From f6f2ac4206d9d90cd13f905ac84fdcfe1b7ed801 Mon Sep 17 00:00:00 2001 From: Nigel Tao Date: Sun, 9 Feb 2020 16:11:23 +1100 Subject: [PATCH] Have std/bmp support 32 bits per pixel color --- internal/cgen/base/image-impl.c | 151 +++++++++++ internal/cgen/data.go | 42 +-- release/c/wuffs-unsupported-snapshot.c | 343 +++++++++++++++++++++++-- std/bmp/decode_bmp.wuffs | 73 ++++-- 4 files changed, 553 insertions(+), 56 deletions(-) diff --git a/internal/cgen/base/image-impl.c b/internal/cgen/base/image-impl.c index 5ca466b1..2dd03397 100644 --- a/internal/cgen/base/image-impl.c +++ b/internal/cgen/base/image-impl.c @@ -29,6 +29,36 @@ wuffs_base__swap_u32_argb_abgr(uint32_t u) { return o | (r >> 16) | (b << 16); } +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. + uint32_t sa = 0x101 * (0xFF & (src_nonpremul >> 24)); + uint32_t sr = 0x101 * (0xFF & (src_nonpremul >> 16)); + uint32_t sg = 0x101 * (0xFF & (src_nonpremul >> 8)); + uint32_t sb = 0x101 * (0xFF & (src_nonpremul >> 0)); + uint32_t da = 0x101 * (0xFF & (dst_premul >> 24)); + uint32_t dr = 0x101 * (0xFF & (dst_premul >> 16)); + uint32_t dg = 0x101 * (0xFF & (dst_premul >> 8)); + uint32_t db = 0x101 * (0xFF & (dst_premul >> 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). + da = sa + ((da * ia) / 0xFFFF); + 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 and combine the components. + da >>= 8; + dr >>= 8; + dg >>= 8; + db >>= 8; + return (db << 0) | (dg << 8) | (dr << 16) | (da << 24); +} + static inline uint32_t // wuffs_base__premul_u32_axxx(uint32_t nonpremul) { // Multiplying by 0x101 (twice, once for alpha and once for color) converts @@ -228,6 +258,61 @@ wuffs_base__pixel_buffer__set_color_u32_at( // -------- +static uint64_t // +wuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul__src( + wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src) { + size_t dst_len4 = dst.len / 4; + size_t src_len4 = src.len / 4; + size_t len = dst_len4 < src_len4 ? dst_len4 : src_len4; + uint8_t* d = dst.ptr; + uint8_t* s = src.ptr; + size_t n = len; + + // TODO: unroll. + + while (n >= 1) { + uint32_t s0 = + wuffs_base__premul_u32_axxx(wuffs_base__load_u32le(s + (0 * 4))); + wuffs_base__store_u32le(d + (0 * 4), s0); + + s += 1 * 4; + d += 1 * 4; + n -= 1; + } + + return len; +} + +static uint64_t // +wuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul__src_over( + wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src) { + size_t dst_len4 = dst.len / 4; + size_t src_len4 = src.len / 4; + size_t len = dst_len4 < src_len4 ? dst_len4 : src_len4; + uint8_t* d = dst.ptr; + uint8_t* s = src.ptr; + size_t n = len; + + // TODO: unroll. + + while (n >= 1) { + uint32_t d0 = wuffs_base__load_u32le(d + (0 * 4)); + uint32_t s0 = wuffs_base__load_u32le(s + (0 * 4)); + wuffs_base__store_u32le( + d + (0 * 4), wuffs_base__composite_premul_nonpremul_u32_axxx(d0, s0)); + + s += 1 * 4; + d += 1 * 4; + n -= 1; + } + + return len; +} + static uint64_t // wuffs_base__pixel_swizzler__copy_1_1(wuffs_base__slice_u8 dst, wuffs_base__slice_u8 dst_palette, @@ -235,6 +320,19 @@ wuffs_base__pixel_swizzler__copy_1_1(wuffs_base__slice_u8 dst, return wuffs_base__slice_u8__copy_from_slice(dst, src); } +static uint64_t // +wuffs_base__pixel_swizzler__copy_4_4(wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src) { + size_t dst_len4 = dst.len / 4; + size_t src_len4 = src.len / 4; + size_t len = dst_len4 < src_len4 ? dst_len4 : src_len4; + if (len > 0) { + memmove(dst.ptr, src.ptr, len * 4); + } + return len; +} + static uint64_t // wuffs_base__pixel_swizzler__xx__index__src(wuffs_base__slice_u8 dst, wuffs_base__slice_u8 dst_palette, @@ -737,6 +835,54 @@ wuffs_base__pixel_swizzler__prepare__bgr(wuffs_base__pixel_swizzler* p, return NULL; } +static wuffs_base__pixel_swizzler__func // +wuffs_base__pixel_swizzler__prepare__bgra_nonpremul( + wuffs_base__pixel_swizzler* p, + wuffs_base__pixel_format dst_format, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src_palette, + wuffs_base__pixel_blend blend) { + switch (dst_format.repr) { + case WUFFS_BASE__PIXEL_FORMAT__BGR_565: + // TODO. + break; + + case WUFFS_BASE__PIXEL_FORMAT__BGR: + case WUFFS_BASE__PIXEL_FORMAT__RGB: + // TODO. + break; + + case WUFFS_BASE__PIXEL_FORMAT__BGRX: + case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL: + switch (blend) { + case WUFFS_BASE__PIXEL_BLEND__SRC: + return wuffs_base__pixel_swizzler__copy_4_4; + } + 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__src; + case WUFFS_BASE__PIXEL_BLEND__SRC_OVER: + return wuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul__src_over; + } + return NULL; + + case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY: + // TODO. + break; + + case WUFFS_BASE__PIXEL_FORMAT__RGBX: + case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL: + case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL: + case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY: + // TODO. + break; + } + return NULL; +} + // -------- wuffs_base__status // @@ -769,6 +915,11 @@ wuffs_base__pixel_swizzler__prepare(wuffs_base__pixel_swizzler* p, func = wuffs_base__pixel_swizzler__prepare__bgr( p, dst_format, dst_palette, src_palette, blend); break; + + case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL: + func = wuffs_base__pixel_swizzler__prepare__bgra_nonpremul( + p, dst_format, dst_palette, src_palette, blend); + break; } p->private_impl.func = func; diff --git a/internal/cgen/data.go b/internal/cgen/data.go index 8eec91a4..5f879232 100644 --- a/internal/cgen/data.go +++ b/internal/cgen/data.go @@ -26,24 +26,27 @@ const baseAllImplC = "" + "" const baseImageImplC = "" + - "// ---------------- Images\n\nconst uint32_t wuffs_base__pixel_format__bits_per_channel[16] = {\n 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07,\n 0x08, 0x0A, 0x0C, 0x10, 0x18, 0x20, 0x30, 0x40,\n};\n\nstatic inline uint32_t //\nwuffs_base__swap_u32_argb_abgr(uint32_t u) {\n uint32_t o = u & 0xFF00FF00;\n uint32_t r = u & 0x00FF0000;\n uint32_t b = u & 0x000000FF;\n return o | (r >> 16) | (b << 16);\n}\n\nstatic inline uint32_t //\nwuffs_base__premul_u32_axxx(uint32_t nonpremul) {\n // Multiplying by 0x101 (twice, once for alpha and once 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 & (nonpremul >> 24);\n uint32_t a16 = a * (0x101 * 0x101);" + - "\n\n uint32_t r = 0xFF & (nonpremul >> 16);\n r = ((r * a16) / 0xFFFF) >> 8;\n uint32_t g = 0xFF & (nonpremul >> 8);\n g = ((g * a16) / 0xFFFF) >> 8;\n uint32_t b = 0xFF & (nonpremul >> 0);\n b = ((b * a16) / 0xFFFF) >> 8;\n\n return (a << 24) | (r << 16) | (g << 8) | (b << 0);\n}\n\nstatic inline uint32_t //\nwuffs_base__nonpremul_u32_axxx(uint32_t premul) {\n uint32_t a = 0xFF & (premul >> 24);\n if (a == 0xFF) {\n return premul;\n } else if (a == 0) {\n return 0;\n }\n uint32_t a16 = a * 0x101;\n\n uint32_t r = 0xFF & (premul >> 16);\n r = ((r * (0x101 * 0xFFFF)) / a16) >> 8;\n uint32_t g = 0xFF & (premul >> 8);\n g = ((g * (0x101 * 0xFFFF)) / a16) >> 8;\n uint32_t b = 0xFF & (premul >> 0);\n b = ((b * (0x101 * 0xFFFF)) / a16) >> 8;\n\n return (a << 24) | (r << 16) | (g << 8) | (b << 0);\n}\n\nwuffs_base__color_u32_argb_premul //\nwuffs_base__pixel_buffer__color_u32_at(const wuffs_base__pixel_buffer* b,\n uint32_t x,\n uint32_t y) {\n if (" + - "!b || (x >= b->pixcfg.private_impl.width) ||\n (y >= b->pixcfg.private_impl.height)) {\n return 0;\n }\n\n if (wuffs_base__pixel_format__is_planar(&b->pixcfg.private_impl.pixfmt)) {\n // TODO: support planar formats.