = wuffs_base__peek_u32le__no_bounds_check(dst_palette_ptr +\n ((size_t)s[0] * 4));\n if (s0) {\n wuffs_base__poke_u32le__no_bounds_check(d + (0 * 4), s0);\n }\n\n s += 1 * 1;\n d += 1 * 4;\n n -= 1;\n }\n\n return len;\n}\n\n#if defined(WUFFS_BASE__CPU_ARCH__X86_64)\n#if defined(__GNUC__)\n__attribute__((target(\"sse4.2\")))\n#endif\nstatic uint64_t //\nwuffs_base__pixel_swizzler__xxxx__y__sse42(uint8_t* dst_ptr,\n size_t dst_len,\n uint8_t* dst_palette_ptr,\n size_t dst_palette_len,\n const ui" + + "nt8_t* src_ptr,\n size_t src_len) {\n size_t dst_len4 = dst_len / 4;\n size_t len = (dst_len4 < src_len) ? dst_len4 : src_len;\n uint8_t* d = dst_ptr;\n const uint8_t* s = src_ptr;\n size_t n = len;\n\n __m128i shuffle = _mm_set_epi8(+0x03, +0x03, +0x03, +0x03, //\n +0x02, +0x02, +0x02, +0x02, //\n +0x01, +0x01, +0x01, +0x01, //\n +0x00, +0x00, +0x00, +0x00);\n __m128i or_ff = _mm_set_epi8(-0x01, +0x00, +0x00, +0x00, //\n -0x01, +0x00, +0x00, +0x00, //\n -0x01, +0x00, +0x00, +0x00, //\n -0x01, +0x00, +0x00, +0x00);\n\n while (n >= 4) {\n __m128i x;\n x = _mm_cvtsi32_si128((int)(wuffs_base__peek_u32le__no_bounds_check(s)));\n x = _mm_shuffle_epi8(x, shuffle);\n x = _mm_or_si128(x, or_ff);\n _mm_storeu_si128((__m128i*)(void*)d, x);\n\n s += 4 * 1;\n d += 4 * 4;\n n -= 4;\n }\n" + + "\n while (n >= 1) {\n wuffs_base__poke_u32le__no_bounds_check(\n d + (0 * 4), 0xFF000000 | (0x010101 * (uint32_t)s[0]));\n\n s += 1 * 1;\n d += 1 * 4;\n n -= 1;\n }\n\n return len;\n}\n#endif // defined(WUFFS_BASE__CPU_ARCH__X86_64)\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__xxxx__y(uint8_t* dst_ptr,\n size_t dst_len,\n uint8_t* dst_palette_ptr,\n size_t dst_palette_len,\n const uint8_t* src_ptr,\n size_t src_len) {\n size_t dst_len4 = dst_len / 4;\n size_t len = (dst_len4 < src_len) ? dst_len4 : src_len;\n uint8_t* d = dst_ptr;\n const uint8_t* s = src_ptr;\n size_t n = len;\n\n while (n >= 1) {\n wuffs_base__poke_u32le__no_bounds_check(\n d + (0 * 4), 0xFF000000 | (0x010101 * (uint32_t)s[0]));\n\n s += 1 * 1;\n d += 1 * 4;\n n -= 1;\n }\n\n return len;\n}\n\n" + "" + "// --------\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__transparent_black_src(\n uint8_t* dst_ptr,\n size_t dst_len,\n uint8_t* dst_palette_ptr,\n size_t dst_palette_len,\n uint64_t num_pixels,\n uint32_t dst_pixfmt_bytes_per_pixel) {\n uint64_t n = ((uint64_t)dst_len) / dst_pixfmt_bytes_per_pixel;\n if (n > num_pixels) {\n n = num_pixels;\n }\n memset(dst_ptr, 0, ((size_t)(n * dst_pixfmt_bytes_per_pixel)));\n return n;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__transparent_black_src_over(\n uint8_t* dst_ptr,\n size_t dst_len,\n uint8_t* dst_palette_ptr,\n size_t dst_palette_len,\n uint64_t num_pixels,\n uint32_t dst_pixfmt_bytes_per_pixel) {\n uint64_t n = ((uint64_t)dst_len) / dst_pixfmt_bytes_per_pixel;\n if (n > num_pixels) {\n n = num_pixels;\n }\n return n;\n}\n\n" + "" + "// --------\n\nstatic wuffs_base__pixel_swizzler__func //\nwuffs_base__pixel_swizzler__prepare__y(wuffs_base__pixel_swizzler* p,\n wuffs_base__pixel_format dst_pixfmt,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src_palette,\n wuffs_base__pixel_blend blend) {\n switch (dst_pixfmt.repr) {\n case WUFFS_BASE__PIXEL_FORMAT__Y:\n return wuffs_base__pixel_swizzler__copy_1_1;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGR_565:\n return wuffs_base__pixel_swizzler__bgr_565__y;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGR:\n case WUFFS_BASE__PIXEL_FORMAT__RGB:\n return wuffs_base__pixel_swizzler__xxx__y;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY:\n case WUFFS_BASE__PIXEL_FORMAT__BGRX:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL:\n case WUFFS_BA" + - "SE__PIXEL_FORMAT__RGBA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY:\n case WUFFS_BASE__PIXEL_FORMAT__RGBX:\n#if defined(WUFFS_BASE__CPU_ARCH__X86_64)\n if (wuffs_base__cpu_arch__have_sse128()) {\n return wuffs_base__pixel_swizzler__xxxx__y__sse128;\n }\n#endif\n return wuffs_base__pixel_swizzler__xxxx__y;\n }\n return NULL;\n}\n\nstatic wuffs_base__pixel_swizzler__func //\nwuffs_base__pixel_swizzler__prepare__indexed__bgra_binary(\n wuffs_base__pixel_swizzler* p,\n wuffs_base__pixel_format dst_pixfmt,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src_palette,\n wuffs_base__pixel_blend blend) {\n switch (dst_pixfmt.repr) {\n case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_NONPREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_BINARY:\n if (wuffs_base__slice_u8__copy_from_slice(dst_palette, src_palette) !=\n 1024) {\n return NULL;\n }\n switch (blend) {\n case WUFFS_BASE_" + - "_PIXEL_BLEND__SRC:\n return wuffs_base__pixel_swizzler__copy_1_1;\n }\n return NULL;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGR_565:\n if (wuffs_base__pixel_swizzler__squash_align4_bgr_565_888(\n dst_palette.ptr, dst_palette.len, NULL, 0, src_palette.ptr,\n src_palette.len) != 256) {\n return NULL;\n }\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n return wuffs_base__pixel_swizzler__bgr_565__index__src;\n case WUFFS_BASE__PIXEL_BLEND__SRC_OVER:\n return wuffs_base__pixel_swizzler__bgr_565__index_binary_alpha__src_over;\n }\n return NULL;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGR:\n if (wuffs_base__slice_u8__copy_from_slice(dst_palette, src_palette) !