From bc546b73830d5eed6f4925a8420ce95e4cb742f8 Mon Sep 17 00:00:00 2001 From: Nigel Tao Date: Tue, 26 May 2020 11:54:36 +1000 Subject: [PATCH] Move more base funcs to the pixconv sub-module $ ls -l gen/lib/c/*/{lib*,*base*} | awk '{print $5 " " $9}' Before: 167976 gen/lib/c/clang-5.0-dynamic/libwuffs.so 13944 gen/lib/c/clang-5.0-dynamic/wuffs-base-core.lo 11640 gen/lib/c/clang-5.0-dynamic/wuffs-base-pixconv.lo 207150 gen/lib/c/clang-5.0-static/libwuffs.a 14240 gen/lib/c/clang-5.0-static/wuffs-base-core.o 12096 gen/lib/c/clang-5.0-static/wuffs-base-pixconv.o 188088 gen/lib/c/gcc-dynamic/libwuffs.so 13296 gen/lib/c/gcc-dynamic/wuffs-base-core.lo 27232 gen/lib/c/gcc-dynamic/wuffs-base-pixconv.lo 234766 gen/lib/c/gcc-static/libwuffs.a 13432 gen/lib/c/gcc-static/wuffs-base-core.o 27216 gen/lib/c/gcc-static/wuffs-base-pixconv.o After: 167976 gen/lib/c/clang-5.0-dynamic/libwuffs.so 11376 gen/lib/c/clang-5.0-dynamic/wuffs-base-core.lo 14328 gen/lib/c/clang-5.0-dynamic/wuffs-base-pixconv.lo 207270 gen/lib/c/clang-5.0-static/libwuffs.a 11696 gen/lib/c/clang-5.0-static/wuffs-base-core.o 14760 gen/lib/c/clang-5.0-static/wuffs-base-pixconv.o 188088 gen/lib/c/gcc-dynamic/libwuffs.so 10680 gen/lib/c/gcc-dynamic/wuffs-base-core.lo 29968 gen/lib/c/gcc-dynamic/wuffs-base-pixconv.lo 234886 gen/lib/c/gcc-static/libwuffs.a 10856 gen/lib/c/gcc-static/wuffs-base-core.o 29912 gen/lib/c/gcc-static/wuffs-base-pixconv.o --- internal/cgen/base/all-impl.c | 7 +- internal/cgen/base/image-impl.c | 258 ------------- internal/cgen/base/pixconv-submodule.c | 236 ++++++++++++ internal/cgen/cgen.go | 7 - internal/cgen/data.go | 45 +-- internal/cgen/gen.go | 1 - release/c/wuffs-unsupported-snapshot.c | 484 ++++++++++++------------- 7 files changed, 502 insertions(+), 536 deletions(-) delete mode 100644 internal/cgen/base/image-impl.c diff --git a/internal/cgen/base/all-impl.c b/internal/cgen/base/all-impl.c index 2bb67b18..f65f3ad4 100644 --- a/internal/cgen/base/all-impl.c +++ b/internal/cgen/base/all-impl.c @@ -108,12 +108,15 @@ const uint64_t wuffs_base__low_bits_mask__u64[65] = { 0x7FFFFFFFFFFFFFFF, 0xFFFFFFFFFFFFFFFF, }; +const uint32_t wuffs_base__pixel_format__bits_per_channel[16] = { + 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, + 0x08, 0x0A, 0x0C, 0x10, 0x18, 0x20, 0x30, 0x40, +}; + // !! INSERT wuffs_base__status strings. // !! INSERT vtable names. -// !! INSERT base/image-impl.c. - // !! INSERT base/strconv-impl.c. #endif // !defined(WUFFS_CONFIG__MODULES) || diff --git a/internal/cgen/base/image-impl.c b/internal/cgen/base/image-impl.c deleted file mode 100644 index 59b53e75..00000000 --- a/internal/cgen/base/image-impl.c +++ /dev/null @@ -1,258 +0,0 @@ -// After editing this file, run "go generate" in the parent directory. - -// Copyright 2017 The Wuffs Authors. -// -// Licensed under the Apache License, Version 2.0 (the "License"); -// you may not use this file except in compliance with the License. -// You may obtain a copy of the License at -// -// https://www.apache.org/licenses/LICENSE-2.0 -// -// Unless required by applicable law or agreed to in writing, software -// distributed under the License is distributed on an "AS IS" BASIS, -// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -// See the License for the specific language governing permissions and -// limitations under the License. - -// ---------------- Images - -const uint32_t wuffs_base__pixel_format__bits_per_channel[16] = { - 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, - 0x08, 0x0A, 0x0C, 0x10, 0x18, 0x20, 0x30, 0x40, -}; - -// -------- - -static inline uint32_t // -wuffs_base__swap_u32_argb_abgr(uint32_t u) { - uint32_t o = u & 0xFF00FF00; - uint32_t r = u & 0x00FF0000; - uint32_t b = u & 0x000000FF; - return o | (r >> 16) | (b << 16); -} - -// -------- - -WUFFS_BASE__MAYBE_STATIC wuffs_base__color_u32_argb_premul // -wuffs_base__pixel_buffer__color_u32_at(const wuffs_base__pixel_buffer* pb, - uint32_t x, - uint32_t y) { - if (!pb || (x >= pb->pixcfg.private_impl.width) || - (y >= pb->pixcfg.private_impl.height)) { - return 0; - } - - if (wuffs_base__pixel_format__is_planar(&pb->pixcfg.private_impl.pixfmt)) { - // TODO: support planar formats. - return 0; - } - - size_t stride = pb->private_impl.planes[0].stride; - uint8_t* row = pb->private_impl.planes[0].ptr + (stride * ((size_t)y)); - - switch (pb->pixcfg.private_impl.pixfmt.repr) { - case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL: - case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY: - return wuffs_base__load_u32le__no_bounds_check(row + (4 * ((size_t)x))); - - case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_PREMUL: - case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_BINARY: { - uint8_t* palette = pb->private_impl.planes[3].ptr; - return wuffs_base__load_u32le__no_bounds_check(palette + - (4 * ((size_t)row[x]))); - } - - // Common formats above. Rarer formats below. - - case WUFFS_BASE__PIXEL_FORMAT__Y: - return 0xFF000000 | (0x00010101 * ((uint32_t)(row[x]))); - - case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_NONPREMUL: { - uint8_t* palette = pb->private_impl.planes[3].ptr; - return wuffs_base__color_u32_argb_nonpremul__as__color_u32_argb_premul( - wuffs_base__load_u32le__no_bounds_check(palette + - (4 * ((size_t)row[x])))); - } - - case WUFFS_BASE__PIXEL_FORMAT__BGR_565: - return wuffs_base__color_u16_rgb_565__as__color_u32_argb_premul( - wuffs_base__load_u16le__no_bounds_check(row + (2 * ((size_t)x)))); - case WUFFS_BASE__PIXEL_FORMAT__BGR: - return 0xFF000000 | - wuffs_base__load_u24le__no_bounds_check(row + (3 * ((size_t)x))); - case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL: - return wuffs_base__color_u32_argb_nonpremul__as__color_u32_argb_premul( - wuffs_base__load_u32le__no_bounds_check(row + (4 * ((size_t)x)))); - case WUFFS_BASE__PIXEL_FORMAT__BGRX: - return 0xFF000000 | - wuffs_base__load_u32le__no_bounds_check(row + (4 * ((size_t)x))); - - case WUFFS_BASE__PIXEL_FORMAT__RGB: - return wuffs_base__swap_u32_argb_abgr( - 0xFF000000 | - wuffs_base__load_u24le__no_bounds_check(row + (3 * ((size_t)x)))); - case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL: - return wuffs_base__swap_u32_argb_abgr( - wuffs_base__color_u32_argb_nonpremul__as__color_u32_argb_premul( - wuffs_base__load_u32le__no_bounds_check(row + - (4 * ((size_t)x))))); - case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL: - case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY: - return wuffs_base__swap_u32_argb_abgr( - wuffs_base__load_u32le__no_bounds_check(row + (4 * ((size_t)x)))); - case WUFFS_BASE__PIXEL_FORMAT__RGBX: - return wuffs_base__swap_u32_argb_abgr( - 0xFF000000 | - wuffs_base__load_u32le__no_bounds_check(row + (4 * ((size_t)x)))); - - default: - // TODO: support more formats. - break; - } - - return 0; -} - -WUFFS_BASE__MAYBE_STATIC wuffs_base__status // -wuffs_base__pixel_buffer__set_color_u32_at( - wuffs_base__pixel_buffer* pb, - uint32_t x, - uint32_t y, - wuffs_base__color_u32_argb_premul color) { - if (!pb) { - return wuffs_base__make_status(wuffs_base__error__bad_receiver); - } - if ((x >= pb->pixcfg.private_impl.width) || - (y >= pb->pixcfg.private_impl.height)) { - return wuffs_base__make_status(wuffs_base__error__bad_argument); - } - - if (wuffs_base__pixel_format__is_planar(&pb->pixcfg.private_impl.pixfmt)) { - // TODO: support planar formats. - return wuffs_base__make_status(wuffs_base__error__unsupported_option); - } - - size_t stride = pb->private_impl.planes[0].stride; - uint8_t* row = pb->private_impl.planes[0].ptr + (stride * ((size_t)y)); - - switch (pb->pixcfg.private_impl.pixfmt.repr) { - case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL: - case WUFFS_BASE__PIXEL_FORMAT__BGRX: - wuffs_base__store_u32le__no_bounds_check(row + (4 * ((size_t)x)), color); - break; - - // Common formats above. Rarer formats below. - - case WUFFS_BASE__PIXEL_FORMAT__Y: - wuffs_base__store_u8__no_bounds_check( - row + ((size_t)x), - wuffs_base__color_u32_argb_premul__as__color_u8_gray(color)); - break; - - case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_BINARY: - wuffs_base__store_u8__no_bounds_check( - row + ((size_t)x), wuffs_base__pixel_palette__closest_element( - wuffs_base__pixel_buffer__palette(pb), - pb->pixcfg.private_impl.pixfmt, color)); - break; - - case WUFFS_BASE__PIXEL_FORMAT__BGR_565: - wuffs_base__store_u16le__no_bounds_check( - row + (2 * ((size_t)x)), - wuffs_base__color_u32_argb_premul__as__color_u16_rgb_565(color)); - break; - case WUFFS_BASE__PIXEL_FORMAT__BGR: - wuffs_base__store_u24le__no_bounds_check(row + (3 * ((size_t)x)), color); - break; - case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL: - wuffs_base__store_u32le__no_bounds_check( - row + (4 * ((size_t)x)), - wuffs_base__color_u32_argb_premul__as__color_u32_argb_nonpremul( - color)); - break; - - case WUFFS_BASE__PIXEL_FORMAT__RGB: - wuffs_base__store_u24le__no_bounds_check( - row + (3 * ((size_t)x)), wuffs_base__swap_u32_argb_abgr(color)); - break; - case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL: - wuffs_base__store_u32le__no_bounds_check( - row + (4 * ((size_t)x)), - wuffs_base__color_u32_argb_premul__as__color_u32_argb_nonpremul( - wuffs_base__swap_u32_argb_abgr(color))); - break; - case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL: - case WUFFS_BASE__PIXEL_FORMAT__RGBX: - wuffs_base__store_u32le__no_bounds_check( - row + (4 * ((size_t)x)), wuffs_base__swap_u32_argb_abgr(color)); - break; - - default: - // TODO: support more formats. - return wuffs_base__make_status(wuffs_base__error__unsupported_option); - } - - return wuffs_base__make_status(NULL); -} - -// -------- - -WUFFS_BASE__MAYBE_STATIC uint8_t // -wuffs_base__pixel_palette__closest_element( - wuffs_base__slice_u8 palette_slice, - wuffs_base__pixel_format palette_format, - wuffs_base__color_u32_argb_premul c) { - size_t n = palette_slice.len / 4; - if (n > 256) { - n = 256; - } - size_t best_index = 0; - uint64_t best_score = 0xFFFFFFFFFFFFFFFF; - - // Work in 16-bit color. - uint32_t ca = 0x101 * (0xFF & (c >> 24)); - uint32_t cr = 0x101 * (0xFF & (c >> 16)); - uint32_t cg = 0x101 * (0xFF & (c >> 8)); - uint32_t cb = 0x101 * (0xFF & (c >> 0)); - - switch (palette_format.repr) { - case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_NONPREMUL: - case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_PREMUL: - case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_BINARY: { - bool nonpremul = palette_format.repr == - WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_NONPREMUL; - - size_t i; - for (i = 0; i < n; i++) { - // Work in 16-bit color. - uint32_t pb = 0x101 * ((uint32_t)(palette_slice.ptr[(4 * i) + 0])); - uint32_t pg = 0x101 * ((uint32_t)(palette_slice.ptr[(4 * i) + 1])); - uint32_t pr = 0x101 * ((uint32_t)(palette_slice.ptr[(4 * i) + 2])); - uint32_t pa = 0x101 * ((uint32_t)(palette_slice.ptr[(4 * i) + 3])); - - // Convert to premultiplied alpha. - if (nonpremul && (pa != 0xFFFF)) { - pb = (pb * pa) / 0xFFFF; - pg = (pg * pa) / 0xFFFF; - pr = (pr * pa) / 0xFFFF; - } - - // These deltas are conceptually int32_t (signed) but after squaring, - // it's equivalent to work in uint32_t (unsigned). - pb -= cb; - pg -= cg; - pr -= cr; - pa -= ca; - uint64_t score = ((uint64_t)(pb * pb)) + ((uint64_t)(pg * pg)) + - ((uint64_t)(pr * pr)) + ((uint64_t)(pa * pa)); - if (best_score > score) { - best_score = score; - best_index = i; - } - } - break; - } - } - - return (uint8_t)best_index; -} diff --git a/internal/cgen/base/pixconv-submodule.c b/internal/cgen/base/pixconv-submodule.c index 630fdfd2..e041e5c8 100644 --- a/internal/cgen/base/pixconv-submodule.c +++ b/internal/cgen/base/pixconv-submodule.c @@ -16,6 +16,242 @@ // ---------------- Pixel Swizzler +static inline uint32_t // +wuffs_base__swap_u32_argb_abgr(uint32_t u) { + uint32_t o = u & 0xFF00FF00; + uint32_t r = u & 0x00FF0000; + uint32_t b = u & 0x000000FF; + return o | (r >> 16) | (b << 16); +} + +// -------- + +WUFFS_BASE__MAYBE_STATIC wuffs_base__color_u32_argb_premul // +wuffs_base__pixel_buffer__color_u32_at(const wuffs_base__pixel_buffer* pb, + uint32_t x, + uint32_t y) { + if (!pb || (x >= pb->pixcfg.private_impl.width) || + (y >= pb->pixcfg.private_impl.height)) { + return 0; + } + + if (wuffs_base__pixel_format__is_planar(&pb->pixcfg.private_impl.pixfmt)) { + // TODO: support planar formats. + return 0; + } + + size_t stride = pb->private_impl.planes[0].stride; + uint8_t* row = pb->private_impl.planes[0].ptr + (stride * ((size_t)y)); + + switch (pb->pixcfg.private_impl.pixfmt.repr) { + case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL: + case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY: + return wuffs_base__load_u32le__no_bounds_check(row + (4 * ((size_t)x))); + + case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_PREMUL: + case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_BINARY: { + uint8_t* palette = pb->private_impl.planes[3].ptr; + return wuffs_base__load_u32le__no_bounds_check(palette + + (4 * ((size_t)row[x]))); + } + + // Common formats above. Rarer formats below. + + case WUFFS_BASE__PIXEL_FORMAT__Y: + return 0xFF000000 | (0x00010101 * ((uint32_t)(row[x]))); + + case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_NONPREMUL: { + uint8_t* palette = pb->private_impl.planes[3].ptr; + return wuffs_base__color_u32_argb_nonpremul__as__color_u32_argb_premul( + wuffs_base__load_u32le__no_bounds_check(palette + + (4 * ((size_t)row[x])))); + } + + case WUFFS_BASE__PIXEL_FORMAT__BGR_565: + return wuffs_base__color_u16_rgb_565__as__color_u32_argb_premul( + wuffs_base__load_u16le__no_bounds_check(row + (2 * ((size_t)x)))); + case WUFFS_BASE__PIXEL_FORMAT__BGR: + return 0xFF000000 | + wuffs_base__load_u24le__no_bounds_check(row + (3 * ((size_t)x))); + case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL: + return wuffs_base__color_u32_argb_nonpremul__as__color_u32_argb_premul( + wuffs_base__load_u32le__no_bounds_check(row + (4 * ((size_t)x)))); + case WUFFS_BASE__PIXEL_FORMAT__BGRX: + return 0xFF000000 | + wuffs_base__load_u32le__no_bounds_check(row + (4 * ((size_t)x))); + + case WUFFS_BASE__PIXEL_FORMAT__RGB: + return wuffs_base__swap_u32_argb_abgr( + 0xFF000000 | + wuffs_base__load_u24le__no_bounds_check(row + (3 * ((size_t)x)))); + case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL: + return wuffs_base__swap_u32_argb_abgr( + wuffs_base__color_u32_argb_nonpremul__as__color_u32_argb_premul( + wuffs_base__load_u32le__no_bounds_check(row + + (4 * ((size_t)x))))); + case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL: + case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY: + return wuffs_base__swap_u32_argb_abgr( + wuffs_base__load_u32le__no_bounds_check(row + (4 * ((size_t)x)))); + case WUFFS_BASE__PIXEL_FORMAT__RGBX: + return wuffs_base__swap_u32_argb_abgr( + 0xFF000000 | + wuffs_base__load_u32le__no_bounds_check(row + (4 * ((size_t)x)))); + + default: + // TODO: support more formats. + break; + } + + return 0; +} + +WUFFS_BASE__MAYBE_STATIC wuffs_base__status // +wuffs_base__pixel_buffer__set_color_u32_at( + wuffs_base__pixel_buffer* pb, + uint32_t x, + uint32_t y, + wuffs_base__color_u32_argb_premul color) { + if (!pb) { + return wuffs_base__make_status(wuffs_base__error__bad_receiver); + } + if ((x >= pb->pixcfg.private_impl.width) || + (y >= pb->pixcfg.private_impl.height)) { + return wuffs_base__make_status(wuffs_base__error__bad_argument); + } + + if (wuffs_base__pixel_format__is_planar(&pb->pixcfg.private_impl.pixfmt)) { + // TODO: support planar formats. + return wuffs_base__make_status(wuffs_base__error__unsupported_option); + } + + size_t stride = pb->private_impl.planes[0].stride; + uint8_t* row = pb->private_impl.planes[0].ptr + (stride * ((size_t)y)); + + switch (pb->pixcfg.private_impl.pixfmt.repr) { + case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL: + case WUFFS_BASE__PIXEL_FORMAT__BGRX: + wuffs_base__store_u32le__no_bounds_check(row + (4 * ((size_t)x)), color); + break; + + // Common formats above. Rarer formats below. + + case WUFFS_BASE__PIXEL_FORMAT__Y: + wuffs_base__store_u8__no_bounds_check( + row + ((size_t)x), + wuffs_base__color_u32_argb_premul__as__color_u8_gray(color)); + break; + + case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_BINARY: + wuffs_base__store_u8__no_bounds_check( + row + ((size_t)x), wuffs_base__pixel_palette__closest_element( + wuffs_base__pixel_buffer__palette(pb), + pb->pixcfg.private_impl.pixfmt, color)); + break; + + case WUFFS_BASE__PIXEL_FORMAT__BGR_565: + wuffs_base__store_u16le__no_bounds_check( + row + (2 * ((size_t)x)), + wuffs_base__color_u32_argb_premul__as__color_u16_rgb_565(color)); + break; + case WUFFS_BASE__PIXEL_FORMAT__BGR: + wuffs_base__store_u24le__no_bounds_check(row + (3 * ((size_t)x)), color); + break; + case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL: + wuffs_base__store_u32le__no_bounds_check( + row + (4 * ((size_t)x)), + wuffs_base__color_u32_argb_premul__as__color_u32_argb_nonpremul( + color)); + break; + + case WUFFS_BASE__PIXEL_FORMAT__RGB: + wuffs_base__store_u24le__no_bounds_check( + row + (3 * ((size_t)x)), wuffs_base__swap_u32_argb_abgr(color)); + break; + case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL: + wuffs_base__store_u32le__no_bounds_check( + row + (4 * ((size_t)x)), + wuffs_base__color_u32_argb_premul__as__color_u32_argb_nonpremul( + wuffs_base__swap_u32_argb_abgr(color))); + break; + case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL: + case WUFFS_BASE__PIXEL_FORMAT__RGBX: + wuffs_base__store_u32le__no_bounds_check( + row + (4 * ((size_t)x)), wuffs_base__swap_u32_argb_abgr(color)); + break; + + default: + // TODO: support more formats. + return wuffs_base__make_status(wuffs_base__error__unsupported_option); + } + + return wuffs_base__make_status(NULL); +} + +// -------- + +WUFFS_BASE__MAYBE_STATIC uint8_t // +wuffs_base__pixel_palette__closest_element( + wuffs_base__slice_u8 palette_slice, + wuffs_base__pixel_format palette_format, + wuffs_base__color_u32_argb_premul c) { + size_t n = palette_slice.len / 4; + if (n > 256) { + n = 256; + } + size_t best_index = 0; + uint64_t best_score = 0xFFFFFFFFFFFFFFFF; + + // Work in 16-bit color. + uint32_t ca = 0x101 * (0xFF & (c >> 24)); + uint32_t cr = 0x101 * (0xFF & (c >> 16)); + uint32_t cg = 0x101 * (0xFF & (c >> 8)); + uint32_t cb = 0x101 * (0xFF & (c >> 0)); + + switch (palette_format.repr) { + case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_NONPREMUL: + case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_PREMUL: + case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_BINARY: { + bool nonpremul = palette_format.repr == + WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_NONPREMUL; + + size_t i; + for (i = 0; i < n; i++) { + // Work in 16-bit color. + uint32_t pb = 0x101 * ((uint32_t)(palette_slice.ptr[(4 * i) + 0])); + uint32_t pg = 0x101 * ((uint32_t)(palette_slice.ptr[(4 * i) + 1])); + uint32_t pr = 0x101 * ((uint32_t)(palette_slice.ptr[(4 * i) + 2])); + uint32_t pa = 0x101 * ((uint32_t)(palette_slice.ptr[(4 * i) + 3])); + + // Convert to premultiplied alpha. + if (nonpremul && (pa != 0xFFFF)) { + pb = (pb * pa) / 0xFFFF; + pg = (pg * pa) / 0xFFFF; + pr = (pr * pa) / 0xFFFF; + } + + // These deltas are conceptually int32_t (signed) but after squaring, + // it's equivalent to work in uint32_t (unsigned). + pb -= cb; + pg -= cg; + pr -= cr; + pa -= ca; + uint64_t score = ((uint64_t)(pb * pb)) + ((uint64_t)(pg * pg)) + + ((uint64_t)(pr * pr)) + ((uint64_t)(pa * pa)); + if (best_score > score) { + best_score = score; + best_index = i; + } + } + break; + } + } + + return (uint8_t)best_index; +} + +// -------- + static inline uint32_t // wuffs_base__composite_nonpremul_nonpremul_u32_axxx(uint32_t dst_nonpremul, uint32_t src_nonpremul) { diff --git a/internal/cgen/cgen.go b/internal/cgen/cgen.go index 313959f2..bf9fb32c 100644 --- a/internal/cgen/cgen.go +++ b/internal/cgen/cgen.go @@ -131,7 +131,6 @@ func Do(args []string) error { "// !! INSERT base/all-public.h.\n": insertBaseAllPublicH, "// !! INSERT base/copyright\n": insertBaseCopyright, "// !! INSERT base/f64conv-submodule.c.\n": insertBaseF64ConvSubmoduleC, - "// !! INSERT base/image-impl.c.\n": insertBaseImageImplC, "// !! INSERT base/pixconv-submodule.c.\n": insertBasePixConvSubmoduleC, "// !! INSERT base/strconv-impl.c.\n": insertBaseStrConvImplC, "// !! INSERT vtable names.\n": func(b *buffer) error { @@ -352,12 +351,6 @@ func insertBaseF64ConvSubmoduleC(buf *buffer) error { return nil } -func insertBaseImageImplC(buf *buffer) error { - buf.writes(baseImageImplC) - buf.writeb('\n') - return nil -} - func insertBasePixConvSubmoduleC(buf *buffer) error { buf.writes(basePixConvSubmoduleC) buf.writeb('\n') diff --git a/internal/cgen/data.go b/internal/cgen/data.go index cf32c5ce..d40a1257 100644 --- a/internal/cgen/data.go +++ b/internal/cgen/data.go @@ -23,25 +23,8 @@ const baseAllImplC = "" + "" + "// ----------------\n\n#if !defined(WUFFS_CONFIG__MODULES) || defined(WUFFS_CONFIG__MODULE__BASE) || \\\n defined(WUFFS_CONFIG__MODULE__BASE__CORE)\n\nconst uint8_t wuffs_base__low_bits_mask__u8[9] = {\n 0x00, 0x01, 0x03, 0x07, 0x0F, 0x1F, 0x3F, 0x7F, 0xFF,\n};\n\nconst uint16_t wuffs_base__low_bits_mask__u16[17] = {\n 0x0000, 0x0001, 0x0003, 0x0007, 0x000F, 0x001F, 0x003F, 0x007F, 0x00FF,\n 0x01FF, 0x03FF, 0x07FF, 0x0FFF, 0x1FFF, 0x3FFF, 0x7FFF, 0xFFFF,\n};\n\nconst uint32_t wuffs_base__low_bits_mask__u32[33] = {\n 0x00000000, 0x00000001, 0x00000003, 0x00000007, 0x0000000F, 0x0000001F,\n 0x0000003F, 0x0000007F, 0x000000FF, 0x000001FF, 0x000003FF, 0x000007FF,\n 0x00000FFF, 0x00001FFF, 0x00003FFF, 0x00007FFF, 0x0000FFFF, 0x0001FFFF,\n 0x0003FFFF, 0x0007FFFF, 0x000FFFFF, 0x001FFFFF, 0x003FFFFF, 0x007FFFFF,\n 0x00FFFFFF, 0x01FFFFFF, 0x03FFFFFF, 0x07FFFFFF, 0x0FFFFFFF, 0x1FFFFFFF,\n 0x3FFFFFFF, 0x7FFFFFFF, 0xFFFFFFFF,\n};\n\nconst uint64_t wuffs_base__low_bits_mask__u64[65] = {\n 0x0000000000000000, 0x000" + "0000000000001, 0x0000000000000003,\n 0x0000000000000007, 0x000000000000000F, 0x000000000000001F,\n 0x000000000000003F, 0x000000000000007F, 0x00000000000000FF,\n 0x00000000000001FF, 0x00000000000003FF, 0x00000000000007FF,\n 0x0000000000000FFF, 0x0000000000001FFF, 0x0000000000003FFF,\n 0x0000000000007FFF, 0x000000000000FFFF, 0x000000000001FFFF,\n 0x000000000003FFFF, 0x000000000007FFFF, 0x00000000000FFFFF,\n 0x00000000001FFFFF, 0x00000000003FFFFF, 0x00000000007FFFFF,\n 0x0000000000FFFFFF, 0x0000000001FFFFFF, 0x0000000003FFFFFF,\n 0x0000000007FFFFFF, 0x000000000FFFFFFF, 0x000000001FFFFFFF,\n 0x000000003FFFFFFF, 0x000000007FFFFFFF, 0x00000000FFFFFFFF,\n 0x00000001FFFFFFFF, 0x00000003FFFFFFFF, 0x00000007FFFFFFFF,\n 0x0000000FFFFFFFFF, 0x0000001FFFFFFFFF, 0x0000003FFFFFFFFF,\n 0x0000007FFFFFFFFF, 0x000000FFFFFFFFFF, 0x000001FFFFFFFFFF,\n 0x000003FFFFFFFFFF, 0x000007FFFFFFFFFF, 0x00000FFFFFFFFFFF,\n 0x00001FFFFFFFFFFF, 0x00003FFFFFFFFFFF, 0x00007FFFFFFFFFFF,\n 0x0000FFFFFFFFFFFF, 0x000" + - "1FFFFFFFFFFFF, 0x0003FFFFFFFFFFFF,\n 0x0007FFFFFFFFFFFF, 0x000FFFFFFFFFFFFF, 0x001FFFFFFFFFFFFF,\n 0x003FFFFFFFFFFFFF, 0x007FFFFFFFFFFFFF, 0x00FFFFFFFFFFFFFF,\n 0x01FFFFFFFFFFFFFF, 0x03FFFFFFFFFFFFFF, 0x07FFFFFFFFFFFFFF,\n 0x0FFFFFFFFFFFFFFF, 0x1FFFFFFFFFFFFFFF, 0x3FFFFFFFFFFFFFFF,\n 0x7FFFFFFFFFFFFFFF, 0xFFFFFFFFFFFFFFFF,\n};\n\n// !! INSERT wuffs_base__status strings.\n\n// !! INSERT vtable names.\n\n// !! INSERT base/image-impl.c.\n\n// !! INSERT base/strconv-impl.c.\n\n#endif // !defined(WUFFS_CONFIG__MODULES) ||\n // defined(WUFFS_CONFIG__MODULE__BASE) ||\n // defined(WUFFS_CONFIG__MODULE__BASE__CORE)\n\n#if !defined(WUFFS_CONFIG__MODULES) || defined(WUFFS_CONFIG__MODULE__BASE) || \\\n defined(WUFFS_CONFIG__MODULE__BASE__INTERFACES)\n\n// !! INSERT InterfaceDefinitions.\n\n#endif // !defined(WUFFS_CONFIG__MODULES) ||\n // defined(WUFFS_CONFIG__MODULE__BASE) ||\n // defined(WUFFS_CONFIG__MODULE__BASE__INTERFACES)\n\n#if !defined(WUFFS_CONFIG__MODULES) || defined(WUFFS_CONFIG__MODULE__BA" + - "SE) || \\\n defined(WUFFS_CONFIG__MODULE__BASE__F64CONV)\n\n// !! INSERT base/f64conv-submodule.c.\n\n#endif // !defined(WUFFS_CONFIG__MODULES) ||\n // defined(WUFFS_CONFIG__MODULE__BASE) ||\n // defined(WUFFS_CONFIG__MODULE__BASE__F64CONV)\n\n#if !defined(WUFFS_CONFIG__MODULES) || defined(WUFFS_CONFIG__MODULE__BASE) || \\\n defined(WUFFS_CONFIG__MODULE__BASE__PIXCONV)\n\n// !! INSERT base/pixconv-submodule.c.\n\n#endif // !defined(WUFFS_CONFIG__MODULES) ||\n // defined(WUFFS_CONFIG__MODULE__BASE) ||\n // defined(WUFFS_CONFIG__MODULE__BASE__PIXCONV)\n\n#ifdef __cplusplus\n} // extern \"C\"\n#endif\n\n#endif // WUFFS_IMPLEMENTATION\n\n// !! WUFFS MONOLITHIC RELEASE DISCARDS EVERYTHING BELOW.