From 4a5df43aeb623e96949228bc0a69bd2d18df768f Mon Sep 17 00:00:00 2001 From: Nigel Tao Date: Mon, 16 Nov 2020 13:22:43 +1100 Subject: [PATCH] Add PIXEL_FORMAT__BGRA_NONPREMUL_4X16LE Binary size, before: 14272 gen/lib/c/clang-9-dynamic/wuffs-base-pixconv.lo 14760 gen/lib/c/clang-9-static/wuffs-base-pixconv.o 22336 gen/lib/c/gcc-dynamic/wuffs-base-pixconv.lo 22280 gen/lib/c/gcc-static/wuffs-base-pixconv.o After: 18712 gen/lib/c/clang-9-dynamic/wuffs-base-pixconv.lo 19392 gen/lib/c/clang-9-static/wuffs-base-pixconv.o 36880 gen/lib/c/gcc-dynamic/wuffs-base-pixconv.lo 36824 gen/lib/c/gcc-static/wuffs-base-pixconv.o --- internal/cgen/base/image-public.h | 66 +++ internal/cgen/base/pixconv-submodule.c | 507 ++++++++++++++++++++-- internal/cgen/data/data.go | 94 ++-- lang/builtin/builtin.go | 4 + release/c/wuffs-unsupported-snapshot.c | 573 +++++++++++++++++++++++-- test/c/std/wbmp.c | 4 + 6 files changed, 1162 insertions(+), 86 deletions(-) diff --git a/internal/cgen/base/image-public.h b/internal/cgen/base/image-public.h index 4ca47dea..002bf2b5 100644 --- a/internal/cgen/base/image-public.h +++ b/internal/cgen/base/image-public.h @@ -109,6 +109,68 @@ wuffs_base__color_u32_argb_premul__as__color_u32_argb_nonpremul( return (a << 24) | (r << 16) | (g << 8) | (b << 0); } +// wuffs_base__color_u64_argb_nonpremul__as__color_u32_argb_premul converts +// from 4x16LE non-premultiplied alpha to 4x8 premultiplied alpha. +static inline wuffs_base__color_u32_argb_premul // +wuffs_base__color_u64_argb_nonpremul__as__color_u32_argb_premul( + uint64_t argb_nonpremul) { + uint32_t a16 = 0xFFFF & (argb_nonpremul >> 48); + + uint32_t r16 = 0xFFFF & (argb_nonpremul >> 32); + r16 = (r16 * a16) / 0xFFFF; + uint32_t g16 = 0xFFFF & (argb_nonpremul >> 16); + g16 = (g16 * a16) / 0xFFFF; + uint32_t b16 = 0xFFFF & (argb_nonpremul >> 0); + b16 = (b16 * a16) / 0xFFFF; + + return ((a16 >> 8) << 24) | ((r16 >> 8) << 16) | ((g16 >> 8) << 8) | + ((b16 >> 8) << 0); +} + +// wuffs_base__color_u32_argb_premul__as__color_u64_argb_nonpremul converts +// from 4x8 premultiplied alpha to 4x16LE non-premultiplied alpha. +static inline uint64_t // +wuffs_base__color_u32_argb_premul__as__color_u64_argb_nonpremul( + wuffs_base__color_u32_argb_premul c) { + uint32_t a = 0xFF & (c >> 24); + if (a == 0xFF) { + uint64_t r16 = 0x101 * (0xFF & (c >> 16)); + uint64_t g16 = 0x101 * (0xFF & (c >> 8)); + uint64_t b16 = 0x101 * (0xFF & (c >> 0)); + return 0xFFFF000000000000u | (r16 << 32) | (g16 << 16) | (b16 << 0); + } else if (a == 0) { + return 0; + } + uint64_t a16 = a * 0x101; + + uint64_t r = 0xFF & (c >> 16); + uint64_t r16 = (r * (0x101 * 0xFFFF)) / a16; + uint64_t g = 0xFF & (c >> 8); + uint64_t g16 = (g * (0x101 * 0xFFFF)) / a16; + uint64_t b = 0xFF & (c >> 0); + uint64_t b16 = (b * (0x101 * 0xFFFF)) / a16; + + return (a16 << 48) | (r16 << 32) | (g16 << 16) | (b16 << 0); +} + +static inline uint64_t // +wuffs_base__color_u32__as__color_u64(uint32_t c) { + uint64_t a16 = 0x101 * (0xFF & (c >> 24)); + uint64_t r16 = 0x101 * (0xFF & (c >> 16)); + uint64_t g16 = 0x101 * (0xFF & (c >> 8)); + uint64_t b16 = 0x101 * (0xFF & (c >> 0)); + return (a16 << 48) | (r16 << 32) | (g16 << 16) | (b16 << 0); +} + +static inline uint32_t // +wuffs_base__color_u64__as__color_u32(uint64_t c) { + uint32_t a = ((uint32_t)(0xFF & (c >> 56))); + uint32_t r = ((uint32_t)(0xFF & (c >> 40))); + uint32_t g = ((uint32_t)(0xFF & (c >> 24))); + uint32_t b = ((uint32_t)(0xFF & (c >> 8))); + return (a << 24) | (r << 16) | (g << 8) | (b << 0); +} + // -------- typedef uint8_t wuffs_base__pixel_blend; @@ -196,13 +258,17 @@ wuffs_base__make_pixel_format(uint32_t repr) { #define WUFFS_BASE__PIXEL_FORMAT__BGR_565 0x80000565 #define WUFFS_BASE__PIXEL_FORMAT__BGR 0x80000888 #define WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL 0x81008888 +#define WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL_4X16LE 0x8100BBBB #define WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL 0x82008888 +#define WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL_4X16LE 0x8200BBBB #define WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY 0x83008888 #define WUFFS_BASE__PIXEL_FORMAT__BGRX 0x90008888 #define WUFFS_BASE__PIXEL_FORMAT__RGB 0xA0000888 #define WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL 0xA1008888 +#define WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL_4X16LE 0xA100BBBB #define WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL 0xA2008888 +#define WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL_4X16LE 0xA200BBBB #define WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY 0xA3008888 #define WUFFS_BASE__PIXEL_FORMAT__RGBX 0xB0008888 diff --git a/internal/cgen/base/pixconv-submodule.c b/internal/cgen/base/pixconv-submodule.c index bde2517d..edcd5353 100644 --- a/internal/cgen/base/pixconv-submodule.c +++ b/internal/cgen/base/pixconv-submodule.c @@ -76,6 +76,9 @@ wuffs_base__pixel_buffer__color_u32_at(const wuffs_base__pixel_buffer* pb, case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL: return wuffs_base__color_u32_argb_nonpremul__as__color_u32_argb_premul( wuffs_base__load_u32le__no_bounds_check(row + (4 * ((size_t)x)))); + case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL_4X16LE: + return wuffs_base__color_u64_argb_nonpremul__as__color_u32_argb_premul( + wuffs_base__load_u64le__no_bounds_check(row + (8 * ((size_t)x)))); case WUFFS_BASE__PIXEL_FORMAT__BGRX: return 0xFF000000 | wuffs_base__load_u32le__no_bounds_check(row + (4 * ((size_t)x))); @@ -163,6 +166,12 @@ wuffs_base__pixel_buffer__set_color_u32_at( wuffs_base__color_u32_argb_premul__as__color_u32_argb_nonpremul( color)); break; + case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL_4X16LE: + wuffs_base__store_u64le__no_bounds_check( + row + (8 * ((size_t)x)), + wuffs_base__color_u32_argb_premul__as__color_u64_argb_nonpremul( + color)); + break; case WUFFS_BASE__PIXEL_FORMAT__RGB: wuffs_base__store_u24le__no_bounds_check( @@ -255,7 +264,7 @@ wuffs_base__pixel_palette__closest_element( static inline uint32_t // wuffs_base__composite_nonpremul_nonpremul_u32_axxx(uint32_t dst_nonpremul, uint32_t src_nonpremul) { - // Convert from 8-bit color to 16-bit color. + // Extract 16-bit color components. uint32_t sa = 0x101 * (0xFF & (src_nonpremul >> 24)); uint32_t sr = 0x101 * (0xFF & (src_nonpremul >> 16)); uint32_t sg = 0x101 * (0xFF & (src_nonpremul >> 8)); @@ -286,18 +295,58 @@ wuffs_base__composite_nonpremul_nonpremul_u32_axxx(uint32_t dst_nonpremul, db = (db * 0xFFFF) / da; } - // Convert from 16-bit color to 8-bit color and combine the components. + // Convert from 16-bit color to 8-bit color. da >>= 8; dr >>= 8; dg >>= 8; db >>= 8; + + // Combine components. return (db << 0) | (dg << 8) | (dr << 16) | (da << 24); } +static inline uint64_t // +wuffs_base__composite_nonpremul_nonpremul_u64_axxx(uint64_t dst_nonpremul, + uint64_t src_nonpremul) { + // Extract components. + uint64_t sa = 0xFFFF & (src_nonpremul >> 48); + uint64_t sr = 0xFFFF & (src_nonpremul >> 32); + uint64_t sg = 0xFFFF & (src_nonpremul >> 16); + uint64_t sb = 0xFFFF & (src_nonpremul >> 0); + uint64_t da = 0xFFFF & (dst_nonpremul >> 48); + uint64_t dr = 0xFFFF & (dst_nonpremul >> 32); + uint64_t dg = 0xFFFF & (dst_nonpremul >> 16); + uint64_t db = 0xFFFF & (dst_nonpremul >> 0); + + // Convert dst from nonpremul to premul. + dr = (dr * da) / 0xFFFF; + dg = (dg * da) / 0xFFFF; + db = (db * da) / 0xFFFF; + + // Calculate the inverse of the src-alpha: how much of the dst to keep. + uint64_t ia = 0xFFFF - sa; + + // Composite src (nonpremul) over dst (premul). + da = sa + ((da * ia) / 0xFFFF); + dr = ((sr * sa) + (dr * ia)) / 0xFFFF; + dg = ((sg * sa) + (dg * ia)) / 0xFFFF; + db = ((sb * sa) + (db * ia)) / 0xFFFF; + + // Convert dst from premul to nonpremul. + if (da != 0) { + dr = (dr * 0xFFFF) / da; + dg = (dg * 0xFFFF) / da; + db = (db * 0xFFFF) / da; + } + + // Combine components. + return (db << 0) | (dg << 16) | (dr << 32) | (da << 48); +} + static inline uint32_t // wuffs_base__composite_nonpremul_premul_u32_axxx(uint32_t dst_nonpremul, uint32_t src_premul) { - // Convert from 8-bit color to 16-bit color. + // Extract 16-bit color components. uint32_t sa = 0x101 * (0xFF & (src_premul >> 24)); uint32_t sr = 0x101 * (0xFF & (src_premul >> 16)); uint32_t sg = 0x101 * (0xFF & (src_premul >> 8)); @@ -328,18 +377,20 @@ wuffs_base__composite_nonpremul_premul_u32_axxx(uint32_t dst_nonpremul, db = (db * 0xFFFF) / da; } - // Convert from 16-bit color to 8-bit color and combine the components. + // Convert from 16-bit color to 8-bit color. da >>= 8; dr >>= 8; dg >>= 8; db >>= 8; + + // Combine components. return (db << 0) | (dg << 8) | (dr << 16) | (da << 24); } static inline uint32_t // wuffs_base__composite_premul_nonpremul_u32_axxx(uint32_t dst_premul, uint32_t src_nonpremul) { - // Convert from 8-bit color to 16-bit color. + // Extract 16-bit color components. uint32_t sa = 0x101 * (0xFF & (src_nonpremul >> 24)); uint32_t sr = 0x101 * (0xFF & (src_nonpremul >> 16)); uint32_t sg = 0x101 * (0xFF & (src_nonpremul >> 8)); @@ -358,18 +409,46 @@ wuffs_base__composite_premul_nonpremul_u32_axxx(uint32_t dst_premul, dg = ((sg * sa) + (dg * ia)) / 0xFFFF; db = ((sb * sa) + (db * ia)) / 0xFFFF; - // Convert from 16-bit color to 8-bit color and combine the components. + // Convert from 16-bit color to 8-bit color. da >>= 8; dr >>= 8; dg >>= 8; db >>= 8; + + // Combine components. return (db << 0) | (dg << 8) | (dr << 16) | (da << 24); } +static inline uint64_t // +wuffs_base__composite_premul_nonpremul_u64_axxx(uint64_t dst_premul, + uint64_t src_nonpremul) { + // Extract components. + uint64_t sa = 0xFFFF & (src_nonpremul >> 48); + uint64_t sr = 0xFFFF & (src_nonpremul >> 32); + uint64_t sg = 0xFFFF & (src_nonpremul >> 16); + uint64_t sb = 0xFFFF & (src_nonpremul >> 0); + uint64_t da = 0xFFFF & (dst_premul >> 48); + uint64_t dr = 0xFFFF & (dst_premul >> 32); + uint64_t dg = 0xFFFF & (dst_premul >> 16); + uint64_t db = 0xFFFF & (dst_premul >> 0); + + // Calculate the inverse of the src-alpha: how much of the dst to keep. + uint64_t ia = 0xFFFF - sa; + + // Composite src (nonpremul) over dst (premul). + da = sa + ((da * ia) / 0xFFFF); + dr = ((sr * sa) + (dr * ia)) / 0xFFFF; + dg = ((sg * sa) + (dg * ia)) / 0xFFFF; + db = ((sb * sa) + (db * ia)) / 0xFFFF; + + // Combine components. + return (db << 0) | (dg << 16) | (dr << 32) | (da << 48); +} + static inline uint32_t // wuffs_base__composite_premul_premul_u32_axxx(uint32_t dst_premul, uint32_t src_premul) { - // Convert from 8-bit color to 16-bit color. + // Extract 16-bit color components. uint32_t sa = 0x101 * (0xFF & (src_premul >> 24)); uint32_t sr = 0x101 * (0xFF & (src_premul >> 16)); uint32_t sg = 0x101 * (0xFF & (src_premul >> 8)); @@ -388,24 +467,27 @@ wuffs_base__composite_premul_premul_u32_axxx(uint32_t dst_premul, dg = sg + ((dg * ia) / 0xFFFF); db = sb + ((db * ia) / 0xFFFF); - // Convert from 16-bit color to 8-bit color and combine the components. + // Convert from 16-bit color to 8-bit color. da >>= 8; dr >>= 8; dg >>= 8; db >>= 8; + + // Combine components. return (db << 0) | (dg << 8) | (dr << 16) | (da << 24); } // -------- static uint64_t // -wuffs_base__pixel_swizzler__squash_bgr_565_888(wuffs_base__slice_u8 dst, - wuffs_base__slice_u8 src) { - size_t len4 = (dst.len < src.len ? dst.len : src.len) / 4; +wuffs_base__pixel_swizzler__squash_align4_bgr_565_888( + wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 src) { + size_t len = (dst.len < src.len ? dst.len : src.len) / 4; uint8_t* d = dst.ptr; const uint8_t* s = src.ptr; - size_t n = len4; + size_t n = len; while (n--) { uint32_t argb = wuffs_base__load_u32le__no_bounds_check(s); uint32_t b5 = 0x1F & (argb >> (8 - 5)); @@ -417,17 +499,17 @@ wuffs_base__pixel_swizzler__squash_bgr_565_888(wuffs_base__slice_u8 dst, s += 4; d += 4; } - return len4 * 4; + return len; } static uint64_t // wuffs_base__pixel_swizzler__swap_rgbx_bgrx(wuffs_base__slice_u8 dst, wuffs_base__slice_u8 src) { - size_t len4 = (dst.len < src.len ? dst.len : src.len) / 4; + size_t len = (dst.len < src.len ? dst.len : src.len) / 4; uint8_t* d = dst.ptr; const uint8_t* s = src.ptr; - size_t n = len4; + size_t n = len; while (n--) { uint8_t b0 = s[0]; uint8_t b1 = s[1]; @@ -440,7 +522,56 @@ wuffs_base__pixel_swizzler__swap_rgbx_bgrx(wuffs_base__slice_u8 dst, s += 4; d += 4; } - return len4 * 4; + return len; +} + +// -------- + +static uint64_t // +wuffs_base__pixel_swizzler__squash_tight_4x8_4x16le(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_len8 = src_len / 8; + size_t len = (dst_len4 < src_len8) ? dst_len4 : src_len8; + uint8_t* d = dst_ptr; + const uint8_t* s = src_ptr; + + const size_t loop_unroll_count = 4; + + size_t n = len; + while (n >= loop_unroll_count) { + wuffs_base__store_u32le__no_bounds_check( + d + (0 * 4), wuffs_base__color_u64__as__color_u32( + wuffs_base__load_u64le__no_bounds_check(s + (0 * 8)))); + wuffs_base__store_u32le__no_bounds_check( + d + (1 * 4), wuffs_base__color_u64__as__color_u32( + wuffs_base__load_u64le__no_bounds_check(s + (1 * 8)))); + wuffs_base__store_u32le__no_bounds_check( + d + (2 * 4), wuffs_base__color_u64__as__color_u32( + wuffs_base__load_u64le__no_bounds_check(s + (2 * 8)))); + wuffs_base__store_u32le__no_bounds_check( + d + (3 * 4), wuffs_base__color_u64__as__color_u32( + wuffs_base__load_u64le__no_bounds_check(s + (3 * 8)))); + + s += loop_unroll_count * 8; + d += loop_unroll_count * 4; + n -= loop_unroll_count; + } + + while (n >= 1) { + wuffs_base__store_u32le__no_bounds_check( + d + (0 * 4), wuffs_base__color_u64__as__color_u32( + wuffs_base__load_u64le__no_bounds_check(s + (0 * 8)))); + + s += 1 * 8; + d += 1 * 4; + n -= 1; + } + return len; } // -------- @@ -556,6 +687,38 @@ wuffs_base__pixel_swizzler__bgr_565__bgra_nonpremul__src( return len; } +static uint64_t // +wuffs_base__pixel_swizzler__bgr_565__bgra_nonpremul_4x16le__src( + 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_len2 = dst_len / 2; + size_t src_len8 = src_len / 8; + size_t len = (dst_len2 < src_len8) ? dst_len2 : src_len8; + uint8_t* d = dst_ptr; + const uint8_t* s = src_ptr; + size_t n = len; + + // TODO: unroll. + + while (n >= 1) { + wuffs_base__store_u16le__no_bounds_check( + d + (0 * 2), + wuffs_base__color_u32_argb_premul__as__color_u16_rgb_565( + wuffs_base__color_u64_argb_nonpremul__as__color_u32_argb_premul( + wuffs_base__load_u64le__no_bounds_check(s + (0 * 8))))); + + s += 1 * 8; + d += 1 * 2; + n -= 1; + } + + return len; +} + static uint64_t // wuffs_base__pixel_swizzler__bgr_565__bgra_nonpremul__src_over( uint8_t* dst_ptr, @@ -574,7 +737,7 @@ wuffs_base__pixel_swizzler__bgr_565__bgra_nonpremul__src_over( // TODO: unroll. while (n >= 1) { - // Convert from 8-bit color to 16-bit color. + // Extract 16-bit color components. uint32_t sa = 0x101 * ((uint32_t)s[3]); uint32_t sr = 0x101 * ((uint32_t)s[2]); uint32_t sg = 0x101 * ((uint32_t)s[1]); @@ -613,6 +776,63 @@ wuffs_base__pixel_swizzler__bgr_565__bgra_nonpremul__src_over( return len; } +static uint64_t // +wuffs_base__pixel_swizzler__bgr_565__bgra_nonpremul_4x16le__src_over( + 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_len2 = dst_len / 2; + size_t src_len8 = src_len / 8; + size_t len = (dst_len2 < src_len8) ? dst_len2 : src_len8; + uint8_t* d = dst_ptr; + const uint8_t* s = src_ptr; + size_t n = len; + + // TODO: unroll. + + while (n >= 1) { + // Extract 16-bit color components. + uint32_t