\n return 0;\n }\n\n size_t stride = b->private_impl.planes[0].stride;\n uint8_t* row = b->private_impl.planes[0].ptr + (stride * ((size_t)y));\n\n switch (b->pixcfg.private_impl.pixfmt.repr) {\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL:\n WUFFS_BASE__FALLTHROUGH;\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY:\n return wuffs_base__load_u32le(row + (4 * ((size_t)x)));\n\n case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_PREMUL:\n WUFFS_BASE__FALLTHROUGH;\n case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_BINARY: {\n uint8_t* palette = b->private_impl.planes[3].ptr;\n return wuffs_base__load_u32le(palette + (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 * ((uin" + - "t32_t)(row[x])));\n\n case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_NONPREMUL: {\n uint8_t* palette = b->private_impl.planes[3].ptr;\n return wuffs_base__premul_u32_axxx(\n wuffs_base__load_u32le(palette + (4 * ((size_t)row[x]))));\n }\n\n case WUFFS_BASE__PIXEL_FORMAT__BGR_565: {\n uint16_t bgr = wuffs_base__load_u16le(row + (2 * ((size_t)x)));\n uint32_t b5 = 0x1F & (bgr >> 0);\n uint32_t b = (b5 << 3) | (b5 >> 2);\n uint32_t g6 = 0x3F & (bgr >> 5);\n uint32_t g = (g6 << 2) | (g6 >> 4);\n uint32_t r5 = 0x1F & (bgr >> 11);\n uint32_t r = (r5 << 3) | (r5 >> 2);\n return 0xFF000000 | (r << 16) | (g << 8) | (b << 0);\n }\n case WUFFS_BASE__PIXEL_FORMAT__BGR:\n return 0xFF000000 | wuffs_base__load_u24le(row + (3 * ((size_t)x)));\n case WUFFS_BASE__PIXEL_FORMAT__BGRX:\n return 0xFF000000 | wuffs_base__load_u32le(row + (4 * ((size_t)x)));\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL:\n return wuffs_base__premul_u32_axxx(\n wuffs_base" + - "__load_u32le(row + (4 * ((size_t)x))));\n\n case WUFFS_BASE__PIXEL_FORMAT__RGB:\n return wuffs_base__swap_u32_argb_abgr(\n 0xFF000000 | wuffs_base__load_u24le(row + (3 * ((size_t)x))));\n case WUFFS_BASE__PIXEL_FORMAT__RGBX:\n return wuffs_base__swap_u32_argb_abgr(\n 0xFF000000 | wuffs_base__load_u32le(row + (4 * ((size_t)x))));\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL:\n return wuffs_base__swap_u32_argb_abgr(wuffs_base__premul_u32_axxx(\n wuffs_base__load_u32le(row + (4 * ((size_t)x)))));\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL:\n WUFFS_BASE__FALLTHROUGH;\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY:\n return wuffs_base__swap_u32_argb_abgr(\n wuffs_base__load_u32le(row + (4 * ((size_t)x))));\n\n default:\n // TODO: support more formats.\n break;\n }\n\n return 0;\n}\n\nwuffs_base__status //\nwuffs_base__pixel_buffer__set_color_u32_at(\n wuffs_base__pixel_buffer* b,\n uint32_t x,\n uint32_t y,\n wuffs_base__color_u32_argb_pr" + - "emul color) {\n if (!b) {\n return wuffs_base__make_status(wuffs_base__error__bad_receiver);\n }\n if ((x >= b->pixcfg.private_impl.width) ||\n (y >= b->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(&b->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 = b->private_impl.planes[0].stride;\n uint8_t* row = b->private_impl.planes[0].ptr + (stride * ((size_t)y));\n\n switch (b->pixcfg.private_impl.pixfmt.repr) {\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL:\n WUFFS_BASE__FALLTHROUGH;\n case WUFFS_BASE__PIXEL_FORMAT__BGRX:\n wuffs_base__store_u32le(row + (4 * ((size_t)x)), color);\n break;\n\n // Common formats above. Rarer formats below.\n\n case WUFFS_BASE__PIXEL_FORMAT__BGR_565: {\n uint32_t b5 = 0x1F & (color >> (8 - 5));\n uint32_t g6 = 0x3F & (color >> (16 - 6));\n " + - " uint32_t r5 = 0x1F & (color >> (24 - 5));\n uint32_t bgr565 = (b5 << 0) | (g6 << 5) | (r5 << 11);\n wuffs_base__store_u16le(row + (2 * ((size_t)x)), (uint16_t)bgr565);\n break;\n }\n case WUFFS_BASE__PIXEL_FORMAT__BGR:\n wuffs_base__store_u24le(row + (3 * ((size_t)x)), color);\n break;\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL:\n wuffs_base__store_u32le(row + (4 * ((size_t)x)),\n wuffs_base__nonpremul_u32_axxx(color));\n break;\n\n case WUFFS_BASE__PIXEL_FORMAT__RGB:\n wuffs_base__store_u24le(row + (3 * ((size_t)x)),\n wuffs_base__swap_u32_argb_abgr(color));\n break;\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL:\n wuffs_base__store_u32le(row + (4 * ((size_t)x)),\n wuffs_base__nonpremul_u32_axxx(\n wuffs_base__swap_u32_argb_abgr(color)));\n break;\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL:\n WUFFS_BASE__FALLTHROUGH;\n c" + - "ase WUFFS_BASE__PIXEL_FORMAT__RGBX:\n wuffs_base__store_u32le(row + (4 * ((size_t)x)),\n 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" + + "// ---------------- Images\n\nconst uint32_t wuffs_base__pixel_format__bits_per_channel[16] = {\n 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07,\n 0x08, 0x0A, 0x0C, 0x10, 0x18, 0x20, 0x30, 0x40,\n};\n\nstatic inline uint32_t //\nwuffs_base__swap_u32_argb_abgr(uint32_t u) {\n uint32_t o = u & 0xFF00FF00;\n uint32_t r = u & 0x00FF0000;\n uint32_t b = u & 0x000000FF;\n return o | (r >> 16) | (b << 16);\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__premul_u32_axxx(uint32_t nonpremul) {\n // Multiplying by 0x101 (twice, once for alpha and once 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 * 0" + + "x8181) / 0xFFFF) >> 8 = 0x4101 >> 8 = 0x41\n uint32_t a = 0xFF & (nonpremul >> 24);\n uint32_t a16 = a * (0x101 * 0x101);\n\n uint32_t r = 0xFF & (nonpremul >> 16);\n r = ((r * a16) / 0xFFFF) >> 8;\n uint32_t g = 0xFF & (nonpremul >> 8);\n g = ((g * a16) / 0xFFFF) >> 8;\n uint32_t b = 0xFF & (nonpremul >> 0);\n b = ((b * a16) / 0xFFFF) >> 8;\n\n return (a << 24) | (r << 16) | (g << 8) | (b << 0);\n}\n\nstatic inline uint32_t //\nwuffs_base__nonpremul_u32_axxx(uint32_t premul) {\n uint32_t a = 0xFF & (premul >> 24);\n if (a == 0xFF) {\n return premul;\n } else if (a == 0) {\n return 0;\n }\n uint32_t a16 = a * 0x101;\n\n uint32_t r = 0xFF & (premul >> 16);\n r = ((r * (0x101 * 0xFFFF)) / a16) >> 8;\n uint32_t g = 0xFF & (premul >> 8);\n g = ((g * (0x101 * 0xFFFF)) / a16) >> 8;\n uint32_t b = 0xFF & (premul >> 0);\n b = ((b * (0x101 * 0xFFFF)) / a16) >> 8;\n\n return (a << 24) | (r << 16) | (g << 8) | (b << 0);\n}\n\nwuffs_base__color_u32_argb_premul //\nwuffs_base__pixel_buffer__color_u32_at(const wuffs_base__pixel_" + + "buffer* b,\n uint32_t x,\n uint32_t y) {\n if (!b || (x >= b->pixcfg.private_impl.width) ||\n (y >= b->pixcfg.private_impl.height)) {\n return 0;\n }\n\n if (wuffs_base__pixel_format__is_planar(&b->pixcfg.private_impl.pixfmt)) {\n // TODO: support planar formats.