=\n 1024) {\n return NULL;\n }\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n return wuffs_base__pixel_swizzler__xxx__index__src;\n case WUFFS_BASE__PIXEL_BLEND__SRC_OVER:\n return wuffs_base__pixel_sw" + - "izzler__xxx__index_binary_alpha__src_over;\n }\n return NULL;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY:\n if (wuffs_base__slice_u8__copy_from_slice(dst_palette, src_palette) !=\n 1024) {\n return NULL;\n }\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n return wuffs_base__pixel_swizzler__xxxx__index__src;\n case WUFFS_BASE__PIXEL_BLEND__SRC_OVER:\n return wuffs_base__pixel_swizzler__xxxx__index_binary_alpha__src_over;\n }\n return NULL;\n\n case WUFFS_BASE__PIXEL_FORMAT__RGB:\n if (wuffs_base__pixel_swizzler__swap_rgbx_bgrx(\n dst_palette.ptr, dst_palette.len, NULL, 0, src_palette.ptr,\n src_palette.len) != 256) {\n return NULL;\n }\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n return wuffs_base__pixel_swizzler__xxx__index__src;\n case WUFFS_BASE__PIXEL" + - "_BLEND__SRC_OVER:\n return wuffs_base__pixel_swizzler__xxx__index_binary_alpha__src_over;\n }\n return NULL;\n\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY:\n if (wuffs_base__pixel_swizzler__swap_rgbx_bgrx(\n dst_palette.ptr, dst_palette.len, NULL, 0, src_palette.ptr,\n src_palette.len) != 256) {\n return NULL;\n }\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n return wuffs_base__pixel_swizzler__xxxx__index__src;\n case WUFFS_BASE__PIXEL_BLEND__SRC_OVER:\n return wuffs_base__pixel_swizzler__xxxx__index_binary_alpha__src_over;\n }\n return NULL;\n }\n return NULL;\n}\n\nstatic wuffs_base__pixel_swizzler__func //\nwuffs_base__pixel_swizzler__prepare__bgr(wuffs_base__pixel_swizzler* p,\n wuffs_base__pixel_format dst_pixfmt,\n wuffs_base_" + - "_slice_u8 dst_palette,\n wuffs_base__slice_u8 src_palette,\n wuffs_base__pixel_blend blend) {\n switch (dst_pixfmt.repr) {\n case WUFFS_BASE__PIXEL_FORMAT__BGR_565:\n return wuffs_base__pixel_swizzler__bgr_565__bgr;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGR:\n return wuffs_base__pixel_swizzler__copy_3_3;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY:\n case WUFFS_BASE__PIXEL_FORMAT__BGRX:\n return wuffs_base__pixel_swizzler__bgrw__bgr;\n\n case WUFFS_BASE__PIXEL_FORMAT__RGB:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY:\n case WUFFS_BASE__PIXEL_FORMAT__RGBX:\n // TODO.\n break;\n }\n return NULL;\n}\n\nstatic wuffs_base__pixel_swizzler__func //\nwuffs_base__pixel_swizzler__prepare__bgra_nonpremul(\n wuffs_base_" + - "_pixel_swizzler* p,\n wuffs_base__pixel_format dst_pixfmt,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src_palette,\n wuffs_base__pixel_blend blend) {\n switch (dst_pixfmt.repr) {\n case WUFFS_BASE__PIXEL_FORMAT__BGR_565:\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n return wuffs_base__pixel_swizzler__bgr_565__bgra_nonpremul__src;\n case WUFFS_BASE__PIXEL_BLEND__SRC_OVER:\n return wuffs_base__pixel_swizzler__bgr_565__bgra_nonpremul__src_over;\n }\n return NULL;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGR:\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n return wuffs_base__pixel_swizzler__bgr__bgra_nonpremul__src;\n case WUFFS_BASE__PIXEL_BLEND__SRC_OVER:\n return wuffs_base__pixel_swizzler__bgr__bgra_nonpremul__src_over;\n }\n return NULL;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL:\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n return wuffs_base__pixel" + - "_swizzler__copy_4_4;\n case WUFFS_BASE__PIXEL_BLEND__SRC_OVER:\n return wuffs_base__pixel_swizzler__bgra_nonpremul__bgra_nonpremul__src_over;\n }\n return NULL;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL:\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n return wuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul__src;\n case WUFFS_BASE__PIXEL_BLEND__SRC_OVER:\n return wuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul__src_over;\n }\n return NULL;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY:\n case WUFFS_BASE__PIXEL_FORMAT__BGRX:\n // TODO.\n break;\n\n case WUFFS_BASE__PIXEL_FORMAT__RGB:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY:\n case WUFFS_BASE__PIXEL_FORMAT__RGBX:\n // TODO.\n break;\n }\n return NULL;\n}\n\nstatic wuffs_base__pixel_swizzler__func //\nwuffs_base__pixel_swizzler__prepare__bgra_nonpr" + - "emul_4x16le(\n wuffs_base__pixel_swizzler* p,\n wuffs_base__pixel_format dst_pixfmt,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src_palette,\n wuffs_base__pixel_blend blend) {\n switch (dst_pixfmt.repr) {\n case WUFFS_BASE__PIXEL_FORMAT__BGR_565:\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n return wuffs_base__pixel_swizzler__bgr_565__bgra_nonpremul_4x16le__src;\n case WUFFS_BASE__PIXEL_BLEND__SRC_OVER:\n return wuffs_base__pixel_swizzler__bgr_565__bgra_nonpremul_4x16le__src_over;\n }\n return NULL;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGR:\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n return wuffs_base__pixel_swizzler__bgr__bgra_nonpremul_4x16le__src;\n case WUFFS_BASE__PIXEL_BLEND__SRC_OVER:\n return wuffs_base__pixel_swizzler__bgr__bgra_nonpremul_4x16le__src_over;\n }\n return NULL;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL:\n switch (blend) {\n case WUFFS_BA" + - "SE__PIXEL_BLEND__SRC:\n return wuffs_base__pixel_swizzler__squash_tight_4x8_4x16le;\n case WUFFS_BASE__PIXEL_BLEND__SRC_OVER:\n return wuffs_base__pixel_swizzler__bgra_nonpremul__bgra_nonpremul_4x16le__src_over;\n }\n return NULL;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL:\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n return wuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul_4x16le__src;\n case WUFFS_BASE__PIXEL_BLEND__SRC_OVER:\n return wuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul_4x16le__src_over;\n }\n return NULL;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY:\n case WUFFS_BASE__PIXEL_FORMAT__BGRX:\n // TODO.