sa = ((uint32_t)wuffs_base__load_u16le__no_bounds_check(s + 6)); + uint32_t sr = ((uint32_t)wuffs_base__load_u16le__no_bounds_check(s + 4)); + uint32_t sg = ((uint32_t)wuffs_base__load_u16le__no_bounds_check(s + 2)); + uint32_t sb = ((uint32_t)wuffs_base__load_u16le__no_bounds_check(s + 0)); + + // Convert from 565 color to 16-bit color. + uint32_t old_rgb_565 = wuffs_base__load_u16le__no_bounds_check(d + (0 * 2)); + uint32_t old_r5 = 0x1F & (old_rgb_565 >> 11); + uint32_t dr = (0x8421 * old_r5) >> 4; + uint32_t old_g6 = 0x3F & (old_rgb_565 >> 5); + uint32_t dg = (0x1041 * old_g6) >> 2; + uint32_t old_b5 = 0x1F & (old_rgb_565 >> 0); + uint32_t db = (0x8421 * old_b5) >> 4; + + // Calculate the inverse of the src-alpha: how much of the dst to keep. + uint32_t ia = 0xFFFF - sa; + + // Composite src (nonpremul) over dst (premul). + dr = ((sr * sa) + (dr * ia)) / 0xFFFF; + dg = ((sg * sa) + (dg * ia)) / 0xFFFF; + db = ((sb * sa) + (db * ia)) / 0xFFFF; + + // Convert from 16-bit color to 565 color and combine the components. + uint32_t new_r5 = 0x1F & (dr >> 11); + uint32_t new_g6 = 0x3F & (dg >> 10); + uint32_t new_b5 = 0x1F & (db >> 11); + uint32_t new_rgb_565 = (new_r5 << 11) | (new_g6 << 5) | (new_b5 << 0); + wuffs_base__store_u16le__no_bounds_check(d + (0 * 2), + (uint16_t)new_rgb_565); + + s += 1 * 8; + d += 1 * 2; + n -= 1; + } + + return len; +} + static uint64_t // wuffs_base__pixel_swizzler__bgr_565__y(uint8_t* dst_ptr, size_t dst_len, @@ -758,6 +978,37 @@ wuffs_base__pixel_swizzler__bgr__bgra_nonpremul__src(uint8_t* dst_ptr, return len; } +static uint64_t // +wuffs_base__pixel_swizzler__bgr__bgra_nonpremul_4x16le__src( + 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_len3 = dst_len / 3; + size_t src_len8 = src_len / 8; + size_t len = (dst_len3 < src_len8) ? dst_len3 : src_len8; + uint8_t* d = dst_ptr; + const uint8_t* s = src_ptr; + size_t n = len; + + // TODO: unroll. + + while (n >= 1) { + uint32_t s0 = + wuffs_base__color_u64_argb_nonpremul__as__color_u32_argb_premul( + wuffs_base__load_u64le__no_bounds_check(s + (0 * 8))); + wuffs_base__store_u24le__no_bounds_check(d + (0 * 3), s0); + + s += 1 * 8; + d += 1 * 3; + n -= 1; + } + + return len; +} + static uint64_t // wuffs_base__pixel_swizzler__bgr__bgra_nonpremul__src_over( uint8_t* dst_ptr, @@ -776,7 +1027,7 @@ wuffs_base__pixel_swizzler__bgr__bgra_nonpremul__src_over( // TODO: unroll. while (n >= 1) { - // Convert from 8-bit color to 16-bit color. + // Extract 16-bit color components. uint32_t sa = 0x101 * ((uint32_t)s[3]); uint32_t sr = 0x101 * ((uint32_t)s[2]); uint32_t sg = 0x101 * ((uint32_t)s[1]); @@ -806,6 +1057,54 @@ wuffs_base__pixel_swizzler__bgr__bgra_nonpremul__src_over( return len; } +static uint64_t // +wuffs_base__pixel_swizzler__bgr__bgra_nonpremul_4x16le__src_over( + 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_len3 = dst_len / 3; + size_t src_len8 = src_len / 8; + size_t len = (dst_len3 < src_len8) ? dst_len3 : src_len8; + uint8_t* d = dst_ptr; + const uint8_t* s = src_ptr; + size_t n = len; + + // TODO: unroll. + + while (n >= 1) { + // Extract 16-bit color components. + uint32_t sa = ((uint32_t)wuffs_base__load_u16le__no_bounds_check(s + 6)); + uint32_t sr = ((uint32_t)wuffs_base__load_u16le__no_bounds_check(s + 4)); + uint32_t sg = ((uint32_t)wuffs_base__load_u16le__no_bounds_check(s + 2)); + uint32_t sb = ((uint32_t)wuffs_base__load_u16le__no_bounds_check(s + 0)); + uint32_t dr = 0x101 * ((uint32_t)d[2]); + uint32_t dg = 0x101 * ((uint32_t)d[1]); + uint32_t db = 0x101 * ((uint32_t)d[0]); + + // Calculate the inverse of the src-alpha: how much of the dst to keep. + uint32_t ia = 0xFFFF - sa; + + // Composite src (nonpremul) over dst (premul). + dr = ((sr * sa) + (dr * ia)) / 0xFFFF; + dg = ((sg * sa) + (dg * ia)) / 0xFFFF; + db = ((sb * sa) + (db * ia)) / 0xFFFF; + + // Convert from 16-bit color to 8-bit color. + d[0] = (uint8_t)(db >> 8); + d[1] = (uint8_t)(dg >> 8); + d[2] = (uint8_t)(dr >> 8); + + s += 1 * 8; + d += 1 * 3; + n -= 1; + } + + return len; +} + // -------- static uint64_t // @@ -823,8 +1122,6 @@ wuffs_base__pixel_swizzler__bgra_nonpremul__bgra_nonpremul__src_over( const uint8_t* s = src_ptr; size_t n = len; - // TODO: unroll. - while (n >= 1) { uint32_t d0 = wuffs_base__load_u32le__no_bounds_check(d + (0 * 4)); uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(s + (0 * 4)); @@ -840,6 +1137,38 @@ wuffs_base__pixel_swizzler__bgra_nonpremul__bgra_nonpremul__src_over( return len; } +static uint64_t // +wuffs_base__pixel_swizzler__bgra_nonpremul__bgra_nonpremul_4x16le__src_over( + 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_len8 = src_len / 8; + size_t len = (dst_len4 < src_len8) ? dst_len4 : src_len8; + uint8_t* d = dst_ptr; + const uint8_t* s = src_ptr; + size_t n = len; + + while (n >= 1) { + uint64_t d0 = wuffs_base__color_u32__as__color_u64( + wuffs_base__load_u32le__no_bounds_check(d + (0 * 4))); + uint64_t s0 = wuffs_base__load_u64le__no_bounds_check(s + (0 * 8)); + wuffs_base__store_u32le__no_bounds_check( + d + (0 * 4), + wuffs_base__color_u64__as__color_u32( + wuffs_base__composite_nonpremul_nonpremul_u64_axxx(d0, s0))); + + s += 1 * 8; + d += 1 * 4; + n -= 1; + } + + return len; +} + // -------- static uint64_t // @@ -873,6 +1202,37 @@ wuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul__src( return len; } +static uint64_t // +wuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul_4x16le__src( + 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_len8 = src_len / 8; + size_t len = (dst_len4 < src_len8) ? dst_len4 : src_len8; + uint8_t* d = dst_ptr; + const uint8_t* s = src_ptr; + size_t n = len; + + // TODO: unroll. + + while (n >= 1) { + uint64_t s0 = wuffs_base__load_u64le__no_bounds_check(s + (0 * 8)); + wuffs_base__store_u32le__no_bounds_check( + d + (0 * 4), + wuffs_base__color_u64_argb_nonpremul__as__color_u32_argb_premul(s0)); + + s += 1 * 8; + d += 1 * 4; + n -= 1; + } + + return len; +} + static uint64_t // wuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul__src_over( uint8_t* dst_ptr, @@ -904,6 +1264,40 @@ wuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul__src_over( return len; } +static uint64_t // +wuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul_4x16le__src_over( + 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_len8 = src_len / 8; + size_t len = (dst_len4 < src_len8) ? dst_len4 : src_len8; + uint8_t* d = dst_ptr; + const uint8_t* s = src_ptr; + size_t n = len; + + // TODO: unroll. + + while (n >= 1) { + uint64_t d0 = wuffs_base__color_u32__as__color_u64( + wuffs_base__load_u32le__no_bounds_check(d + (0 * 4))); + uint64_t s0 = wuffs_base__load_u64le__no_bounds_check(s + (0 * 8)); + wuffs_base__store_u32le__no_bounds_check( + d + (0 * 4), + wuffs_base__color_u64__as__color_u32( + wuffs_base__composite_premul_nonpremul_u64_axxx(d0, s0))); + + s += 1 * 8; + d += 1 * 4; + n -= 1; + } + + return len; +} + // -------- static uint64_t // @@ -1273,8 +1667,8 @@ wuffs_base__pixel_swizzler__prepare__indexed__bgra_binary( return NULL; case WUFFS_BASE__PIXEL_FORMAT__BGR_565: - if (wuffs_base__pixel_swizzler__squash_bgr_565_888(dst_palette, - src_palette) != 1024) { + if (wuffs_base__pixel_swizzler__squash_align4_bgr_565_888( + dst_palette, src_palette) != 256) { return NULL; } switch (blend) { @@ -1315,7 +1709,7 @@ wuffs_base__pixel_swizzler__prepare__indexed__bgra_binary( case WUFFS_BASE__PIXEL_FORMAT__RGB: if (wuffs_base__pixel_swizzler__swap_rgbx_bgrx(dst_palette, - src_palette) != 1024) { + src_palette) != 256) { return NULL; } switch (blend) { @@ -1330,7 +1724,7 @@ wuffs_base__pixel_swizzler__prepare__indexed__bgra_binary( case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL: case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY: if (wuffs_base__pixel_swizzler__swap_rgbx_bgrx(dst_palette, - src_palette) != 1024) { + src_palette) != 256) { return NULL; } switch (blend) { @@ -1434,6 +1828,66 @@ wuffs_base__pixel_swizzler__prepare__bgra_nonpremul( return NULL; } +static wuffs_base__pixel_swizzler__func // +wuffs_base__pixel_swizzler__prepare__bgra_nonpremul_4x16le( + wuffs_base__pixel_swizzler* p, + wuffs_base__pixel_format dst_pixfmt, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src_palette, + wuffs_base__pixel_blend blend) { + switch (dst_pixfmt.repr) { + case WUFFS_BASE__PIXEL_FORMAT__BGR_565: + switch (blend) { + case WUFFS_BASE__PIXEL_BLEND__SRC: + return wuffs_base__pixel_swizzler__bgr_565__bgra_nonpremul_4x16le__src; + case WUFFS_BASE__PIXEL_BLEND__SRC_OVER: + return wuffs_base__pixel_swizzler__bgr_565__bgra_nonpremul_4x16le__src_over; + } + return NULL; + + case WUFFS_BASE__PIXEL_FORMAT__BGR: + switch (blend) { + case WUFFS_BASE__PIXEL_BLEND__SRC: + return wuffs_base__pixel_swizzler__bgr__bgra_nonpremul_4x16le__src; + case WUFFS_BASE__PIXEL_BLEND__SRC_OVER: + return wuffs_base__pixel_swizzler__bgr__bgra_nonpremul_4x16le__src_over; + } + return NULL; + + case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL: + switch (blend) { + case WUFFS_BASE__PIXEL_BLEND__SRC: + return wuffs_base__pixel_swizzler__squash_tight_4x8_4x16le; + case WUFFS_BASE__PIXEL_BLEND__SRC_OVER: + return wuffs_base__pixel_swizzler__bgra_nonpremul__bgra_nonpremul_4x16le__src_over; + } + return NULL; + + case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL: + switch (blend) { + case WUFFS_BASE__PIXEL_BLEND__SRC: + return wuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul_4x16le__src; + case WUFFS_BASE__PIXEL_BLEND__SRC_OVER: + return wuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul_4x16le__src_over; + } + return NULL; + + case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY: + case WUFFS_BASE__PIXEL_FORMAT__BGRX: + // TODO. + break; + + case WUFFS_BASE__PIXEL_FORMAT__RGB: + case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL: + case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL: + case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY: + case WUFFS_BASE__PIXEL_FORMAT__RGBX: + // TODO. + break; + } + return NULL; +} + // -------- WUFFS_BASE__MAYBE_STATIC wuffs_base__status // @@ -1480,6 +1934,11 @@ wuffs_base__pixel_swizzler__prepare(wuffs_base__pixel_swizzler* p, func = wuffs_base__pixel_swizzler__prepare__bgra_nonpremul( p, dst_pixfmt, dst_palette, src_palette, blend); break; + + case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL_4X16LE: + func = wuffs_base__pixel_swizzler__prepare__bgra_nonpremul_4x16le( + p, dst_pixfmt, dst_palette, src_palette, blend); + break; } p->private_impl.func = func; diff --git a/internal/cgen/data/data.go b/internal/cgen/data/data.go index d4a5cbb8..a1138935 100644 --- a/internal/cgen/data/data.go +++ b/internal/cgen/data/data.go @@ -118,7 +118,9 @@ const BaseImagePublicH = "" + "// ---------------- Images\n\n// wuffs_base__color_u32_argb_premul is an 8 bit per channel premultiplied\n// Alpha, Red, Green, Blue color, as a uint32_t value. Its value is always\n// 0xAARRGGBB (Alpha most significant, Blue least), regardless of endianness.\ntypedef uint32_t wuffs_base__color_u32_argb_premul;\n\nstatic inline uint16_t //\nwuffs_base__color_u32_argb_premul__as__color_u16_rgb_565(\n wuffs_base__color_u32_argb_premul c) {\n uint32_t r5 = 0xF800 & (c >> 8);\n uint32_t g6 = 0x07E0 & (c >> 5);\n uint32_t b5 = 0x001F & (c >> 3);\n return (uint16_t)(r5 | g6 | b5);\n}\n\nstatic inline wuffs_base__color_u32_argb_premul //\nwuffs_base__color_u16_rgb_565__as__color_u32_argb_premul(uint16_t rgb_565) {\n uint32_t b5 = 0x1F & (rgb_565 >> 0);\n uint32_t b = (b5 << 3) | (b5 >> 2);\n uint32_t g6 = 0x3F & (rgb_565 >> 5);\n uint32_t g = (g6 << 2) | (g6 >> 4);\n uint32_t r5 = 0x1F & (rgb_565 >> 11);\n uint32_t r = (r5 << 3) | (r5 >> 2);\n return 0xFF000000 | (r << 16) | (g << 8) | (b << 0);\n}\n\nstatic inline uint8_t //" + "\nwuffs_base__color_u32_argb_premul__as__color_u8_gray(\n wuffs_base__color_u32_argb_premul c) {\n // Work in 16-bit color.\n uint32_t cr = 0x101 * (0xFF & (c >> 16));\n uint32_t cg = 0x101 * (0xFF & (c >> 8));\n uint32_t cb = 0x101 * (0xFF & (c >> 0));\n\n // These coefficients (the fractions 0.299, 0.587 and 0.114) are the same\n // as those given by the JFIF specification.\n //\n // Note that 19595 + 38470 + 7471 equals 65536, also known as (1 << 16). We\n // shift by 24, not just by 16, because the return value is 8-bit color, not\n // 16-bit color.\n uint32_t weighted_average = (19595 * cr) + (38470 * cg) + (7471 * cb) + 32768;\n return (uint8_t)(weighted_average >> 24);\n}\n\n// wuffs_base__color_u32_argb_nonpremul__as__color_u32_argb_premul converts\n// from non-premultiplied alpha to premultiplied alpha.\nstatic inline wuffs_base__color_u32_argb_premul //\nwuffs_base__color_u32_argb_nonpremul__as__color_u32_argb_premul(\n uint32_t argb_nonpremul) {\n // Multiplying by 0x101 (twice, once for alpha and onc" + "e for color) converts\n // from 8-bit to 16-bit color. Shifting right by 8 undoes that.\n //\n // Working in the higher bit depth can produce slightly different (and\n // arguably slightly more accurate) results. For example, given 8-bit blue\n // and alpha of 0x80 and 0x81:\n //\n // - ((0x80 * 0x81 ) / 0xFF ) = 0x40 = 0x40\n // - ((0x8080 * 0x8181) / 0xFFFF) >> 8 = 0x4101 >> 8 = 0x41\n uint32_t a = 0xFF & (argb_nonpremul >> 24);\n uint32_t a16 = a * (0x101 * 0x101);\n\n uint32_t r = 0xFF & (argb_nonpremul >> 16);\n r = ((r * a16) / 0xFFFF) >> 8;\n uint32_t g = 0xFF & (argb_nonpremul >> 8);\n g = ((g * a16) / 0xFFFF) >> 8;\n uint32_t b = 0xFF & (argb_nonpremul >> 0);\n b = ((b * a16) / 0xFFFF) >> 8;\n\n return (a << 24) | (r << 16) | (g << 8) | (b << 0);\n}\n\n// wuffs_base__color_u32_argb_premul__as__color_u32_argb_nonpremul converts\n// from premultiplied alpha to non-premultiplied alpha.\nstatic inline uint32_t //\nwuffs_base__color_u32_argb_premul__as__color_u32_argb_nonpremul(\n wuffs_base_" + - "_color_u32_argb_premul c) {\n uint32_t a = 0xFF & (c >> 24);\n if (a == 0xFF) {\n return c;\n } else if (a == 0) {\n return 0;\n }\n uint32_t a16 = a * 0x101;\n\n uint32_t r = 0xFF & (c >> 16);\n r = ((r * (0x101 * 0xFFFF)) / a16) >> 8;\n uint32_t g = 0xFF & (c >> 8);\n g = ((g * (0x101 * 0xFFFF)) / a16) >> 8;\n uint32_t b = 0xFF & (c >> 0);\n b = ((b * (0x101 * 0xFFFF)) / a16) >> 8;\n\n return (a << 24) | (r << 16) | (g << 8) | (b << 0);\n}\n\n" + + "_color_u32_argb_premul c) {\n uint32_t a = 0xFF & (c >> 24);\n if (a == 0xFF) {\n return c;\n } else if (a == 0) {\n return 0;\n }\n uint32_t a16 = a * 0x101;\n\n uint32_t r = 0xFF & (c >> 16);\n r = ((r * (0x101 * 0xFFFF)) / a16) >> 8;\n uint32_t g = 0xFF & (c >> 8);\n g = ((g * (0x101 * 0xFFFF)) / a16) >> 8;\n uint32_t b = 0xFF & (c >> 0);\n b = ((b * (0x101 * 0xFFFF)) / a16) >> 8;\n\n return (a << 24) | (r << 16) | (g << 8) | (b << 0);\n}\n\n// wuffs_base__color_u64_argb_nonpremul__as__color_u32_argb_premul converts\n// from 4x16LE non-premultiplied alpha to 4x8 premultiplied alpha.\nstatic inline wuffs_base__color_u32_argb_premul //\nwuffs_base__color_u64_argb_nonpremul__as__color_u32_argb_premul(\n uint64_t argb_nonpremul) {\n uint32_t a16 = 0xFFFF & (argb_nonpremul >> 48);\n\n uint32_t r16 = 0xFFFF & (argb_nonpremul >> 32);\n r16 = (r16 * a16) / 0xFFFF;\n uint32_t g16 = 0xFFFF & (argb_nonpremul >> 16);\n g16 = (g16 * a16) / 0xFFFF;\n uint32_t b16 = 0xFFFF & (argb_nonpremul >> 0);\n b16 = (b16 * a16) / 0xF" + + "FFF;\n\n return ((a16 >> 8) << 24) | ((r16 >> 8) << 16) | ((g16 >> 8) << 8) |\n ((b16 >> 8) << 0);\n}\n\n// wuffs_base__color_u32_argb_premul__as__color_u64_argb_nonpremul converts\n// from 4x8 premultiplied alpha to 4x16LE non-premultiplied alpha.