\n return 0;\n }\n\n size_t stride = b->private_impl.planes[0].stride;\n uint8_t* row = b->private_impl.planes[0].ptr + (stride * ((size_t)y));\n\n switch (b->pixcfg.private_impl.pixfmt.repr) {\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL:\n WUFFS_BASE__FALLTHROUGH;\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY:\n return wuffs_base__load_u32le(row + (4 * ((size_t)x)));\n\n case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_PREMUL:\n WUFFS_BASE__FALLTHROUGH;\n case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_BINARY: {\n uint8_t* palette = b->private_impl.planes[3].ptr;\n return wuffs_base__load_u32le(palette + (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 = b->private_impl.planes[3].ptr;\n return wuffs_base__premul_u32_axxx(\n wuffs_base__load_u32le(palette + (4 * ((size_t)row[x]))));\n }\n\n case WUFFS_BASE__PIXEL_FORMAT__BGR_565: {\n uint16_t bgr = wuffs_base__load_u16le(row + (2 * ((size_t)x)));\n uint32_t b5 = 0x1F & (bgr >> 0);\n uint32_t b = (b5 << 3) | (b5 >> 2);\n uint32_t g6 = 0x3F & (bgr >> 5);\n uint32_t g = (g6 << 2) | (g6 >> 4);\n uint32_t r5 = 0x1F & (bgr >> 11);\n uint32_t r = (r5 << 3) | (r5 >> 2);\n return 0xFF000000 | (r << 16) | (g << 8) | (b << 0);\n }\n case WUFFS_BASE__PIXEL_FORMAT__BGR:\n return 0xFF000000 | wuffs_base__load_u24le(row + (3 * ((size_t)x)));\n case WUFFS_BASE__PIXEL_FORMAT__BGRX:\n return 0xFF000000 | wuffs_base__load_u32le(row + (4 * ((size_" + + "t)x)));\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL:\n return wuffs_base__premul_u32_axxx(\n wuffs_base__load_u32le(row + (4 * ((size_t)x))));\n\n case WUFFS_BASE__PIXEL_FORMAT__RGB:\n return wuffs_base__swap_u32_argb_abgr(\n 0xFF000000 | wuffs_base__load_u24le(row + (3 * ((size_t)x))));\n case WUFFS_BASE__PIXEL_FORMAT__RGBX:\n return wuffs_base__swap_u32_argb_abgr(\n 0xFF000000 | wuffs_base__load_u32le(row + (4 * ((size_t)x))));\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL:\n return wuffs_base__swap_u32_argb_abgr(wuffs_base__premul_u32_axxx(\n wuffs_base__load_u32le(row + (4 * ((size_t)x)))));\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL:\n WUFFS_BASE__FALLTHROUGH;\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY:\n return wuffs_base__swap_u32_argb_abgr(\n wuffs_base__load_u32le(row + (4 * ((size_t)x))));\n\n default:\n // TODO: support more formats.\n break;\n }\n\n return 0;\n}\n\nwuffs_base__status //\nwuffs_base__pixel_buf" + + "fer__set_color_u32_at(\n wuffs_base__pixel_buffer* b,\n uint32_t x,\n uint32_t y,\n wuffs_base__color_u32_argb_premul color) {\n if (!b) {\n return wuffs_base__make_status(wuffs_base__error__bad_receiver);\n }\n if ((x >= b->pixcfg.private_impl.width) ||\n (y >= b->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(&b->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 = b->private_impl.planes[0].stride;\n uint8_t* row = b->private_impl.planes[0].ptr + (stride * ((size_t)y));\n\n switch (b->pixcfg.private_impl.pixfmt.repr) {\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL:\n WUFFS_BASE__FALLTHROUGH;\n case WUFFS_BASE__PIXEL_FORMAT__BGRX:\n wuffs_base__store_u32le(row + (4 * ((size_t)x)), color);\n break;\n\n // Common formats above. Rarer formats below.\n\n case WUFFS_BASE__" + + "PIXEL_FORMAT__BGR_565: {\n uint32_t b5 = 0x1F & (color >> (8 - 5));\n uint32_t g6 = 0x3F & (color >> (16 - 6));\n uint32_t r5 = 0x1F & (color >> (24 - 5));\n uint32_t bgr565 = (b5 << 0) | (g6 << 5) | (r5 << 11);\n wuffs_base__store_u16le(row + (2 * ((size_t)x)), (uint16_t)bgr565);\n break;\n }\n case WUFFS_BASE__PIXEL_FORMAT__BGR:\n wuffs_base__store_u24le(row + (3 * ((size_t)x)), color);\n break;\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL:\n wuffs_base__store_u32le(row + (4 * ((size_t)x)),\n wuffs_base__nonpremul_u32_axxx(color));\n break;\n\n case WUFFS_BASE__PIXEL_FORMAT__RGB:\n wuffs_base__store_u24le(row + (3 * ((size_t)x)),\n wuffs_base__swap_u32_argb_abgr(color));\n break;\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL:\n wuffs_base__store_u32le(row + (4 * ((size_t)x)),\n wuffs_base__nonpremul_u32_axxx(\n wuffs_base__swap_" + + "u32_argb_abgr(color)));\n break;\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL:\n WUFFS_BASE__FALLTHROUGH;\n case WUFFS_BASE__PIXEL_FORMAT__RGBX:\n wuffs_base__store_u32le(row + (4 * ((size_t)x)),\n 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\nstatic uint64_t //\nwuffs_base__pixel_swizzler__copy_1_1(wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n return wuffs_base__slice_u8__copy_from_slice(dst, src);\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__xx__index__src(wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\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 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(\n d + (0 * 2),\n wuffs_base__load_u16le(dst_palette.ptr + ((size_t)s[0] * 4)));\n wuffs_base__store_u16le(\n d + (1 * 2),\n wuffs_base__load_u16le(dst_palette" + - ".ptr + ((size_t)s[1] * 4)));\n wuffs_base__store_u16le(\n d + (2 * 2),\n wuffs_base__load_u16le(dst_palette.ptr + ((size_t)s[2] * 4)));\n wuffs_base__store_u16le(\n d + (3 * 2),\n wuffs_base__load_u16le(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(\n d + (0 * 2),\n wuffs_base__load_u16le(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__xxx__index__src(wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\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 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 will\n // overwrite that with the correct byte. There is always another store,\n // whether a 4-byte store in this loop or a 1-byte store in the next loop.