\n\n#endif // WUFFS_INCLUDE_GUARD__BASE\n" + - "" - -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\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\n" + - "" + - "// --------\n\nWUFFS_BASE__MAYBE_STATIC wuffs_base__color_u32_argb_premul //\nwuffs_base__pixel_buffer__color_u32_at(const wuffs_base__pixel_buffer* pb,\n uint32_t x,\n uint32_t y) {\n if (!pb || (x >= pb->pixcfg.private_impl.width) ||\n (y >= pb->pixcfg.private_impl.height)) {\n return 0;\n }\n\n if (wuffs_base__pixel_format__is_planar(&pb->pixcfg.private_impl.pixfmt)) {\n // TODO: support planar formats.\n return 0;\n }\n\n size_t stride = pb->private_impl.planes[0].stride;\n 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.plane" + - "s[3].ptr;\n return wuffs_base__load_u32le__no_bounds_check(palette +\n (4 * ((size_t)row[x])));\n }\n\n // Common formats above. Rarer formats below.\n\n case WUFFS_BASE__PIXEL_FORMAT__Y:\n return 0xFF000000 | (0x00010101 * ((uint32_t)(row[x])));\n\n case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_NONPREMUL: {\n uint8_t* palette = pb->private_impl.planes[3].ptr;\n return wuffs_base__color_u32_argb_nonpremul__as__color_u32_argb_premul(\n wuffs_base__load_u32le__no_bounds_check(palette +\n (4 * ((size_t)row[x]))));\n }\n\n case WUFFS_BASE__PIXEL_FORMAT__BGR_565:\n return wuffs_base__color_u16_rgb_565__as__color_u32_argb_premul(\n wuffs_base__load_u16le__no_bounds_check(row + (2 * ((size_t)x))));\n case WUFFS_BASE__PIXEL_FORMAT__BGR:\n return 0xFF000000 |\n wuffs_base__load_u24le__no_bounds_check(row + (3 * ((size_t)x)));\n case WUFFS_BASE__PIXEL_FORMAT_" + - "_BGRA_NONPREMUL:\n return wuffs_base__color_u32_argb_nonpremul__as__color_u32_argb_premul(\n wuffs_base__load_u32le__no_bounds_check(row + (4 * ((size_t)x))));\n case WUFFS_BASE__PIXEL_FORMAT__BGRX:\n return 0xFF000000 |\n wuffs_base__load_u32le__no_bounds_check(row + (4 * ((size_t)x)));\n\n case WUFFS_BASE__PIXEL_FORMAT__RGB:\n return wuffs_base__swap_u32_argb_abgr(\n 0xFF000000 |\n wuffs_base__load_u24le__no_bounds_check(row + (3 * ((size_t)x))));\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL:\n return wuffs_base__swap_u32_argb_abgr(\n wuffs_base__color_u32_argb_nonpremul__as__color_u32_argb_premul(\n wuffs_base__load_u32le__no_bounds_check(row +\n (4 * ((size_t)x)))));\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY:\n return wuffs_base__swap_u32_argb_abgr(\n wuffs_base__load_u32le__no_bounds_check(row + (4 * ((size_t)x)" + - ")));\n case WUFFS_BASE__PIXEL_FORMAT__RGBX:\n return wuffs_base__swap_u32_argb_abgr(\n 0xFF000000 |\n wuffs_base__load_u32le__no_bounds_check(row + (4 * ((size_t)x))));\n\n default:\n // TODO: support more formats.\n break;\n }\n\n return 0;\n}\n\nWUFFS_BASE__MAYBE_STATIC wuffs_base__status //\nwuffs_base__pixel_buffer__set_color_u32_at(\n wuffs_base__pixel_buffer* pb,\n uint32_t x,\n uint32_t y,\n wuffs_base__color_u32_argb_premul color) {\n if (!pb) {\n return wuffs_base__make_status(wuffs_base__error__bad_receiver);\n }\n if ((x >= pb->pixcfg.private_impl.width) ||\n (y >= pb->pixcfg.private_impl.height)) {\n return wuffs_base__make_status(wuffs_base__error__bad_argument);\n }\n\n if (wuffs_base__pixel_format__is_planar(&pb->pixcfg.private_impl.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),\n wuffs_base__color_u32_argb_premul__as__color_u8_gray(color));\n break;\n\n case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_BINARY:\n wuffs_base__store_u8__no_bounds_check(\n row + ((size_t)x), wuffs_base__pixel_palette__closest_element(\n wuffs_base__pixel_buffer__palette(pb),\n pb->pixcfg.private_impl.pixfmt, color));\n break;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGR_565:\n wuffs_base__store_u16le__no_bounds_check(\n row + (2 * ((size_t)x)),\n wuffs_base__color_u32_argb_premul__as" + - "__color_u16_rgb_565(color));\n break;\n case WUFFS_BASE__PIXEL_FORMAT__BGR:\n wuffs_base__store_u24le__no_bounds_check(row + (3 * ((size_t)x)), color);\n break;\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL:\n wuffs_base__store_u32le__no_bounds_check(\n row + (4 * ((size_t)x)),\n wuffs_base__color_u32_argb_premul__as__color_u32_argb_nonpremul(\n color));\n break;\n\n case WUFFS_BASE__PIXEL_FORMAT__RGB:\n wuffs_base__store_u24le__no_bounds_check(\n row + (3 * ((size_t)x)), wuffs_base__swap_u32_argb_abgr(color));\n break;\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL:\n wuffs_base__store_u32le__no_bounds_check(\n row + (4 * ((size_t)x)),\n wuffs_base__color_u32_argb_premul__as__color_u32_argb_nonpremul(\n wuffs_base__swap_u32_argb_abgr(color)));\n break;\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBX:\n wuffs_base__store_u32le__no_bounds_check(\n row" + - " + (4 * ((size_t)x)), wuffs_base__swap_u32_argb_abgr(color));\n break;\n\n default:\n // TODO: support more formats.\n return wuffs_base__make_status(wuffs_base__error__unsupported_option);\n }\n\n return wuffs_base__make_status(NULL);\n}\n\n" + - "" + - "// --------\n\nWUFFS_BASE__MAYBE_STATIC uint8_t //\nwuffs_base__pixel_palette__closest_element(\n wuffs_base__slice_u8 palette_slice,\n wuffs_base__pixel_format palette_format,\n wuffs_base__color_u32_argb_premul c) {\n size_t n = palette_slice.len / 4;\n if (n > 256) {\n n = 256;\n }\n size_t best_index = 0;\n uint64_t best_score = 0xFFFFFFFFFFFFFFFF;\n\n // Work in 16-bit color.\n uint32_t ca = 0x101 * (0xFF & (c >> 24));\n uint32_t cr = 0x101 * (0xFF & (c >> 16));\n uint32_t cg = 0x101 * (0xFF & (c >> 8));\n uint32_t cb = 0x101 * (0xFF & (c >> 0));\n\n switch (palette_format.repr) {\n case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_NONPREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_BINARY: {\n bool nonpremul = palette_format.repr ==\n WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_NONPREMUL;\n\n size_t i;\n for (i = 0; i < n; i++) {\n // Work in 16-bit color.\n uint32_t pb = 0x101 * ((uint32_t)(palette" + - "_slice.ptr[(4 * i) + 0]));\n uint32_t pg = 0x101 * ((uint32_t)(palette_slice.ptr[(4 * i) + 1]));\n uint32_t pr = 0x101 * ((uint32_t)(palette_slice.ptr[(4 * i) + 2]));\n uint32_t pa = 0x101 * ((uint32_t)(palette_slice.ptr[(4 * i) + 3]));\n\n // Convert to premultiplied alpha.\n if (nonpremul && (pa != 0xFFFF)) {\n pb = (pb * pa) / 0xFFFF;\n pg = (pg * pa) / 0xFFFF;\n pr = (pr * pa) / 0xFFFF;\n }\n\n // These deltas are conceptually int32_t (signed) but after squaring,\n // it's equivalent to work in uint32_t (unsigned).\n pb -= cb;\n pg -= cg;\n pr -= cr;\n pa -= ca;\n uint64_t score = ((uint64_t)(pb * pb)) + ((uint64_t)(pg * pg)) +\n ((uint64_t)(pr * pr)) + ((uint64_t)(pa * pa));\n if (best_score > score) {\n best_score = score;\n best_index = i;\n }\n }\n break;\n }\n }\n\n return (uint8_t)best_index;\n}\n" + + "1FFFFFFFFFFFF, 0x0003FFFFFFFFFFFF,\n 0x0007FFFFFFFFFFFF, 0x000FFFFFFFFFFFFF, 0x001FFFFFFFFFFFFF,\n 0x003FFFFFFFFFFFFF, 0x007FFFFFFFFFFFFF, 0x00FFFFFFFFFFFFFF,\n 0x01FFFFFFFFFFFFFF, 0x03FFFFFFFFFFFFFF, 0x07FFFFFFFFFFFFFF,\n 0x0FFFFFFFFFFFFFFF, 0x1FFFFFFFFFFFFFFF, 0x3FFFFFFFFFFFFFFF,\n 0x7FFFFFFFFFFFFFFF, 0xFFFFFFFFFFFFFFFF,\n};\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\n// !! INSERT wuffs_base__status strings.\n\n// !! INSERT vtable names.\n\n// !! INSERT base/strconv-impl.c.\n\n#endif // !defined(WUFFS_CONFIG__MODULES) ||\n // defined(WUFFS_CONFIG__MODULE__BASE) ||\n // defined(WUFFS_CONFIG__MODULE__BASE__CORE)\n\n#if !defined(WUFFS_CONFIG__MODULES) || defined(WUFFS_CONFIG__MODULE__BASE) || \\\n defined(WUFFS_CONFIG__MODULE__BASE__INTERFACES)\n\n// !! INSERT InterfaceDefinitions.\n\n#endif // !defined(WUFFS_CONFIG__MODULES) ||\n // defined(WUFFS_CONFIG__MODULE_" + + "_BASE) ||\n // defined(WUFFS_CONFIG__MODULE__BASE__INTERFACES)\n\n#if !defined(WUFFS_CONFIG__MODULES) || defined(WUFFS_CONFIG__MODULE__BASE) || \\\n defined(WUFFS_CONFIG__MODULE__BASE__F64CONV)\n\n// !! INSERT base/f64conv-submodule.c.\n\n#endif // !defined(WUFFS_CONFIG__MODULES) ||\n // defined(WUFFS_CONFIG__MODULE__BASE) ||\n // defined(WUFFS_CONFIG__MODULE__BASE__F64CONV)\n\n#if !defined(WUFFS_CONFIG__MODULES) || defined(WUFFS_CONFIG__MODULE__BASE) || \\\n defined(WUFFS_CONFIG__MODULE__BASE__PIXCONV)\n\n// !! INSERT base/pixconv-submodule.c.