\nstatic inline uint64_t //\nwuffs_base__color_u32_argb_premul__as__color_u64_argb_nonpremul(\n wuffs_base__color_u32_argb_premul c) {\n uint32_t a = 0xFF & (c >> 24);\n if (a == 0xFF) {\n uint64_t r16 = 0x101 * (0xFF & (c >> 16));\n uint64_t g16 = 0x101 * (0xFF & (c >> 8));\n uint64_t b16 = 0x101 * (0xFF & (c >> 0));\n return 0xFFFF000000000000u | (r16 << 32) | (g16 << 16) | (b16 << 0);\n } else if (a == 0) {\n return 0;\n }\n uint64_t a16 = a * 0x101;\n\n uint64_t r = 0xFF & (c >> 16);\n uint64_t r16 = (r * (0x101 * 0xFFFF)) / a16;\n uint64_t g = 0xFF & (c >> 8);\n uint64_t g16 = (g * (0x101 * 0xFFFF)) / a16;\n uint64_t b = 0xFF & (c >> 0);\n uint64_t b16 = (b * (0x101 * 0xFFFF)) / a16;\n\n return (a16 << 48) | (r16 << 32) | (g16 << 16) | (b16 << 0);\n}\n" + + "\nstatic inline uint64_t //\nwuffs_base__color_u32__as__color_u64(uint32_t c) {\n uint64_t a16 = 0x101 * (0xFF & (c >> 24));\n uint64_t r16 = 0x101 * (0xFF & (c >> 16));\n uint64_t g16 = 0x101 * (0xFF & (c >> 8));\n uint64_t b16 = 0x101 * (0xFF & (c >> 0));\n return (a16 << 48) | (r16 << 32) | (g16 << 16) | (b16 << 0);\n}\n\nstatic inline uint32_t //\nwuffs_base__color_u64__as__color_u32(uint64_t c) {\n uint32_t a = ((uint32_t)(0xFF & (c >> 56)));\n uint32_t r = ((uint32_t)(0xFF & (c >> 40)));\n uint32_t g = ((uint32_t)(0xFF & (c >> 24)));\n uint32_t b = ((uint32_t)(0xFF & (c >> 8)));\n return (a << 24) | (r << 16) | (g << 8) | (b << 0);\n}\n\n" + "" + "// --------\n\ntypedef uint8_t wuffs_base__pixel_blend;\n\n// wuffs_base__pixel_blend encodes how to blend source and destination pixels,\n// accounting for transparency. It encompasses the Porter-Duff compositing\n// operators as well as the other blending modes defined by PDF.\n//\n// TODO: implement the other modes.\n#define WUFFS_BASE__PIXEL_BLEND__SRC ((wuffs_base__pixel_blend)0)\n#define WUFFS_BASE__PIXEL_BLEND__SRC_OVER ((wuffs_base__pixel_blend)1)\n\n" + "" + @@ -126,10 +128,10 @@ const BaseImagePublicH = "" + "" + "// --------\n\n#define WUFFS_BASE__PIXEL_FORMAT__NUM_PLANES_MAX 4\n\n#define WUFFS_BASE__PIXEL_FORMAT__INDEXED__INDEX_PLANE 0\n#define WUFFS_BASE__PIXEL_FORMAT__INDEXED__COLOR_PLANE 3\n\n// wuffs_base__pixel_format encodes the format of the bytes that constitute an\n// image frame's pixel data.\n//\n// See https://github.com/google/wuffs/blob/master/doc/note/pixel-formats.md\n//\n// Do not manipulate its bits directly; they are private implementation\n// details. Use methods such as wuffs_base__pixel_format__num_planes instead.\ntypedef struct wuffs_base__pixel_format__struct {\n uint32_t repr;\n\n#ifdef __cplusplus\n inline bool is_valid() const;\n inline uint32_t bits_per_pixel() const;\n inline bool is_direct() const;\n inline bool is_indexed() const;\n inline bool is_interleaved() const;\n inline bool is_planar() const;\n inline uint32_t num_planes() const;\n inline wuffs_base__pixel_alpha_transparency transparency() const;\n#endif // __cplusplus\n\n} wuffs_base__pixel_format;\n\nstatic inline wuffs_base__pixel_format //\nwu" + "ffs_base__make_pixel_format(uint32_t repr) {\n wuffs_base__pixel_format f;\n f.repr = repr;\n return f;\n}\n\n// Common 8-bit-depth pixel formats. This list is not exhaustive; not all valid\n// wuffs_base__pixel_format values are present.\n\n#define WUFFS_BASE__PIXEL_FORMAT__INVALID 0x00000000\n\n#define WUFFS_BASE__PIXEL_FORMAT__A 0x02000008\n\n#define WUFFS_BASE__PIXEL_FORMAT__Y 0x20000008\n#define WUFFS_BASE__PIXEL_FORMAT__YA_NONPREMUL 0x21000008\n#define WUFFS_BASE__PIXEL_FORMAT__YA_PREMUL 0x22000008\n\n#define WUFFS_BASE__PIXEL_FORMAT__YCBCR 0x40020888\n#define WUFFS_BASE__PIXEL_FORMAT__YCBCRA_NONPREMUL 0x41038888\n#define WUFFS_BASE__PIXEL_FORMAT__YCBCRK 0x50038888\n\n#define WUFFS_BASE__PIXEL_FORMAT__YCOCG 0x60020888\n#define WUFFS_BASE__PIXEL_FORMAT__YCOCGA_NONPREMUL 0x61038888\n#define WUFFS_BASE__PIXEL_FORMAT__YCOCGK 0x70038888\n\n#define WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_NONPREMUL 0x81040008\n#define WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_PREMUL 0x82040008\n#define WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_BINARY 0x8" + - "3040008\n\n#define WUFFS_BASE__PIXEL_FORMAT__BGR_565 0x80000565\n#define WUFFS_BASE__PIXEL_FORMAT__BGR 0x80000888\n#define WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL 0x81008888\n#define WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL 0x82008888\n#define WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY 0x83008888\n#define WUFFS_BASE__PIXEL_FORMAT__BGRX 0x90008888\n\n#define WUFFS_BASE__PIXEL_FORMAT__RGB 0xA0000888\n#define WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL 0xA1008888\n#define WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL 0xA2008888\n#define WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY 0xA3008888\n#define WUFFS_BASE__PIXEL_FORMAT__RGBX 0xB0008888\n\n#define WUFFS_BASE__PIXEL_FORMAT__CMY 0xC0020888\n#define WUFFS_BASE__PIXEL_FORMAT__CMYK 0xD0038888\n\nextern const uint32_t wuffs_base__pixel_format__bits_per_channel[16];\n\nstatic inline bool //\nwuffs_base__pixel_format__is_valid(const wuffs_base__pixel_format* f) {\n return f->repr != 0;\n}\n\n// wuffs_base__pixel_format__bits_per_pixel returns the number of bits per\n// pixel for interleaved pixel formats, and ret" + - "urns 0 for planar pixel formats.\nstatic inline uint32_t //\nwuffs_base__pixel_format__bits_per_pixel(const wuffs_base__pixel_format* f) {\n if (((f->repr >> 16) & 0x03) != 0) {\n return 0;\n }\n return wuffs_base__pixel_format__bits_per_channel[0x0F & (f->repr >> 0)] +\n wuffs_base__pixel_format__bits_per_channel[0x0F & (f->repr >> 4)] +\n wuffs_base__pixel_format__bits_per_channel[0x0F & (f->repr >> 8)] +\n wuffs_base__pixel_format__bits_per_channel[0x0F & (f->repr >> 12)];\n}\n\nstatic inline bool //\nwuffs_base__pixel_format__is_direct(const wuffs_base__pixel_format* f) {\n return ((f->repr >> 18) & 0x01) == 0;\n}\n\nstatic inline bool //\nwuffs_base__pixel_format__is_indexed(const wuffs_base__pixel_format* f) {\n return ((f->repr >> 18) & 0x01) != 0;\n}\n\nstatic inline bool //\nwuffs_base__pixel_format__is_interleaved(const wuffs_base__pixel_format* f) {\n return ((f->repr >> 16) & 0x03) == 0;\n}\n\nstatic inline bool //\nwuffs_base__pixel_format__is_planar(const wuffs_base__pixel_format* f) {" + - "\n return ((f->repr >> 16) & 0x03) != 0;\n}\n\nstatic inline uint32_t //\nwuffs_base__pixel_format__num_planes(const wuffs_base__pixel_format* f) {\n return ((f->repr >> 16) & 0x03) + 1;\n}\n\nstatic inline wuffs_base__pixel_alpha_transparency //\nwuffs_base__pixel_format__transparency(const wuffs_base__pixel_format* f) {\n return (wuffs_base__pixel_alpha_transparency)((f->repr >> 24) & 0x03);\n}\n\n#ifdef __cplusplus\n\ninline bool //\nwuffs_base__pixel_format::is_valid() const {\n return wuffs_base__pixel_format__is_valid(this);\n}\n\ninline uint32_t //\nwuffs_base__pixel_format::bits_per_pixel() const {\n return wuffs_base__pixel_format__bits_per_pixel(this);\n}\n\ninline bool //\nwuffs_base__pixel_format::is_direct() const {\n return wuffs_base__pixel_format__is_direct(this);\n}\n\ninline bool //\nwuffs_base__pixel_format::is_indexed() const {\n return wuffs_base__pixel_format__is_indexed(this);\n}\n\ninline bool //\nwuffs_base__pixel_format::is_interleaved() const {\n return wuffs_base__pixel_format__is_interleaved(this);\n}\n\ni" + - "nline bool //\nwuffs_base__pixel_format::is_planar() const {\n return wuffs_base__pixel_format__is_planar(this);\n}\n\ninline uint32_t //\nwuffs_base__pixel_format::num_planes() const {\n return wuffs_base__pixel_format__num_planes(this);\n}\n\ninline wuffs_base__pixel_alpha_transparency //\nwuffs_base__pixel_format::transparency() const {\n return wuffs_base__pixel_format__transparency(this);\n}\n\n#endif // __cplusplus\n\n" + + "3040008\n\n#define WUFFS_BASE__PIXEL_FORMAT__BGR_565 0x80000565\n#define WUFFS_BASE__PIXEL_FORMAT__BGR 0x80000888\n#define WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL 0x81008888\n#define WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL_4X16LE 0x8100BBBB\n#define WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL 0x82008888\n#define WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL_4X16LE 0x8200BBBB\n#define WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY 0x83008888\n#define WUFFS_BASE__PIXEL_FORMAT__BGRX 0x90008888\n\n#define WUFFS_BASE__PIXEL_FORMAT__RGB 0xA0000888\n#define WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL 0xA1008888\n#define WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL_4X16LE 0xA100BBBB\n#define WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL 0xA2008888\n#define WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL_4X16LE 0xA200BBBB\n#define WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY 0xA3008888\n#define WUFFS_BASE__PIXEL_FORMAT__RGBX 0xB0008888\n\n#define WUFFS_BASE__PIXEL_FORMAT__CMY 0xC0020888\n#define WUFFS_BASE__PIXEL_FORMAT__CMYK 0xD0038888\n\nextern const uint32_t wuffs_base__pixel_format__bits" + + "_per_channel[16];\n\nstatic inline bool //\nwuffs_base__pixel_format__is_valid(const wuffs_base__pixel_format* f) {\n return f->repr != 0;\n}\n\n// wuffs_base__pixel_format__bits_per_pixel returns the number of bits per\n// pixel for interleaved pixel formats, and returns 0 for planar pixel formats.\nstatic inline uint32_t //\nwuffs_base__pixel_format__bits_per_pixel(const wuffs_base__pixel_format* f) {\n if (((f->repr >> 16) & 0x03) != 0) {\n return 0;\n }\n return wuffs_base__pixel_format__bits_per_channel[0x0F & (f->repr >> 0)] +\n wuffs_base__pixel_format__bits_per_channel[0x0F & (f->repr >> 4)] +\n wuffs_base__pixel_format__bits_per_channel[0x0F & (f->repr >> 8)] +\n wuffs_base__pixel_format__bits_per_channel[0x0F & (f->repr >> 12)];\n}\n\nstatic inline bool //\nwuffs_base__pixel_format__is_direct(const wuffs_base__pixel_format* f) {\n return ((f->repr >> 18) & 0x01) == 0;\n}\n\nstatic inline bool //\nwuffs_base__pixel_format__is_indexed(const wuffs_base__pixel_format* f) {\n return ((f->repr >" + + "> 18) & 0x01) != 0;\n}\n\nstatic inline bool //\nwuffs_base__pixel_format__is_interleaved(const wuffs_base__pixel_format* f) {\n return ((f->repr >> 16) & 0x03) == 0;\n}\n\nstatic inline bool //\nwuffs_base__pixel_format__is_planar(const wuffs_base__pixel_format* f) {\n return ((f->repr >> 16) & 0x03) != 0;\n}\n\nstatic inline uint32_t //\nwuffs_base__pixel_format__num_planes(const wuffs_base__pixel_format* f) {\n return ((f->repr >> 16) & 0x03) + 1;\n}\n\nstatic inline wuffs_base__pixel_alpha_transparency //\nwuffs_base__pixel_format__transparency(const wuffs_base__pixel_format* f) {\n return (wuffs_base__pixel_alpha_transparency)((f->repr >> 24) & 0x03);\n}\n\n#ifdef __cplusplus\n\ninline bool //\nwuffs_base__pixel_format::is_valid() const {\n return wuffs_base__pixel_format__is_valid(this);\n}\n\ninline uint32_t //\nwuffs_base__pixel_format::bits_per_pixel() const {\n return wuffs_base__pixel_format__bits_per_pixel(this);\n}\n\ninline bool //\nwuffs_base__pixel_format::is_direct() const {\n return wuffs_base__pixel_format__is_di" + + "rect(this);\n}\n\ninline bool //\nwuffs_base__pixel_format::is_indexed() const {\n return wuffs_base__pixel_format__is_indexed(this);\n}\n\ninline bool //\nwuffs_base__pixel_format::is_interleaved() const {\n return wuffs_base__pixel_format__is_interleaved(this);\n}\n\ninline bool //\nwuffs_base__pixel_format::is_planar() const {\n return wuffs_base__pixel_format__is_planar(this);\n}\n\ninline uint32_t //\nwuffs_base__pixel_format::num_planes() const {\n return wuffs_base__pixel_format__num_planes(this);\n}\n\ninline wuffs_base__pixel_alpha_transparency //\nwuffs_base__pixel_format::transparency() const {\n return wuffs_base__pixel_format__transparency(this);\n}\n\n#endif // __cplusplus\n\n" + "" + "// --------\n\n// wuffs_base__pixel_subsampling encodes whether sample values cover one pixel\n// or cover multiple pixels.\n//\n// See https://github.com/google/wuffs/blob/master/doc/note/pixel-subsampling.md\n//\n// Do not manipulate its bits directly; they are private implementation\n// details. Use methods such as wuffs_base__pixel_subsampling__bias_x instead.\ntypedef struct wuffs_base__pixel_subsampling__struct {\n uint32_t repr;\n\n#ifdef __cplusplus\n inline uint32_t bias_x(uint32_t plane) const;\n inline uint32_t denominator_x(uint32_t plane) const;\n inline uint32_t bias_y(uint32_t plane) const;\n inline uint32_t denominator_y(uint32_t plane) const;\n#endif // __cplusplus\n\n} wuffs_base__pixel_subsampling;\n\nstatic inline wuffs_base__pixel_subsampling //\nwuffs_base__make_pixel_subsampling(uint32_t repr) {\n wuffs_base__pixel_subsampling s;\n s.repr = repr;\n return s;\n}\n\n#define WUFFS_BASE__PIXEL_SUBSAMPLING__NONE 0x00000000\n\n#define WUFFS_BASE__PIXEL_SUBSAMPLING__444 0x000000\n#define WUFFS_BASE__PIXEL_SUBSAMPL" + "ING__440 0x010100\n#define WUFFS_BASE__PIXEL_SUBSAMPLING__422 0x101000\n#define WUFFS_BASE__PIXEL_SUBSAMPLING__420 0x111100\n#define WUFFS_BASE__PIXEL_SUBSAMPLING__411 0x303000\n#define WUFFS_BASE__PIXEL_SUBSAMPLING__410 0x313100\n\nstatic inline uint32_t //\nwuffs_base__pixel_subsampling__bias_x(const wuffs_base__pixel_subsampling* s,\n uint32_t plane) {\n uint32_t shift = ((plane & 0x03) * 8) + 6;\n return (s->repr >> shift) & 0x03;\n}\n\nstatic inline uint32_t //\nwuffs_base__pixel_subsampling__denominator_x(\n const wuffs_base__pixel_subsampling* s,\n uint32_t plane) {\n uint32_t shift = ((plane & 0x03) * 8) + 4;\n return ((s->repr >> shift) & 0x03) + 1;\n}\n\nstatic inline uint32_t //\nwuffs_base__pixel_subsampling__bias_y(const wuffs_base__pixel_subsampling* s,\n uint32_t plane) {\n uint32_t shift = ((plane & 0x03) * 8) + 2;\n return (s->repr >> shift) & 0x03;\n}\n\nstatic inline uint32_t //\nwuffs_base__pixel_subsampling__denominator_y(\n c" + @@ -517,43 +519,57 @@ const BasePixConvSubmoduleC = "" + "" + "// --------\n\nWUFFS_BASE__MAYBE_STATIC wuffs_base__color_u32_argb_premul //\nwuffs_base__pixel_buffer__color_u32_at(const wuffs_base__pixel_buffer* pb,\n uint32_t x,\n uint32_t y) {\n if (!pb || (x >= pb->pixcfg.private_impl.width) ||\n (y >= pb->pixcfg.private_impl.height)) {\n return 0;\n }\n\n if (wuffs_base__pixel_format__is_planar(&pb->pixcfg.private_impl.pixfmt)) {\n // TODO: support planar formats.\n return 0;\n }\n\n size_t stride = pb->private_impl.planes[0].stride;\n const uint8_t* row = pb->private_impl.planes[0].ptr + (stride * ((size_t)y));\n\n switch (pb->pixcfg.private_impl.pixfmt.repr) {\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY:\n return wuffs_base__load_u32le__no_bounds_check(row + (4 * ((size_t)x)));\n\n case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_BINARY: {\n uint8_t* palette = pb->private_impl" + ".planes[3].ptr;\n return wuffs_base__load_u32le__no_bounds_check(palette +\n (4 * ((size_t)row[x])));\n }\n\n // Common formats above. Rarer formats below.