\n break;\n\n case WUFFS_BASE__PIXEL_FORMAT__RGB:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY:\n case WUFFS_BASE__PIXEL_FORMAT__RGBX:\n // TODO.\n break;\n }\n return NULL;\n}\n" + - "\nstatic wuffs_base__pixel_swizzler__func //\nwuffs_base__pixel_swizzler__prepare__bgrx(wuffs_base__pixel_swizzler* p,\n wuffs_base__pixel_format dst_pixfmt,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src_palette,\n wuffs_base__pixel_blend blend) {\n switch (dst_pixfmt.repr) {\n case WUFFS_BASE__PIXEL_FORMAT__BGR_565:\n return wuffs_base__pixel_swizzler__bgr_565__bgrx;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGR:\n return wuffs_base__pixel_swizzler__xxx__xxxx;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY:\n return wuffs_base__pixel_swizzler__bgrw__bgrx;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGRX:\n return wuffs_base__pixel_swizzler__copy_4_4;\n\n case WUFFS_BASE__PIXEL_FORMAT__RGB:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_NON" + - "PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY:\n case WUFFS_BASE__PIXEL_FORMAT__RGBX:\n // TODO.\n break;\n }\n return NULL;\n}\n\nstatic wuffs_base__pixel_swizzler__func //\nwuffs_base__pixel_swizzler__prepare__rgb(wuffs_base__pixel_swizzler* p,\n wuffs_base__pixel_format dst_pixfmt,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src_palette,\n wuffs_base__pixel_blend blend) {\n switch (dst_pixfmt.repr) {\n case WUFFS_BASE__PIXEL_FORMAT__BGR_565:\n return wuffs_base__pixel_swizzler__bgr_565__rgb;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGR:\n return wuffs_base__pixel_swizzler__swap_rgb_bgr;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY:\n case WUFFS_BASE__PIXEL_FORMAT__BGRX" + - ":\n#if defined(WUFFS_BASE__CPU_ARCH__X86_64)\n if (wuffs_base__cpu_arch__have_sse128()) {\n return wuffs_base__pixel_swizzler__bgrw__rgb__sse128;\n }\n#endif\n return wuffs_base__pixel_swizzler__bgrw__rgb;\n\n case WUFFS_BASE__PIXEL_FORMAT__RGB:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY:\n case WUFFS_BASE__PIXEL_FORMAT__RGBX:\n // TODO.\n break;\n }\n return NULL;\n}\n\nstatic wuffs_base__pixel_swizzler__func //\nwuffs_base__pixel_swizzler__prepare__rgba_nonpremul(\n wuffs_base__pixel_swizzler* p,\n wuffs_base__pixel_format dst_pixfmt,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src_palette,\n wuffs_base__pixel_blend blend) {\n switch (dst_pixfmt.repr) {\n case WUFFS_BASE__PIXEL_FORMAT__BGR_565:\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n return wuffs_base__pixel_swizzler__bgr_565__rgba_nonpremul__src;\n case WUFFS_BASE" + - "__PIXEL_BLEND__SRC_OVER:\n return wuffs_base__pixel_swizzler__bgr_565__rgba_nonpremul__src_over;\n }\n return NULL;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGR:\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n return wuffs_base__pixel_swizzler__bgr__rgba_nonpremul__src;\n case WUFFS_BASE__PIXEL_BLEND__SRC_OVER:\n return wuffs_base__pixel_swizzler__bgr__rgba_nonpremul__src_over;\n }\n return NULL;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL:\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n#if defined(WUFFS_BASE__CPU_ARCH__X86_64)\n if (wuffs_base__cpu_arch__have_sse128()) {\n return wuffs_base__pixel_swizzler__swap_rgbx_bgrx__sse128;\n }\n#endif\n return wuffs_base__pixel_swizzler__swap_rgbx_bgrx;\n case WUFFS_BASE__PIXEL_BLEND__SRC_OVER:\n return wuffs_base__pixel_swizzler__bgra_nonpremul__rgba_nonpremul__src_over;\n }\n return NULL;\n\n case WUFFS_BASE__PIXEL_FORMAT__B" + - "GRA_PREMUL:\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n return wuffs_base__pixel_swizzler__bgra_premul__rgba_nonpremul__src;\n case WUFFS_BASE__PIXEL_BLEND__SRC_OVER:\n return wuffs_base__pixel_swizzler__bgra_premul__rgba_nonpremul__src_over;\n }\n return NULL;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY:\n case WUFFS_BASE__PIXEL_FORMAT__BGRX:\n // TODO.\n break;\n\n case WUFFS_BASE__PIXEL_FORMAT__RGB:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY:\n case WUFFS_BASE__PIXEL_FORMAT__RGBX:\n // TODO.\n break;\n }\n return NULL;\n}\n\n" + + "SE__PIXEL_FORMAT__RGBA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY:\n case WUFFS_BASE__PIXEL_FORMAT__RGBX:\n#if defined(WUFFS_BASE__CPU_ARCH__X86_64)\n if (wuffs_base__cpu_arch__have_sse42()) {\n return wuffs_base__pixel_swizzler__xxxx__y__sse42;\n }\n#endif\n return wuffs_base__pixel_swizzler__xxxx__y;\n }\n return NULL;\n}\n\nstatic wuffs_base__pixel_swizzler__func //\nwuffs_base__pixel_swizzler__prepare__indexed__bgra_binary(\n wuffs_base__pixel_swizzler* p,\n wuffs_base__pixel_format dst_pixfmt,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src_palette,\n wuffs_base__pixel_blend blend) {\n switch (dst_pixfmt.repr) {\n case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_NONPREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_BINARY:\n if (wuffs_base__slice_u8__copy_from_slice(dst_palette, src_palette) !