\n while (n > loop_unroll_count) {\n wuffs_base__store_u32le(\n d + (0 * 3),\n wuffs_base__load_u32le(dst_palette.ptr + ((size_t)s[0] * 4)));\n wuffs_base__store_u32le(\n d + (1 * 3),\n wuffs_base__load_u32le(dst_palette.ptr + ((size_t)s[1] * 4)));\n wuffs_base__store_u32le(\n d + (2 * 3),\n wuffs_base__load_u32le(dst_palette.ptr + ((size_t)s[2] * 4)));\n wuffs_base__store_u32le(\n d + (3 * 3),\n wuffs_base__load_u32le(dst_palette.ptr + ((size_t)s[3] * 4)));\n\n s += loop_unroll_count * 1;\n d += loop_unroll_count * 3;\n n -= loop_unrol" + - "l_count;\n }\n\n while (n >= 1) {\n uint32_t s0 = wuffs_base__load_u32le(dst_palette.ptr + ((size_t)s[0] * 4));\n d[0] = (uint8_t)(s0 >> 0);\n d[1] = (uint8_t)(s0 >> 8);\n d[2] = (uint8_t)(s0 >> 16);\n\n s += 1 * 1;\n d += 1 * 3;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__xxx__index_binary_alpha__src_over(\n wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\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 uint8_t* s = src.ptr;\n size_t n = len;\n\n const size_t loop_unroll_count = 4;\n\n while (n >= loop_unroll_count) {\n uint32_t s0 = wuffs_base__load_u32le(dst_palette.ptr + ((size_t)s[0] * 4));\n if (s0) {\n wuffs_base__store_u24le(d + (0 * 4), s0);\n }\n uint32_t s1 = wuffs_base__load_u32le(dst_palette.ptr + ((size_t)s[1] * 4));\n if (s1) {\n wuffs_base__store_u24le(d + (1 * 4), s" + - "1);\n }\n uint32_t s2 = wuffs_base__load_u32le(dst_palette.ptr + ((size_t)s[2] * 4));\n if (s2) {\n wuffs_base__store_u24le(d + (2 * 4), s2);\n }\n uint32_t s3 = wuffs_base__load_u32le(dst_palette.ptr + ((size_t)s[3] * 4));\n if (s3) {\n wuffs_base__store_u24le(d + (3 * 4), s3);\n }\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(dst_palette.ptr + ((size_t)s[0] * 4));\n if (s0) {\n wuffs_base__store_u24le(d + (0 * 4), s0);\n }\n\n s += 1 * 1;\n d += 1 * 3;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__xxxx__index__src(wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n if (dst_palette.len != 1024) {\n return 0;\n }\n size_t dst_len4 = dst.len / 4;\n size_t len = dst_len4 < src.len ? dst_len4 : src.len;\n u" + - "int8_t* d = dst.ptr;\n 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_u32le(\n d + (0 * 4),\n wuffs_base__load_u32le(dst_palette.ptr + ((size_t)s[0] * 4)));\n wuffs_base__store_u32le(\n d + (1 * 4),\n wuffs_base__load_u32le(dst_palette.ptr + ((size_t)s[1] * 4)));\n wuffs_base__store_u32le(\n d + (2 * 4),\n wuffs_base__load_u32le(dst_palette.ptr + ((size_t)s[2] * 4)));\n wuffs_base__store_u32le(\n d + (3 * 4),\n wuffs_base__load_u32le(dst_palette.ptr + ((size_t)s[3] * 4)));\n\n s += loop_unroll_count * 1;\n d += loop_unroll_count * 4;\n n -= loop_unroll_count;\n }\n\n while (n >= 1) {\n wuffs_base__store_u32le(\n d + (0 * 4),\n wuffs_base__load_u32le(dst_palette.ptr + ((size_t)s[0] * 4)));\n\n s += 1 * 1;\n d += 1 * 4;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__xxxx__index_binary_alpha__src_over(\n wuffs_bas" + - "e__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n if (dst_palette.len != 1024) {\n return 0;\n }\n size_t dst_len4 = dst.len / 4;\n size_t len = dst_len4 < src.len ? dst_len4 : src.len;\n uint8_t* d = dst.ptr;\n uint8_t* s = src.ptr;\n size_t n = len;\n\n const size_t loop_unroll_count = 4;\n\n while (n >= loop_unroll_count) {\n uint32_t s0 = wuffs_base__load_u32le(dst_palette.ptr + ((size_t)s[0] * 4));\n if (s0) {\n wuffs_base__store_u32le(d + (0 * 4), s0);\n }\n uint32_t s1 = wuffs_base__load_u32le(dst_palette.ptr + ((size_t)s[1] * 4));\n if (s1) {\n wuffs_base__store_u32le(d + (1 * 4), s1);\n }\n uint32_t s2 = wuffs_base__load_u32le(dst_palette.ptr + ((size_t)s[2] * 4));\n if (s2) {\n wuffs_base__store_u32le(d + (2 * 4), s2);\n }\n uint32_t s3 = wuffs_base__load_u32le(dst_palette.ptr + ((size_t)s[3] * 4));\n if (s3) {\n wuffs_base__store_u32le(d + (3 * 4), s3);\n }\n\n s += loop_unroll_count * 1;\n d += loop_unroll_count * 4;" + - "\n n -= loop_unroll_count;\n }\n\n while (n >= 1) {\n uint32_t s0 = wuffs_base__load_u32le(dst_palette.ptr + ((size_t)s[0] * 4));\n if (s0) {\n wuffs_base__store_u32le(d + (0 * 4), s0);\n }\n\n s += 1 * 1;\n d += 1 * 4;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__xxxx__xxx(wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n size_t dst_len4 = dst.len / 4;\n size_t src_len3 = src.len / 3;\n size_t len = dst_len4 < src_len3 ? dst_len4 : src_len3;\n uint8_t* d = dst.ptr;\n uint8_t* s = src.ptr;\n size_t n = len;\n\n // TODO: unroll.\n\n while (n >= 1) {\n wuffs_base__store_u32le(d + (0 * 4),\n 0xFF000000 | wuffs_base__load_u24le(s + (0 * 3)));\n\n s += 1 * 3;\n d += 1 * 4;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__xxxx__y(wuffs_base__slice_u8 dst,\n " + - " wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n size_t dst_len4 = dst.len / 4;\n size_t len = dst_len4 < src.len ? dst_len4 : src.len;\n uint8_t* d = dst.ptr;\n uint8_t* s = src.ptr;\n size_t n = len;\n\n // TODO: unroll.\n\n while (n >= 1) {\n wuffs_base__store_u32le(d + (0 * 4),\n 0xFF000000 | (0x010101 * (uint32_t)s[0]));\n\n s += 1 * 1;\n d += 1 * 4;\n n -= 1;\n }\n\n return len;\n}\n\n" + + "// --------\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul__src(\n wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\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 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__premul_u32_axxx(wuffs_base__load_u32le(s + (0 * 4)));\n wuffs_base__store_u32le(d + (0 * 4), 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 wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\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 uint8_t* s = src.ptr;\n size_t n = len;\n\n // TODO: unrol" + + "l.\n\n while (n >= 1) {\n uint32_t d0 = wuffs_base__load_u32le(d + (0 * 4));\n uint32_t s0 = wuffs_base__load_u32le(s + (0 * 4));\n wuffs_base__store_u32le(\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__copy_1_1(wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n return wuffs_base__slice_u8__copy_from_slice(dst, src);\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__copy_4_4(wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\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\nstatic uint64_t" + + " //\nwuffs_base__pixel_swizzler__xx__index__src(wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\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 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(\n d + (0 * 2),\n wuffs_base__load_u16le(dst_palette.ptr + ((size_t)s[0] * 4)));\n wuffs_base__store_u16le(\n d + (1 * 2),\n wuffs_base__load_u16le(dst_palette.ptr + ((size_t)s[1] * 4)));\n wuffs_base__store_u16le(\n d + (2 * 2),\n wuffs_base__load_u16le(dst_palette.ptr + ((size_t)s[2] * 4)));\n wuffs_base__store_u16le(\n d + (3 * 2),\n wuffs_base__load_u16le(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(\n d + (0 * 2),\n wuffs_base__load_u16le(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__xxx__index__src(wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\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 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 will\n // overwrite that with the correct " + + "byte. There is always another store,\n // whether a 4-byte store in this loop or a 1-byte store in the next loop.