\n\n case WUFFS_BASE__PIXEL_FORMAT__Y:\n return 0xFF000000 | (0x00010101 * ((uint32_t)(row[x])));\n\n case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_NONPREMUL: {\n uint8_t* palette = pb->private_impl.planes[3].ptr;\n return wuffs_base__color_u32_argb_nonpremul__as__color_u32_argb_premul(\n wuffs_base__load_u32le__no_bounds_check(palette +\n (4 * ((size_t)row[x]))));\n }\n\n case WUFFS_BASE__PIXEL_FORMAT__BGR_565:\n return wuffs_base__color_u16_rgb_565__as__color_u32_argb_premul(\n wuffs_base__load_u16le__no_bounds_check(row + (2 * ((size_t)x))));\n case WUFFS_BASE__PIXEL_FORMAT__BGR:\n return 0xFF000000 |\n wuffs_base__load_u24le__no_bounds_check(row + (3 * ((size_t)x)));\n case WUFFS_BASE__PIXEL_F" + - "ORMAT__BGRA_NONPREMUL:\n return wuffs_base__color_u32_argb_nonpremul__as__color_u32_argb_premul(\n wuffs_base__load_u32le__no_bounds_check(row + (4 * ((size_t)x))));\n case WUFFS_BASE__PIXEL_FORMAT__BGRX:\n return 0xFF000000 |\n wuffs_base__load_u32le__no_bounds_check(row + (4 * ((size_t)x)));\n\n case WUFFS_BASE__PIXEL_FORMAT__RGB:\n return wuffs_base__swap_u32_argb_abgr(\n 0xFF000000 |\n wuffs_base__load_u24le__no_bounds_check(row + (3 * ((size_t)x))));\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL:\n return wuffs_base__swap_u32_argb_abgr(\n wuffs_base__color_u32_argb_nonpremul__as__color_u32_argb_premul(\n wuffs_base__load_u32le__no_bounds_check(row +\n (4 * ((size_t)x)))));\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY:\n return wuffs_base__swap_u32_argb_abgr(\n wuffs_base__load_u32le__no_bounds_check(row + (4 * ((siz" + - "e_t)x))));\n case WUFFS_BASE__PIXEL_FORMAT__RGBX:\n return wuffs_base__swap_u32_argb_abgr(\n 0xFF000000 |\n wuffs_base__load_u32le__no_bounds_check(row + (4 * ((size_t)x))));\n\n default:\n // TODO: support more formats.\n break;\n }\n\n return 0;\n}\n\nWUFFS_BASE__MAYBE_STATIC wuffs_base__status //\nwuffs_base__pixel_buffer__set_color_u32_at(\n wuffs_base__pixel_buffer* pb,\n uint32_t x,\n uint32_t y,\n wuffs_base__color_u32_argb_premul color) {\n if (!pb) {\n return wuffs_base__make_status(wuffs_base__error__bad_receiver);\n }\n if ((x >= pb->pixcfg.private_impl.width) ||\n (y >= pb->pixcfg.private_impl.height)) {\n return wuffs_base__make_status(wuffs_base__error__bad_argument);\n }\n\n if (wuffs_base__pixel_format__is_planar(&pb->pixcfg.private_impl.pixfmt)) {\n // TODO: support planar formats.\n return wuffs_base__make_status(wuffs_base__error__unsupported_option);\n }\n\n size_t stride = pb->private_impl.planes[0].stride;\n uint8_t* row = pb->private_impl.pla" + - "nes[0].ptr + (stride * ((size_t)y));\n\n switch (pb->pixcfg.private_impl.pixfmt.repr) {\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__BGRX:\n wuffs_base__store_u32le__no_bounds_check(row + (4 * ((size_t)x)), color);\n break;\n\n // Common formats above. Rarer formats below.\n\n case WUFFS_BASE__PIXEL_FORMAT__Y:\n wuffs_base__store_u8__no_bounds_check(\n row + ((size_t)x),\n wuffs_base__color_u32_argb_premul__as__color_u8_gray(color));\n break;\n\n case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_BINARY:\n wuffs_base__store_u8__no_bounds_check(\n row + ((size_t)x), wuffs_base__pixel_palette__closest_element(\n wuffs_base__pixel_buffer__palette(pb),\n pb->pixcfg.private_impl.pixfmt, color));\n break;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGR_565:\n wuffs_base__store_u16le__no_bounds_check(\n row + (2 * ((size_t)x)),\n wuffs_base__color_u32_argb_prem" + - "ul__as__color_u16_rgb_565(color));\n break;\n case WUFFS_BASE__PIXEL_FORMAT__BGR:\n wuffs_base__store_u24le__no_bounds_check(row + (3 * ((size_t)x)), color);\n break;\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL:\n wuffs_base__store_u32le__no_bounds_check(\n row + (4 * ((size_t)x)),\n wuffs_base__color_u32_argb_premul__as__color_u32_argb_nonpremul(\n color));\n break;\n\n case WUFFS_BASE__PIXEL_FORMAT__RGB:\n wuffs_base__store_u24le__no_bounds_check(\n row + (3 * ((size_t)x)), wuffs_base__swap_u32_argb_abgr(color));\n break;\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL:\n wuffs_base__store_u32le__no_bounds_check(\n row + (4 * ((size_t)x)),\n wuffs_base__color_u32_argb_premul__as__color_u32_argb_nonpremul(\n wuffs_base__swap_u32_argb_abgr(color)));\n break;\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBX:\n wuffs_base__store_u32le__no_bounds_check(\n " + - " row + (4 * ((size_t)x)), wuffs_base__swap_u32_argb_abgr(color));\n break;\n\n default:\n // TODO: support more formats.\n return wuffs_base__make_status(wuffs_base__error__unsupported_option);\n }\n\n return wuffs_base__make_status(NULL);\n}\n\n" + + "ORMAT__BGRA_NONPREMUL:\n return wuffs_base__color_u32_argb_nonpremul__as__color_u32_argb_premul(\n wuffs_base__load_u32le__no_bounds_check(row + (4 * ((size_t)x))));\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL_4X16LE:\n return wuffs_base__color_u64_argb_nonpremul__as__color_u32_argb_premul(\n wuffs_base__load_u64le__no_bounds_check(row + (8 * ((size_t)x))));\n case WUFFS_BASE__PIXEL_FORMAT__BGRX:\n return 0xFF000000 |\n wuffs_base__load_u32le__no_bounds_check(row + (4 * ((size_t)x)));\n\n case WUFFS_BASE__PIXEL_FORMAT__RGB:\n return wuffs_base__swap_u32_argb_abgr(\n 0xFF000000 |\n wuffs_base__load_u24le__no_bounds_check(row + (3 * ((size_t)x))));\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL:\n return wuffs_base__swap_u32_argb_abgr(\n wuffs_base__color_u32_argb_nonpremul__as__color_u32_argb_premul(\n wuffs_base__load_u32le__no_bounds_check(row +\n (4 * ((size_t)x)" + + "))));\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY:\n return wuffs_base__swap_u32_argb_abgr(\n wuffs_base__load_u32le__no_bounds_check(row + (4 * ((size_t)x))));\n case WUFFS_BASE__PIXEL_FORMAT__RGBX:\n return wuffs_base__swap_u32_argb_abgr(\n 0xFF000000 |\n wuffs_base__load_u32le__no_bounds_check(row + (4 * ((size_t)x))));\n\n default:\n // TODO: support more formats.\n break;\n }\n\n return 0;\n}\n\nWUFFS_BASE__MAYBE_STATIC wuffs_base__status //\nwuffs_base__pixel_buffer__set_color_u32_at(\n wuffs_base__pixel_buffer* pb,\n uint32_t x,\n uint32_t y,\n wuffs_base__color_u32_argb_premul color) {\n if (!pb) {\n return wuffs_base__make_status(wuffs_base__error__bad_receiver);\n }\n if ((x >= pb->pixcfg.private_impl.width) ||\n (y >= pb->pixcfg.private_impl.height)) {\n return wuffs_base__make_status(wuffs_base__error__bad_argument);\n }\n\n if (wuffs_base__pixel_format__is_planar(&pb->pixcfg.private_impl.pixfm" + + "t)) {\n // TODO: support planar formats.\n return wuffs_base__make_status(wuffs_base__error__unsupported_option);\n }\n\n size_t stride = pb->private_impl.planes[0].stride;\n uint8_t* row = pb->private_impl.planes[0].ptr + (stride * ((size_t)y));\n\n switch (pb->pixcfg.private_impl.pixfmt.repr) {\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__BGRX:\n wuffs_base__store_u32le__no_bounds_check(row + (4 * ((size_t)x)), color);\n break;\n\n // Common formats above. Rarer formats below.\n\n case WUFFS_BASE__PIXEL_FORMAT__Y:\n wuffs_base__store_u8__no_bounds_check(\n row + ((size_t)x),\n wuffs_base__color_u32_argb_premul__as__color_u8_gray(color));\n break;\n\n case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_BINARY:\n wuffs_base__store_u8__no_bounds_check(\n row + ((size_t)x), wuffs_base__pixel_palette__closest_element(\n wuffs_base__pixel_buffer__palette(pb),\n pb->pixcfg" + + ".private_impl.pixfmt, color));\n break;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGR_565:\n wuffs_base__store_u16le__no_bounds_check(\n row + (2 * ((size_t)x)),\n wuffs_base__color_u32_argb_premul__as__color_u16_rgb_565(color));\n break;\n case WUFFS_BASE__PIXEL_FORMAT__BGR:\n wuffs_base__store_u24le__no_bounds_check(row + (3 * ((size_t)x)), color);\n break;\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL:\n wuffs_base__store_u32le__no_bounds_check(\n row + (4 * ((size_t)x)),\n wuffs_base__color_u32_argb_premul__as__color_u32_argb_nonpremul(\n color));\n break;\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL_4X16LE:\n wuffs_base__store_u64le__no_bounds_check(\n row + (8 * ((size_t)x)),\n wuffs_base__color_u32_argb_premul__as__color_u64_argb_nonpremul(\n color));\n break;\n\n case WUFFS_BASE__PIXEL_FORMAT__RGB:\n wuffs_base__store_u24le__no_bounds_check(\n row + (3 * ((size_t)x)), wuffs_bas" + + "e__swap_u32_argb_abgr(color));\n break;\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL:\n wuffs_base__store_u32le__no_bounds_check(\n row + (4 * ((size_t)x)),\n wuffs_base__color_u32_argb_premul__as__color_u32_argb_nonpremul(\n wuffs_base__swap_u32_argb_abgr(color)));\n break;\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBX:\n wuffs_base__store_u32le__no_bounds_check(\n row + (4 * ((size_t)x)), wuffs_base__swap_u32_argb_abgr(color));\n break;\n\n default:\n // TODO: support more formats.\n return wuffs_base__make_status(wuffs_base__error__unsupported_option);\n }\n\n return wuffs_base__make_status(NULL);\n}\n\n" + "" + "// --------\n\nWUFFS_BASE__MAYBE_STATIC uint8_t //\nwuffs_base__pixel_palette__closest_element(\n wuffs_base__slice_u8 palette_slice,\n wuffs_base__pixel_format palette_format,\n wuffs_base__color_u32_argb_premul c) {\n size_t n = palette_slice.len / 4;\n if (n > 256) {\n n = 256;\n }\n size_t best_index = 0;\n uint64_t best_score = 0xFFFFFFFFFFFFFFFF;\n\n // Work in 16-bit color.\n uint32_t ca = 0x101 * (0xFF & (c >> 24));\n uint32_t cr = 0x101 * (0xFF & (c >> 16));\n uint32_t cg = 0x101 * (0xFF & (c >> 8));\n uint32_t cb = 0x101 * (0xFF & (c >> 0));\n\n switch (palette_format.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 bool nonpremul = palette_format.repr ==\n WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_NONPREMUL;\n\n size_t i;\n for (i = 0; i < n; i++) {\n // Work in 16-bit color.\n uint32_t pb = 0x101 * ((uint32_t)(palette" + "_slice.ptr[(4 * i) + 0]));\n uint32_t pg = 0x101 * ((uint32_t)(palette_slice.ptr[(4 * i) + 1]));\n uint32_t pr = 0x101 * ((uint32_t)(palette_slice.ptr[(4 * i) + 2]));\n uint32_t pa = 0x101 * ((uint32_t)(palette_slice.ptr[(4 * i) + 3]));\n\n // Convert to premultiplied alpha.\n if (nonpremul && (pa != 0xFFFF)) {\n pb = (pb * pa) / 0xFFFF;\n pg = (pg * pa) / 0xFFFF;\n pr = (pr * pa) / 0xFFFF;\n }\n\n // These deltas are conceptually int32_t (signed) but after squaring,\n // it's equivalent to work in uint32_t (unsigned).\n pb -= cb;\n pg -= cg;\n pr -= cr;\n pa -= ca;\n uint64_t score = ((uint64_t)(pb * pb)) + ((uint64_t)(pg * pg)) +\n ((uint64_t)(pr * pr)) + ((uint64_t)(pa * pa));\n if (best_score > score) {\n best_score = score;\n best_index = i;\n }\n }\n break;\n }\n }\n\n return (uint8_t)best_index;\n}\n\n" + "" + - "// --------\n\nstatic inline uint32_t //\nwuffs_base__composite_nonpremul_nonpremul_u32_axxx(uint32_t dst_nonpremul,\n uint32_t src_nonpremul) {\n // Convert from 8-bit color to 16-bit color.\n uint32_t sa = 0x101 * (0xFF & (src_nonpremul >> 24));\n uint32_t sr = 0x101 * (0xFF & (src_nonpremul >> 16));\n uint32_t sg = 0x101 * (0xFF & (src_nonpremul >> 8));\n uint32_t sb = 0x101 * (0xFF & (src_nonpremul >> 0));\n uint32_t da = 0x101 * (0xFF & (dst_nonpremul >> 24));\n uint32_t dr = 0x101 * (0xFF & (dst_nonpremul >> 16));\n uint32_t dg = 0x101 * (0xFF & (dst_nonpremul >> 8));\n uint32_t db = 0x101 * (0xFF & (dst_nonpremul >> 0));\n\n // Convert dst from nonpremul to premul.\n dr = (dr * da) / 0xFFFF;\n dg = (dg * da) / 0xFFFF;\n db = (db * da) / 0xFFFF;\n\n // Calculate the inverse of the src-alpha: how much of the dst to keep.\n uint32_t ia = 0xFFFF - sa;\n\n // Composite src (nonpremul) over dst (premul).\n da = sa + ((da * ia) / 0xFFFF);\n dr = ((sr * sa) + (dr * i" + - "a)) / 0xFFFF;\n dg = ((sg * sa) + (dg * ia)) / 0xFFFF;\n db = ((sb * sa) + (db * ia)) / 0xFFFF;\n\n // Convert dst from premul to nonpremul.\n if (da != 0) {\n dr = (dr * 0xFFFF) / da;\n dg = (dg * 0xFFFF) / da;\n db = (db * 0xFFFF) / da;\n }\n\n // Convert from 16-bit color to 8-bit color and combine the components.\n da >>= 8;\n dr >>= 8;\n dg >>= 8;\n db >>= 8;\n return (db << 0) | (dg << 8) | (dr << 16) | (da << 24);\n}\n\nstatic inline uint32_t //\nwuffs_base__composite_nonpremul_premul_u32_axxx(uint32_t dst_nonpremul,\n uint32_t src_premul) {\n // Convert from 8-bit color to 16-bit color.\n uint32_t sa = 0x101 * (0xFF & (src_premul >> 24));\n uint32_t sr = 0x101 * (0xFF & (src_premul >> 16));\n uint32_t sg = 0x101 * (0xFF & (src_premul >> 8));\n uint32_t sb = 0x101 * (0xFF & (src_premul >> 0));\n uint32_t da = 0x101 * (0xFF & (dst_nonpremul >> 24));\n uint32_t dr = 0x101 * (0xFF & (dst_nonpremul >> 16));\n uint32_t dg = 0x101 * (0xFF & (dst_nonpremul >> 8))" + - ";\n uint32_t db = 0x101 * (0xFF & (dst_nonpremul >> 0));\n\n // Convert dst from nonpremul to premul.\n dr = (dr * da) / 0xFFFF;\n dg = (dg * da) / 0xFFFF;\n db = (db * da) / 0xFFFF;\n\n // Calculate the inverse of the src-alpha: how much of the dst to keep.\n uint32_t ia = 0xFFFF - sa;\n\n // Composite src (premul) over dst (premul).\n da = sa + ((da * ia) / 0xFFFF);\n dr = sr + ((dr * ia) / 0xFFFF);\n dg = sg + ((dg * ia) / 0xFFFF);\n db = sb + ((db * ia) / 0xFFFF);\n\n // Convert dst from premul to nonpremul.\n if (da != 0) {\n dr = (dr * 0xFFFF) / da;\n dg = (dg * 0xFFFF) / da;\n db = (db * 0xFFFF) / da;\n }\n\n // Convert from 16-bit color to 8-bit color and combine the components.\n da >>= 8;\n dr >>= 8;\n dg >>= 8;\n db >>= 8;\n return (db << 0) | (dg << 8) | (dr << 16) | (da << 24);\n}\n\nstatic inline uint32_t //\nwuffs_base__composite_premul_nonpremul_u32_axxx(uint32_t dst_premul,\n uint32_t src_nonpremul) {\n // Convert from 8-bit color to 16-bit color.\n " + - " uint32_t sa = 0x101 * (0xFF & (src_nonpremul >> 24));\n uint32_t sr = 0x101 * (0xFF & (src_nonpremul >> 16));\n uint32_t sg = 0x101 * (0xFF & (src_nonpremul >> 8));\n uint32_t sb = 0x101 * (0xFF & (src_nonpremul >> 0));\n uint32_t da = 0x101 * (0xFF & (dst_premul >> 24));\n uint32_t dr = 0x101 * (0xFF & (dst_premul >> 16));\n uint32_t dg = 0x101 * (0xFF & (dst_premul >> 8));\n uint32_t db = 0x101 * (0xFF & (dst_premul >> 0));\n\n // Calculate the inverse of the src-alpha: how much of the dst to keep.\n uint32_t ia = 0xFFFF - sa;\n\n // Composite src (nonpremul) over dst (premul).\n da = sa + ((da * ia) / 0xFFFF);\n dr = ((sr * sa) + (dr * ia)) / 0xFFFF;\n dg = ((sg * sa) + (dg * ia)) / 0xFFFF;\n db = ((sb * sa) + (db * ia)) / 0xFFFF;\n\n // Convert from 16-bit color to 8-bit color and combine the components.\n da >>= 8;\n dr >>= 8;\n dg >>= 8;\n db >>= 8;\n return (db << 0) | (dg << 8) | (dr << 16) | (da << 24);\n}\n\nstatic inline uint32_t //\nwuffs_base__composite_premul_premul_u32_axxx(uint32_t dst_premul,\n " + - " uint32_t src_premul) {\n // Convert from 8-bit color to 16-bit color.\n uint32_t sa = 0x101 * (0xFF & (src_premul >> 24));\n uint32_t sr = 0x101 * (0xFF & (src_premul >> 16));\n uint32_t sg = 0x101 * (0xFF & (src_premul >> 8));\n uint32_t sb = 0x101 * (0xFF & (src_premul >> 0));\n uint32_t da = 0x101 * (0xFF & (dst_premul >> 24));\n uint32_t dr = 0x101 * (0xFF & (dst_premul >> 16));\n uint32_t dg = 0x101 * (0xFF & (dst_premul >> 8));\n uint32_t db = 0x101 * (0xFF & (dst_premul >> 0));\n\n // Calculate the inverse of the src-alpha: how much of the dst to keep.\n uint32_t ia = 0xFFFF - sa;\n\n // Composite src (premul) over dst (premul).\n da = sa + ((da * ia) / 0xFFFF);\n dr = sr + ((dr * ia) / 0xFFFF);\n dg = sg + ((dg * ia) / 0xFFFF);\n db = sb + ((db * ia) / 0xFFFF);\n\n // Convert from 16-bit color to 8-bit color and combine the components.\n da >>= 8;\n dr >>= 8;\n dg >>= 8;\n db >>= 8;\n return (db << 0) | (dg << 8) | (dr << 16) | (da << 24);\n}\n\n" + + "// --------\n\nstatic inline uint32_t //\nwuffs_base__composite_nonpremul_nonpremul_u32_axxx(uint32_t dst_nonpremul,\n uint32_t src_nonpremul) {\n // Extract 16-bit color components.