diff --git a/internal/cgen/base/image-impl.c b/internal/cgen/base/image-impl.c index 9fdf7ca8..59217088 100644 --- a/internal/cgen/base/image-impl.c +++ b/internal/cgen/base/image-impl.c @@ -29,6 +29,90 @@ wuffs_base__swap_u32_argb_abgr(uint32_t u) { return o | (r >> 16) | (b << 16); } +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. + 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_nonpremul >> 24)); + uint32_t dr = 0x101 * (0xFF & (dst_nonpremul >> 16)); + uint32_t dg = 0x101 * (0xFF & (dst_nonpremul >> 8)); + uint32_t db = 0x101 * (0xFF & (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. + 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 dst from premul to nonpremul. + if (da != 0) { + dr = (dr * 0xFFFF) / da; + dg = (dg * 0xFFFF) / da; + db = (db * 0xFFFF) / da; + } + + // 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__composite_nonpremul_premul_u32_axxx(uint32_t dst_nonpremul, + uint32_t src_premul) { + // Convert from 8-bit color to 16-bit color. + uint32_t sa = 0x101 * (0xFF & (src_premul >> 24)); + uint32_t sr = 0x101 * (0xFF & (src_premul >> 16)); + uint32_t sg = 0x101 * (0xFF & (src_premul >> 8)); + uint32_t sb = 0x101 * (0xFF & (src_premul >> 0)); + uint32_t da = 0x101 * (0xFF & (dst_nonpremul >> 24)); + uint32_t dr = 0x101 * (0xFF & (dst_nonpremul >> 16)); + uint32_t dg = 0x101 * (0xFF & (dst_nonpremul >> 8)); + uint32_t db = 0x101 * (0xFF & (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. + uint32_t ia = 0xFFFF - sa; + + // Composite src (premul) over dst (premul). + da = sa + ((da * ia) / 0xFFFF); + dr = sr + ((dr * ia) / 0xFFFF); + dg = sg + ((dg * ia) / 0xFFFF); + db = sb + ((db * ia) / 0xFFFF); + + // Convert dst from premul to nonpremul. + if (da != 0) { + dr = (dr * 0xFFFF) / da; + dg = (dg * 0xFFFF) / da; + db = (db * 0xFFFF) / da; + } + + // 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__composite_premul_nonpremul_u32_axxx(uint32_t dst_premul, uint32_t src_nonpremul) { @@ -59,6 +143,36 @@ wuffs_base__composite_premul_nonpremul_u32_axxx(uint32_t dst_premul, return (db << 0) | (dg << 8) | (dr << 16) | (da << 24); } +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. + uint32_t sa = 0x101 * (0xFF & (src_premul >> 24)); + uint32_t sr = 0x101 * (0xFF & (src_premul >> 16)); + uint32_t sg = 0x101 * (0xFF & (src_premul >> 8)); + uint32_t sb = 0x101 * (0xFF & (src_premul >> 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 (premul) over dst (premul). + da = sa + ((da * ia) / 0xFFFF); + dr = sr + ((dr * ia) / 0xFFFF); + dg = sg + ((dg * ia) / 0xFFFF); + db = sb + ((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); +} + wuffs_base__color_u32_argb_premul // wuffs_base__pixel_buffer__color_u32_at(const wuffs_base__pixel_buffer* pb, uint32_t x, @@ -292,6 +406,35 @@ wuffs_base__pixel_palette__closest_element( // -------- +static uint64_t // +wuffs_base__pixel_swizzler__bgra_nonpremul__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__no_bounds_check(d + (0 * 4)); + uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(s + (0 * 4)); + wuffs_base__store_u32le__no_bounds_check( + d + (0 * 4), + wuffs_base__composite_nonpremul_nonpremul_u32_axxx(d0, s0)); + + s += 1 * 4; + d += 1 * 4; + n -= 1; + } + + return len; +} + static uint64_t // wuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul__src( wuffs_base__slice_u8 dst, @@ -904,6 +1047,8 @@ wuffs_base__pixel_swizzler__prepare__bgra_nonpremul( switch (blend) { case WUFFS_BASE__PIXEL_BLEND__SRC: return wuffs_base__pixel_swizzler__copy_4_4; + case WUFFS_BASE__PIXEL_BLEND__SRC_OVER: + return wuffs_base__pixel_swizzler__bgra_nonpremul__bgra_nonpremul__src_over; } return NULL; diff --git a/internal/cgen/data.go b/internal/cgen/data.go index de410ef6..6f49c3f9 100644 --- a/internal/cgen/data.go +++ b/internal/cgen/data.go @@ -27,30 +27,35 @@ 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__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\nwuffs_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->priva" + - "te_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__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__premul_u32_axxx(\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 uint16_t bgr =\n wuffs_base__load_u16le__no_bounds_check(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 |\n wuffs_base__load_u24le__no_bounds_check(row + (3 * ((size_t)x)));\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL:\n return wuffs_base__premul_u32_axxx(\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_F" + - "ORMAT__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(wuffs_base__premul_u32_axxx(\n wuffs_base__load_u32le__no_bounds_check(row + (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__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.