\n while (n > loop_unroll_count) {\n wuffs_base__store_u32le(\n d + (0 * 3),\n wuffs_base__load_u32le(dst_palette.ptr + ((size_t)s[0] * 4)));\n wuffs_base__store_u32le(\n d + (1 * 3),\n wuffs_base__load_u32le(dst_palette.ptr + ((size_t)s[1] * 4)));\n wuffs_base__store_u32le(\n d + (2 * 3),\n wuffs_base__load_u32le(dst_palette.ptr + ((size_t)s[2] * 4)));\n wuffs_base__store_u32le(\n d + (3 * 3),\n wuffs_base__load_u32le(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(dst_palette.ptr + ((size_t)s[0] * 4));\n d[0] = (uint8_t)(s0 >> 0);\n d[1] = (uint8_t)(s0 >> 8);\n d[2] = (uint8_t)(s0 >> 16);\n\n s += 1 * 1;\n d += 1 * 3;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pi" + + "xel_swizzler__xxx__index_binary_alpha__src_over(\n wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\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 uint8_t* s = src.ptr;\n size_t n = len;\n\n const size_t loop_unroll_count = 4;\n\n while (n >= loop_unroll_count) {\n uint32_t s0 = wuffs_base__load_u32le(dst_palette.ptr + ((size_t)s[0] * 4));\n if (s0) {\n wuffs_base__store_u24le(d + (0 * 4), s0);\n }\n uint32_t s1 = wuffs_base__load_u32le(dst_palette.ptr + ((size_t)s[1] * 4));\n if (s1) {\n wuffs_base__store_u24le(d + (1 * 4), s1);\n }\n uint32_t s2 = wuffs_base__load_u32le(dst_palette.ptr + ((size_t)s[2] * 4));\n if (s2) {\n wuffs_base__store_u24le(d + (2 * 4), s2);\n }\n uint32_t s3 = wuffs_base__load_u32le(dst_palette.ptr + ((size_t)s[3] * 4));\n if (s3) {\n wuffs_base__store_u24le(d + (3 * 4), s3);\n }\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(dst_palette.ptr + ((size_t)s[0] * 4));\n if (s0) {\n wuffs_base__store_u24le(d + (0 * 4), s0);\n }\n\n s += 1 * 1;\n d += 1 * 3;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__xxxx__index__src(wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n if (dst_palette.len != 1024) {\n return 0;\n }\n size_t dst_len4 = dst.len / 4;\n size_t len = dst_len4 < src.len ? dst_len4 : src.len;\n uint8_t* d = dst.ptr;\n 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_u32le(\n d + (0 * 4),\n wuffs_base__load_u32le(dst_palette.ptr + ((size_t)s[0] * 4)));\n wuffs_base__store_u32le(\n d + (1 * 4)," + + "\n wuffs_base__load_u32le(dst_palette.ptr + ((size_t)s[1] * 4)));\n wuffs_base__store_u32le(\n d + (2 * 4),\n wuffs_base__load_u32le(dst_palette.ptr + ((size_t)s[2] * 4)));\n wuffs_base__store_u32le(\n d + (3 * 4),\n wuffs_base__load_u32le(dst_palette.ptr + ((size_t)s[3] * 4)));\n\n s += loop_unroll_count * 1;\n d += loop_unroll_count * 4;\n n -= loop_unroll_count;\n }\n\n while (n >= 1) {\n wuffs_base__store_u32le(\n d + (0 * 4),\n wuffs_base__load_u32le(dst_palette.ptr + ((size_t)s[0] * 4)));\n\n s += 1 * 1;\n d += 1 * 4;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__xxxx__index_binary_alpha__src_over(\n wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n if (dst_palette.len != 1024) {\n return 0;\n }\n size_t dst_len4 = dst.len / 4;\n size_t len = dst_len4 < src.len ? dst_len4 : src.len;\n uint8_t* d = dst.ptr;\n uint8_t* s = src.ptr;\n size_t n = len;\n\n const size_t" + + " loop_unroll_count = 4;\n\n while (n >= loop_unroll_count) {\n uint32_t s0 = wuffs_base__load_u32le(dst_palette.ptr + ((size_t)s[0] * 4));\n if (s0) {\n wuffs_base__store_u32le(d + (0 * 4), s0);\n }\n uint32_t s1 = wuffs_base__load_u32le(dst_palette.ptr + ((size_t)s[1] * 4));\n if (s1) {\n wuffs_base__store_u32le(d + (1 * 4), s1);\n }\n uint32_t s2 = wuffs_base__load_u32le(dst_palette.ptr + ((size_t)s[2] * 4));\n if (s2) {\n wuffs_base__store_u32le(d + (2 * 4), s2);\n }\n uint32_t s3 = wuffs_base__load_u32le(dst_palette.ptr + ((size_t)s[3] * 4));\n if (s3) {\n wuffs_base__store_u32le(d + (3 * 4), s3);\n }\n\n s += loop_unroll_count * 1;\n d += loop_unroll_count * 4;\n n -= loop_unroll_count;\n }\n\n while (n >= 1) {\n uint32_t s0 = wuffs_base__load_u32le(dst_palette.ptr + ((size_t)s[0] * 4));\n if (s0) {\n wuffs_base__store_u32le(d + (0 * 4), s0);\n }\n\n s += 1 * 1;\n d += 1 * 4;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_sw" + + "izzler__xxxx__xxx(wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n size_t dst_len4 = dst.len / 4;\n size_t src_len3 = src.len / 3;\n size_t len = dst_len4 < src_len3 ? dst_len4 : src_len3;\n uint8_t* d = dst.ptr;\n uint8_t* s = src.ptr;\n size_t n = len;\n\n // TODO: unroll.\n\n while (n >= 1) {\n wuffs_base__store_u32le(d + (0 * 4),\n 0xFF000000 | wuffs_base__load_u24le(s + (0 * 3)));\n\n s += 1 * 3;\n d += 1 * 4;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__xxxx__y(wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src) {\n size_t dst_len4 = dst.len / 4;\n size_t len = dst_len4 < src.len ? dst_len4 : src.len;\n uint8_t* d = dst.ptr;\n uint8_t* s = src.ptr;\n size_t n = len;\n\n // TODO: unroll.\n\n while (n >= 1) {\n wuffs_base" + + "__store_u32le(d + (0 * 4),\n 0xFF000000 | (0x010101 * (uint32_t)s[0]));\n\n s += 1 * 1;\n d += 1 * 4;\n n -= 1;\n }\n\n return len;\n}\n\n" + "" + "// --------\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__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 uint8_t* s = src.ptr;\n\n size_t n = len4;\n while (n--) {\n uint32_t argb = wuffs_base__load_u32le(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 wuffs_base__store_u32le(d, (b5 << 0) | (g6 << 5) | (r5 << 11));\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 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" + @@ -53,10 +56,11 @@ const baseImageImplC = "" + " return NULL;\n }\n switch (blend) {\n case WUFFS_BASE__PIXEL_BLEND__SRC:\n return wuffs_base__pixel_swizzler__xx__index__src;\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__sr" + "c;\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__P" + "IXEL_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_format,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src_palette,\n wuffs_base__pixel_blend blend) {\n switch (dst_format.repr) {\n case WUFFS_BASE__PIXEL_FORMAT__BGR_565:\n // TODO.\n break;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGR:\n case WUFFS_BASE__PIXEL_FORMAT__RGB:\n // TODO.\n break;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGRX:\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMU" + - "L:\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY:\n return wuffs_base__pixel_swizzler__xxxx__xxx;\n\n case WUFFS_BASE__PIXEL_FORMAT__RGBX:\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 // TODO.\n break;\n }\n return NULL;\n}\n\n" + + "L:\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY:\n return wuffs_base__pixel_swizzler__xxxx__xxx;\n\n case WUFFS_BASE__PIXEL_FORMAT__RGBX:\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 // 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_format,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__slice_u8 src_palette,\n wuffs_base__pixel_blend blend) {\n switch (dst_format.repr) {\n case WUFFS_BASE__PIXEL_FORMAT__BGR_565:\n // TODO.