\n uint32_t sa = 0x101 * (0xFF & (src_nonpremul >> 24));\n uint32_t sr = 0x101 * (0xFF & (src_nonpremul >> 16));\n uint32_t sg = 0x101 * (0xFF & (src_nonpremul >> 8));\n uint32_t sb = 0x101 * (0xFF & (src_nonpremul >> 0));\n uint32_t da = 0x101 * (0xFF & (dst_nonpremul >> 24));\n uint32_t dr = 0x101 * (0xFF & (dst_nonpremul >> 16));\n uint32_t dg = 0x101 * (0xFF & (dst_nonpremul >> 8));\n uint32_t db = 0x101 * (0xFF & (dst_nonpremul >> 0));\n\n // Convert dst from nonpremul to premul.\n dr = (dr * da) / 0xFFFF;\n dg = (dg * da) / 0xFFFF;\n db = (db * da) / 0xFFFF;\n\n // Calculate the inverse of the src-alpha: how much of the dst to keep.\n uint32_t ia = 0xFFFF - sa;\n\n // Composite src (nonpremul) over dst (premul).\n da = sa + ((da * ia) / 0xFFFF);\n dr = ((sr * sa) + (dr * ia)) / 0xF" + + "FFF;\n dg = ((sg * sa) + (dg * ia)) / 0xFFFF;\n db = ((sb * sa) + (db * ia)) / 0xFFFF;\n\n // Convert dst from premul to nonpremul.\n if (da != 0) {\n dr = (dr * 0xFFFF) / da;\n dg = (dg * 0xFFFF) / da;\n db = (db * 0xFFFF) / da;\n }\n\n // Convert from 16-bit color to 8-bit color.\n da >>= 8;\n dr >>= 8;\n dg >>= 8;\n db >>= 8;\n\n // Combine components.\n return (db << 0) | (dg << 8) | (dr << 16) | (da << 24);\n}\n\nstatic inline uint64_t //\nwuffs_base__composite_nonpremul_nonpremul_u64_axxx(uint64_t dst_nonpremul,\n uint64_t src_nonpremul) {\n // Extract components.\n uint64_t sa = 0xFFFF & (src_nonpremul >> 48);\n uint64_t sr = 0xFFFF & (src_nonpremul >> 32);\n uint64_t sg = 0xFFFF & (src_nonpremul >> 16);\n uint64_t sb = 0xFFFF & (src_nonpremul >> 0);\n uint64_t da = 0xFFFF & (dst_nonpremul >> 48);\n uint64_t dr = 0xFFFF & (dst_nonpremul >> 32);\n uint64_t dg = 0xFFFF & (dst_nonpremul >> 16);\n uint64_t db = 0xFFFF & (dst_nonpremul >> 0);\n\n // Convert ds" + + "t from nonpremul to premul.\n dr = (dr * da) / 0xFFFF;\n dg = (dg * da) / 0xFFFF;\n db = (db * da) / 0xFFFF;\n\n // Calculate the inverse of the src-alpha: how much of the dst to keep.\n uint64_t ia = 0xFFFF - sa;\n\n // Composite src (nonpremul) over dst (premul).\n da = sa + ((da * ia) / 0xFFFF);\n dr = ((sr * sa) + (dr * ia)) / 0xFFFF;\n dg = ((sg * sa) + (dg * ia)) / 0xFFFF;\n db = ((sb * sa) + (db * ia)) / 0xFFFF;\n\n // Convert dst from premul to nonpremul.\n if (da != 0) {\n dr = (dr * 0xFFFF) / da;\n dg = (dg * 0xFFFF) / da;\n db = (db * 0xFFFF) / da;\n }\n\n // Combine components.\n return (db << 0) | (dg << 16) | (dr << 32) | (da << 48);\n}\n\nstatic inline uint32_t //\nwuffs_base__composite_nonpremul_premul_u32_axxx(uint32_t dst_nonpremul,\n uint32_t src_premul) {\n // Extract 16-bit color components.\n uint32_t sa = 0x101 * (0xFF & (src_premul >> 24));\n uint32_t sr = 0x101 * (0xFF & (src_premul >> 16));\n uint32_t sg = 0x101 * (0xFF & (src_premul >> 8)" + + ");\n uint32_t sb = 0x101 * (0xFF & (src_premul >> 0));\n uint32_t da = 0x101 * (0xFF & (dst_nonpremul >> 24));\n uint32_t dr = 0x101 * (0xFF & (dst_nonpremul >> 16));\n uint32_t dg = 0x101 * (0xFF & (dst_nonpremul >> 8));\n uint32_t db = 0x101 * (0xFF & (dst_nonpremul >> 0));\n\n // Convert dst from nonpremul to premul.\n dr = (dr * da) / 0xFFFF;\n dg = (dg * da) / 0xFFFF;\n db = (db * da) / 0xFFFF;\n\n // Calculate the inverse of the src-alpha: how much of the dst to keep.\n uint32_t ia = 0xFFFF - sa;\n\n // Composite src (premul) over dst (premul).\n da = sa + ((da * ia) / 0xFFFF);\n dr = sr + ((dr * ia) / 0xFFFF);\n dg = sg + ((dg * ia) / 0xFFFF);\n db = sb + ((db * ia) / 0xFFFF);\n\n // Convert dst from premul to nonpremul.\n if (da != 0) {\n dr = (dr * 0xFFFF) / da;\n dg = (dg * 0xFFFF) / da;\n db = (db * 0xFFFF) / da;\n }\n\n // Convert from 16-bit color to 8-bit color.\n da >>= 8;\n dr >>= 8;\n dg >>= 8;\n db >>= 8;\n\n // Combine components.\n return (db << 0) | (dg << 8) | (dr << 16) | (da << 24);\n}\n" + + "\nstatic inline uint32_t //\nwuffs_base__composite_premul_nonpremul_u32_axxx(uint32_t dst_premul,\n uint32_t src_nonpremul) {\n // Extract 16-bit color components.\n uint32_t sa = 0x101 * (0xFF & (src_nonpremul >> 24));\n uint32_t sr = 0x101 * (0xFF & (src_nonpremul >> 16));\n uint32_t sg = 0x101 * (0xFF & (src_nonpremul >> 8));\n uint32_t sb = 0x101 * (0xFF & (src_nonpremul >> 0));\n uint32_t da = 0x101 * (0xFF & (dst_premul >> 24));\n uint32_t dr = 0x101 * (0xFF & (dst_premul >> 16));\n uint32_t dg = 0x101 * (0xFF & (dst_premul >> 8));\n uint32_t db = 0x101 * (0xFF & (dst_premul >> 0));\n\n // Calculate the inverse of the src-alpha: how much of the dst to keep.\n uint32_t ia = 0xFFFF - sa;\n\n // Composite src (nonpremul) over dst (premul).\n da = sa + ((da * ia) / 0xFFFF);\n dr = ((sr * sa) + (dr * ia)) / 0xFFFF;\n dg = ((sg * sa) + (dg * ia)) / 0xFFFF;\n db = ((sb * sa) + (db * ia)) / 0xFFFF;\n\n // Convert from 16-bit color to 8-bit color.\n da >>= 8;\n dr >>= 8;" + + "\n dg >>= 8;\n db >>= 8;\n\n // Combine components.\n return (db << 0) | (dg << 8) | (dr << 16) | (da << 24);\n}\n\nstatic inline uint64_t //\nwuffs_base__composite_premul_nonpremul_u64_axxx(uint64_t dst_premul,\n uint64_t src_nonpremul) {\n // Extract components.\n uint64_t sa = 0xFFFF & (src_nonpremul >> 48);\n uint64_t sr = 0xFFFF & (src_nonpremul >> 32);\n uint64_t sg = 0xFFFF & (src_nonpremul >> 16);\n uint64_t sb = 0xFFFF & (src_nonpremul >> 0);\n uint64_t da = 0xFFFF & (dst_premul >> 48);\n uint64_t dr = 0xFFFF & (dst_premul >> 32);\n uint64_t dg = 0xFFFF & (dst_premul >> 16);\n uint64_t db = 0xFFFF & (dst_premul >> 0);\n\n // Calculate the inverse of the src-alpha: how much of the dst to keep.\n uint64_t ia = 0xFFFF - sa;\n\n // Composite src (nonpremul) over dst (premul).\n da = sa + ((da * ia) / 0xFFFF);\n dr = ((sr * sa) + (dr * ia)) / 0xFFFF;\n dg = ((sg * sa) + (dg * ia)) / 0xFFFF;\n db = ((sb * sa) + (db * ia)) / 0xFFFF;\n\n // Combine components.\n return " + + "(db << 0) | (dg << 16) | (dr << 32) | (da << 48);\n}\n\nstatic inline uint32_t //\nwuffs_base__composite_premul_premul_u32_axxx(uint32_t dst_premul,\n uint32_t src_premul) {\n // Extract 16-bit color components.\n uint32_t sa = 0x101 * (0xFF & (src_premul >> 24));\n uint32_t sr = 0x101 * (0xFF & (src_premul >> 16));\n uint32_t sg = 0x101 * (0xFF & (src_premul >> 8));\n uint32_t sb = 0x101 * (0xFF & (src_premul >> 0));\n uint32_t da = 0x101 * (0xFF & (dst_premul >> 24));\n uint32_t dr = 0x101 * (0xFF & (dst_premul >> 16));\n uint32_t dg = 0x101 * (0xFF & (dst_premul >> 8));\n uint32_t db = 0x101 * (0xFF & (dst_premul >> 0));\n\n // Calculate the inverse of the src-alpha: how much of the dst to keep.\n uint32_t ia = 0xFFFF - sa;\n\n // Composite src (premul) over dst (premul).\n da = sa + ((da * ia) / 0xFFFF);\n dr = sr + ((dr * ia) / 0xFFFF);\n dg = sg + ((dg * ia) / 0xFFFF);\n db = sb + ((db * ia) / 0xFFFF);\n\n // Convert from 16-bit color to 8-bit color.\n da >>= 8;\n dr" + + " >>= 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_bgr_565_888(wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 src) {\n size_t len4 = (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 = len4;\n while (n--) {\n uint32_t argb = wuffs_base__load_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__store_u32le__no_bounds_check(\n d, alpha | (r5 << 11) | (g6 << 5) | (b5 << 0));\n s += 4;\n d += 4;\n }\n return len4 * 4;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__swap_rgbx_bgrx(wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 src) {\n size_t len4 = (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 = len4;\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 len4 * 4;\n}\n\n" + + "// --------\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__squash_align4_bgr_565_888(\n wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 src) {\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__load_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__store_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 uint64_t //\nwuffs_base__pixel_swizzler__swap_rgbx_bgrx(wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 src) {\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 uint" + + "8_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 const size_t loop_unroll_count = 4;\n\n size_t n = len;\n while (n >= loop_unroll_count) {\n wuffs_base__store_u32le__no_bounds_check(\n d + (0 * 4), wuffs_base__color_u64__as__color_u32(\n wuffs_base__load_u64le__no_bounds_check(s + (0 * 8))));\n wuffs_base__store_u32le__no_bounds_check(\n d + (1 * 4), wuffs_base__color_u64__as__" + + "color_u32(\n wuffs_base__load_u64le__no_bounds_check(s + (1 * 8))));\n wuffs_base__store_u32le__no_bounds_check(\n d + (2 * 4), wuffs_base__color_u64__as__color_u32(\n wuffs_base__load_u64le__no_bounds_check(s + (2 * 8))));\n wuffs_base__store_u32le__no_bounds_check(\n d + (3 * 4), wuffs_base__color_u64__as__color_u32(\n wuffs_base__load_u64le__no_bounds_check(s + (3 * 8))));\n\n s += loop_unroll_count * 8;\n d += loop_unroll_count * 4;\n n -= loop_unroll_count;\n }\n\n while (n >= 1) {\n wuffs_base__store_u32le__no_bounds_check(\n d + (0 * 4), wuffs_base__color_u64__as__color_u32(\n wuffs_base__load_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" + "" + "// --------\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__copy_1_1(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;\n if (len > 0) {\n memmove(dst_ptr, src_ptr, len);\n }\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__copy_3_3(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_len3 = dst_len / 3;\n size_t src_len3 = src_len / 3;\n size_t len = (dst_len3 < src_len3) ? dst_len3 : src_len3;\n if (len > 0) {\n " + " memmove(dst_ptr, src_ptr, len * 3);\n }\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__copy_4_4(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 if (len > 0) {\n memmove(dst_ptr, src_ptr, len * 4);\n }\n return len;\n}\n\n" + "" + "// --------\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__bgr_565__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_len2 = dst_len / 2;\n size_t src_len3 = src_len / 3;\n size_t len = (dst_len2 < src_len3) ? dst_len2 : 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 uint32_t b5 = s[0] >> 3;\n uint32_t g6 = s[1] >> 2;\n uint32_t r5 = s[2] >> 3;\n uint32_t rgb_565 = (r5 << 11) | (g6 << 5) | (b5 << 0);\n wuffs_base__store_u16le__no_bounds_check(d + (0 * 2), (uint16_t)rgb_565);\n\n s += 1 * 3;\n d += 1 * 2;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__bgr_565__bgra_nonpremul__src(\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_len2 = dst_len / 2;\n size_t src_len4 = src_len / 4;\n size_t len = (dst_len2 < src_len4) ? dst_len2 : 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__store_u16le__no_bounds_check(\n d + (0 * 2),\n wuffs_base__color_u32_argb_premul__as__color_u16_rgb_565(\n wuffs_base__color_u32_argb_nonpremul__as__color_u32_argb_premul(\n wuffs_base__load_u32le__no_bounds_check(s + (0 * 4)))));\n\n s += 1 * 4;\n d += 1 * 2;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__bgr_565__bgra_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_len2 = dst_len / 2;\n size_t src_len4 = src_len / 4;" + - "\n size_t len = (dst_len2 < src_len4) ? dst_len2 : 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 // Convert from 8-bit color to 16-bit color.\n uint32_t sa = 0x101 * ((uint32_t)s[3]);\n uint32_t sr = 0x101 * ((uint32_t)s[2]);\n uint32_t sg = 0x101 * ((uint32_t)s[1]);\n uint32_t sb = 0x101 * ((uint32_t)s[0]);\n\n // Convert from 565 color to 16-bit color.\n uint32_t old_rgb_565 = wuffs_base__load_u16le__no_bounds_check(d + (0 * 2));\n uint32_t old_r5 = 0x1F & (old_rgb_565 >> 11);\n uint32_t dr = (0x8421 * old_r5) >> 4;\n uint32_t old_g6 = 0x3F & (old_rgb_565 >> 5);\n uint32_t dg = (0x1041 * old_g6) >> 2;\n uint32_t old_b5 = 0x1F & (old_rgb_565 >> 0);\n uint32_t db = (0x8421 * old_b5) >> 4;\n\n // Calculate the inverse of the src-alpha: how much of the dst to keep.\n uint32_t ia = 0xFFFF - sa;\n\n // Composite src (nonpremul) over dst (premul).\n dr = ((sr * sa) + (dr * ia)) / 0xFFFF;\n dg = ((sg * sa)" + - " + (dg * ia)) / 0xFFFF;\n db = ((sb * sa) + (db * ia)) / 0xFFFF;\n\n // Convert from 16-bit color to 565 color and combine the components.\n uint32_t new_r5 = 0x1F & (dr >> 11);\n uint32_t new_g6 = 0x3F & (dg >> 10);\n uint32_t new_b5 = 0x1F & (db >> 11);\n uint32_t new_rgb_565 = (new_r5 << 11) | (new_g6 << 5) | (new_b5 << 0);\n wuffs_base__store_u16le__no_bounds_check(d + (0 * 2),\n (uint16_t)new_rgb_565);\n\n s += 1 * 4;\n d += 1 * 2;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__bgr_565__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_len2 = dst_len / 2;\n size_t len = (dst_len2 < src_len) ? dst_len2 : src_len;\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 y5 = s[0] >> 3;\n uint32_t y6 = s[0] >> 2;\n uint32_t rgb_565 = (y5 << 11) | (y6 << 5) | (y5 << 0);\n wuffs_base__store_u16le__no_bounds_check(d + (0 * 2), (uint16_t)rgb_565);\n\n s += 1 * 1;\n d += 1 * 2;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__bgr_565__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_len2 = dst_len / 2;\n size_t len = (dst_len2 < src_len) ? dst_len2 : 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__store_u16le__no_bounds_check(\n d + (0 * 2), wuffs_base__load_u16le__no_bounds_check(\n dst_palette_ptr + ((size_t)s[0] * 4)));\n wuffs_base__store_u16le__no_bounds_check(\n d + (1 * 2), wuffs_base__load_u16le__no_bounds_check(\n dst_palette_ptr + ((size_t)s[1] * 4)));\n wuffs_base__store_u16le__no_bounds_check(\n d + (2 * 2), wuffs_base__load_u16le__no_bounds_check(\n dst_palette_ptr + ((size_t)s[2] * 4)));\n wuffs_base__store_u16le__no_bounds_check(\n d + (3 * 2), wuffs_base__load_u16le__no_bounds_check(\n dst_palette_ptr + ((size_t)s[3] * 4)));\n\n s += loop_unroll_count * 1;\n d += loop_unroll_count * 2;\n n -= loop_unroll_count;\n }\n\n while (n >= 1) {\n wuffs_base__store_u16le__no_bounds_check(\n d + (0 * 2), wuffs_base__load_u16le__no_bounds_check(\n dst_palette_ptr + ((size_t)s" + - "[0] * 4)));\n\n s += 1 * 1;\n d += 1 * 2;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__bgr_565__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_len2 = dst_len / 2;\n size_t len = (dst_len2 < src_len) ? dst_len2 : src_len;\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 s0 = wuffs_base__load_u32le__no_bounds_check(dst_palette_ptr +\n ((size_t)s[0] * 4));\n if (s0) {\n wuffs_base__store_u16le__no_bounds_check(d + (0 * 2), (uint16_t)s0);\n }\n\n s += 1 * 1;\n d += 1 * 2;\n n -= 1;\n }\n\n return len;\n}\n\n" + + " 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_len2 = dst_len / 2;\n size_t src_len4 = src_len / 4;\n size_t len = (dst_len2 < src_len4) ? dst_len2 : 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__store_u16le__no_bounds_check(\n d + (0 * 2),\n wuffs_base__color_u32_argb_premul__as__color_u16_rgb_565(\n wuffs_base__color_u32_argb_nonpremul__as__color_u32_argb_premul(\n wuffs_base__load_u32le__no_bounds_check(s + (0 * 4)))));\n\n s += 1 * 4;\n d += 1 * 2;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__bgr_565__bgra_nonpremul_4x16le__src(\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_len2 = dst_len / 2;\n size_t src_len8 = src_len / " + + "8;\n size_t len = (dst_len2 < src_len8) ? dst_len2 : src_len8;\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__store_u16le__no_bounds_check(\n d + (0 * 2),\n wuffs_base__color_u32_argb_premul__as__color_u16_rgb_565(\n wuffs_base__color_u64_argb_nonpremul__as__color_u32_argb_premul(\n wuffs_base__load_u64le__no_bounds_check(s + (0 * 8)))));\n\n s += 1 * 8;\n d += 1 * 2;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__bgr_565__bgra_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_len2 = dst_len / 2;\n size_t src_len4 = src_len / 4;\n size_t len = (dst_len2 < src_len4) ? dst_len2 : 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 // Extract 16-bit color " + + "components.