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), wuffs_base__color_u32_argb_premul__as_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 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__no_bounds_check(row + (2 * ((size_t)x)),\n (uint16_t)bgr565);\n break;\n }\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))" + - ", wuffs_base__nonpremul_u32_axxx(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)), 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__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" + + "// ---------------- 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_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 u" + + "int32_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)) / 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 // Con" + + "vert 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\nwuffs_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 uint8_t* row = pb->private_impl.planes[0].ptr + (stride * ((size_t)y));\n\n switch (pb->pixcfg.private_impl.pixfmt.repr) {\n c" + + "ase 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__premul_u32_axxx(\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 uint16_t bgr =\n " + + " wuffs_base__load_u16le__no_bounds_check(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 |\n wuffs_base__load_u24le__no_bounds_check(row + (3 * ((size_t)x)));\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL:\n return wuffs_base__premul_u32_axxx(\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(wuffs_base__premul_u32_axxx(\n wuffs_base__load_u32le__no_bounds_check(row + (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__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->pixc" + + "fg.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.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), wuffs_base__color_u32_argb_premul__as_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 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__no_bounds_check(row + (2 * ((size_t)x)),\n (uint16_t)bgr565);\n break;\n }\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)), wuffs_base__nonpremul_u32_axxx(color));\n break;\n\n case WUFFS_BASE__PIXEL_FORMAT__RGB:\n wuffs_base__store_u24le__no_bounds_check(\n r" + + "ow + (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)), 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__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\nuint8_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 best_index;\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 = wuffs_base__premul_u32_axxx(\n wuffs_base__load_u32le__no_bounds_check(s + (0 * 4)));\n wuffs_base__store_u32le__no_bounds_check(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: 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__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 (l" + - "en > 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__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__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 uint" + - "8_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__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_b" + - "ase__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 ((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__no_bounds_check(dst_palette.ptr +\n ((size_t)s[0] * 4));\n if (s0) {\n wuffs_base__store_u24le__no_bounds_check(d + (0 * 4), s0);\n }\n uint32_t s1 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr +\n ((size_t)s[1] * 4));\n if (s1) {\n wuffs_base__store_u24le__no_bounds_check(d + (1 * 4), s1);\n }\n uint32_t s2 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr +\n ((size_t)s[2] * 4));\n if (s2) {\n wuffs_base__store_u24le__no_bounds_check(d + (2 * 4), s2);\n }\n uint32_t s3 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr +\n ((size_t)s[3] * 4));\n if (s3) {\n