\n break;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGR:\n case WUFFS_BASE__PIXEL_FORMAT__RGB:\n // TODO.\n break;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGRX:\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 }\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 // TODO.\n break;\n\n case WUFFS_BASE__PIXEL_FORMAT__RGBX:\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 // TODO.\n break;\n }\n return NULL;\n}\n\n" + "" + "// --------\n\nwuffs_base__status //\nwuffs_base__pixel_swizzler__prepare(wuffs_base__pixel_swizzler* p,\n wuffs_base__pixel_format dst_format,\n wuffs_base__slice_u8 dst_palette,\n wuffs_base__pixel_format src_format,\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\n // TODO: support many more formats.\n\n wuffs_base__pixel_swizzler__func func = NULL;\n\n switch (src_format.repr) {\n case WUFFS_BASE__PIXEL_FORMAT__Y:\n func = wuffs_base__pixel_swizzler__prepare__y(p, dst_format, 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_format, 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_format, dst_palette, src_palette, blend);\n break;\n }\n\n p->private_impl.func = func;\n return wuffs_base__make_status(\n func ? NULL : wuffs_base__error__unsupported_pixel_swizzler_option);\n}\n\nuint64_t //\nwuffs_base__pixel_swizzler__swizzle_interleaved(\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, dst_palette, src);\n }\n return 0;\n}\n" + + "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_format, 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_format, dst_palette, src_palette, blend);\n break;\n }\n\n p->private_impl.func = func;\n return wuffs_base__make_status(\n func ? NULL : wuffs_base__error__unsupported_pixel_swizzler_option);\n}\n\nuint64_t //\nwuffs_base__pixel_swizzler__swizzle_interleaved(\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, dst_palette, src);\n }\n return 0;\n}\n" + "" const baseCorePrivateH = "" + diff --git a/release/c/wuffs-unsupported-snapshot.c b/release/c/wuffs-unsupported-snapshot.c index 86495e7b..24139960 100644 --- a/release/c/wuffs-unsupported-snapshot.c +++ b/release/c/wuffs-unsupported-snapshot.c @@ -3255,6 +3255,10 @@ struct wuffs_bmp__decoder__struct { uint64_t f_bytes_total; wuffs_base__pixel_format f_pixfmt; uint32_t f_padding; + uint32_t f_mask_r; + uint32_t f_mask_g; + uint32_t f_mask_b; + uint32_t f_mask_a; uint64_t f_frame_config_io_position; uint32_t f_dst_x; uint32_t f_dst_y; @@ -3273,7 +3277,8 @@ struct wuffs_bmp__decoder__struct { struct { struct { - uint32_t v_bitmapInfoLen; + uint32_t v_bitmap_info_len; + uint32_t v_compression; uint64_t scratch; } s_decode_image_config[1]; struct { @@ -6515,6 +6520,36 @@ wuffs_base__swap_u32_argb_abgr(uint32_t u) { return o | (r >> 16) | (b << 16); } +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. + uint32_t sa = 0x101 * (0xFF & (src_nonpremul >> 24)); + uint32_t sr = 0x101 * (0xFF & (src_nonpremul >> 16)); + uint32_t sg = 0x101 * (0xFF & (src_nonpremul >> 8)); + uint32_t sb = 0x101 * (0xFF & (src_nonpremul >> 0)); + uint32_t da = 0x101 * (0xFF & (dst_premul >> 24)); + uint32_t dr = 0x101 * (0xFF & (dst_premul >> 16)); + uint32_t dg = 0x101 * (0xFF & (dst_premul >> 8)); + uint32_t db = 0x101 * (0xFF & (dst_premul >> 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). + da = sa + ((da * ia) / 0xFFFF); + 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 and combine the components. + da >>= 8; + dr >>= 8; + dg >>= 8; + db >>= 8; + return (db << 0) | (dg << 8) | (dr << 16) | (da << 24); +} + static inline uint32_t // wuffs_base__premul_u32_axxx(uint32_t nonpremul) { // Multiplying by 0x101 (twice, once for alpha and once for color) converts @@ -6714,6 +6749,61 @@ wuffs_base__pixel_buffer__set_color_u32_at( // -------- +static uint64_t // +wuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul__src( + wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src) { + size_t dst_len4 = dst.len / 4; + size_t src_len4 = src.len / 4; + size_t len = dst_len4 < src_len4 ? dst_len4 : src_len4; + uint8_t* d = dst.ptr; + uint8_t* s = src.ptr; + size_t n = len; + + // TODO: unroll. + + while (n >= 1) { + uint32_t s0 = + wuffs_base__premul_u32_axxx(wuffs_base__load_u32le(s + (0 * 4))); + wuffs_base__store_u32le(d + (0 * 4), s0); + + s += 1 * 4; + d += 1 * 4; + n -= 1; + } + + return len; +} + +static uint64_t // +wuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul__src_over( + wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src) { + size_t dst_len4 = dst.len / 4; + size_t src_len4 = src.len / 4; + size_t len = dst_len4 < src_len4 ? dst_len4 : src_len4; + uint8_t* d = dst.ptr; + uint8_t* s = src.ptr; + size_t n = len; + + // TODO: unroll. + + while (n >= 1) { + uint32_t d0 = wuffs_base__load_u32le(d + (0 * 4)); + uint32_t s0 = wuffs_base__load_u32le(s + (0 * 4)); + wuffs_base__store_u32le( + d + (0 * 4), wuffs_base__composite_premul_nonpremul_u32_axxx(d0, s0)); + + s += 1 * 4; + d += 1 * 4; + n -= 1; + } + + return len; +} + static uint64_t // wuffs_base__pixel_swizzler__copy_1_1(wuffs_base__slice_u8 dst, wuffs_base__slice_u8 dst_palette, @@ -6721,6 +6811,19 @@ wuffs_base__pixel_swizzler__copy_1_1(wuffs_base__slice_u8 dst, return wuffs_base__slice_u8__copy_from_slice(dst, src); } +static uint64_t // +wuffs_base__pixel_swizzler__copy_4_4(wuffs_base__slice_u8 dst, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src) { + size_t dst_len4 = dst.len / 4; + size_t src_len4 = src.len / 4; + size_t len = dst_len4 < src_len4 ? dst_len4 : src_len4; + if (len > 0) { + memmove(dst.ptr, src.ptr, len * 4); + } + return len; +} + static uint64_t // wuffs_base__pixel_swizzler__xx__index__src(wuffs_base__slice_u8 dst, wuffs_base__slice_u8 dst_palette, @@ -7223,6 +7326,54 @@ wuffs_base__pixel_swizzler__prepare__bgr(wuffs_base__pixel_swizzler* p, return NULL; } +static wuffs_base__pixel_swizzler__func // +wuffs_base__pixel_swizzler__prepare__bgra_nonpremul( + wuffs_base__pixel_swizzler* p, + wuffs_base__pixel_format dst_format, + wuffs_base__slice_u8 dst_palette, + wuffs_base__slice_u8 src_palette, + wuffs_base__pixel_blend blend) { + switch (dst_format.repr) { + case WUFFS_BASE__PIXEL_FORMAT__BGR_565: + // TODO. + break; + + case WUFFS_BASE__PIXEL_FORMAT__BGR: + case WUFFS_BASE__PIXEL_FORMAT__RGB: + // TODO. + break; + + case WUFFS_BASE__PIXEL_FORMAT__BGRX: + case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL: + switch (blend) { + case WUFFS_BASE__PIXEL_BLEND__SRC: + return wuffs_base__pixel_swizzler__copy_4_4; + } + 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__src; + case WUFFS_BASE__PIXEL_BLEND__SRC_OVER: + return wuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul__src_over; + } + return NULL; + + case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY: + // TODO. + break; + + case WUFFS_BASE__PIXEL_FORMAT__RGBX: + case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL: + case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL: + case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY: + // TODO. + break; + } + return NULL; +} + // -------- wuffs_base__status // @@ -7255,6 +7406,11 @@ wuffs_base__pixel_swizzler__prepare(wuffs_base__pixel_swizzler* p, func = wuffs_base__pixel_swizzler__prepare__bgr( p, dst_format, dst_palette, src_palette, blend); break; + + case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL: + func = wuffs_base__pixel_swizzler__prepare__bgra_nonpremul( + p, dst_format, dst_palette, src_palette, blend); + break; } p->private_impl.func = func; @@ -7587,7 +7743,7 @@ wuffs_bmp__decoder__decode_image_config(wuffs_bmp__decoder* self, wuffs_base__status status = wuffs_base__make_status(NULL); uint32_t v_magic = 0; - uint32_t v_bitmapInfoLen = 0; + uint32_t v_bitmap_info_len = 0; uint32_t v_width = 0; uint32_t v_height = 0; uint32_t v_planes = 0; @@ -7607,8 +7763,9 @@ wuffs_bmp__decoder__decode_image_config(wuffs_bmp__decoder* self, uint32_t coro_susp_point = self->private_impl.p_decode_image_config[0]; if (coro_susp_point) { - v_bitmapInfoLen = - self->private_data.s_decode_image_config[0].v_bitmapInfoLen; + v_bitmap_info_len = + self->private_data.s_decode_image_config[0].v_bitmap_info_len; + v_compression = self->private_data.s_decode_image_config[0].v_compression; } switch (coro_susp_point) { WUFFS_BASE__COROUTINE_SUSPENSION_POINT_0; @@ -7728,17 +7885,18 @@ wuffs_bmp__decoder__decode_image_config(wuffs_bmp__decoder* self, *scratch |= ((uint64_t)(num_bits_2)) << 56; } } - v_bitmapInfoLen = t_2; + v_bitmap_info_len = t_2; } - if (v_bitmapInfoLen != 124) { + if ((v_bitmap_info_len != 40) && (v_bitmap_info_len != 108) && + (v_bitmap_info_len != 124)) { status = wuffs_base__make_status(wuffs_bmp__error__unsupported_bmp_file); goto exit; } - if (self->private_impl.f_padding < v_bitmapInfoLen) { + if (self->private_impl.f_padding < v_bitmap_info_len) { status = wuffs_base__make_status(wuffs_bmp__error__bad_header); goto exit; } - self->private_impl.f_padding -= v_bitmapInfoLen; + self->private_impl.f_padding -= v_bitmap_info_len; { WUFFS_BASE__COROUTINE_SUSPENSION_POINT(8); uint32_t t_3; @@ -7810,8 +7968,7 @@ wuffs_bmp__decoder__decode_image_config(wuffs_bmp__decoder* self, status = wuffs_base__make_status(wuffs_bmp__error__bad_header); goto exit; } else if (v_height >= 2147483648) { - self->private_impl.f_height = - ((0 - self->private_impl.f_height) & 2147483647); + self->private_impl.f_height = ((0 - v_height) & 2147483647); self->private_impl.f_top_down = true; } else { self->private_impl.f_height = v_height; @@ -7889,6 +8046,13 @@ wuffs_bmp__decoder__decode_image_config(wuffs_bmp__decoder* self, self->private_impl.f_pad_per_row = (self->private_impl.f_width & 3); self->private_impl.f_pixfmt = wuffs_base__utility__make_pixel_format(1073744008); + } else if (v_bits_per_pixel == 32) { + self->private_impl.f_bits_per_pixel = 32; + self->private_impl.f_bytes_per_row = + (((uint64_t)(self->private_impl.f_width)) * 4); + self->private_impl.f_pad_per_row = 0; + self->private_impl.f_pixfmt = + wuffs_base__utility__make_pixel_format(1157662856); } else { status = wuffs_base__make_status(wuffs_bmp__error__unsupported_bmp_file); goto exit; @@ -7927,12 +8091,7 @@ wuffs_bmp__decoder__decode_image_config(wuffs_bmp__decoder* self, } v_compression = t_7; } - if (v_compression != 0) { - status = wuffs_base__make_status(wuffs_bmp__error__unsupported_bmp_file); - goto exit; - } - self->private_data.s_decode_image_config[0].scratch = - (v_bitmapInfoLen - 20); + self->private_data.s_decode_image_config[0].scratch = 20; WUFFS_BASE__COROUTINE_SUSPENSION_POINT(18); if (self->private_data.s_decode_image_config[0].scratch > ((uint64_t)(io2_a_src - iop_a_src))) { @@ -7943,6 +8102,154 @@ wuffs_bmp__decoder__decode_image_config(wuffs_bmp__decoder* self, goto suspend; } iop_a_src += self->private_data.s_decode_image_config[0].scratch; + if (v_bitmap_info_len >= 108) { + { + WUFFS_BASE__COROUTINE_SUSPENSION_POINT(19); + uint32_t t_8; + if (WUFFS_BASE__LIKELY(io2_a_src - iop_a_src >= 4)) { + t_8 = wuffs_base__load_u32le(iop_a_src); + iop_a_src += 4; + } else { + self->private_data.s_decode_image_config[0].scratch = 0; + WUFFS_BASE__COROUTINE_SUSPENSION_POINT(20); + while (true) { + if (WUFFS_BASE__UNLIKELY(iop_a_src == io2_a_src)) { + status = + wuffs_base__make_status(wuffs_base__suspension__short_read); + goto suspend; + } + uint64_t* scratch = + &self->private_data.s_decode_image_config[0].scratch; + uint32_t num_bits_8 = ((uint32_t)(*scratch >> 56)); + *scratch <<= 8; + *scratch >>= 8; + *scratch |= ((uint64_t)(*iop_a_src++)) << num_bits_8; + if (num_bits_8 == 24) { + t_8 = ((uint32_t)(*scratch)); + break; + } + num_bits_8 += 8; + *scratch |= ((uint64_t)(num_bits_8)) << 56; + } + } + self->private_impl.f_mask_r = t_8; + } + { + WUFFS_BASE__COROUTINE_SUSPENSION_POINT(21); + uint32_t t_9; + if (WUFFS_BASE__LIKELY(io2_a_src - iop_a_src >= 4)) { + t_9 = wuffs_base__load_u32le(iop_a_src); + iop_a_src += 4; + } else { + self->private_data.s_decode_image_config[0].scratch = 0; + WUFFS_BASE__COROUTINE_SUSPENSION_POINT(22); + while (true) { + if (WUFFS_BASE__UNLIKELY(iop_a_src == io2_a_src)) { + status = + wuffs_base__make_status(wuffs_base__suspension__short_read); + goto suspend; + } + uint64_t* scratch = + &self->private_data.s_decode_image_config[0].scratch; + uint32_t num_bits_9 = ((uint32_t)(*scratch >> 56)); + *scratch <<= 8; + *scratch >>= 8; + *scratch |= ((uint64_t)(*iop_a_src++)) << num_bits_9; + if (num_bits_9 == 24) { + t_9 = ((uint32_t)(*scratch)); + break; + } + num_bits_9 += 8; + *scratch |= ((uint64_t)(num_bits_9)) << 56; + } + } + self->private_impl.f_mask_g = t_9; + } + { + WUFFS_BASE__COROUTINE_SUSPENSION_POINT(23); + uint32_t t_10; + if (WUFFS_BASE__LIKELY(io2_a_src - iop_a_src >= 4)) { + t_10 = wuffs_base__load_u32le(iop_a_src); + iop_a_src += 4; + } else { + self->private_data.s_decode_image_config[0].scratch = 0; + WUFFS_BASE__COROUTINE_SUSPENSION_POINT(24); + while (true) { + if (WUFFS_BASE__UNLIKELY(iop_a_src == io2_a_src)) { + status = + wuffs_base__make_status(wuffs_base__suspension__short_read); + goto suspend; + } + uint64_t* scratch = + &self->private_data.s_decode_image_config[0].scratch; + uint32_t num_bits_10 = ((uint32_t)(*scratch >> 56)); + *scratch <<= 8; + *scratch >>= 8; + *scratch |= ((uint64_t)(*iop_a_src++)) << num_bits_10; + if (num_bits_10 == 24) { + t_10 = ((uint32_t)(*scratch)); + break; + } + num_bits_10 += 8; + *scratch |= ((uint64_t)(num_bits_10)) << 56; + } + } + self->private_impl.f_mask_b = t_10; + } + { + WUFFS_BASE__COROUTINE_SUSPENSION_POINT(25); + uint32_t t_11; + if (WUFFS_BASE__LIKELY(io2_a_src - iop_a_src >= 4)) { + t_11 = wuffs_base__load_u32le(iop_a_src); + iop_a_src += 4; + } else { + self->private_data.s_decode_image_config[0].scratch = 0; + WUFFS_BASE__COROUTINE_SUSPENSION_POINT(26); + while (true) { + if (WUFFS_BASE__UNLIKELY(iop_a_src == io2_a_src)) { + status = + wuffs_base__make_status(wuffs_base__suspension__short_read); + goto suspend; + } + uint64_t* scratch = + &self->private_data.s_decode_image_config[0].scratch; + uint32_t num_bits_11 = ((uint32_t)(*scratch >> 56)); + *scratch <<= 8; + *scratch >>= 8; + *scratch |= ((uint64_t)(*iop_a_src++)) << num_bits_11; + if (num_bits_11 == 24) { + t_11 = ((uint32_t)(*scratch)); + break; + } + num_bits_11 += 8; + *scratch |= ((uint64_t)(num_bits_11)) << 56; + } + } + self->private_impl.f_mask_a = t_11; + } + if ((v_compression == 3) && (self->private_impl.f_mask_r == 16711680) && + (self->private_impl.f_mask_g == 65280) && + (self->private_impl.f_mask_b == 255) && + (self->private_impl.f_mask_a == 4278190080)) { + v_compression = 0; + } + self->private_data.s_decode_image_config[0].scratch = + (v_bitmap_info_len - 56); + WUFFS_BASE__COROUTINE_SUSPENSION_POINT(27); + if (self->private_data.s_decode_image_config[0].scratch > + ((uint64_t)(io2_a_src - iop_a_src))) { + self->private_data.s_decode_image_config[0].scratch -= + ((uint64_t)(io2_a_src - iop_a_src)); + iop_a_src = io2_a_src; + status = wuffs_base__make_status(wuffs_base__suspension__short_read); + goto suspend; + } + iop_a_src += self->private_data.s_decode_image_config[0].scratch; + } + if (v_compression != 0) { + status = wuffs_base__make_status(wuffs_bmp__error__unsupported_bmp_file); + goto exit; + } self->private_impl.f_frame_config_io_position = wuffs_base__u64__sat_add( a_src->meta.pos, ((uint64_t)(iop_a_src - io0_a_src))); if (a_dst != NULL) { @@ -7965,7 +8272,9 @@ suspend: wuffs_base__status__is_suspension(&status) ? coro_susp_point : 0; self->private_impl.active_coroutine = wuffs_base__status__is_suspension(&status) ? 1 : 0; - self->private_data.s_decode_image_config[0].v_bitmapInfoLen = v_bitmapInfoLen; + self->private_data.s_decode_image_config[0].v_bitmap_info_len = + v_bitmap_info_len; + self->private_data.s_decode_image_config[0].v_compression = v_compression; goto exit; exit: diff --git a/std/bmp/decode_bmp.wuffs b/std/bmp/decode_bmp.wuffs index 96f0ad5b..032f798c 100644 --- a/std/bmp/decode_bmp.wuffs +++ b/std/bmp/decode_bmp.wuffs @@ -34,6 +34,11 @@ pub struct decoder? implements base.image_decoder( padding : base.u32, + mask_r : base.u32, + mask_g : base.u32, + mask_b : base.u32, + mask_a : base.u32, + frame_config_io_position : base.u64, dst_x : base.u32, @@ -50,13 +55,13 @@ pub struct decoder? implements base.image_decoder( ) pub func decoder.decode_image_config?(dst: nptr base.image_config, src: base.io_reader) { - var magic : base.u32 - var bitmapInfoLen : base.u32 - var width : base.u32 - var height : base.u32 - var planes : base.u32 - var bits_per_pixel : base.u32 - var compression : base.u32 + var magic : base.u32 + var bitmap_info_len : base.u32 + var width : base.u32 + var height : base.u32 + var planes : base.u32 + var bits_per_pixel : base.u32 + var compression : base.u32 if this.call_sequence <> 0 { return base."#bad call sequence" @@ -77,18 +82,16 @@ pub func decoder.decode_image_config?(dst: nptr base.image_config, src: base.io_ } this.padding -= 14 - // Read the BITMAPINFOHEADER (version 5 is 124 bytes). - // - // TODO: support other versions. + // Read the BITMAPINFOHEADER (version 3 / 4 / 5 is 40 / 108 / 124 bytes). - bitmapInfoLen = args.src.read_u32le?() - if bitmapInfoLen <> 124 { + bitmap_info_len = args.src.read_u32le?() + if (bitmap_info_len <> 40) and (bitmap_info_len <> 108) and (bitmap_info_len <> 124) { return "#unsupported BMP file" } - if this.padding < bitmapInfoLen { + if this.padding < bitmap_info_len { return "#bad header" } - this.padding -= bitmapInfoLen + this.padding -= bitmap_info_len width = args.src.read_u32le?() if width >= 0x8000_0000 { @@ -102,7 +105,7 @@ pub func decoder.decode_image_config?(dst: nptr base.image_config, src: base.io_ } else if height >= 0x8000_0000 { // The &0x7FFF_FFFF is redundant, but proves to the compiler that the // result is within this.height's refined bounds. - this.height = (0 ~mod- this.height) & 0x7FFF_FFFF + this.height = (0 ~mod- height) & 0x7FFF_FFFF this.top_down = true } else { this.height = height @@ -123,22 +126,52 @@ pub func decoder.decode_image_config?(dst: nptr base.image_config, src: base.io_ // TODO: a Wuffs (not just C) name for the // WUFFS_BASE__PIXEL_FORMAT__BGR magic pixfmt constant. this.pixfmt = this.util.make_pixel_format(repr: 0x4000_0888) + } else if bits_per_pixel == 32 { + this.bits_per_pixel = 32 + this.bytes_per_row = (this.width as base.u64) * 4 + this.pad_per_row = 0 + // TODO: a Wuffs (not just C) name for the + // WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL magic pixfmt constant. + this.pixfmt = this.util.make_pixel_format(repr: 0x4500_8888) } else { - // TODO: support 8bpp, 32bpp, possibly more. + // TODO: support other bits_per_pixel's. return "#unsupported BMP file" } this.bytes_total = this.bytes_per_row * (this.height as base.u64) compression = args.src.read_u32le?() + + // We've already read 20 bytes from the BITMAPINFOHEADER: size (4), width + // (4), height (4), planes (2), bpp (2), compression (4). Skip the rest of + // the version 3 BITMAPINFOHEADER (whose total size is 40). + args.src.skip32?(n: 40 - 20) + + if bitmap_info_len >= 108 { + this.mask_r = args.src.read_u32le?() + this.mask_g = args.src.read_u32le?() + this.mask_b = args.src.read_u32le?() + this.mask_a = args.src.read_u32le?() + + // If compression is 3 (BITFIELDS) but the explicit masks are what the + // implicit masks are for no compression, treat it as no compression. + if (compression == 3) and + (this.mask_r == 0x00FF_0000) and + (this.mask_g == 0x0000_FF00) and + (this.mask_b == 0x0000_00FF) and + (this.mask_a == 0xFF00_0000) { + compression = 0 + } + + // Skip the rest of the BITMAPINFOHEADER. We've already read (40 + (4 * + // 4)) bytes. + args.src.skip32?(n: bitmap_info_len - 56) + } + if compression <> 0 { // TODO: support compression. return "#unsupported BMP file" } - // We've already read 20 bytes from the BITMAPINFOHEADER: size (4), width - // (4), height (4), planes (2), bpp (2), compression (4). Skip the rest. - args.src.skip32?(n: bitmapInfoLen - 20) - this.frame_config_io_position = args.src.position() if args.dst <> nullptr {