=\n 1024) {\n return NULL;\n }\n switch (blend) {\n case WUFFS_BASE__P" + + "IXEL_BLEND__SRC:\n return wuffs_base__pixel_swizzler__copy_1_1;\n }\n return NULL;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGR_565:\n if (wuffs_base__pixel_swizzler__squash_align4_bgr_565_888(\n dst_palette.ptr, dst_palette.len, NULL, 0, src_palette.ptr,\n src_palette.len) != 256) {\n return NULL;\n }\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n return wuffs_base__pixel_swizzler__bgr_565__index__src;\n case WUFFS_BASE__PIXEL_BLEND__SRC_OVER:\n return wuffs_base__pixel_swizzler__bgr_565__index_binary_alpha__src_over;\n }\n return NULL;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGR:\n if (wuffs_base__slice_u8__copy_from_slice(dst_palette, src_palette) !=\n 1024) {\n return NULL;\n }\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n return wuffs_base__pixel_swizzler__xxx__index__src;\n case WUFFS_BASE__PIXEL_BLEND__SRC_OVER:\n return wuffs_base__pixel_swiz" + + "zler__xxx__index_binary_alpha__src_over;\n }\n return NULL;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY:\n if (wuffs_base__slice_u8__copy_from_slice(dst_palette, src_palette) !=\n 1024) {\n return NULL;\n }\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n return wuffs_base__pixel_swizzler__xxxx__index__src;\n case WUFFS_BASE__PIXEL_BLEND__SRC_OVER:\n return wuffs_base__pixel_swizzler__xxxx__index_binary_alpha__src_over;\n }\n return NULL;\n\n case WUFFS_BASE__PIXEL_FORMAT__RGB:\n if (wuffs_base__pixel_swizzler__swap_rgbx_bgrx(\n dst_palette.ptr, dst_palette.len, NULL, 0, src_palette.ptr,\n src_palette.len) != 256) {\n return NULL;\n }\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n return wuffs_base__pixel_swizzler__xxx__index__src;\n case WUFFS_BASE__PIXEL_B" + + "LEND__SRC_OVER:\n return wuffs_base__pixel_swizzler__xxx__index_binary_alpha__src_over;\n }\n return NULL;\n\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY:\n if (wuffs_base__pixel_swizzler__swap_rgbx_bgrx(\n dst_palette.ptr, dst_palette.len, NULL, 0, src_palette.ptr,\n src_palette.len) != 256) {\n return NULL;\n }\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n return wuffs_base__pixel_swizzler__xxxx__index__src;\n case WUFFS_BASE__PIXEL_BLEND__SRC_OVER:\n return wuffs_base__pixel_swizzler__xxxx__index_binary_alpha__src_over;\n }\n return NULL;\n }\n return NULL;\n}\n\nstatic wuffs_base__pixel_swizzler__func //\nwuffs_base__pixel_swizzler__prepare__bgr(wuffs_base__pixel_swizzler* p,\n wuffs_base__pixel_format dst_pixfmt,\n wuffs_base__s" + + "lice_u8 dst_palette,\n wuffs_base__slice_u8 src_palette,\n wuffs_base__pixel_blend blend) {\n switch (dst_pixfmt.repr) {\n case WUFFS_BASE__PIXEL_FORMAT__BGR_565:\n return wuffs_base__pixel_swizzler__bgr_565__bgr;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGR:\n return wuffs_base__pixel_swizzler__copy_3_3;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY:\n case WUFFS_BASE__PIXEL_FORMAT__BGRX:\n return wuffs_base__pixel_swizzler__bgrw__bgr;\n\n case WUFFS_BASE__PIXEL_FORMAT__RGB:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY:\n case WUFFS_BASE__PIXEL_FORMAT__RGBX:\n // TODO.\n break;\n }\n return NULL;\n}\n\nstatic wuffs_base__pixel_swizzler__func //\nwuffs_base__pixel_swizzler__prepare__bgra_nonpremul(\n wuffs_base__p" + + "ixel_swizzler* p,\n wuffs_base__pixel_format dst_pixfmt,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src_palette,\n wuffs_base__pixel_blend blend) {\n switch (dst_pixfmt.repr) {\n case WUFFS_BASE__PIXEL_FORMAT__BGR_565:\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n return wuffs_base__pixel_swizzler__bgr_565__bgra_nonpremul__src;\n case WUFFS_BASE__PIXEL_BLEND__SRC_OVER:\n return wuffs_base__pixel_swizzler__bgr_565__bgra_nonpremul__src_over;\n }\n return NULL;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGR:\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n return wuffs_base__pixel_swizzler__bgr__bgra_nonpremul__src;\n case WUFFS_BASE__PIXEL_BLEND__SRC_OVER:\n return wuffs_base__pixel_swizzler__bgr__bgra_nonpremul__src_over;\n }\n return NULL;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL:\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n return wuffs_base__pixel_s" + + "wizzler__copy_4_4;\n case WUFFS_BASE__PIXEL_BLEND__SRC_OVER:\n return wuffs_base__pixel_swizzler__bgra_nonpremul__bgra_nonpremul__src_over;\n }\n return NULL;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL:\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n return wuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul__src;\n case WUFFS_BASE__PIXEL_BLEND__SRC_OVER:\n return wuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul__src_over;\n }\n return NULL;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY:\n case WUFFS_BASE__PIXEL_FORMAT__BGRX:\n // TODO.\n break;\n\n case WUFFS_BASE__PIXEL_FORMAT__RGB:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY:\n case WUFFS_BASE__PIXEL_FORMAT__RGBX:\n // TODO.