wuffs_base__store_u24le__no_bounds_check(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__no_bounds_check(dst_palette.ptr +\n ((size_t)s[0] * 4));\n if (s0) {\n wuffs_base__store_u24le__no_bounds_check(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__no_bounds_check(\n d + (0 * 4), wuffs_base__load_u32le__no_bounds_chec" + - "k(\n dst_palette.ptr + ((size_t)s[0] * 4)));\n wuffs_base__store_u32le__no_bounds_check(\n d + (1 * 4), 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 * 4), 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 * 4), 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 * 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__load_u32le__no_bounds_check(\n dst_palette.ptr + ((size_t)s[0] * 4)));\n\n s += 1 * 1;\n d += 1 * 4;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__xxxx__index_binary_alpha__src_ov" + - "er(\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__no_bounds_check(dst_palette.ptr +\n ((size_t)s[0] * 4));\n if (s0) {\n wuffs_base__store_u32le__no_bounds_check(d + (0 * 4), s0);\n }\n uint32_t s1 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr +\n ((size_t)s[1] * 4));\n if (s1) {\n wuffs_base__store_u32le__no_bounds_check(d + (1 * 4), s1);\n }\n uint32_t s2 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr +\n ((size_t)s[2] * 4));\n " + - "if (s2) {\n wuffs_base__store_u32le__no_bounds_check(d + (2 * 4), s2);\n }\n uint32_t s3 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr +\n ((size_t)s[3] * 4));\n if (s3) {\n wuffs_base__store_u32le__no_bounds_check(d + (3 * 4), s3);\n }\n\n s += loop_unroll_count * 1;\n d += loop_unroll_count * 4;\n n -= loop_unroll_count;\n }\n\n while (n >= 1) {\n uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr +\n ((size_t)s[0] * 4));\n if (s0) {\n wuffs_base__store_u32le__no_bounds_check(d + (0 * 4), s0);\n }\n\n s += 1 * 1;\n d += 1 * 4;\n n -= 1;\n }\n\n return len;\n}\n\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__no_bounds_check(\n d + (0 * 4),\n 0xFF000000 | wuffs_base__load_u24le__no_bounds_check(s + (0 * 3)));\n\n s += 1 * 3;\n d += 1 * 4;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__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__no_bounds_check(\n d + (0 * 4), 0xFF000000 | (0x010101 * (uint32_t)s[0]));\n\n s += 1 * 1;\n d += 1 * 4;\n n -= 1;\n }\n\n return len;\n}\n\n" + + "// --------\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__bgra_nonpremul__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: 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\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 = wuffs_base__premul_u32_axxx(\n wuffs_base__load_u32le__no_bounds_check(s + (0 * 4)));\n wuffs_base__store_u32le__no_bounds_check(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: 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__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__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__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 wri" + + "tes 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__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 ((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__no_bounds_check(dst_palette.ptr +\n ((size_t)s[0] * 4));\n if (s0) {\n wuffs_base__store_u24le__no_bounds_check(d + (0" + + " * 4), s0);\n }\n uint32_t s1 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr +\n ((size_t)s[1] * 4));\n if (s1) {\n wuffs_base__store_u24le__no_bounds_check(d + (1 * 4), s1);\n }\n uint32_t s2 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr +\n ((size_t)s[2] * 4));\n if (s2) {\n wuffs_base__store_u24le__no_bounds_check(d + (2 * 4), s2);\n }\n uint32_t s3 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr +\n ((size_t)s[3] * 4));\n if (s3) {\n wuffs_base__store_u24le__no_bounds_check(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__no_bounds_check(dst_palette.ptr +\n ((size_t)s[0] * 4));\n if (s0) {\n" + + " wuffs_base__store_u24le__no_bounds_check(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__no_bounds_check(\n d + (0 * 4), 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 * 4), 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 * 4), 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 * 4), 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 * 