\n uint32_t sa = 0x101 * ((uint32_t)s[3]);\n uint32_t sr = 0x101 * ((uint32_t)s[2]);\n uint32_t sg = 0x101 * ((uint32_t)s[1]);\n uint32_t sb = 0x101 * ((uint32_t)s[0]);\n\n // Convert from 565 color to 16-bit color.\n uint32_t old_rgb_565 = wuffs_base__load_u16le__no_bounds_check(d + (0 * 2));\n uint32_t old_r5 = 0x1F & (old_rgb_565 >> 11);\n uint32_t dr = (0x8421 * old_r5) >> 4;\n uint32_t old_g6 = 0x3F & (old_rgb_565 >> 5);\n uint32_t dg = (0x1041 * old_g6) >> 2;\n uint32_t old_b5 = 0x1F & (old_rgb_565 >> 0);\n uint32_t db = (0x8421 * old_b5) >> 4;\n\n // Calculate the inverse of the src-alpha: how much of the dst to keep.\n uint32_t ia = 0xFFFF - sa;\n\n // Composite src (nonpremul) over dst (premul).\n dr = ((sr * sa) + (dr * ia)) / 0xFFFF;\n dg = ((sg * sa) + (dg * ia)) / 0xFFFF;\n db = ((sb * sa) + (db * ia)) / 0xFFFF;\n\n // Convert from 16-bit color to 565 color and combine the components.\n uint32_t new_r5 = 0x1F & (dr >> 11);\n uint32_t new_g6 = 0x3F " + + "& (dg >> 10);\n uint32_t new_b5 = 0x1F & (db >> 11);\n uint32_t new_rgb_565 = (new_r5 << 11) | (new_g6 << 5) | (new_b5 << 0);\n wuffs_base__store_u16le__no_bounds_check(d + (0 * 2),\n (uint16_t)new_rgb_565);\n\n s += 1 * 4;\n d += 1 * 2;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__bgr_565__bgra_nonpremul_4x16le__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_len2 = dst_len / 2;\n size_t src_len8 = src_len / 8;\n size_t len = (dst_len2 < src_len8) ? dst_len2 : src_len8;\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 // Extract 16-bit color components.\n uint32_t sa = ((uint32_t)wuffs_base__load_u16le__no_bounds_check(s + 6));\n uint32_t sr = ((uint32_t)wuffs_base__load_u16le__no_bounds_check(s + 4));\n uint32_t sg = ((uint32" + + "_t)wuffs_base__load_u16le__no_bounds_check(s + 2));\n uint32_t sb = ((uint32_t)wuffs_base__load_u16le__no_bounds_check(s + 0));\n\n // Convert from 565 color to 16-bit color.\n uint32_t old_rgb_565 = wuffs_base__load_u16le__no_bounds_check(d + (0 * 2));\n uint32_t old_r5 = 0x1F & (old_rgb_565 >> 11);\n uint32_t dr = (0x8421 * old_r5) >> 4;\n uint32_t old_g6 = 0x3F & (old_rgb_565 >> 5);\n uint32_t dg = (0x1041 * old_g6) >> 2;\n uint32_t old_b5 = 0x1F & (old_rgb_565 >> 0);\n uint32_t db = (0x8421 * old_b5) >> 4;\n\n // Calculate the inverse of the src-alpha: how much of the dst to keep.\n uint32_t ia = 0xFFFF - sa;\n\n // Composite src (nonpremul) over dst (premul).\n dr = ((sr * sa) + (dr * ia)) / 0xFFFF;\n dg = ((sg * sa) + (dg * ia)) / 0xFFFF;\n db = ((sb * sa) + (db * ia)) / 0xFFFF;\n\n // Convert from 16-bit color to 565 color and combine the components.\n uint32_t new_r5 = 0x1F & (dr >> 11);\n uint32_t new_g6 = 0x3F & (dg >> 10);\n uint32_t new_b5 = 0x1F & (db >> 11);\n " + + " uint32_t new_rgb_565 = (new_r5 << 11) | (new_g6 << 5) | (new_b5 << 0);\n wuffs_base__store_u16le__no_bounds_check(d + (0 * 2),\n (uint16_t)new_rgb_565);\n\n s += 1 * 8;\n d += 1 * 2;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__bgr_565__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_len2 = dst_len / 2;\n size_t len = (dst_len2 < src_len) ? dst_len2 : src_len;\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 y5 = s[0] >> 3;\n uint32_t y6 = s[0] >> 2;\n uint32_t rgb_565 = (y5 << 11) | (y6 << 5) | (y5 << 0);\n wuffs_base__store_u16le__no_bounds_check(d + (0 * 2), " + + "(uint16_t)rgb_565);\n\n s += 1 * 1;\n d += 1 * 2;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__bgr_565__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_len2 = dst_len / 2;\n size_t len = (dst_len2 < src_len) ? dst_len2 : 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__store_u16le__no_bounds_check(\n d + (0 * 2), wuffs_base__load_u16le__no_bounds_check(\n dst_palette_ptr + ((size_t)s[0] * 4)));\n wuffs_base__store_u16le__no_bounds_ch" + + "eck(\n d + (1 * 2), wuffs_base__load_u16le__no_bounds_check(\n dst_palette_ptr + ((size_t)s[1] * 4)));\n wuffs_base__store_u16le__no_bounds_check(\n d + (2 * 2), wuffs_base__load_u16le__no_bounds_check(\n dst_palette_ptr + ((size_t)s[2] * 4)));\n wuffs_base__store_u16le__no_bounds_check(\n d + (3 * 2), wuffs_base__load_u16le__no_bounds_check(\n dst_palette_ptr + ((size_t)s[3] * 4)));\n\n s += loop_unroll_count * 1;\n d += loop_unroll_count * 2;\n n -= loop_unroll_count;\n }\n\n while (n >= 1) {\n wuffs_base__store_u16le__no_bounds_check(\n d + (0 * 2), wuffs_base__load_u16le__no_bounds_check(\n dst_palette_ptr + ((size_t)s[0] * 4)));\n\n s += 1 * 1;\n d += 1 * 2;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__bgr_565__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_len2 = dst_len / 2;\n size_t len = (dst_len2 < src_len) ? dst_len2 : src_len;\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 s0 = wuffs_base__load_u32le__no_bounds_check(dst_palette_ptr +\n ((size_t)s[0] * 4));\n if (s0) {\n wuffs_base__store_u16le__no_bounds_check(d + (0 * 2), (uint16_t)s0);\n }\n\n s += 1 * 1;\n d += 1 * 2;\n n -= 1;\n }\n\n return len;\n}\n\n" + "" + "// --------\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__bgr__bgra_nonpremul__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 size_t dst_len3 = dst_len / 3;\n size_t src_len4 = src_len / 4;\n size_t len = (dst_len3 < src_len4) ? dst_len3 : 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 s0 =\n wuffs_base__color_u32_argb_nonpremul__as__color_u32_argb_premul(\n wuffs_base__load_u32le__no_bounds_check(s + (0 * 4)));\n wuffs_base__store_u24le__no_bounds_check(d + (0 * 3), s0);\n\n s += 1 * 4;\n d += 1 * 3;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nw" + - "uffs_base__pixel_swizzler__bgr__bgra_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_len3 = dst_len / 3;\n size_t src_len4 = src_len / 4;\n size_t len = (dst_len3 < src_len4) ? dst_len3 : 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 // Convert from 8-bit color to 16-bit color.\n uint32_t sa = 0x101 * ((uint32_t)s[3]);\n uint32_t sr = 0x101 * ((uint32_t)s[2]);\n uint32_t sg = 0x101 * ((uint32_t)s[1]);\n uint32_t sb = 0x101 * ((uint32_t)s[0]);\n uint32_t dr = 0x101 * ((uint32_t)d[2]);\n uint32_t dg = 0x101 * ((uint32_t)d[1]);\n uint32_t db = 0x101 * ((uint32_t)d[0]);\n\n // Calculate the inverse of the src-alpha: how much of the dst to keep.\n uint32_t ia = 0xFFFF - sa;\n\n // Composite src (nonpremul) over dst (premul).\n dr = ((sr * sa) + (dr * ia)) / 0xFFFF;\n dg = ((sg * s" + - "a) + (dg * ia)) / 0xFFFF;\n db = ((sb * sa) + (db * ia)) / 0xFFFF;\n\n // Convert from 16-bit color to 8-bit color.\n d[0] = (uint8_t)(db >> 8);\n d[1] = (uint8_t)(dg >> 8);\n d[2] = (uint8_t)(dr >> 8);\n\n s += 1 * 4;\n d += 1 * 3;\n n -= 1;\n }\n\n return len;\n}\n\n" + + "uffs_base__pixel_swizzler__bgr__bgra_nonpremul_4x16le__src(\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_len3 = dst_len / 3;\n size_t src_len8 = src_len / 8;\n size_t len = (dst_len3 < src_len8) ? dst_len3 : src_len8;\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 s0 =\n wuffs_base__color_u64_argb_nonpremul__as__color_u32_argb_premul(\n wuffs_base__load_u64le__no_bounds_check(s + (0 * 8)));\n wuffs_base__store_u24le__no_bounds_check(d + (0 * 3), s0);\n\n s += 1 * 8;\n d += 1 * 3;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__bgr__bgra_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_len3 = dst_len / 3;\n size_t src_len4 = src_le" + + "n / 4;\n size_t len = (dst_len3 < src_len4) ? dst_len3 : 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 // Extract 16-bit color components.\n uint32_t sa = 0x101 * ((uint32_t)s[3]);\n uint32_t sr = 0x101 * ((uint32_t)s[2]);\n uint32_t sg = 0x101 * ((uint32_t)s[1]);\n uint32_t sb = 0x101 * ((uint32_t)s[0]);\n uint32_t dr = 0x101 * ((uint32_t)d[2]);\n uint32_t dg = 0x101 * ((uint32_t)d[1]);\n uint32_t db = 0x101 * ((uint32_t)d[0]);\n\n // Calculate the inverse of the src-alpha: how much of the dst to keep.\n uint32_t ia = 0xFFFF - sa;\n\n // Composite src (nonpremul) over dst (premul).\n dr = ((sr * sa) + (dr * ia)) / 0xFFFF;\n dg = ((sg * sa) + (dg * ia)) / 0xFFFF;\n db = ((sb * sa) + (db * ia)) / 0xFFFF;\n\n // Convert from 16-bit color to 8-bit color.\n d[0] = (uint8_t)(db >> 8);\n d[1] = (uint8_t)(dg >> 8);\n d[2] = (uint8_t)(dr >> 8);\n\n s += 1 * 4;\n d += 1 * 3;\n n -= 1;\n }\n\n return len;\n" + + "}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__bgr__bgra_nonpremul_4x16le__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_len3 = dst_len / 3;\n size_t src_len8 = src_len / 8;\n size_t len = (dst_len3 < src_len8) ? dst_len3 : src_len8;\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 // Extract 16-bit color components.\n uint32_t sa = ((uint32_t)wuffs_base__load_u16le__no_bounds_check(s + 6));\n uint32_t sr = ((uint32_t)wuffs_base__load_u16le__no_bounds_check(s + 4));\n uint32_t sg = ((uint32_t)wuffs_base__load_u16le__no_bounds_check(s + 2));\n uint32_t sb = ((uint32_t)wuffs_base__load_u16le__no_bounds_check(s + 0));\n uint32_t dr = 0x101 * ((uint32_t)d[2]);\n uint32_t dg = 0x101 * ((uint32_t)d[1]);\n uint32_t db = 0x101 * ((uint32_t)d[0]);\n\n // Calculate the inverse of the src-alpha: how much of the" + + " dst to keep.\n uint32_t ia = 0xFFFF - sa;\n\n // Composite src (nonpremul) over dst (premul).\n dr = ((sr * sa) + (dr * ia)) / 0xFFFF;\n dg = ((sg * sa) + (dg * ia)) / 0xFFFF;\n db = ((sb * sa) + (db * ia)) / 0xFFFF;\n\n // Convert from 16-bit color to 8-bit color.\n d[0] = (uint8_t)(db >> 8);\n d[1] = (uint8_t)(dg >> 8);\n d[2] = (uint8_t)(dr >> 8);\n\n s += 1 * 8;\n d += 1 * 3;\n n -= 1;\n }\n\n return len;\n}\n\n" + "" + - "// --------\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__bgra_nonpremul__bgra_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__load_u32le__no_bounds_check(d + (0 * 4));\n uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(s + (0 * 4));\n wuffs_base__store_u32le__no_bounds_check(\n d + (0 * 4),\n wuffs_base__composite_nonpremul_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__bgra_nonpremul__bgra_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 while (n >= 1) {\n uint32_t d0 = wuffs_base__load_u32le__no_bounds_check(d + (0 * 4));\n uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(s + (0 * 4));\n wuffs_base__store_u32le__no_bounds_check(\n d + (0 * 4),\n wuffs_base__composite_nonpremul_nonpremul_u32_axxx(d0, s0));\n\n s += 1 * 4;\n d += 1 * 4;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__bgra_nonpremul__bgra_nonpremul_4x16le__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_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 size_t n = len;\n\n while (n >= 1) {\n uint64_t d0 = wuffs_base__color_u32__as__color_u64(\n wuffs_base__load_u32le__no_bounds_check(d + (0 * 4)));\n uint64_t s0 = wuffs_base__load_u64le__no_bounds_check(s + (0 * 8));\n wuffs_base__store_u32le__no_bounds_check(\n d + (0 * 4),\n wuffs_base__color_u64__as__color_u32(\n wuffs_base__composite_nonpremul_nonpremul_u64_axxx(d0, s0)));\n\n s += 1 * 8;\n d += 1 * 4;\n n -= 1;\n }\n\n return len;\n}\n\n" + "" + - "// --------\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul__src(\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 s0 = wuffs_base__load_u32le__no_bounds_check(s + (0 * 4));\n wuffs_base__store_u32le__no_bounds_check(\n d + (0 * 4),\n wuffs_base__color_u32_argb_nonpremul__as__color_u32_argb_premul(s0));\n\n s += 1 * 4;\n d += 1 * 4;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__bgra_premul__bgra_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__load_u32le__no_bounds_check(d + (0 * 4));\n uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(s + (0 * 4));\n wuffs_base__store_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__bgra_premul__bgra_nonpremul__src(\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 s0 = wuffs_base__load_u32le__no_bounds_check(s + (0 * 4));\n wuffs_base__store_u32le__no_bounds_check(\n d + (0 * 4),\n wuffs_base__color_u32_argb_nonpremul__as__color_u32_argb_premul(s0));\n\n s += 1 * 4;\n d += 1 * 4;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul_4x16le__src(\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_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 size_t n = len;\n\n // TODO: unroll.\n\n while (n >= 1) {\n uint64_t s0 = wuffs_base__load_u64le__no_bounds_check(s + (0 * 8));\n wuffs_base__store_u32le__no_bounds_check(\n d + (0 * 4),\n wuffs_base__color_u64_argb_nonpremul__as__color_u32_argb_premul(s0));\n\n s += 1 * 8;\n d += 1 * 4;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__bgra_premul__bgra_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__load_u32le__no_" + + "bounds_check(d + (0 * 4));\n uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(s + (0 * 4));\n wuffs_base__store_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\nstatic uint64_t //\nwuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul_4x16le__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_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 size_t n = len;\n\n // TODO: unroll.\n\n while (n >= 1) {\n uint64_t d0 = wuffs_base__color_u32__as__color_u64(\n wuffs_base__load_u32le__no_bounds_check(d + (0 * 4)));\n uint64_t s0 = wuffs_base__load_u64le__no_bounds_check(s + (0 * 8));\n wuffs_base__store_u32le__no_bounds_check(\n d + (0" + + " * 4),\n wuffs_base__color_u64__as__color_u32(\n wuffs_base__composite_premul_nonpremul_u64_axxx(d0, s0)));\n\n s += 1 * 8;\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__store_u32le__no_bounds_check(\n d + (0 * 3), wuffs_base__load_u32le__no_bounds_check(\n dst_palette_ptr + ((size_t)s[0] * 4)));\n wuffs_base__store_u32le__no_bounds_check(\n d + (1 * 3), wuffs_base__load_u32le__no_bounds_check(\n dst_palette_ptr + ((size_t)s[1] * 4)));\n wuffs_base__store_u32le__no_bounds_check(\n d + (2 * 3), wuffs_base__load_u32le__no_bounds_check(\n dst_palette_ptr + ((size_t)s[2] * 4)));\n wuffs_base__store_u32le__no_bounds_check(\n d + (3 * 3), wuffs_base__load_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__load_u32le__no_bounds_check(dst_palette_ptr +\n ((siz" + @@ -569,18 +585,20 @@ const BasePixConvSubmoduleC = "" + "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 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__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_BASE__PIXEL_FORMAT__RGBA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY:\n case WU" + - "FFS_BASE__PIXEL_FORMAT__RGBX:\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_bgr_565_888(dst_palette,\n " + - " src_palette) != 1024) {\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_swizzler__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_pa" + - "lette) !