\n break;\n }\n return NULL;\n}\n\nstatic wuffs_base__pixel_swizzler__func //\nwuffs_base__pixel_swizzler__prepare__bgra_nonprem" + + "ul_4x16le(\n wuffs_base__pixel_swizzler* p,\n wuffs_base__pixel_format dst_pixfmt,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src_palette,\n wuffs_base__pixel_blend blend) {\n switch (dst_pixfmt.repr) {\n case WUFFS_BASE__PIXEL_FORMAT__BGR_565:\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n return wuffs_base__pixel_swizzler__bgr_565__bgra_nonpremul_4x16le__src;\n case WUFFS_BASE__PIXEL_BLEND__SRC_OVER:\n return wuffs_base__pixel_swizzler__bgr_565__bgra_nonpremul_4x16le__src_over;\n }\n return NULL;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGR:\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n return wuffs_base__pixel_swizzler__bgr__bgra_nonpremul_4x16le__src;\n case WUFFS_BASE__PIXEL_BLEND__SRC_OVER:\n return wuffs_base__pixel_swizzler__bgr__bgra_nonpremul_4x16le__src_over;\n }\n return NULL;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL:\n switch (blend) {\n case WUFFS_BASE" + + "__PIXEL_BLEND__SRC:\n return wuffs_base__pixel_swizzler__squash_tight_4x8_4x16le;\n case WUFFS_BASE__PIXEL_BLEND__SRC_OVER:\n return wuffs_base__pixel_swizzler__bgra_nonpremul__bgra_nonpremul_4x16le__src_over;\n }\n return NULL;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL:\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n return wuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul_4x16le__src;\n case WUFFS_BASE__PIXEL_BLEND__SRC_OVER:\n return wuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul_4x16le__src_over;\n }\n return NULL;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY:\n case WUFFS_BASE__PIXEL_FORMAT__BGRX:\n // TODO.\n break;\n\n case WUFFS_BASE__PIXEL_FORMAT__RGB:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY:\n case WUFFS_BASE__PIXEL_FORMAT__RGBX:\n // TODO.\n break;\n }\n return NULL;\n}\n\ns" + + "tatic wuffs_base__pixel_swizzler__func //\nwuffs_base__pixel_swizzler__prepare__bgrx(wuffs_base__pixel_swizzler* p,\n wuffs_base__pixel_format dst_pixfmt,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src_palette,\n wuffs_base__pixel_blend blend) {\n switch (dst_pixfmt.repr) {\n case WUFFS_BASE__PIXEL_FORMAT__BGR_565:\n return wuffs_base__pixel_swizzler__bgr_565__bgrx;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGR:\n return wuffs_base__pixel_swizzler__xxx__xxxx;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY:\n return wuffs_base__pixel_swizzler__bgrw__bgrx;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGRX:\n return wuffs_base__pixel_swizzler__copy_4_4;\n\n case WUFFS_BASE__PIXEL_FORMAT__RGB:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPR" + + "EMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY:\n case WUFFS_BASE__PIXEL_FORMAT__RGBX:\n // TODO.\n break;\n }\n return NULL;\n}\n\nstatic wuffs_base__pixel_swizzler__func //\nwuffs_base__pixel_swizzler__prepare__rgb(wuffs_base__pixel_swizzler* p,\n wuffs_base__pixel_format dst_pixfmt,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src_palette,\n wuffs_base__pixel_blend blend) {\n switch (dst_pixfmt.repr) {\n case WUFFS_BASE__PIXEL_FORMAT__BGR_565:\n return wuffs_base__pixel_swizzler__bgr_565__rgb;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGR:\n return wuffs_base__pixel_swizzler__swap_rgb_bgr;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY:\n case WUFFS_BASE__PIXEL_FORMAT__BGRX:\n" + + "#if defined(WUFFS_BASE__CPU_ARCH__X86_64)\n if (wuffs_base__cpu_arch__have_sse42()) {\n return wuffs_base__pixel_swizzler__bgrw__rgb__sse42;\n }\n#endif\n return wuffs_base__pixel_swizzler__bgrw__rgb;\n\n case WUFFS_BASE__PIXEL_FORMAT__RGB:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY:\n case WUFFS_BASE__PIXEL_FORMAT__RGBX:\n // TODO.\n break;\n }\n return NULL;\n}\n\nstatic wuffs_base__pixel_swizzler__func //\nwuffs_base__pixel_swizzler__prepare__rgba_nonpremul(\n wuffs_base__pixel_swizzler* p,\n wuffs_base__pixel_format dst_pixfmt,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src_palette,\n wuffs_base__pixel_blend blend) {\n switch (dst_pixfmt.repr) {\n case WUFFS_BASE__PIXEL_FORMAT__BGR_565:\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n return wuffs_base__pixel_swizzler__bgr_565__rgba_nonpremul__src;\n case WUFFS_BASE__PI" + + "XEL_BLEND__SRC_OVER:\n return wuffs_base__pixel_swizzler__bgr_565__rgba_nonpremul__src_over;\n }\n return NULL;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGR:\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n return wuffs_base__pixel_swizzler__bgr__rgba_nonpremul__src;\n case WUFFS_BASE__PIXEL_BLEND__SRC_OVER:\n return wuffs_base__pixel_swizzler__bgr__rgba_nonpremul__src_over;\n }\n return NULL;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL:\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n#if defined(WUFFS_BASE__CPU_ARCH__X86_64)\n if (wuffs_base__cpu_arch__have_sse42()) {\n return wuffs_base__pixel_swizzler__swap_rgbx_bgrx__sse42;\n }\n#endif\n return wuffs_base__pixel_swizzler__swap_rgbx_bgrx;\n case WUFFS_BASE__PIXEL_BLEND__SRC_OVER:\n return wuffs_base__pixel_swizzler__bgra_nonpremul__rgba_nonpremul__src_over;\n }\n return NULL;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_PR" + + "EMUL:\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n return wuffs_base__pixel_swizzler__bgra_premul__rgba_nonpremul__src;\n case WUFFS_BASE__PIXEL_BLEND__SRC_OVER:\n return wuffs_base__pixel_swizzler__bgra_premul__rgba_nonpremul__src_over;\n }\n return NULL;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY:\n case WUFFS_BASE__PIXEL_FORMAT__BGRX:\n // TODO.\n break;\n\n case WUFFS_BASE__PIXEL_FORMAT__RGB:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY:\n case WUFFS_BASE__PIXEL_FORMAT__RGBX:\n // TODO.