\n\n#endif // !defined(WUFFS_CONFIG__MODULES) ||\n // defined(WUFFS_CONFIG__MODULE__BASE) ||\n // defined(WUFFS_CONFIG__MODULE__BASE__PIXCONV)\n\n#ifdef __cplusplus\n} // extern \"C\"\n#endif\n\n#endif // WUFFS_IMPLEMENTATION\n\n// !! WUFFS MONOLITHIC RELEASE DISCARDS EVERYTHING BELOW.\n\n#endif // WUFFS_INCLUDE_GUARD__BASE\n" + "" const baseStrConvImplC = "" + @@ -130,12 +113,24 @@ const baseF64ConvSubmoduleC = "" + "" const basePixConvSubmoduleC = "" + - "// ---------------- Pixel Swizzler\n\nstatic inline uint32_t //\nwuffs_base__composite_nonpremul_nonpremul_u32_axxx(uint32_t dst_nonpremul,\n uint32_t src_nonpremul) {\n // Convert from 8-bit color to 16-bit color.\n uint32_t sa = 0x101 * (0xFF & (src_nonpremul >> 24));\n uint32_t sr = 0x101 * (0xFF & (src_nonpremul >> 16));\n uint32_t sg = 0x101 * (0xFF & (src_nonpremul >> 8));\n uint32_t sb = 0x101 * (0xFF & (src_nonpremul >> 0));\n uint32_t da = 0x101 * (0xFF & (dst_nonpremul >> 24));\n uint32_t dr = 0x101 * (0xFF & (dst_nonpremul >> 16));\n uint32_t dg = 0x101 * (0xFF & (dst_nonpremul >> 8));\n uint32_t db = 0x101 * (0xFF & (dst_nonpremul >> 0));\n\n // Convert dst from nonpremul to premul.\n dr = (dr * da) / 0xFFFF;\n dg = (dg * da) / 0xFFFF;\n db = (db * da) / 0xFFFF;\n\n // Calculate the inverse of the src-alpha: how much of the dst to keep.\n uint32_t ia = 0xFFFF - sa;\n\n // Composite src (nonpremul) over dst (premul).\n da = sa + ((da * ia) / 0xFFFF);\n dr" + - " = ((sr * sa) + (dr * 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 // Convert from 16-bit color to 8-bit color and combine the components.\n da >>= 8;\n dr >>= 8;\n dg >>= 8;\n db >>= 8;\n return (db << 0) | (dg << 8) | (dr << 16) | (da << 24);\n}\n\nstatic inline uint32_t //\nwuffs_base__composite_premul_nonpremul_u32_axxx(uint32_t dst_premul,\n uint32_t src_nonpremul) {\n // Convert from 8-bit c" + - "olor 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(ui" + - "nt32_t dst_premul,\n uint32_t src_premul) {\n // Convert from 8-bit color to 16-bit color.\n uint32_t sa = 0x101 * (0xFF & (src_premul >> 24));\n uint32_t sr = 0x101 * (0xFF & (src_premul >> 16));\n uint32_t sg = 0x101 * (0xFF & (src_premul >> 8));\n uint32_t sb = 0x101 * (0xFF & (src_premul >> 0));\n uint32_t da = 0x101 * (0xFF & (dst_premul >> 24));\n uint32_t dr = 0x101 * (0xFF & (dst_premul >> 16));\n uint32_t dg = 0x101 * (0xFF & (dst_premul >> 8));\n uint32_t db = 0x101 * (0xFF & (dst_premul >> 0));\n\n // Calculate the inverse of the src-alpha: how much of the dst to keep.\n uint32_t ia = 0xFFFF - sa;\n\n // Composite src (premul) over dst (premul).\n da = sa + ((da * ia) / 0xFFFF);\n dr = sr + ((dr * ia) / 0xFFFF);\n dg = sg + ((dg * ia) / 0xFFFF);\n db = sb + ((db * ia) / 0xFFFF);\n\n // Convert from 16-bit color to 8-bit color and combine the components.\n da >>= 8;\n dr >>= 8;\n dg >>= 8;\n db >>= 8;\n return (db << 0) | (dg << 8) | (dr << 16) | (da << 24);\n}" + - "\n\n" + + "// ---------------- Pixel Swizzler\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\n" + + "" + + "// --------\n\nWUFFS_BASE__MAYBE_STATIC wuffs_base__color_u32_argb_premul //\nwuffs_base__pixel_buffer__color_u32_at(const wuffs_base__pixel_buffer* pb,\n uint32_t x,\n uint32_t y) {\n if (!pb || (x >= pb->pixcfg.private_impl.width) ||\n (y >= pb->pixcfg.private_impl.height)) {\n return 0;\n }\n\n if (wuffs_base__pixel_format__is_planar(&pb->pixcfg.private_impl.pixfmt)) {\n // TODO: support planar formats.\n return 0;\n }\n\n size_t stride = pb->private_impl.planes[0].stride;\n 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.plane" + + "s[3].ptr;\n return wuffs_base__load_u32le__no_bounds_check(palette +\n (4 * ((size_t)row[x])));\n }\n\n // Common formats above. Rarer formats below.\n\n case WUFFS_BASE__PIXEL_FORMAT__Y:\n return 0xFF000000 | (0x00010101 * ((uint32_t)(row[x])));\n\n case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_NONPREMUL: {\n uint8_t* palette = pb->private_impl.planes[3].ptr;\n return wuffs_base__color_u32_argb_nonpremul__as__color_u32_argb_premul(\n wuffs_base__load_u32le__no_bounds_check(palette +\n (4 * ((size_t)row[x]))));\n }\n\n case WUFFS_BASE__PIXEL_FORMAT__BGR_565:\n return wuffs_base__color_u16_rgb_565__as__color_u32_argb_premul(\n wuffs_base__load_u16le__no_bounds_check(row + (2 * ((size_t)x))));\n case WUFFS_BASE__PIXEL_FORMAT__BGR:\n return 0xFF000000 |\n wuffs_base__load_u24le__no_bounds_check(row + (3 * ((size_t)x)));\n case WUFFS_BASE__PIXEL_FORMAT_" + + "_BGRA_NONPREMUL:\n return wuffs_base__color_u32_argb_nonpremul__as__color_u32_argb_premul(\n wuffs_base__load_u32le__no_bounds_check(row + (4 * ((size_t)x))));\n case WUFFS_BASE__PIXEL_FORMAT__BGRX:\n return 0xFF000000 |\n wuffs_base__load_u32le__no_bounds_check(row + (4 * ((size_t)x)));\n\n case WUFFS_BASE__PIXEL_FORMAT__RGB:\n return wuffs_base__swap_u32_argb_abgr(\n 0xFF000000 |\n wuffs_base__load_u24le__no_bounds_check(row + (3 * ((size_t)x))));\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL:\n return wuffs_base__swap_u32_argb_abgr(\n wuffs_base__color_u32_argb_nonpremul__as__color_u32_argb_premul(\n wuffs_base__load_u32le__no_bounds_check(row +\n (4 * ((size_t)x)))));\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY:\n return wuffs_base__swap_u32_argb_abgr(\n wuffs_base__load_u32le__no_bounds_check(row + (4 * ((size_t)x)" + + ")));\n case WUFFS_BASE__PIXEL_FORMAT__RGBX:\n return wuffs_base__swap_u32_argb_abgr(\n 0xFF000000 |\n wuffs_base__load_u32le__no_bounds_check(row + (4 * ((size_t)x))));\n\n default:\n // TODO: support more formats.\n break;\n }\n\n return 0;\n}\n\nWUFFS_BASE__MAYBE_STATIC wuffs_base__status //\nwuffs_base__pixel_buffer__set_color_u32_at(\n wuffs_base__pixel_buffer* pb,\n uint32_t x,\n uint32_t y,\n wuffs_base__color_u32_argb_premul color) {\n if (!pb) {\n return wuffs_base__make_status(wuffs_base__error__bad_receiver);\n }\n if ((x >= pb->pixcfg.private_impl.width) ||\n (y >= pb->pixcfg.private_impl.height)) {\n return wuffs_base__make_status(wuffs_base__error__bad_argument);\n }\n\n if (wuffs_base__pixel_format__is_planar(&pb->pixcfg.private_impl.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),\n wuffs_base__color_u32_argb_premul__as__color_u8_gray(color));\n break;\n\n case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_BINARY:\n wuffs_base__store_u8__no_bounds_check(\n row + ((size_t)x), wuffs_base__pixel_palette__closest_element(\n wuffs_base__pixel_buffer__palette(pb),\n pb->pixcfg.private_impl.pixfmt, color));\n break;\n\n case WUFFS_BASE__PIXEL_FORMAT__BGR_565:\n wuffs_base__store_u16le__no_bounds_check(\n row + (2 * ((size_t)x)),\n wuffs_base__color_u32_argb_premul__as" + + "__color_u16_rgb_565(color));\n break;\n case WUFFS_BASE__PIXEL_FORMAT__BGR:\n wuffs_base__store_u24le__no_bounds_check(row + (3 * ((size_t)x)), color);\n break;\n case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL:\n wuffs_base__store_u32le__no_bounds_check(\n row + (4 * ((size_t)x)),\n wuffs_base__color_u32_argb_premul__as__color_u32_argb_nonpremul(\n color));\n break;\n\n case WUFFS_BASE__PIXEL_FORMAT__RGB:\n wuffs_base__store_u24le__no_bounds_check(\n row + (3 * ((size_t)x)), wuffs_base__swap_u32_argb_abgr(color));\n break;\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL:\n wuffs_base__store_u32le__no_bounds_check(\n row + (4 * ((size_t)x)),\n wuffs_base__color_u32_argb_premul__as__color_u32_argb_nonpremul(\n wuffs_base__swap_u32_argb_abgr(color)));\n break;\n case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__RGBX:\n wuffs_base__store_u32le__no_bounds_check(\n row" + + " + (4 * ((size_t)x)), wuffs_base__swap_u32_argb_abgr(color));\n break;\n\n default:\n // TODO: support more formats.\n return wuffs_base__make_status(wuffs_base__error__unsupported_option);\n }\n\n return wuffs_base__make_status(NULL);\n}\n\n" + + "" + + "// --------\n\nWUFFS_BASE__MAYBE_STATIC uint8_t //\nwuffs_base__pixel_palette__closest_element(\n wuffs_base__slice_u8 palette_slice,\n wuffs_base__pixel_format palette_format,\n wuffs_base__color_u32_argb_premul c) {\n size_t n = palette_slice.len / 4;\n if (n > 256) {\n n = 256;\n }\n size_t best_index = 0;\n uint64_t best_score = 0xFFFFFFFFFFFFFFFF;\n\n // Work in 16-bit color.