=\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(dst_palette,\n src_palette) != 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_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(dst_palette,\n src_palette) != 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 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__B" + - "GR:\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__xxxx__xxx;\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_b" + - "ase__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_swiz" + - "zler__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\n" + + "FFS_BASE__PIXEL_FORMAT__RGBX:\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, src_palette) != 25" + + "6) {\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_swizzler__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(dst_palette,\n src_palette) != 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(dst_palette,\n " + + " src_palette) != 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_swi" + + "zzler__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__xxxx__xxx;\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_non" + + "premul__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_nonpremul_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\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.src_pixfmt_bytes_per_pixel = 0;\n\n wuffs_base__pixel_swizzler__func func = NULL;\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 for" + - "mats.\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, blend);\n break;\n\n case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_BINARY:\n func = wuffs_base__pixel_swizzler__prepare__indexed__bgra_binary(\n p, dst_pixfmt, dst_palette, src_palette, blend);\n break;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGR:\n func = wuffs_base__pixel_swizzler__prepare__bgr(\n p, dst_pixfmt, dst_palette, src_palette, blend);\n break;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL:\n func = wuffs_base__pixel_swizzler__prepare__bgra_nonpremul(\n p, dst_pixfmt, dst_palette, src_palette, blend);\n break;\n }\n\n p->private_impl.func = func;\n p->private_impl.src_pixfmt_bytes_per_pixel = src_pixfmt_bits_per_pixel / 8;\n return wuffs_base__make_status(\n func ? NULL : wuffs_base__error__unsupported_pixel_swizzler_option);\n}" + - "\n\nWUFFS_BASE__MAYBE_STATIC uint64_t //\nwuffs_base__pixel_swizzler__swizzle_interleaved_from_reader(\n const wuffs_base__pixel_swizzler* p,\n wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n const uint8_t** ptr_iop_r,\n const uint8_t* io2_r) {\n if (p && p->private_impl.func) {\n const uint8_t* iop_r = *ptr_iop_r;\n uint64_t n = (*p->private_impl.func)(dst.ptr, dst.len, dst_palette.ptr,\n dst_palette.len, iop_r,\n (size_t)(io2_r - iop_r));\n *ptr_iop_r += n * p->private_impl.src_pixfmt_bytes_per_pixel;\n return n;\n }\n return 0;\n}\n\nWUFFS_BASE__MAYBE_STATIC uint64_t //\nwuffs_base__pixel_swizzler__swizzle_interleaved_from_slice(\n const wuffs_base__pixel_swizzler* p,\n wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n if (p && p->private_impl.func) {\n return (*p->private_impl.func)(dst.ptr, dst.len, dst_palette.ptr,\n " + - " dst_palette.len, src.ptr, src.len);\n }\n return 0;\n}\n" + + "mats.\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, blend);\n break;\n\n case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_BINARY:\n func = wuffs_base__pixel_swizzler__prepare__indexed__bgra_binary(\n p, dst_pixfmt, dst_palette, src_palette, blend);\n break;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGR:\n func = wuffs_base__pixel_swizzler__prepare__bgr(\n p, dst_pixfmt, dst_palette, src_palette, blend);\n break;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL:\n func = wuffs_base__pixel_swizzler__prepare__bgra_nonpremul(\n p, dst_pixfmt, dst_palette, src_palette, blend);\n break;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL_4X16LE:\n func = wuffs_base__pixel_swizzler__prepare__bgra_nonpremul_4x16le(\n p, dst_pixfmt, dst_palette, src_palette, blend);\n break;\n }\n\n p->private_im" + + "pl.func = func;\n p->private_impl.src_pixfmt_bytes_per_pixel = src_pixfmt_bits_per_pixel / 8;\n return wuffs_base__make_status(\n func ? NULL : wuffs_base__error__unsupported_pixel_swizzler_option);\n}\n\nWUFFS_BASE__MAYBE_STATIC uint64_t //\nwuffs_base__pixel_swizzler__swizzle_interleaved_from_reader(\n const wuffs_base__pixel_swizzler* p,\n wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n const uint8_t** ptr_iop_r,\n const uint8_t* io2_r) {\n if (p && p->private_impl.func) {\n const uint8_t* iop_r = *ptr_iop_r;\n uint64_t n = (*p->private_impl.func)(dst.ptr, dst.len, dst_palette.ptr,\n dst_palette.len, iop_r,\n (size_t)(io2_r - iop_r));\n *ptr_iop_r += n * p->private_impl.src_pixfmt_bytes_per_pixel;\n return n;\n }\n return 0;\n}\n\nWUFFS_BASE__MAYBE_STATIC uint64_t //\nwuffs_base__pixel_swizzler__swizzle_interleaved_from_slice(\n const wuffs_base__pixel_swizzler* p,\n wuffs_base__slice_u8 dst,\n " + + " wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n if (p && p->private_impl.func) {\n return (*p->private_impl.func)(dst.ptr, dst.len, dst_palette.ptr,\n dst_palette.len, src.ptr, src.len);\n }\n return 0;\n}\n" + "" const BaseUTF8SubmoduleC = "" + diff --git a/lang/builtin/builtin.go b/lang/builtin/builtin.go index 6ff5a120..b4145776 100644 --- a/lang/builtin/builtin.go +++ b/lang/builtin/builtin.go @@ -60,13 +60,17 @@ var Consts = [...]struct { {t.IDU32, "0x80000565", "PIXEL_FORMAT__BGR_565"}, {t.IDU32, "0x80000888", "PIXEL_FORMAT__BGR"}, {t.IDU32, "0x81008888", "PIXEL_FORMAT__BGRA_NONPREMUL"}, + {t.IDU32, "0x8100BBBB", "PIXEL_FORMAT__BGRA_NONPREMUL_4X16LE"}, {t.IDU32, "0x82008888", "PIXEL_FORMAT__BGRA_PREMUL"}, + {t.IDU32, "0x8200BBBB", "PIXEL_FORMAT__BGRA_PREMUL_4X16LE"}, {t.IDU32, "0x83008888", "PIXEL_FORMAT__BGRA_BINARY"}, {t.IDU32, "0x90008888", "PIXEL_FORMAT__BGRX"}, {t.IDU32, "0xA0000888", "PIXEL_FORMAT__RGB"}, {t.IDU32, "0xA1008888", "PIXEL_FORMAT__RGBA_NONPREMUL"}, + {t.IDU32, "0xA100BBBB", "PIXEL_FORMAT__RGBA_NONPREMUL_4X16LE"}, {t.IDU32, "0xA2008888", "PIXEL_FORMAT__RGBA_PREMUL"}, + {t.IDU32, "0xA200BBBB", "PIXEL_FORMAT__RGBA_PREMUL_4X16LE"}, {t.IDU32, "0xA3008888", "PIXEL_FORMAT__RGBA_BINARY"}, {t.IDU32, "0xB0008888", "PIXEL_FORMAT__RGBX"}, diff --git a/release/c/wuffs-unsupported-snapshot.c b/release/c/wuffs-unsupported-snapshot.c index 8c6dfa47..3987662a 100644 --- a/release/c/wuffs-unsupported-snapshot.c +++ b/release/c/wuffs-unsupported-snapshot.c @@ -2866,6 +2866,68 @@ wuffs_base__color_u32_argb_premul__as__color_u32_argb_nonpremul( return (a << 24) | (r << 16) | (g << 8) | (b << 0); } +// wuffs_base__color_u64_argb_nonpremul__as__color_u32_argb_premul converts +// from 4x16LE non-premultiplied alpha to 4x8 premultiplied alpha. +static inline wuffs_base__color_u32_argb_premul // +wuffs_base__color_u64_argb_nonpremul__as__color_u32_argb_premul( + uint64_t argb_nonpremul) { + uint32_t a16 = 0xFFFF & (argb_nonpremul >> 48); + + uint32_t r16 = 0xFFFF & (argb_nonpremul >> 32); + r16 = (r16 * a16) / 0xFFFF; + uint32_t g16 = 0xFFFF & (argb_nonpremul >> 16); + g16 = (g16 * a16) / 0xFFFF; + uint32_t b16 = 0xFFFF & (argb_nonpremul >> 0); + b16 = (b16 * a16) / 0xFFFF; + + return ((a16 >> 8) << 24) | ((r16 >> 8) << 16) | ((g16 >> 8) << 8) | + ((b16 >> 8) << 0); +} + +// wuffs_base__color_u32_argb_premul__as__color_u64_argb_nonpremul converts +// from 4x8 premultiplied alpha to 4x16LE non-premultiplied alpha. +static inline uint64_t // +wuffs_base__color_u32_argb_premul__as__color_u64_argb_nonpremul( + wuffs_base__color_u32_argb_premul c) { + uint32_t a = 0xFF & (c >> 24); + if (a == 0xFF) { + uint64_t r16 = 0x101 * (0xFF & (c >> 16)); + uint64_t g16 = 0x101 * (0xFF & (c >> 8)); + uint64_t b16 = 0x101 * (0xFF & (c >> 0)); + return 0xFFFF000000000000u | (r16 << 32) | (g16 << 16) | (b16 << 0); + } else if (a == 0) { + return 0; + } + uint64_t a16 = a * 0x101; + + uint64_t r = 0xFF & (c >> 16); + uint64_t r16 = (r * (0x101 * 0xFFFF)) / a16; + uint64_t g = 0xFF & (c >> 8); + uint64_t g16 = (g * (0x101 * 0xFFFF)) / a16; + uint64_t b = 0xFF & (c >> 0); + uint64_t b16 = (b * (0x101 * 0xFFFF)) / a16; + + return (a16 << 48) | (r16 << 32) | (g16 << 16) | (b16 << 0); +} + +static inline uint64_t // +wuffs_base__color_u32__as__color_u64(uint32_t c) { + uint64_t a16 = 0x101 * (0xFF & (c >> 24)); + uint64_t r16 = 0x101 * (0xFF & (c >> 16)); + uint64_t g16 = 0x101 * (0xFF & (c >> 8)); + uint64_t b16 = 0x101 * (0xFF & (c >> 0)); + return (a16 << 48) | (r16 << 32) | (g16 << 16) | (b16 << 0); +} + +static inline uint32_t // +wuffs_base__color_u64__as__color_u32(uint64_t c) { + uint32_t a = ((uint32_t)(0xFF & (c >> 56))); + uint32_t r = ((uint32_t)(0xFF & (c >> 40))); + uint32_t g = ((uint32_t)(0xFF & (c >> 24))); + uint32_t b = ((uint32_t)(0xFF & (c >> 8))); + return (a << 24) | (r << 16) | (g << 8) | (b << 0); +} + // -------- typedef uint8_t wuffs_base__pixel_blend; @@ -2953,13 +3015,17 @@ wuffs_base__make_pixel_format(uint32_t repr) { #define WUFFS_BASE__PIXEL_FORMAT__BGR_565 0x80000565 #define WUFFS_BASE__PIXEL_FORMAT__BGR 0x80000888 #define WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL 0x81008888 +#define WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL_4X16LE 0x8100BBBB #define WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL 0x82008888 +#define WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL_4X16LE 0x8200BBBB #define WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY 0x83008888 #define WUFFS_BASE__PIXEL_FORMAT__BGRX 0x90008888 #define WUFFS_BASE__PIXEL_FORMAT__RGB 0xA0000888 #define WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL 0xA1008888 +#define WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL_4X16LE 0xA100BBBB #define WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL 0xA2008888 +#define WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL_4X16LE 0xA200BBBB #define WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY 0xA3008888 #define WUFFS_BASE__PIXEL_FORMAT__RGBX 0xB0008888 @@ -13466,6 +13532,9 @@ wuffs_base__pixel_buffer__color_u32_at(const wuffs_base__pixel_buffer* pb, case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL: return wuffs_base__color_u32_argb_nonpremul__as__color_u32_argb_premul( wuffs_base__load_u32le__no_bounds_check(row + (4 * ((size_t)x)))); + case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL_4X16LE: + return wuffs_base__color_u64_argb_nonpremul__as__color_u32_argb_premul( + wuffs_base__load_u64le__no_bounds_check(row + (8 * ((size_t)x)))); case WUFFS_BASE__PIXEL_FORMAT__BGRX: return 0xFF000000 | wuffs_base__load_u32le__no_bounds_check(row + (4 * ((size_t)x))); @@ -13553,6 +13622,12 @@ wuffs_base__pixel_buffer__set_color_u32_at( wuffs_base__color_u32_argb_premul__as__color_u32_argb_nonpremul( color)); break; + case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL_4X16LE: + wuffs_base__store_u64le__no_bounds_check( + row + (8 * ((size_t)x)), + wuffs_base__color_u32_argb_premul__as__color_u64_argb_nonpremul( + color)); + break; case WUFFS_BASE__PIXEL_FORMAT__RGB: wuffs_base__store_u24le__no_bounds_check( @@ -13645,7 +13720,7 @@ wuffs_base__pixel_palette__closest_element( static inline uint32_t // wuffs_base__composite_nonpremul_nonpremul_u32_axxx(uint32_t dst_nonpremul, uint32_t src_nonpremul) { - // Convert from 8-bit color to 16-bit color. + // Extract 16-bit color components. uint32_t sa = 0x101 * (0xFF & (src_nonpremul >> 24)); uint32_t sr = 0x101 * (0xFF & (src_nonpremul >> 16)); uint32_t sg = 0x101 * (0xFF & (src_nonpremul >> 8)); @@ -13676,18 +13751,58 @@ wuffs_base__composite_nonpremul_nonpremul_u32_axxx(uint32_t dst_nonpremul, db = (db * 0xFFFF) / da; } - // Convert from 16-bit color to 8-bit color and combine the components. + // Convert from 16-bit color to 8-bit color. da >>= 8; dr >>= 8; dg >>= 8; db >>= 8; + + // Combine components. return (db << 0) | (dg << 8) | (dr << 16) | (da << 24); } +static inline uint64_t // +wuffs_base__composite_nonpremul_nonpremul_u64_axxx(uint64_t dst_nonpremul, + uint64_t src_nonpremul) { + // Extract components. + uint64_t sa = 0xFFFF & (src_nonpremul >> 48); + uint64_t sr = 0xFFFF & (src_nonpremul >> 32); + uint64_t sg = 0xFFFF & (src_nonpremul >> 16); + uint64_t sb = 0xFFFF & (src_nonpremul >> 0); + uint64_t da = 0xFFFF & (dst_nonpremul >> 48); + uint64_t dr = 0xFFFF & (dst_nonpremul >> 32); + uint64_t dg = 0xFFFF & (dst_nonpremul >> 16); + uint64_t db = 0xFFFF & (dst_nonpremul >> 0); + + // Convert dst from nonpremul to premul. + dr = (dr * da) / 0xFFFF; + dg = (dg * da) / 0xFFFF; + db = (db * da) / 0xFFFF; + + // Calculate the inverse of the src-alpha: how much of the dst to keep. + uint64_t ia = 0xFFFF - sa; + + // Composite src (nonpremul) over dst (premul). + da = sa + ((da * ia) / 0xFFFF); + dr = ((sr * sa) + (dr * ia)) / 0xFFFF; + dg = ((sg * sa) + (dg * ia)) / 0xFFFF; + db = ((sb * sa) + (db * ia)) / 0xFFFF; + + // Convert dst from premul to nonpremul. + if (da != 0) { + dr = (dr * 0xFFFF) / da; + dg = (dg * 0xFFFF) / da; + db = (db * 0xFFFF) / da; + } + + // Combine components. + return (db << 0) | (dg << 16) | (dr << 32) | (da << 48); +} + static inline uint32_t // wuffs_base__composite_nonpremul_premul_u32_axxx(uint32_t dst_nonpremul, uint32_t src_premul) { - // Convert from 8-bit color to 16-bit color. + // Extract 16-bit color components. uint32_t sa = 0x101 * (0xFF & (src_premul >> 24)); uint32_t sr = 0x101 * (0xFF & (src_premul >> 16)); uint32_t sg = 0x101 * (0xFF & (src_premul >> 8)); @@ -13718,18 +13833,20 @@ wuffs_base__composite_nonpremul_premul_u32_axxx(uint32_t dst_nonpremul, db = (db * 0xFFFF) / da; } - // Convert from 16-bit color to 8-bit color and combine the components. + // Convert from 16-bit color to 8-bit color. da >>= 8; dr >>= 8; dg >>= 8; db >>= 8; + + // Combine components. return (db << 0) | (dg << 8) | (dr << 16) | (da << 24); } static inline uint32_t // wuffs_base__composite_premul_nonpremul_u32_axxx(uint32_t dst_premul, uint32_t src_nonpremul) { - // Convert from 8-bit color to 16-bit color. + // Extract 16-bit color components. uint32_t sa = 0x101 * (0xFF & (src_nonpremul >> 24)); uint32_t sr = 0x101 * (0xFF & (src_nonpremul >> 16)); uint32_t sg = 0x101 * (0xFF & (src_nonpremul >> 8)); @@ -13748,18 +13865,46 @@ wuffs_base__composite_premul_nonpremul_u32_axxx(uint32_t dst_premul, dg = ((sg * sa) + (dg * ia)) / 0xFFFF; db = ((sb * sa) + (db * ia)) / 0xFFFF; - // Convert from 16-bit color to 8-bit color and combine the components. + // Convert from 16-bit color to 8-bit color. da >>= 8; dr >>= 8; dg >>= 8; db >>= 8; + + // Combine components. return (db << 0) | (dg << 8) | (dr << 16) | (da << 24); } +static inline uint64_t // +wuffs_base__composite_premul_nonpremul_u64_axxx(uint64_t dst_premul, + uint64_t src_nonpremul) { + // Extract components. + uint64_t sa = 0xFFFF & (src_nonpremul >> 48); + uint64_t sr = 0xFFFF & (src_nonpremul >> 32); + uint64_t sg = 0xFFFF & (src_nonpremul >> 16); + uint64_t sb = 0xFFFF & (src_nonpremul >> 0); + uint64_t da = 0xFFFF & (dst_premul >> 48); + uint64_t dr = 0xFFFF & (dst_premul >> 32); + uint64_t dg = 0xFFFF & (dst_premul >> 16); + uint64_t db = 0xFFFF & (dst_premul >> 0); + + // Calculate the inverse of the src-alpha: how much of the dst to keep. + uint64_t ia = 0xFFFF - sa; + + // Composite src (nonpremul) over dst (premul). + da = sa + ((da * ia) / 0xFFFF); + dr = ((sr * sa) + (dr * ia)) / 0xFFFF; + dg = ((sg * sa) + (dg * ia)) / 0xFFFF; + db = ((sb * sa) + (db * ia)) / 0xFFFF; + + // Combine components. + return (db << 0) | (dg << 16) | (dr << 32) | (da << 48); +} + static inline uint32_t // wuffs_base__composite_premul_premul_u32_axxx(uint32_t dst_premul, uint32_t src_premul) { - // Convert from 8-bit color to 16-bit color. + // Extract 16-bit color components. uint32_t sa = 0x101 * (0xFF & (src_premul >> 24)); uint32_t sr = 0x101 * (0xFF & (src_premul >> 16)); uint32_t sg = 0x101 * (0xFF & (src_premul >> 8)); @@ -13778,24 +13923,27 @@ wuffs_base__composite_premul_premul_u32_axxx(uint32_t dst_premul, dg = sg + ((dg * ia) / 0xFFFF); db = sb + ((db * ia) / 0xFFFF); - // Convert from 16-bit color to 8-bit color and combine