\n break;\n }\n return NULL;\n}\n\n" + "" + "// --------\n\nWUFFS_BASE__MAYBE_STATIC wuffs_base__status //\nwuffs_base__pixel_swizzler__prepare(wuffs_base__pixel_swizzler* p,\n wuffs_base__pixel_format dst_pixfmt,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__pixel_format src_pixfmt,\n wuffs_base__slice_u8 src_palette,\n wuffs_base__pixel_blend blend) {\n if (!p) {\n return wuffs_base__make_status(wuffs_base__error__bad_receiver);\n }\n p->private_impl.func = NULL;\n p->private_impl.transparent_black_func = NULL;\n p->private_impl.dst_pixfmt_bytes_per_pixel = 0;\n p->private_impl.src_pixfmt_bytes_per_pixel = 0;\n\n wuffs_base__pixel_swizzler__func func = NULL;\n wuffs_base__pixel_swizzler__transparent_black_func transparent_black_func =\n NULL;\n\n uint32_t dst_pixfmt_bits_per_pixel =\n wuffs_base__pixel_format__bits_per_pixel(&dst_pixfmt);\n if ((dst_pixfmt_bits_per_pixel == " + "0) ||\n ((dst_pixfmt_bits_per_pixel & 7) != 0)) {\n return wuffs_base__make_status(\n wuffs_base__error__unsupported_pixel_swizzler_option);\n }\n\n uint32_t src_pixfmt_bits_per_pixel =\n wuffs_base__pixel_format__bits_per_pixel(&src_pixfmt);\n if ((src_pixfmt_bits_per_pixel == 0) ||\n ((src_pixfmt_bits_per_pixel & 7) != 0)) {\n return wuffs_base__make_status(\n wuffs_base__error__unsupported_pixel_swizzler_option);\n }\n\n // TODO: support many more formats.\n\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n transparent_black_func =\n wuffs_base__pixel_swizzler__transparent_black_src;\n break;\n\n case WUFFS_BASE__PIXEL_BLEND__SRC_OVER:\n transparent_black_func =\n wuffs_base__pixel_swizzler__transparent_black_src_over;\n break;\n }\n\n switch (src_pixfmt.repr) {\n case WUFFS_BASE__PIXEL_FORMAT__Y:\n func = wuffs_base__pixel_swizzler__prepare__y(p, dst_pixfmt, dst_palette,\n src_palette" + diff --git a/internal/cgen/statement.go b/internal/cgen/statement.go index b77b55cf..fa8ac8cd 100644 --- a/internal/cgen/statement.go +++ b/internal/cgen/statement.go @@ -270,7 +270,7 @@ func cpuArchCNames(asserts []*a.Node) (caMacro string, caName string, caAttribut } } if sse42 { - return "X86_64", "sse128", "__attribute__((target(\"sse4.2\")))" + return "X86_64", "sse42", "__attribute__((target(\"sse4.2\")))" } return "", "", "" } diff --git a/release/c/wuffs-unsupported-snapshot.c b/release/c/wuffs-unsupported-snapshot.c index 606fa8fb..8cb8e3a3 100644 --- a/release/c/wuffs-unsupported-snapshot.c +++ b/release/c/wuffs-unsupported-snapshot.c @@ -104,12 +104,12 @@ extern "C" { // WUFFS_BASE__CPU_ARCH__X86_64__ETC are bits returned by // wuffs_base__cpu_arch__x86_64__capabilities. -// - "SSE128" means all of SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2 and POPCNT. -// - "AVX256" means all of AVX and AVX2. +// - "SSE42" means all of SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2 and POPCNT. +// - "AVX2" means all of AVX and AVX2. // - "AVX512ETC" is reserved, pending need. Note that AVX-512 consists of // multiple extensions that may be implemented independently. -#define WUFFS_BASE__CPU_ARCH__X86_64__SSE128 0x01 -#define WUFFS_BASE__CPU_ARCH__X86_64__AVX256 0x02 +#define WUFFS_BASE__CPU_ARCH__X86_64__SSE42 0x01 +#define WUFFS_BASE__CPU_ARCH__X86_64__AVX2 0x02 static inline uint32_t // wuffs_base__cpu_arch__x86_64__capabilities() { @@ -121,9 +121,9 @@ wuffs_base__cpu_arch__x86_64__capabilities() { unsigned int ecx1 = 0; unsigned int edx1 = 0; if (__get_cpuid(1, &eax1, &ebx1, &ecx1, &edx1)) { - const unsigned int sse128_ecx1 = bit_SSE4_2 | bit_POPCNT; - if ((ecx1 & sse128_ecx1) == sse128_ecx1) { - ret |= WUFFS_BASE__CPU_ARCH__X86_64__SSE128; + const unsigned int sse42_ecx1 = bit_SSE4_2 | bit_POPCNT; + if ((ecx1 & sse42_ecx1) == sse42_ecx1) { + ret |= WUFFS_BASE__CPU_ARCH__X86_64__SSE42; } } @@ -134,7 +134,7 @@ wuffs_base__cpu_arch__x86_64__capabilities() { if (__get_cpuid_count(7, 0, &eax7, &ebx7, &ecx7, &edx7)) { const unsigned int avx256_ebx7 = bit_AVX2; if ((ebx7 & avx256_ebx7) == avx256_ebx7) { - ret |= WUFFS_BASE__CPU_ARCH__X86_64__AVX256; + ret |= WUFFS_BASE__CPU_ARCH__X86_64__AVX2; } } @@ -145,15 +145,15 @@ wuffs_base__cpu_arch__x86_64__capabilities() { } static inline bool // -wuffs_base__cpu_arch__have_sse128() { +wuffs_base__cpu_arch__have_sse42() { #if defined(WUFFS_BASE__CPU_ARCH__X86_64) unsigned int eax1 = 0; unsigned int ebx1 = 0; unsigned int ecx1 = 0; unsigned int edx1 = 0; if (__get_cpuid(1, &eax1, &ebx1, &ecx1, &edx1)) { - const unsigned int sse128_ecx1 = bit_SSE4_2 | bit_POPCNT; - return (ecx1 & sse128_ecx1) == sse128_ecx1; + const unsigned int sse42_ecx1 = bit_SSE4_2 | bit_POPCNT; + return (ecx1 & sse42_ecx1) == sse42_ecx1; } #endif // defined(WUFFS_BASE__CPU_ARCH__X86_64) return false; @@ -14640,12 +14640,12 @@ wuffs_base__pixel_swizzler__swap_rgb_bgr(uint8_t* dst_ptr, __attribute__((target("sse4.2"))) #endif static uint64_t // -wuffs_base__pixel_swizzler__swap_rgbx_bgrx__sse128(uint8_t* dst_ptr, - size_t dst_len, - uint8_t* dst_palette_ptr, - size_t dst_palette_len, - const uint8_t* src_ptr, - size_t src_len) { +wuffs_base__pixel_swizzler__swap_rgbx_bgrx__sse42(uint8_t* dst_ptr, + size_t dst_len, + uint8_t* dst_palette_ptr, + size_t dst_palette_len, + const uint8_t* src_ptr, + size_t src_len) { size_t len = (dst_len < src_len ? dst_len : src_len) / 4; uint8_t* d = dst_ptr; const uint8_t* s = src_ptr; @@ -15849,12 +15849,12 @@ wuffs_base__pixel_swizzler__bgrw__bgrx(uint8_t* dst_ptr, __attribute__((target("sse4.2"))) #endif static uint64_t // -wuffs_base__pixel_swizzler__bgrw__rgb__sse128(uint8_t* dst_ptr, - size_t dst_len, - uint8_t* dst_palette_ptr, - size_t dst_palette_len, - const uint8_t* src_ptr, - size_t src_len) { +wuffs_base__pixel_swizzler__bgrw__rgb__sse42(uint8_t* dst_ptr, + size_t dst_len, + uint8_t* dst_palette_ptr, + size_t dst_palette_len, + const uint8_t* src_ptr, + size_t src_len) { size_t dst_len4 = dst_len / 4; size_t src_len3 = src_len / 3; size_t len = (dst_len4 < src_len3) ? dst_len4 : src_len3; @@ -16227,12 +16227,12 @@ wuffs_base__pixel_swizzler__xxxx__index_binary_alpha__src_over( __attribute__((target("sse4.2"))) #endif static uint64_t // -wuffs_base__pixel_swizzler__xxxx__y__sse128(uint8_t* dst_ptr, - size_t dst_len, - uint8_t* dst_palette_ptr, - size_t dst_palette_len, - const uint8_t* src_ptr, - size_t src_len) { +wuffs_base__pixel_swizzler__xxxx__y__sse42(uint8_t* dst_ptr, + size_t dst_len, + uint8_t* dst_palette_ptr, + size_t dst_palette_len, + const uint8_t* src_ptr, + size_t src_len) { size_t dst_len4 = dst_len / 4; size_t len = (dst_len4 < src_len) ? dst_len4 : src_len; uint8_t* d = dst_ptr; @@ -16359,8 +16359,8 @@ wuffs_base__pixel_swizzler__prepare__y(wuffs_base__pixel_swizzler* p, case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY: case WUFFS_BASE__PIXEL_FORMAT__RGBX: #if defined(WUFFS_BASE__CPU_ARCH__X86_64) - if (wuffs_base__cpu_arch__have_sse128()) { - return wuffs_base__pixel_swizzler__xxxx__y__sse128; + if (wuffs_base__cpu_arch__have_sse42()) { + return wuffs_base__pixel_swizzler__xxxx__y__sse42; } #endif return wuffs_base__pixel_swizzler__xxxx__y; @@ -16664,8 +16664,8 @@ wuffs_base__pixel_swizzler__prepare__rgb(wuffs_base__pixel_swizzler* p, case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY: case WUFFS_BASE__PIXEL_FORMAT__BGRX: #if defined(WUFFS_BASE__CPU_ARCH__X86_64) - if (wuffs_base__cpu_arch__have_sse128()) { - return wuffs_base__pixel_swizzler__bgrw__rgb__sse128; + if (wuffs_base__cpu_arch__have_sse42()) { + return wuffs_base__pixel_swizzler__bgrw__rgb__sse42; } #endif return wuffs_base__pixel_swizzler__bgrw__rgb; @@ -16711,8 +16711,8 @@ wuffs_base__pixel_swizzler__prepare__rgba_nonpremul( switch (blend) { case WUFFS_BASE__PIXEL_BLEND__SRC: #if defined(WUFFS_BASE__CPU_ARCH__X86_64) - if (wuffs_base__cpu_arch__have_sse128()) { - return wuffs_base__pixel_swizzler__swap_rgbx_bgrx__sse128; + if (wuffs_base__cpu_arch__have_sse42()) { + return wuffs_base__pixel_swizzler__swap_rgbx_bgrx__sse42; } #endif return wuffs_base__pixel_swizzler__swap_rgbx_bgrx; @@ -30289,14 +30289,14 @@ wuffs_png__decoder__filter_4_distance_4_fallback( #if defined(WUFFS_BASE__CPU_ARCH__X86_64) static wuffs_base__empty_struct -wuffs_png__decoder__filter_1_distance_4_sse128( +wuffs_png__decoder__filter_1_distance_4_sse42( wuffs_png__decoder* self, wuffs_base__slice_u8 a_curr); #endif // defined(WUFFS_BASE__CPU_ARCH__X86_64) #if defined(WUFFS_BASE__CPU_ARCH__X86_64) static wuffs_base__empty_struct -wuffs_png__decoder__filter_3_distance_4_sse128( +wuffs_png__decoder__filter_3_distance_4_sse42( wuffs_png__decoder* self, wuffs_base__slice_u8 a_curr, wuffs_base__slice_u8 a_prev); @@ -30304,7 +30304,7 @@ wuffs_png__decoder__filter_3_distance_4_sse128( #if defined(WUFFS_BASE__CPU_ARCH__X86_64) static wuffs_base__empty_struct -wuffs_png__decoder__filter_4_distance_3_sse128( +wuffs_png__decoder__filter_4_distance_3_sse42( wuffs_png__decoder* self, wuffs_base__slice_u8 a_curr, wuffs_base__slice_u8 a_prev); @@ -30312,7 +30312,7 @@ wuffs_png__decoder__filter_4_distance_3_sse128( #if defined(WUFFS_BASE__CPU_ARCH__X86_64) static wuffs_base__empty_struct -wuffs_png__decoder__filter_4_distance_4_sse128( +wuffs_png__decoder__filter_4_distance_4_sse42( wuffs_png__decoder* self, wuffs_base__slice_u8 a_curr, wuffs_base__slice_u8 a_prev); @@ -31141,14 +31141,14 @@ wuffs_png__decoder__filter_4_distance_4_fallback( return wuffs_base__make_empty_struct(); } -// -------- func png.decoder.filter_1_distance_4_sse128 +// -------- func png.decoder.filter_1_distance_4_sse42 #if defined(WUFFS_BASE__CPU_ARCH__X86_64) #if defined(__GNUC__) __attribute__((target("sse4.2"))) #endif static wuffs_base__empty_struct -wuffs_png__decoder__filter_1_distance_4_sse128( +wuffs_png__decoder__filter_1_distance_4_sse42( wuffs_png__decoder* self, wuffs_base__slice_u8 a_curr) { wuffs_base__slice_u8 v_c = {0}; @@ -31187,14 +31187,14 @@ wuffs_png__decoder__filter_1_distance_4_sse128( } #endif // defined(WUFFS_BASE__CPU_ARCH__X86_64) -// -------- func png.decoder.filter_3_distance_4_sse128 +// -------- func png.decoder.filter_3_distance_4_sse42 #if defined(WUFFS_BASE__CPU_ARCH__X86_64) #if defined(__GNUC__) __attribute__((target("sse4.2"))) #endif static wuffs_base__empty_struct -wuffs_png__decoder__filter_3_distance_4_sse128( +wuffs_png__decoder__filter_3_distance_4_sse42( wuffs_png__decoder* self, wuffs_base__slice_u8 a_curr, wuffs_base__slice_u8 a_prev) { @@ -31292,14 +31292,14 @@ wuffs_png__decoder__filter_3_distance_4_sse128( } #endif // defined(WUFFS_BASE__CPU_ARCH__X86_64) -// -------- func png.decoder.filter_4_distance_3_sse128 +// -------- func png.decoder.filter_4_distance_3_sse42 #if defined(WUFFS_BASE__CPU_ARCH__X86_64) #if defined(__GNUC__) __attribute__((target("sse4.2"))) #endif static wuffs_base__empty_struct -wuffs_png__decoder__filter_4_distance_3_sse128( +wuffs_png__decoder__filter_4_distance_3_sse42( wuffs_png__decoder* self, wuffs_base__slice_u8 a_curr, wuffs_base__slice_u8 