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__load_u32le__no_bounds_check(\n dst_palette.ptr + ((size_t)s[0] * 4)));\n\n s += 1 * 1;\n d += 1 * 4;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__xxxx__index_binary_alpha__src_over(\n 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__no_bounds_check(dst_palette.ptr +\n ((size_t)s[0] * 4));\n if (s0) {\n wuffs_base__store_u32le__no_bounds_check(d + (0 * 4), s0);\n }\n uint32_t s1 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr +\n ((size_t)s[1] * 4));\n if (s1) {\n wuffs_base__store_u32le__no_bounds_check(d + (1 * 4), s1);\n }\n uint32_t s2 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr +\n ((size_t)s[2] * 4));\n if (s2) {\n wuffs_base__store_u32le__no_bounds_check(d + (2 * 4), s2);\n }\n uint32_t s3 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr +\n ((size_t)s[3] * 4));\n if (s3) {\n w" + + "uffs_base__store_u32le__no_bounds_check(d + (3 * 4), s3);\n }\n\n s += loop_unroll_count * 1;\n d += loop_unroll_count * 4;\n n -= loop_unroll_count;\n }\n\n while (n >= 1) {\n uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(dst_palette.ptr +\n ((size_t)s[0] * 4));\n if (s0) {\n wuffs_base__store_u32le__no_bounds_check(d + (0 * 4), s0);\n }\n\n s += 1 * 1;\n d += 1 * 4;\n n -= 1;\n }\n\n return len;\n}\n\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__no_bounds_check(\n d + (0 * 4),\n 0xFF00000" + + "0 | wuffs_base__load_u24le__no_bounds_check(s + (0 * 3)));\n\n s += 1 * 3;\n d += 1 * 4;\n n -= 1;\n }\n\n return len;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__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__no_bounds_check(\n d + (0 * 4), 0xFF000000 | (0x010101 * (uint32_t)s[0]));\n\n s += 1 * 1;\n d += 1 * 4;\n n -= 1;\n }\n\n return len;\n}\n\n" + "" + "// --------\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__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__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 wuffs_base__store_u32le__no_bounds_check(\n 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 u" + "int8_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" + @@ -61,7 +66,7 @@ const baseImageImplC = "" + "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__BGRA_NONPREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__BGR" + "A_BINARY:\n case WUFFS_BASE__PIXEL_FORMAT__BGRX:\n return wuffs_base__pixel_swizzler__xxxx__xxx;\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 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_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__BGRA_NONPREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__BGRX:\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__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" + + "__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 // TODO.\n break;\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 case WUFFS_BASE__PIXEL_FORMAT__RGBX:\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 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" + diff --git a/release/c/wuffs-unsupported-snapshot.c b/release/c/wuffs-unsupported-snapshot.c index ff2e2feb..876c908f 100644 --- a/release/c/wuffs-unsupported-snapshot.c +++ b/release/c/wuffs-unsupported-snapshot.c @@ -8514,6 +8514,90 @@ wuffs_base__swap_u32_argb_abgr(uint32_t u) { return o | (r >> 16) | (b << 16); } +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. + 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_nonpremul >> 24)); + uint32_t dr = 0x101 * (0xFF & (dst_nonpremul >> 16)); + uint32_t dg = 0x101 * (0xFF & (dst_nonpremul >> 8)); + uint32_t db = 0x101 * (0xFF & (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. + 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 dst from premul to nonpremul. + if (da != 0) { + dr = (dr * 0xFFFF) / da; + dg = (dg * 0xFFFF) / da; + db = (db * 0xFFFF) / da; + } + + // 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__composite_nonpremul_premul_u32_axxx(uint32_t dst_nonpremul, + uint32_t src_premul) { + // Convert from 8-bit color to 16-bit color. + uint32_t sa = 0x101 * (0xFF & (src_premul >> 24)); + uint32_t sr = 0x101 * (0xFF & (src_premul >> 16)); + uint32_t sg = 0x101 * (0xFF & (src_premul >> 8)); + uint32_t sb = 0x101 * (0xFF & (src_premul >> 0)); + uint32_t da = 0x101 * (0xFF & (dst_nonpremul >> 24)); + uint32_t dr = 0x101 * (0xFF & (dst_nonpremul >> 16)); + uint32_t dg = 0x101 * (0xFF & (dst_nonpremul >> 8)); + uint32_t db = 0x101 * (0xFF & (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. + uint32_t ia = 0xFFFF - sa; + + // Composite src (premul) over dst (premul). + da = sa + ((da * ia) / 0xFFFF); + dr = sr + ((dr * ia) / 0xFFFF); + dg = sg + ((dg * ia) / 0xFFFF); + db = sb + ((db * ia) / 0xFFFF); + + // Convert dst from premul to nonpremul. + if (da != 0) { + dr = (dr * 0xFFFF) / da; + dg = (dg * 0xFFFF) / da; + db = (db * 0xFFFF) / da; + } + + // 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__composite_premul_nonpremul_u32_axxx(uint32_t dst_premul, uint32_t src_nonpremul) { @@ -8544,6 +8628,36 @@ wuffs_base__composite_premul_nonpremul_u32_axxx(uint32_t dst_premul, return (db << 0) | (dg << 8) | (dr << 16) | (da << 24); } +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. + uint32_t sa = 0x101 * (0xFF & (src_premul >> 24)); + uint32_t sr = 0x101 * (0xFF & (src_premul >> 16)); + uint32_t sg = 0x101 * (0xFF & (src_premul >> 8)); + uint32_t sb = 0x101 * (0xFF & (src_premul >> 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 (premul) over dst (premul). + da = sa + ((da * ia) / 0xFFFF); + dr = sr + ((dr * ia) / 0xFFFF); + dg = sg + ((dg * ia) / 0xFFFF); + db = sb + ((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); +} + wuffs_base__color_u32_argb_premul // wuffs_base__pixel_buffer__color_u32_at(const wuffs_base__pixel_buffer* pb, uint32_t x, @@ -8777,6 +8891,35 @@ wuffs_base__pixel_palette__closest_element( // -------- +static uint64_t // +wuffs_base__pixel_swizzler__bgra_nonpremul__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__no_bounds_check(d + (0 * 4)); + uint32_t s0 = wuffs_base__load_u32le__no_bounds_check(s + (0 * 4)); + wuffs_base__store_u32le__no_bounds_check( + d + (0 * 4), + wuffs_base__composite_nonpremul_nonpremul_u32_axxx(d0, s0)); + + s += 1 * 4; + d += 1 * 4; + n -= 1; + } + + return len; +} + static uint64_t // wuffs_base__pixel_swizzler__bgra_premul__bgra_nonpremul__src( wuffs_base__slice_u8 dst, @@ -9389,6 +9532,8 @@ wuffs_base__pixel_swizzler__prepare__bgra_nonpremul( switch (blend) { case WUFFS_BASE__PIXEL_BLEND__SRC: return wuffs_base__pixel_swizzler__copy_4_4; + case WUFFS_BASE__PIXEL_BLEND__SRC_OVER: + return wuffs_base__pixel_swizzler__bgra_nonpremul__bgra_nonpremul__src_over; } return NULL; diff --git a/test/c/std/wbmp.c b/test/c/std/wbmp.c index 11c51629..34705244 100644 --- a/test/c/std/wbmp.c +++ b/test/c/std/wbmp.c @@ -84,6 +84,23 @@ fill_palette_with_grays(wuffs_base__pixel_buffer* pb) { } } +bool // +colors_differ(wuffs_base__color_u32_argb_premul color0, + wuffs_base__color_u32_argb_premul color1, + uint32_t per_channel_tolerance) { + uint32_t shift; + for (shift = 0; shift < 32; shift += 8) { + uint32_t channel0 = 0xFF & (color0 >> shift); + uint32_t channel1 = 0xFF & (color1 >> shift); + uint32_t delta = + (channel0 > channel1) ? (channel0 - channel1) : (channel1 - channel0); + if (delta > per_channel_tolerance) { + return true; + } + } + return false; +} + const char* // test_wuffs_pixel_swizzler_swizzle() { CHECK_FOCUS(__func__); @@ -127,6 +144,7 @@ test_wuffs_pixel_swizzler_swizzle() { const wuffs_base__pixel_blend blends[] = { WUFFS_BASE__PIXEL_BLEND__SRC, + WUFFS_BASE__PIXEL_BLEND__SRC_OVER, }; int s; @@ -198,19 +216,24 @@ test_wuffs_pixel_swizzler_swizzle() { wuffs_base__pixel_buffer__plane(&src_pixbuf, 0), height / 2)); // Check the middle dst pixel. + uint32_t tolerance = 0; wuffs_base__color_u32_argb_premul want_dst_pixel = 0; if (blends[b] == WUFFS_BASE__PIXEL_BLEND__SRC) { want_dst_pixel = srcs[s].pixel; + } else if (blends[b] == WUFFS_BASE__PIXEL_BLEND__SRC_OVER) { + tolerance = 1; + want_dst_pixel = wuffs_base__composite_premul_premul_u32_axxx( + dsts[d].pixel, srcs[s].pixel); } else { return "unsupported blend"; } wuffs_base__color_u32_argb_premul have_dst_pixel = wuffs_base__pixel_buffer__color_u32_at(&dst_pixbuf, width / 2, height / 2); - if (have_dst_pixel != want_dst_pixel) { + if (colors_differ(have_dst_pixel, want_dst_pixel, tolerance)) { RETURN_FAIL("s=%d, d=%d, b=%d: dst_pixel: have 0x%08" PRIX32 - ", want 0x%08" PRIX32, - s, d, b, have_dst_pixel, want_dst_pixel); + ", want 0x%08" PRIX32 ", per-channel tolerance=%" PRId32, + s, d, b, have_dst_pixel, want_dst_pixel, tolerance); } } }