the components. + // Convert from 16-bit color to 8-bit color. da >>= 8; dr >>= 8; dg >>= 8; db >>= 8; + + // Combine components. return (db << 0) | (dg << 8) | (dr << 16) | (da << 24); } // -------- static uint64_t // -wuffs_base__pixel_swizzler__squash_bgr_565_888(wuffs_base__slice_u8 dst, - wuffs_base__slice_u8 src) { - size_t len4 = (dst.len < src.len ? dst.len : src.len) / 4; +wuffs_base__pixel_swizzler__squash_align4_bgr_565_888( + wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 src) { + size_t len = (dst.len < src.len ? dst.len : src.len) / 4; uint8_t* d = dst.ptr; const uint8_t* s = src.ptr; - size_t n = len4; + size_t n = len; while (n--) { uint32_t argb = wuffs_base__load_u32le__no_bounds_check(s); uint32_t b5 = 0x1F & (argb >> (8 - 5)); @@ -13807,17 +13955,17 @@ wuffs_base__pixel_swizzler__squash_bgr_565_888(wuffs_base__slice_u8 dst, s += 4; d += 4; } - return len4 * 4; + return len; } static uint64_t // wuffs_base__pixel_swizzler__swap_rgbx_bgrx(wuffs_base__slice_u8 dst, wuffs_base__slice_u8 src) { - size_t len4 = (dst.len < src.len ? dst.len : src.len) / 4; + size_t len = (dst.len < src.len ? dst.len : src.len) / 4; uint8_t* d = dst.ptr; const uint8_t* s = src.ptr; - size_t n = len4; + size_t n = len; while (n--) { uint8_t b0 = s[0]; uint8_t b1 = s[1]; @@ -13830,7 +13978,56 @@ wuffs_base__pixel_swizzler__swap_rgbx_bgrx(wuffs_base__slice_u8 dst, s += 4; d += 4; } - return len4 * 4; + return len; +} + +// -------- + +static uint64_t // +wuffs_base__pixel_swizzler__squash_tight_4x8_4x16le(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_len8 = src_len / 8; + size_t len = (dst_len4 < src_len8) ? dst_len4 : src_len8; + uint8_t* d = dst_ptr; + const uint8_t* s = src_ptr; + + const size_t loop_unroll_count = 4; + + size_t n = len; + while (n >= loop_unroll_count) { + wuffs_base__store_u32le__no_bounds_check( + d + (0 * 4), wuffs_base__color_u64__as__color_u32( + wuffs_base__load_u64le__no_bounds_check(s + (0 * 8)))); + wuffs_base__store_u32le__no_bounds_check( + d + (1 * 4), wuffs_base__color_u64__as__color_u32( + wuffs_base__load_u64le__no_bounds_check(s + (1 * 8)))); + wuffs_base__store_u32le__no_bounds_check( + d + (2 * 4), wuffs_base__color_u64__as__color_u32( + wuffs_base__load_u64le__no_bounds_check(s + (2 * 8)))); + wuffs_base__store_u32le__no_bounds_check( + d + (3 * 4), wuffs_base__color_u64__as__color_u32( + wuffs_base__load_u64le__no_bounds_check(s + (3 * 8)))); + + s += loop_unroll_count * 8; + d += loop_unroll_count * 4; + n -= loop_unroll_count; + } + + while (n >= 1) { + wuffs_base__store_u32le__no_bounds_check( + d + (0 * 4), wuffs_base__color_u64__as__color_u32( + wuffs_base__load_u64le__no_bounds_check(s + (0 * 8)))); + + s += 1 * 8; + d += 1 * 4; + n -= 1; + } + return len; } // -------- @@ -13946,6 +14143,38 @@ wuffs_base__pixel_swizzler__bgr_565__bgra_nonpremul__src( return len; } +static uint64_t // +wuffs_base__pixel_swizzler__bgr_565__bgra_nonpremul_4x16le__src( + 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_len2 = dst_len / 2; + size_t src_len8 = src_len / 8; + size_t len = (dst_len2 < src_len8) ? dst_len2 : src_len8; + uint8_t* d = dst_ptr; + const uint8_t* s = src_ptr; + size_t n = len; + + // TODO: unroll. + + while (n >= 1) { + wuffs_base__store_u16le__no_bounds_check( + d + (0 * 2), + wuffs_base__color_u32_argb_premul__as__color_u16_rgb_565( + wuffs_base__color_u64_argb_nonpremul__as__color_u32_argb_premul( + wuffs_base__load_u64le__no_bounds_check(s + (0 * 8))))); + + s += 1 * 8; + d += 1 * 2; + n -= 1; + } + + return len; +} + static uint64_t // wuffs_base__pixel_swizzler__bgr_565__bgra_nonpremul__src_over( uint8_t* dst_ptr, @@ -13964,7 +14193,7 @@ wuffs_base__pixel_swizzler__bgr_565__bgra_nonpremul__src_over( // TODO: unroll. while (n >= 1) { - // Convert from 8-bit color to 16-bit color. + // Extract 16-bit color components. uint32_t sa = 0x101 * ((uint32_t)s[3]); uint32_t sr = 0x101 * ((uint32_t)s[2]); uint32_t sg = 0x101 * ((uint32_t)s[1]); @@ -14003,6 +14232,63 @@ wuffs_base__pixel_swizzler__bgr_565__bgra_nonpremul__src_over( return len; } +static uint64_t // +wuffs_base__pixel_swizzler__bgr_565__bgra_nonpremul_4x16le__src_over( + 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_len2 = dst_len / 2; + size_t src_len8 = src_len / 8; + size_t len = (dst_len2 < src_len8) ? dst_len2 : src_len8; + uint8_t* d = dst_ptr; + const uint8_t* s = src_ptr; + size_t n = len; + + // TODO: unroll. + + while (n >= 1) { + // Extract 16-bit color components. + uint32_t sa = ((uint32_t)wuffs_base__load_u16le__no_bounds_check(s + 6)); + uint32_t sr = ((uint32_t)wuffs_base__load_u16le__no_bounds_check(s + 4)); + uint32_t sg = ((uint32_t)wuffs_base__load_u16le__no_bounds_check(s + 2)); + uint32_t sb = ((uint32_t)wuffs_base__load_u16le__no_bounds_check(s + 0)); + + // Convert from 565 color to 16-bit color. + uint32_t old_rgb_565 = wuffs_base__load_u16le__no_bounds_check(d + (0 * 2)); + uint32_t old_r5 = 0x1F & (old_rgb_565 >> 11); + uint32_t dr = (0x8421 * old_r5) >> 4; + uint32_t old_g6 = 0x3F & (old_rgb_565 >> 5); + uint32_t dg = (0x1041 * old_g6) >> 2; + uint32_t old_b5 = 0x1F & (old_rgb_565 >> 0); + uint32_t db = (0x8421 * old_b5) >> 4; + + // Calculate the inverse of the src-alpha: how much of the dst to keep. + uint32_t ia = 0xFFFF - sa; + + // Composite src (nonpremul) over dst (premul). + dr = ((sr * sa) + (dr * ia)) / 0xFFFF; + dg = ((sg * sa) + (dg * ia)) / 0xFFFF; + db = ((sb * sa) + (db * ia)) / 0xFFFF; + + // Convert from 16-bit color to 565 color and combine the components. + uint32_t new_r5 = 0x1F & (dr >> 11); + uint32_t new_g6 = 0x3F & (dg >> 10); + uint32_t new_b5 = 0x1F & (db >> 11); + uint32_t new_rgb_565 = (new_r5 << 11) | (new_g6 << 5) | (new_b5 << 0); + wuffs_base__store_u16le__no_bounds_check(d + (0 * 2), + (uint16_t)new_rgb_565); + + s += 1 * 8; + d += 1 * 2; + n -= 1; + } + + return len; +} + static uint64_t // wuffs_base__pixel_swizzler__bgr_565__y(uint8_t* dst_ptr, size_t dst_len, @@ -14148,6 +14434,37 @@ wuffs_base__pixel_swizzler__bgr__bgra_nonpremul__src(uint8_t* dst_ptr, return len; } +static uint64_t // +wuffs_base__pixel_swizzler__bgr__bgra_nonpremul_4x16le__src( + 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_len3 = dst_len / 3; + size_t src_len8 = src_len / 8; + size_t len = (dst_len3 < src_len8) ? dst_len3 : src_len8; + uint8_t* d = dst_ptr; + const uint8_t* s = src_ptr; + size_t n = len; + + // TODO: unroll. + + while (n >= 1) { + uint32_t s0 = + wuffs_base__color_u64_argb_nonpremul__as__color_u32_argb_premul( + wuffs_base__load_u64le__no_bounds_check(s + (0 * 8))); + wuffs_base__store_u24le__no_bounds_check(d + (0 * 3), s0); + + s += 1 * 8; + d += 1 * 3; + n -= 1; + } + + return len; +} + static uint64_t // wuffs_base__pixel_swizzler__bgr__bgra_nonpremul__src_over( uint8_t* dst_ptr, @@ -14166,7 +14483,7 @@ wuffs_base__pixel_swizzler__bgr__bgra_nonpremul__src_over( // TODO: unroll. while (n >= 1) { - // Convert from 8-bit color to 16-bit color. + // Extract 16-bit color components. uint32_t sa = 0x101 * ((uint32_t)s[3]); uint32_t sr = 0x101 * ((uint32_t)s[2]); uint32_t sg = 0x101 * ((uint32_t)s[1]); @@ -14196,6 +14513,54 @@ wuffs_base__pixel_swizzler__bgr__bgra_nonpremul__src_over( return len; } +static uint64_t // +wuffs_base__pixel_swizzler__bgr__bgra_nonpremul_4x16le__src_over( + 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_len3 = dst_len / 3; + size_t src_len8 = src_len / 8; + size_t len = (dst_len3 < src_len8) ? dst_len3 : src_len8; + uint8_t* d = dst_ptr; + const uint8_t* s = src_ptr; + size_t n = len; + + // TODO: unroll. + + while (n >= 1) { + // Extract 16-bit color components. + uint32_t sa = ((uint32_t)wuffs_base__load_u16le__no_bounds_check(s + 6)); + uint32_t sr = ((uint32_t)wuffs_base__load_u16le__no_bounds_check(s + 4)); + uint32_t sg = ((uint32_t)wuffs_base__load_u16le__no_bounds_check(s + 2)); + uint32_t sb = ((uint32_t)wuffs_base__load_u16le__no_bounds_check(s + 0)); + uint32_t dr = 0x101 * ((uint32_t)d[2]); + uint32_t dg = 0x101 * ((uint32_t)d[1]); + uint32_t db = 0x101 * ((uint32_t)d[0]); + + // Calculate the inverse of the src-alpha: how much of the dst to keep. + uint32_t ia = 0xFFFF - sa; + + // Composite src (nonpremul) over dst (premul). + dr = ((sr * sa) + (dr * ia)) / 0xFFFF; + dg = ((sg * sa) + (dg * ia)) / 0xFFFF; + db = ((sb * sa) + (db * ia)) / 0xFFFF; + + // Convert from 16-bit color to 8-bit color. + d[0] = (uint8_t)(db >> 8); + d[1] = (uint8_t)(dg >> 8); + d[2] = (uint8_t)(dr >> 8); + + s += 1 * 8; + d += 1 * 3; + n -= 1; + } + + return len; +} + // -------- static uint64_t // @@ -14213,8 +14578,6 @@ wuffs_base__pixel_swizzler__bgra_nonpremul__bgra_nonpremul__src_over( const uint8_t* s = src_ptr; size_t n = len; - // TODO: unroll. - while (n >= 1) { uint32_t d0 = wuffs_base__load_u32le__no_bounds_check(d + (0 * 4)); uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(s + (0 * 4)); @@ -14230,6 +14593,38 @@ wuffs_base__pixel_swizzler__bgra_nonpremul__bgra_nonpremul__src_over( return len; } +static uint64_t // +wuffs_base__pixel_swizzler__bgra_nonpremul__bgra_nonpremul_4x16le__src_over( + 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_len8 = src_len / 8; + size_t len = (dst_len4 < src_len8) ? dst_len4 : src_len8; + uint8_t* d = dst_ptr; + const uint8_t* s = src_ptr; + size_t n = len; + + while (n >= 1) { + uint64_t d0 = wuffs_base__color_u32__as__color_u64( + wuffs_base__load_u32le__no_bounds_check(d + (0 * 4))); + uint64_t s0 = wuffs_base__load_u64le__no_bounds_check(s + (0 * 8)); + wuffs_base__store_u32le__no_bounds_check( + d + (0 * 4), + wuffs_base__color_u64__as__color_u32( + wuffs_base__composite_nonpremul_nonpremul_u64_axxx(d0, s0))); + + s += 1 * 8; + d += 1 * 4; + n -= 1; + } + + return len; +} + // -------- static uint64_t // @@ -14263,6 +14658,37 @@ wuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul__src( return len; } +static uint64_t // +wuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul_4x16le__src( + 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_len8 = src_len / 8; + size_t len = (dst_len4 < src_len8) ? dst_len4 : src_len8; + uint8_t* d = dst_ptr; + const uint8_t* s = src_ptr; + size_t n = len; + + // TODO: unroll. + + while (n >= 1) { + uint64_t s0 = wuffs_base__load_u64le__no_bounds_check(s + (0 * 8)); + wuffs_base__store_u32le__no_bounds_check( + d + (0 * 4), + wuffs_base__color_u64_argb_nonpremul__as__color_u32_argb_premul(s0)); + + s += 1 * 8; + d += 1 * 4; + n -= 1; + } + + return len; +} + static uint64_t // wuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul__src_over( uint8_t* dst_ptr, @@ -14294,6 +14720,40 @@ wuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul__src_over( return len; } +static uint64_t // +wuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul_4x16le__src_over( + 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_len8 = src_len / 8; + size_t len = (dst_len4 < src_len8) ? dst_len4 : src_len8; + uint8_t* d = dst_ptr; + const uint8_t* s = src_ptr; + size_t n = len; + + // TODO: unroll. + + while (n >= 1) { + uint64_t d0 = wuffs_base__color_u32__as__color_u64( + wuffs_base__load_u32le__no_bounds_check(d + (0 * 4))); + uint64_t s0 = wuffs_base__load_u64le__no_bounds_check(s + (0 * 8)); + wuffs_base__store_u32le__no_bounds_check( + d + (0 * 4), + wuffs_base__color_u64__as__color_u32( + wuffs_base__composite_premul_nonpremul_u64_axxx(d0, s0))); + + s += 1 * 8; + d += 1 * 4; + n -= 1; + } + + return len; +} + // -------- static uint64_t // @@ -14663,8 +15123,8 @@ wuffs_base__pixel_swizzler__prepare__indexed__bgra_binary( return NULL; case WUFFS_BASE__PIXEL_FORMAT__BGR_565: - if (wuffs_base__pixel_swizzler__squash_bgr_565_888(dst_palette, - src_palette) != 1024) { + if (wuffs_base__pixel_swizzler__squash_align4_bgr_565_888( + dst_palette, src_palette) != 256) { return NULL; } switch (blend) { @@ -14705,7 +15165,7 @@ wuffs_base__pixel_swizzler__prepare__indexed__bgra_binary( case WUFFS_BASE__PIXEL_FORMAT__RGB: if (wuffs_base__pixel_swizzler__swap_rgbx_bgrx(dst_palette, - src_palette) != 1024) { + src_palette) != 256) { return NULL; } switch (blend) { @@ -14720,7 +15180,7 @@ wuffs_base__pixel_swizzler__prepare__indexed__bgra_binary( case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL: case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY: if (wuffs_base__pixel_swizzler__swap_rgbx_bgrx(dst_palette, - src_palette) != 1024) { + src_palette) != 256) { return NULL; } switch (blend) { @@ -14824,6 +15284,66 @@ wuffs_base__pixel_swizzler__prepare__bgra_nonpremul( return NULL; } +static wuffs_base__pixel_swizzler__func // +wuffs_base__pixel_swizzler__prepare__bgra_nonpremul_4x16le( + wuffs_base__pixel_swizzler* p, + wuffs_base__pixel_format dst_pixfmt, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src_palette, + wuffs_base__pixel_blend blend) { + switch (dst_pixfmt.repr) { + case WUFFS_BASE__PIXEL_FORMAT__BGR_565: + switch (blend) { + case WUFFS_BASE__PIXEL_BLEND__SRC: + return wuffs_base__pixel_swizzler__bgr_565__bgra_nonpremul_4x16le__src; + case WUFFS_BASE__PIXEL_BLEND__SRC_OVER: + return wuffs_base__pixel_swizzler__bgr_565__bgra_nonpremul_4x16le__src_over; + } + return NULL; + + case WUFFS_BASE__PIXEL_FORMAT__BGR: + switch (blend) { + case WUFFS_BASE__PIXEL_BLEND__SRC: + return wuffs_base__pixel_swizzler__bgr__bgra_nonpremul_4x16le__src; + case WUFFS_BASE__PIXEL_BLEND__SRC_OVER: + return wuffs_base__pixel_swizzler__bgr__bgra_nonpremul_4x16le__src_over; + } + return NULL; + + case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL: + switch (blend) { + case WUFFS_BASE__PIXEL_BLEND__SRC: + return wuffs_base__pixel_swizzler__squash_tight_4x8_4x16le; + case WUFFS_BASE__PIXEL_BLEND__SRC_OVER: + return wuffs_base__pixel_swizzler__bgra_nonpremul__bgra_nonpremul_4x16le__src_over; + } + return NULL; + + case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL: + switch (blend) { + case WUFFS_BASE__PIXEL_BLEND__SRC: + return wuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul_4x16le__src; + case WUFFS_BASE__PIXEL_BLEND__SRC_OVER: + return wuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul_4x16le__src_over; + } + return NULL; + + case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY: + case WUFFS_BASE__PIXEL_FORMAT__BGRX: + // TODO. + break; + + case WUFFS_BASE__PIXEL_FORMAT__RGB: + case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL: + case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL: + case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY: + case WUFFS_BASE__PIXEL_FORMAT__RGBX: + // TODO. + break; + } + return NULL; +} + // -------- WUFFS_BASE__MAYBE_STATIC wuffs_base__status // @@ -14870,6 +15390,11 @@ wuffs_base__pixel_swizzler__prepare(wuffs_base__pixel_swizzler* p, func = wuffs_base__pixel_swizzler__prepare__bgra_nonpremul( p, dst_pixfmt, dst_palette, src_palette, blend); break; + + case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL_4X16LE: + func = wuffs_base__pixel_swizzler__prepare__bgra_nonpremul_4x16le( + p, dst_pixfmt, dst_palette, src_palette, blend); + break; } p->private_impl.func = func; diff --git a/test/c/std/wbmp.c b/test/c/std/wbmp.c index eb9d3626..3308adde 100644 --- a/test/c/std/wbmp.c +++ b/test/c/std/wbmp.c @@ -130,6 +130,10 @@ test_wuffs_pixel_swizzler_swizzle() { .pixel = 0x55443300, .pixfmt_repr = WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL, }, + { + .pixel = 0x55443300, + .pixfmt_repr = WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL_4X16LE, + }, }; const struct {