a_prev) { @@ -31418,14 +31418,14 @@ wuffs_png__decoder__filter_4_distance_3_sse128( } #endif // defined(WUFFS_BASE__CPU_ARCH__X86_64) -// -------- func png.decoder.filter_4_distance_4_sse128 +// -------- func png.decoder.filter_4_distance_4_sse42 #if defined(WUFFS_BASE__CPU_ARCH__X86_64) #if defined(__GNUC__) __attribute__((target("sse4.2"))) #endif static wuffs_base__empty_struct -wuffs_png__decoder__filter_4_distance_4_sse128( +wuffs_png__decoder__filter_4_distance_4_sse42( wuffs_png__decoder* self, wuffs_base__slice_u8 a_curr, wuffs_base__slice_u8 a_prev) { @@ -32138,23 +32138,23 @@ wuffs_png__decoder__choose_filter_implementations( &wuffs_png__decoder__filter_3_distance_3_fallback); self->private_impl.choosy_filter_4 = ( #if defined(WUFFS_BASE__CPU_ARCH__X86_64) - wuffs_base__cpu_arch__have_sse128() ? &wuffs_png__decoder__filter_4_distance_3_sse128 : + wuffs_base__cpu_arch__have_sse42() ? &wuffs_png__decoder__filter_4_distance_3_sse42 : #endif &wuffs_png__decoder__filter_4_distance_3_fallback); } else if (self->private_impl.f_filter_distance == 4) { self->private_impl.choosy_filter_1 = ( #if defined(WUFFS_BASE__CPU_ARCH__X86_64) - wuffs_base__cpu_arch__have_sse128() ? &wuffs_png__decoder__filter_1_distance_4_sse128 : + wuffs_base__cpu_arch__have_sse42() ? &wuffs_png__decoder__filter_1_distance_4_sse42 : #endif &wuffs_png__decoder__filter_1_distance_4_fallback); self->private_impl.choosy_filter_3 = ( #if defined(WUFFS_BASE__CPU_ARCH__X86_64) - wuffs_base__cpu_arch__have_sse128() ? &wuffs_png__decoder__filter_3_distance_4_sse128 : + wuffs_base__cpu_arch__have_sse42() ? &wuffs_png__decoder__filter_3_distance_4_sse42 : #endif &wuffs_png__decoder__filter_3_distance_4_fallback); self->private_impl.choosy_filter_4 = ( #if defined(WUFFS_BASE__CPU_ARCH__X86_64) - wuffs_base__cpu_arch__have_sse128() ? &wuffs_png__decoder__filter_4_distance_4_sse128 : + wuffs_base__cpu_arch__have_sse42() ? &wuffs_png__decoder__filter_4_distance_4_sse42 : #endif &wuffs_png__decoder__filter_4_distance_4_fallback); } diff --git a/std/png/decode_filter_sse128.wuffs b/std/png/decode_filter_sse128.wuffs index ac5661c4..e96a3048 100644 --- a/std/png/decode_filter_sse128.wuffs +++ b/std/png/decode_filter_sse128.wuffs @@ -19,7 +19,7 @@ // This (filter = 1, distance = 3) implementation doesn't actually bench faster // than the non-SIMD one. // -// pri func decoder.filter_1_distance_3_sse128!(curr: slice base.u8), +// pri func decoder.filter_1_distance_3_sse42!(curr: slice base.u8), // choose cpu_arch >= x86_sse42, // { // var c : slice base.u8 @@ -38,7 +38,7 @@ // } // } -pri func decoder.filter_1_distance_4_sse128!(curr: slice base.u8), +pri func decoder.filter_1_distance_4_sse42!(curr: slice base.u8), choose cpu_arch >= x86_sse42, { var c : slice base.u8 @@ -57,16 +57,16 @@ pri func decoder.filter_1_distance_4_sse128!(curr: slice base.u8), // Filter 3: Average. -// Similar to filter_1_distance_3_sse128, the SIMD implementation for (filter = +// Similar to filter_1_distance_3_sse42, the SIMD implementation for (filter = // 3, distance = 3) doesn't actually bench faster than the non-SIMD one. // -// pri func decoder.filter_3_distance_3_sse128!(curr: slice base.u8, prev: slice base.u8), +// pri func decoder.filter_3_distance_3_sse42!(curr: slice base.u8, prev: slice base.u8), // choose cpu_arch >= x86_sse42, // { // etc // } -pri func decoder.filter_3_distance_4_sse128!(curr: slice base.u8, prev: slice base.u8), +pri func decoder.filter_3_distance_4_sse42!(curr: slice base.u8, prev: slice base.u8), choose cpu_arch >= x86_sse42, { var c : slice base.u8 @@ -123,10 +123,10 @@ pri func decoder.filter_3_distance_4_sse128!(curr: slice base.u8, prev: slice ba // Filter 4: Paeth. -pri func decoder.filter_4_distance_3_sse128!(curr: slice base.u8, prev: slice base.u8), +pri func decoder.filter_4_distance_3_sse42!(curr: slice base.u8, prev: slice base.u8), choose cpu_arch >= x86_sse42, { - // See the comments in filter_4_distance_4_sse128 for an explanation of how + // See the comments in filter_4_distance_4_sse42 for an explanation of how // this works. That function's single loop is copied twice here, once with // "length: 4" and once with "length: 3". It's generally faster to load 4 // bytes at a time instead of 3. @@ -201,7 +201,7 @@ pri func decoder.filter_4_distance_3_sse128!(curr: slice base.u8, prev: slice ba } } -pri func decoder.filter_4_distance_4_sse128!(curr: slice base.u8, prev: slice base.u8), +pri func decoder.filter_4_distance_4_sse42!(curr: slice base.u8, prev: slice base.u8), choose cpu_arch >= x86_sse42, { var c : slice base.u8 diff --git a/std/png/decode_png.wuffs b/std/png/decode_png.wuffs index a6a6ef76..d2f0cb2d 100644 --- a/std/png/decode_png.wuffs +++ b/std/png/decode_png.wuffs @@ -313,17 +313,17 @@ pri func decoder.choose_filter_implementations!() { choose filter_1 = [filter_1_distance_3_fallback] choose filter_3 = [filter_3_distance_3_fallback] choose filter_4 = [ - filter_4_distance_3_sse128, + filter_4_distance_3_sse42, filter_4_distance_3_fallback] } else if this.filter_distance == 4 { choose filter_1 = [ - filter_1_distance_4_sse128, + filter_1_distance_4_sse42, filter_1_distance_4_fallback] choose filter_3 = [ - filter_3_distance_4_sse128, + filter_3_distance_4_sse42, filter_3_distance_4_fallback] choose filter_4 = [ - filter_4_distance_4_sse128, + filter_4_distance_4_sse42, filter_4_distance_4_fallback] } }