\n uint32_t ca = 0x101 * (0xFF & (c >> 24));\n uint32_t cr = 0x101 * (0xFF & (c >> 16));\n uint32_t cg = 0x101 * (0xFF & (c >> 8));\n uint32_t cb = 0x101 * (0xFF & (c >> 0));\n\n switch (palette_format.repr) {\n case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_NONPREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_PREMUL:\n case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_BINARY: {\n bool nonpremul = palette_format.repr ==\n WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_NONPREMUL;\n\n size_t i;\n for (i = 0; i < n; i++) {\n // Work in 16-bit color.\n uint32_t pb = 0x101 * ((uint32_t)(palette" + + "_slice.ptr[(4 * i) + 0]));\n uint32_t pg = 0x101 * ((uint32_t)(palette_slice.ptr[(4 * i) + 1]));\n uint32_t pr = 0x101 * ((uint32_t)(palette_slice.ptr[(4 * i) + 2]));\n uint32_t pa = 0x101 * ((uint32_t)(palette_slice.ptr[(4 * i) + 3]));\n\n // Convert to premultiplied alpha.\n if (nonpremul && (pa != 0xFFFF)) {\n pb = (pb * pa) / 0xFFFF;\n pg = (pg * pa) / 0xFFFF;\n pr = (pr * pa) / 0xFFFF;\n }\n\n // These deltas are conceptually int32_t (signed) but after squaring,\n // it's equivalent to work in uint32_t (unsigned).\n pb -= cb;\n pg -= cg;\n pr -= cr;\n pa -= ca;\n uint64_t score = ((uint64_t)(pb * pb)) + ((uint64_t)(pg * pg)) +\n ((uint64_t)(pr * pr)) + ((uint64_t)(pa * pa));\n if (best_score > score) {\n best_score = score;\n best_index = i;\n }\n }\n break;\n }\n }\n\n return (uint8_t)best_index;\n}\n\n" + + "" + + "// --------\n\nstatic inline uint32_t //\nwuffs_base__composite_nonpremul_nonpremul_u32_axxx(uint32_t dst_nonpremul,\n uint32_t src_nonpremul) {\n // Convert from 8-bit color to 16-bit color.\n uint32_t sa = 0x101 * (0xFF & (src_nonpremul >> 24));\n uint32_t sr = 0x101 * (0xFF & (src_nonpremul >> 16));\n uint32_t sg = 0x101 * (0xFF & (src_nonpremul >> 8));\n uint32_t sb = 0x101 * (0xFF & (src_nonpremul >> 0));\n uint32_t da = 0x101 * (0xFF & (dst_nonpremul >> 24));\n uint32_t dr = 0x101 * (0xFF & (dst_nonpremul >> 16));\n uint32_t dg = 0x101 * (0xFF & (dst_nonpremul >> 8));\n uint32_t db = 0x101 * (0xFF & (dst_nonpremul >> 0));\n\n // Convert dst from nonpremul to premul.\n dr = (dr * da) / 0xFFFF;\n dg = (dg * da) / 0xFFFF;\n db = (db * da) / 0xFFFF;\n\n // Calculate the inverse of the src-alpha: how much of the dst to keep.\n uint32_t ia = 0xFFFF - sa;\n\n // Composite src (nonpremul) over dst (premul).\n da = sa + ((da * ia) / 0xFFFF);\n dr = ((sr * sa) + (dr * i" + + "a)) / 0xFFFF;\n dg = ((sg * sa) + (dg * ia)) / 0xFFFF;\n db = ((sb * sa) + (db * ia)) / 0xFFFF;\n\n // Convert dst from premul to nonpremul.\n if (da != 0) {\n dr = (dr * 0xFFFF) / da;\n dg = (dg * 0xFFFF) / da;\n db = (db * 0xFFFF) / da;\n }\n\n // Convert from 16-bit color to 8-bit color and combine the components.\n da >>= 8;\n dr >>= 8;\n dg >>= 8;\n db >>= 8;\n return (db << 0) | (dg << 8) | (dr << 16) | (da << 24);\n}\n\nstatic inline uint32_t //\nwuffs_base__composite_nonpremul_premul_u32_axxx(uint32_t dst_nonpremul,\n uint32_t src_premul) {\n // Convert from 8-bit color to 16-bit color.\n uint32_t sa = 0x101 * (0xFF & (src_premul >> 24));\n uint32_t sr = 0x101 * (0xFF & (src_premul >> 16));\n uint32_t sg = 0x101 * (0xFF & (src_premul >> 8));\n uint32_t sb = 0x101 * (0xFF & (src_premul >> 0));\n uint32_t da = 0x101 * (0xFF & (dst_nonpremul >> 24));\n uint32_t dr = 0x101 * (0xFF & (dst_nonpremul >> 16));\n uint32_t dg = 0x101 * (0xFF & (dst_nonpremul >> 8))" + + ";\n uint32_t db = 0x101 * (0xFF & (dst_nonpremul >> 0));\n\n // Convert dst from nonpremul to premul.\n dr = (dr * da) / 0xFFFF;\n dg = (dg * da) / 0xFFFF;\n db = (db * da) / 0xFFFF;\n\n // Calculate the inverse of the src-alpha: how much of the dst to keep.\n uint32_t ia = 0xFFFF - sa;\n\n // Composite src (premul) over dst (premul).\n da = sa + ((da * ia) / 0xFFFF);\n dr = sr + ((dr * ia) / 0xFFFF);\n dg = sg + ((dg * ia) / 0xFFFF);\n db = sb + ((db * ia) / 0xFFFF);\n\n // Convert dst from premul to nonpremul.\n if (da != 0) {\n dr = (dr * 0xFFFF) / da;\n dg = (dg * 0xFFFF) / da;\n db = (db * 0xFFFF) / da;\n }\n\n // Convert from 16-bit color to 8-bit color and combine the components.\n da >>= 8;\n dr >>= 8;\n dg >>= 8;\n db >>= 8;\n return (db << 0) | (dg << 8) | (dr << 16) | (da << 24);\n}\n\nstatic inline uint32_t //\nwuffs_base__composite_premul_nonpremul_u32_axxx(uint32_t dst_premul,\n uint32_t src_nonpremul) {\n // Convert from 8-bit color to 16-bit color.\n " + + " uint32_t sa = 0x101 * (0xFF & (src_nonpremul >> 24));\n uint32_t sr = 0x101 * (0xFF & (src_nonpremul >> 16));\n uint32_t sg = 0x101 * (0xFF & (src_nonpremul >> 8));\n uint32_t sb = 0x101 * (0xFF & (src_nonpremul >> 0));\n uint32_t da = 0x101 * (0xFF & (dst_premul >> 24));\n uint32_t dr = 0x101 * (0xFF & (dst_premul >> 16));\n uint32_t dg = 0x101 * (0xFF & (dst_premul >> 8));\n uint32_t db = 0x101 * (0xFF & (dst_premul >> 0));\n\n // Calculate the inverse of the src-alpha: how much of the dst to keep.\n uint32_t ia = 0xFFFF - sa;\n\n // Composite src (nonpremul) over dst (premul).\n da = sa + ((da * ia) / 0xFFFF);\n dr = ((sr * sa) + (dr * ia)) / 0xFFFF;\n dg = ((sg * sa) + (dg * ia)) / 0xFFFF;\n db = ((sb * sa) + (db * ia)) / 0xFFFF;\n\n // Convert from 16-bit color to 8-bit color and combine the components.\n da >>= 8;\n dr >>= 8;\n dg >>= 8;\n db >>= 8;\n return (db << 0) | (dg << 8) | (dr << 16) | (da << 24);\n}\n\nstatic inline uint32_t //\nwuffs_base__composite_premul_premul_u32_axxx(uint32_t dst_premul,\n " + + " uint32_t src_premul) {\n // Convert from 8-bit color to 16-bit color.\n uint32_t sa = 0x101 * (0xFF & (src_premul >> 24));\n uint32_t sr = 0x101 * (0xFF & (src_premul >> 16));\n uint32_t sg = 0x101 * (0xFF & (src_premul >> 8));\n uint32_t sb = 0x101 * (0xFF & (src_premul >> 0));\n uint32_t da = 0x101 * (0xFF & (dst_premul >> 24));\n uint32_t dr = 0x101 * (0xFF & (dst_premul >> 16));\n uint32_t dg = 0x101 * (0xFF & (dst_premul >> 8));\n uint32_t db = 0x101 * (0xFF & (dst_premul >> 0));\n\n // Calculate the inverse of the src-alpha: how much of the dst to keep.\n uint32_t ia = 0xFFFF - sa;\n\n // Composite src (premul) over dst (premul).\n da = sa + ((da * ia) / 0xFFFF);\n dr = sr + ((dr * ia) / 0xFFFF);\n dg = sg + ((dg * ia) / 0xFFFF);\n db = sb + ((db * ia) / 0xFFFF);\n\n // Convert from 16-bit color to 8-bit color and combine the components.\n da >>= 8;\n dr >>= 8;\n dg >>= 8;\n db >>= 8;\n return (db << 0) | (dg << 8) | (dr << 16) | (da << 24);\n}\n\n" + "" + "// --------\n\nstatic 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 uint32_t alpha = argb & 0xFF000000;\n wuffs_base__store_u32le__no_bounds_check(\n d, alpha | (r5 << 11) | (g6 << 5) | (b5 << 0));\n s += 4;\n d += 4;\n }\n return len4 * 4;\n}\n\nstatic uint64_t //\nwuffs_base__pixel_swizzler__swap_rgbx_bgrx(wuffs_base__slice_u8 dst,\n wuffs_base__slice_u8 src) {\n size_t len4 = (dst.len < src.len ? dst.len : src.len) / 4;\n uint8_t* d = dst.ptr;\n 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" + diff --git a/internal/cgen/gen.go b/internal/cgen/gen.go index b946828f..8ce09e0b 100644 --- a/internal/cgen/gen.go +++ b/internal/cgen/gen.go @@ -74,7 +74,6 @@ func genBase(out *bytes.Buffer) error { filename, varname string }{ {"base/all-impl.c", "baseAllImplC"}, - {"base/image-impl.c", "baseImageImplC"}, {"base/strconv-impl.c", "baseStrConvImplC"}, {"base/f64conv-submodule.c", "baseF64ConvSubmoduleC"}, diff --git a/release/c/wuffs-unsupported-snapshot.c b/release/c/wuffs-unsupported-snapshot.c index cba1b028..ca40912c 100644 --- a/release/c/wuffs-unsupported-snapshot.c +++ b/release/c/wuffs-unsupported-snapshot.c @@ -7917,6 +7917,11 @@ const uint64_t wuffs_base__low_bits_mask__u64[65] = { 0x7FFFFFFFFFFFFFFF, 0xFFFFFFFFFFFFFFFF, }; +const uint32_t wuffs_base__pixel_format__bits_per_channel[16] = { + 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, + 0x08, 0x0A, 0x0C, 0x10, 0x18, 0x20, 0x30, 0x40, +}; + const char* wuffs_base__note__end_of_data = "@base: end of data"; const char* wuffs_base__note__metadata_reported = "@base: metadata reported"; const char* wuffs_base__suspension__even_more_information = @@ -7971,249 +7976,6 @@ const char* wuffs_base__io_transformer__vtable_name = const char* wuffs_base__token_decoder__vtable_name = "{vtable}wuffs_base__token_decoder"; -// ---------------- Images - -const uint32_t wuffs_base__pixel_format__bits_per_channel[16] = { - 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, - 0x08, 0x0A, 0x0C, 0x10, 0x18, 0x20, 0x30, 0x40, -}; - -// -------- - -static inline uint32_t // -wuffs_base__swap_u32_argb_abgr(uint32_t u) { - uint32_t o = u & 0xFF00FF00; - uint32_t r = u & 0x00FF0000; - uint32_t b = u & 0x000000FF; - return o | (r >> 16) | (b << 16); -} - -// -------- - -WUFFS_BASE__MAYBE_STATIC wuffs_base__color_u32_argb_premul // -wuffs_base__pixel_buffer__color_u32_at(const wuffs_base__pixel_buffer* pb, - uint32_t x, - uint32_t y) { - if (!pb || (x >= pb->pixcfg.private_impl.width) || - (y >= pb->pixcfg.private_impl.height)) { - return 0; - } - - if (wuffs_base__pixel_format__is_planar(&pb->pixcfg.private_impl.pixfmt)) { - // TODO: support planar formats. - return 0; - } - - size_t stride = pb->private_impl.planes[0].stride; - uint8_t* row = pb->private_impl.planes[0].ptr + (stride * ((size_t)y)); - - switch (pb->pixcfg.private_impl.pixfmt.repr) { - case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL: - case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY: - return wuffs_base__load_u32le__no_bounds_check(row + (4 * ((size_t)x))); - - case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_PREMUL: - case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_BINARY: { - uint8_t* palette = pb->private_impl.planes[3].ptr; - return wuffs_base__load_u32le__no_bounds_check(palette + - (4 * ((size_t)row[x]))); - } - - // Common formats above. Rarer formats below. - - case WUFFS_BASE__PIXEL_FORMAT__Y: - return 0xFF000000 | (0x00010101 * ((uint32_t)(row[x]))); - - case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_NONPREMUL: { - uint8_t* palette = pb->private_impl.planes[3].ptr; - return wuffs_base__color_u32_argb_nonpremul__as__color_u32_argb_premul( - wuffs_base__load_u32le__no_bounds_check(palette + - (4 * ((size_t)row[x])))); - } - - case WUFFS_BASE__PIXEL_FORMAT__BGR_565: - return wuffs_base__color_u16_rgb_565__as__color_u32_argb_premul( - wuffs_base__load_u16le__no_bounds_check(row + (2 * ((size_t)x)))); - case WUFFS_BASE__PIXEL_FORMAT__BGR: - return 0xFF000000 | - wuffs_base__load_u24le__no_bounds_check(row + (3 * ((size_t)x))); - case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL: - return wuffs_base__color_u32_argb_nonpremul__as__color_u32_argb_premul( - wuffs_base__load_u32le__no_bounds_check(row + (4 * ((size_t)x)))); - case WUFFS_BASE__PIXEL_FORMAT__BGRX: - return 0xFF000000 | - wuffs_base__load_u32le__no_bounds_check(row + (4 * ((size_t)x))); - - case WUFFS_BASE__PIXEL_FORMAT__RGB: - return wuffs_base__swap_u32_argb_abgr( - 0xFF000000 | - wuffs_base__load_u24le__no_bounds_check(row + (3 * ((size_t)x)))); - case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL: - return wuffs_base__swap_u32_argb_abgr( - wuffs_base__color_u32_argb_nonpremul__as__color_u32_argb_premul( - wuffs_base__load_u32le__no_bounds_check(row + - (4 * ((size_t)x))))); - case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL: - case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY: - return wuffs_base__swap_u32_argb_abgr( - wuffs_base__load_u32le__no_bounds_check(row + (4 * ((size_t)x)))); - case WUFFS_BASE__PIXEL_FORMAT__RGBX: - return wuffs_base__swap_u32_argb_abgr( - 0xFF000000 | - wuffs_base__load_u32le__no_bounds_check(row + (4 * ((size_t)x)))); - - default: - // TODO: support more formats. - break; - } - - return 0; -} - -WUFFS_BASE__MAYBE_STATIC wuffs_base__status // -wuffs_base__pixel_buffer__set_color_u32_at( - wuffs_base__pixel_buffer* pb, - uint32_t x, - uint32_t y, - wuffs_base__color_u32_argb_premul color) { - if (!pb) { - return wuffs_base__make_status(wuffs_base__error__bad_receiver); - } - if ((x >= pb->pixcfg.private_impl.width) || - (y >= pb->pixcfg.private_impl.height)) { - return wuffs_base__make_status(wuffs_base__error__bad_argument); - } - - if (wuffs_base__pixel_format__is_planar(&pb->pixcfg.private_impl.pixfmt)) { - // TODO: support planar formats. - return wuffs_base__make_status(wuffs_base__error__unsupported_option); - } - - size_t stride = pb->private_impl.planes[0].stride; - uint8_t* row = pb->private_impl.planes[0].ptr + (stride * ((size_t)y)); - - switch (pb->pixcfg.private_impl.pixfmt.repr) { - case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL: - case WUFFS_BASE__PIXEL_FORMAT__BGRX: - wuffs_base__store_u32le__no_bounds_check(row + (4 * ((size_t)x)), color); - break; - - // Common formats above. Rarer formats below. - - case WUFFS_BASE__PIXEL_FORMAT__Y: - wuffs_base__store_u8__no_bounds_check( - row + ((size_t)x), - wuffs_base__color_u32_argb_premul__as__color_u8_gray(color)); - break; - - case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_BINARY: - wuffs_base__store_u8__no_bounds_check( - row + ((size_t)x), wuffs_base__pixel_palette__closest_element( - wuffs_base__pixel_buffer__palette(pb), - pb->pixcfg.private_impl.pixfmt, color)); - break; - - case WUFFS_BASE__PIXEL_FORMAT__BGR_565: - wuffs_base__store_u16le__no_bounds_check( - row + (2 * ((size_t)x)), - wuffs_base__color_u32_argb_premul__as__color_u16_rgb_565(color)); - break; - case WUFFS_BASE__PIXEL_FORMAT__BGR: - wuffs_base__store_u24le__no_bounds_check(row + (3 * ((size_t)x)), color); - break; - case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL: - wuffs_base__store_u32le__no_bounds_check( - row + (4 * ((size_t)x)), - wuffs_base__color_u32_argb_premul__as__color_u32_argb_nonpremul( - color)); - break; - - case WUFFS_BASE__PIXEL_FORMAT__RGB: - wuffs_base__store_u24le__no_bounds_check( - row + (3 * ((size_t)x)), wuffs_base__swap_u32_argb_abgr(color)); - break; - case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL: - wuffs_base__store_u32le__no_bounds_check( - row + (4 * ((size_t)x)), - wuffs_base__color_u32_argb_premul__as__color_u32_argb_nonpremul( - wuffs_base__swap_u32_argb_abgr(color))); - break; - case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL: - case WUFFS_BASE__PIXEL_FORMAT__RGBX: - wuffs_base__store_u32le__no_bounds_check( - row + (4 * ((size_t)x)), wuffs_base__swap_u32_argb_abgr(color)); - break; - - default: - // TODO: support more formats. - return wuffs_base__make_status(wuffs_base__error__unsupported_option); - } - - return wuffs_base__make_status(NULL); -} - -// -------- - -WUFFS_BASE__MAYBE_STATIC uint8_t // -wuffs_base__pixel_palette__closest_element( - wuffs_base__slice_u8 palette_slice, - wuffs_base__pixel_format palette_format, - wuffs_base__color_u32_argb_premul c) { - size_t n = palette_slice.len / 4; - if (n > 256) { - n = 256; - } - size_t best_index = 0; - uint64_t best_score = 0xFFFFFFFFFFFFFFFF; - - // Work in 16-bit color. - uint32_t ca = 0x101 * (0xFF & (c >> 24)); - uint32_t cr = 0x101 * (0xFF & (c >> 16)); - uint32_t cg = 0x101 * (0xFF & (c >> 8)); - uint32_t cb = 0x101 * (0xFF & (c >> 0)); - - switch (palette_format.repr) { - case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_NONPREMUL: - case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_PREMUL: - case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_BINARY: { - bool nonpremul = palette_format.repr == - WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_NONPREMUL; - - size_t i; - for (i = 0; i < n; i++) { - // Work in 16-bit color. - uint32_t pb = 0x101 * ((uint32_t)(palette_slice.ptr[(4 * i) + 0])); - uint32_t pg = 0x101 * ((uint32_t)(palette_slice.ptr[(4 * i) + 1])); - uint32_t pr = 0x101 * ((uint32_t)(palette_slice.ptr[(4 * i) + 2])); - uint32_t pa = 0x101 * ((uint32_t)(palette_slice.ptr[(4 * i) + 3])); - - // Convert to premultiplied alpha. - if (nonpremul && (pa != 0xFFFF)) { - pb = (pb * pa) / 0xFFFF; - pg = (pg * pa) / 0xFFFF; - pr = (pr * pa) / 0xFFFF; - } - - // These deltas are conceptually int32_t (signed) but after squaring, - // it's equivalent to work in uint32_t (unsigned). - pb -= cb; - pg -= cg; - pr -= cr; - pa -= ca; - uint64_t score = ((uint64_t)(pb * pb)) + ((uint64_t)(pg * pg)) + - ((uint64_t)(pr * pr)) + ((uint64_t)(pa * pa)); - if (best_score > score) { - best_score = score; - best_index = i; - } - } - break; - } - } - - return (uint8_t)best_index; -} - // ---------------- String Conversions // wuffs_base__parse_number__foo_digits entries are 0x00 for invalid digits, @@ -10567,6 +10329,242 @@ infinity: // ---------------- Pixel Swizzler +static inline uint32_t // +wuffs_base__swap_u32_argb_abgr(uint32_t u) { + uint32_t o = u & 0xFF00FF00; + uint32_t r = u & 0x00FF0000; + uint32_t b = u & 0x000000FF; + return o | (r >> 16) | (b << 16); +} + +// -------- + +WUFFS_BASE__MAYBE_STATIC wuffs_base__color_u32_argb_premul // +wuffs_base__pixel_buffer__color_u32_at(const wuffs_base__pixel_buffer* pb, + uint32_t x, + uint32_t y) { + if (!pb || (x >= pb->pixcfg.private_impl.width) || + (y >= pb->pixcfg.private_impl.height)) { + return 0; + } + + if (wuffs_base__pixel_format__is_planar(&pb->pixcfg.private_impl.pixfmt)) { + // TODO: support planar formats. + return 0; + } + + size_t stride = pb->private_impl.planes[0].stride; + uint8_t* row = pb->private_impl.planes[0].ptr + (stride * ((size_t)y)); + + switch (pb->pixcfg.private_impl.pixfmt.repr) { + case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL: + case WUFFS_BASE__PIXEL_FORMAT__BGRA_BINARY: + return wuffs_base__load_u32le__no_bounds_check(row + (4 * ((size_t)x))); + + case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_PREMUL: + case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_BINARY: { + uint8_t* palette = pb->private_impl.planes[3].ptr; + return wuffs_base__load_u32le__no_bounds_check(palette + + (4 * ((size_t)row[x]))); + } + + // Common formats above. Rarer formats below. + + case WUFFS_BASE__PIXEL_FORMAT__Y: + return 0xFF000000 | (0x00010101 * ((uint32_t)(row[x]))); + + case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_NONPREMUL: { + uint8_t* palette = pb->private_impl.planes[3].ptr; + return wuffs_base__color_u32_argb_nonpremul__as__color_u32_argb_premul( + wuffs_base__load_u32le__no_bounds_check(palette + + (4 * ((size_t)row[x])))); + } + + case WUFFS_BASE__PIXEL_FORMAT__BGR_565: + return wuffs_base__color_u16_rgb_565__as__color_u32_argb_premul( + wuffs_base__load_u16le__no_bounds_check(row + (2 * ((size_t)x)))); + case WUFFS_BASE__PIXEL_FORMAT__BGR: + return 0xFF000000 | + wuffs_base__load_u24le__no_bounds_check(row + (3 * ((size_t)x))); + case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL: + return wuffs_base__color_u32_argb_nonpremul__as__color_u32_argb_premul( + wuffs_base__load_u32le__no_bounds_check(row + (4 * ((size_t)x)))); + case WUFFS_BASE__PIXEL_FORMAT__BGRX: + return 0xFF000000 | + wuffs_base__load_u32le__no_bounds_check(row + (4 * ((size_t)x))); + + case WUFFS_BASE__PIXEL_FORMAT__RGB: + return wuffs_base__swap_u32_argb_abgr( + 0xFF000000 | + wuffs_base__load_u24le__no_bounds_check(row + (3 * ((size_t)x)))); + case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL: + return wuffs_base__swap_u32_argb_abgr( + wuffs_base__color_u32_argb_nonpremul__as__color_u32_argb_premul( + wuffs_base__load_u32le__no_bounds_check(row + + (4 * ((size_t)x))))); + case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL: + case WUFFS_BASE__PIXEL_FORMAT__RGBA_BINARY: + return wuffs_base__swap_u32_argb_abgr( + wuffs_base__load_u32le__no_bounds_check(row + (4 * ((size_t)x)))); + case WUFFS_BASE__PIXEL_FORMAT__RGBX: + return wuffs_base__swap_u32_argb_abgr( + 0xFF000000 | + wuffs_base__load_u32le__no_bounds_check(row + (4 * ((size_t)x)))); + + default: + // TODO: support more formats. + break; + } + + return 0; +} + +WUFFS_BASE__MAYBE_STATIC wuffs_base__status // +wuffs_base__pixel_buffer__set_color_u32_at( + wuffs_base__pixel_buffer* pb, + uint32_t x, + uint32_t y, + wuffs_base__color_u32_argb_premul color) { + if (!pb) { + return wuffs_base__make_status(wuffs_base__error__bad_receiver); + } + if ((x >= pb->pixcfg.private_impl.width) || + (y >= pb->pixcfg.private_impl.height)) { + return wuffs_base__make_status(wuffs_base__error__bad_argument); + } + + if (wuffs_base__pixel_format__is_planar(&pb->pixcfg.private_impl.pixfmt)) { + // TODO: support planar formats. + return wuffs_base__make_status(wuffs_base__error__unsupported_option); + } + + size_t stride = pb->private_impl.planes[0].stride; + uint8_t* row = pb->private_impl.planes[0].ptr + (stride * ((size_t)y)); + + switch (pb->pixcfg.private_impl.pixfmt.repr) { + case WUFFS_BASE__PIXEL_FORMAT__BGRA_PREMUL: + case WUFFS_BASE__PIXEL_FORMAT__BGRX: + wuffs_base__store_u32le__no_bounds_check(row + (4 * ((size_t)x)), color); + break; + + // Common formats above. Rarer formats below. + + case WUFFS_BASE__PIXEL_FORMAT__Y: + wuffs_base__store_u8__no_bounds_check( + row + ((size_t)x), + wuffs_base__color_u32_argb_premul__as__color_u8_gray(color)); + break; + + case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_BINARY: + wuffs_base__store_u8__no_bounds_check( + row + ((size_t)x), wuffs_base__pixel_palette__closest_element( + wuffs_base__pixel_buffer__palette(pb), + pb->pixcfg.private_impl.pixfmt, color)); + break; + + case WUFFS_BASE__PIXEL_FORMAT__BGR_565: + wuffs_base__store_u16le__no_bounds_check( + row + (2 * ((size_t)x)), + wuffs_base__color_u32_argb_premul__as__color_u16_rgb_565(color)); + break; + case WUFFS_BASE__PIXEL_FORMAT__BGR: + wuffs_base__store_u24le__no_bounds_check(row + (3 * ((size_t)x)), color); + break; + case WUFFS_BASE__PIXEL_FORMAT__BGRA_NONPREMUL: + wuffs_base__store_u32le__no_bounds_check( + row + (4 * ((size_t)x)), + wuffs_base__color_u32_argb_premul__as__color_u32_argb_nonpremul( + color)); + break; + + case WUFFS_BASE__PIXEL_FORMAT__RGB: + wuffs_base__store_u24le__no_bounds_check( + row + (3 * ((size_t)x)), wuffs_base__swap_u32_argb_abgr(color)); + break; + case WUFFS_BASE__PIXEL_FORMAT__RGBA_NONPREMUL: + wuffs_base__store_u32le__no_bounds_check( + row + (4 * ((size_t)x)), + wuffs_base__color_u32_argb_premul__as__color_u32_argb_nonpremul( + wuffs_base__swap_u32_argb_abgr(color))); + break; + case WUFFS_BASE__PIXEL_FORMAT__RGBA_PREMUL: + case WUFFS_BASE__PIXEL_FORMAT__RGBX: + wuffs_base__store_u32le__no_bounds_check( + row + (4 * ((size_t)x)), wuffs_base__swap_u32_argb_abgr(color)); + break; + + default: + // TODO: support more formats. + return wuffs_base__make_status(wuffs_base__error__unsupported_option); + } + + return wuffs_base__make_status(NULL); +} + +// -------- + +WUFFS_BASE__MAYBE_STATIC uint8_t // +wuffs_base__pixel_palette__closest_element( + wuffs_base__slice_u8 palette_slice, + wuffs_base__pixel_format palette_format, + wuffs_base__color_u32_argb_premul c) { + size_t n = palette_slice.len / 4; + if (n > 256) { + n = 256; + } + size_t best_index = 0; + uint64_t best_score = 0xFFFFFFFFFFFFFFFF; + + // Work in 16-bit color. + uint32_t ca = 0x101 * (0xFF & (c >> 24)); + uint32_t cr = 0x101 * (0xFF & (c >> 16)); + uint32_t cg = 0x101 * (0xFF & (c >> 8)); + uint32_t cb = 0x101 * (0xFF & (c >> 0)); + + switch (palette_format.repr) { + case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_NONPREMUL: + case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_PREMUL: + case WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_BINARY: { + bool nonpremul = palette_format.repr == + WUFFS_BASE__PIXEL_FORMAT__INDEXED__BGRA_NONPREMUL; + + size_t i; + for (i = 0; i < n; i++) { + // Work in 16-bit color. + uint32_t pb = 0x101 * ((uint32_t)(palette_slice.ptr[(4 * i) + 0])); + uint32_t pg = 0x101 * ((uint32_t)(palette_slice.ptr[(4 * i) + 1])); + uint32_t pr = 0x101 * ((uint32_t)(palette_slice.ptr[(4 * i) + 2])); + uint32_t pa = 0x101 * ((uint32_t)(palette_slice.ptr[(4 * i) + 3])); + + // Convert to premultiplied alpha. + if (nonpremul && (pa != 0xFFFF)) { + pb = (pb * pa) / 0xFFFF; + pg = (pg * pa) / 0xFFFF; + pr = (pr * pa) / 0xFFFF; + } + + // These deltas are conceptually int32_t (signed) but after squaring, + // it's equivalent to work in uint32_t (unsigned). + pb -= cb; + pg -= cg; + pr -= cr; + pa -= ca; + uint64_t score = ((uint64_t)(pb * pb)) + ((uint64_t)(pg * pg)) + + ((uint64_t)(pr * pr)) + ((uint64_t)(pa * pa)); + if (best_score > score) { + best_score = score; + best_index = i; + } + } + break; + } + } + + return (uint8_t)best_index; +} + +// -------- + static inline uint32_t // wuffs_base__composite_nonpremul_nonpremul_u32_axxx(uint32_t dst_nonpremul, uint32_t src_nonpremul) {