base: tweak swizzle_ycck ycc_upsampling parameter

This moves from a 2-valued "basic (box) or fancy (triangle)" upsampling
parameter to a 3-valued one, which can now represent the fact that
libjpeg's and libwebp's fancy (triangle) upsampling is ever-so-slightly
different (see script/script/replicate-libwebp-chroma-upsampling.go and
its comments).

This is not an optimization or pessimization per se, but for the record,
the performance impact is mixed.

name                                             old speed     new speed     delta

wuffs_jpeg_decode_19k_8bpp/clang14               317MB/s ± 0%  314MB/s ± 0%  -0.99%  (p=0.008 n=5+5)
wuffs_jpeg_decode_30k_24bpp_progressive/clang14  318MB/s ± 0%  328MB/s ± 0%  +3.13%  (p=0.008 n=5+5)
wuffs_jpeg_decode_30k_24bpp_sequential/clang14   879MB/s ± 0%  871MB/s ± 0%  -0.83%  (p=0.008 n=5+5)
wuffs_jpeg_decode_77k_24bpp/clang14              719MB/s ± 0%  713MB/s ± 0%  -0.89%  (p=0.008 n=5+5)
wuffs_jpeg_decode_552k_24bpp_420/clang14         779MB/s ± 0%  771MB/s ± 0%  -1.01%  (p=0.008 n=5+5)
wuffs_jpeg_decode_552k_24bpp_444/clang14         902MB/s ± 0%  890MB/s ± 0%  -1.32%  (p=0.008 n=5+5)
wuffs_jpeg_decode_4002k_24bpp/clang14            784MB/s ± 0%  779MB/s ± 0%  -0.66%  (p=0.008 n=5+5)

wuffs_jpeg_decode_19k_8bpp/gcc12                 358MB/s ± 0%  362MB/s ± 0%  +0.93%  (p=0.008 n=5+5)
wuffs_jpeg_decode_30k_24bpp_progressive/gcc12    305MB/s ± 0%  311MB/s ± 0%  +1.91%  (p=0.008 n=5+5)
wuffs_jpeg_decode_30k_24bpp_sequential/gcc12     912MB/s ± 0%  918MB/s ± 0%  +0.73%  (p=0.016 n=4+5)
wuffs_jpeg_decode_77k_24bpp/gcc12                749MB/s ± 0%  763MB/s ± 0%  +1.83%  (p=0.008 n=5+5)
wuffs_jpeg_decode_552k_24bpp_420/gcc12           813MB/s ± 0%  824MB/s ± 0%  +1.37%  (p=0.008 n=5+5)
wuffs_jpeg_decode_552k_24bpp_444/gcc12           969MB/s ± 0%  984MB/s ± 0%  +1.62%  (p=0.008 n=5+5)
wuffs_jpeg_decode_4002k_24bpp/gcc12              815MB/s ± 0%  832MB/s ± 0%  +2.06%  (p=0.008 n=5+5)
This commit is contained in:
Nigel Tao committed 2026-07-03 21:38:28 +10:00
1 parent e3b5a7d506
commit 85a0283cfc
8 files changed
+211 -61

No files matched your search

+5 -5
View File
@@ -237,11 +237,11 @@ fuzz_swizzle_ycck(wuffs_base__io_buffer* src, uint64_t hash) {
uint32_t height_in_mcus = (3 & (hash >> 34)) + 1;
uint32_t allow_hv3 = 1 & (hash >> 36);
bool is_rgb_or_cmyk = 1 & (hash >> 37);
bool triangle_filter_for_2to1 = 1 & (hash >> 38);
uint8_t ycc_model = (1 & (hash >> 37)) ? 0x80 : 0x00;
uint8_t ycc_upsampling = 1 & (hash >> 38);
bool use_4th_component = 1 & (hash >> 39);
if (triangle_filter_for_2to1) {
if (ycc_upsampling) {
x_min_incl = 0;
y_min_incl = 0;
} else {
@@ -411,8 +411,8 @@ fuzz_swizzle_ycck(wuffs_base__io_buffer* src, uint64_t hash) {
width0, width1, width2, width3, //
h0, h1, h2, h3, //
v0, v1, v2, v3, //
is_rgb_or_cmyk, //
triangle_filter_for_2to1, //
ycc_model, //
ycc_upsampling, //
wuffs_base__make_slice_u8(scratch_buffer, sizeof(scratch_buffer)));
if (status.repr) {
return wuffs_base__status__message(&status);
+1 -1
View File
@@ -68,7 +68,7 @@ wuffs_base__pixel_swizzler__swizzle_ycck(
uint8_t v2,
uint8_t v3,
uint8_t ycc_model,
bool triangle_filter_for_2to1,
uint8_t ycc_upsampling,
wuffs_base__slice_u8 scratch_buffer_2k);
// ---------------- Images (Utility)
+13 -11
View File
@@ -445,10 +445,12 @@ wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle_x86_avx2(
size_t src_len,
uint32_t h1v2_bias_ignored,
bool first_column,
bool last_column) {
bool last_column,
uint8_t odd_column_rounding) {
uint8_t* dp = dst_ptr;
const uint8_t* sp_major = src_ptr_major;
const uint8_t* sp_minor = src_ptr_minor;
const uint8_t oc = odd_column_rounding;
if (first_column) {
src_len--;
@@ -456,7 +458,7 @@ wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle_x86_avx2(
uint32_t sv = (12u * ((uint32_t)(*sp_major++))) + //
(4u * ((uint32_t)(*sp_minor++)));
*dp++ = (uint8_t)((sv + 8u) >> 4u);
*dp++ = (uint8_t)((sv + 7u) >> 4u);
*dp++ = (uint8_t)((sv + oc) >> 4u);
return dst_ptr;
}
@@ -468,7 +470,7 @@ wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle_x86_avx2(
uint32_t sv = (9u * ((uint32_t)(*sp_major++))) + //
(3u * ((uint32_t)(*sp_minor++)));
*dp++ = (uint8_t)((sv + (3u * sv_major_m1) + (sv_minor_m1) + 8u) >> 4u);
*dp++ = (uint8_t)((sv + (3u * sv_major_p1) + (sv_minor_p1) + 7u) >> 4u);
*dp++ = (uint8_t)((sv + (3u * sv_major_p1) + (sv_minor_p1) + oc) >> 4u);
if (src_len <= 0u) {
return dst_ptr;
}
@@ -489,7 +491,7 @@ wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle_x86_avx2(
uint32_t sv = (9u * ((uint32_t)(*sp_major++))) + //
(3u * ((uint32_t)(*sp_minor++)));
*dp++ = (uint8_t)((sv + (3u * sv_major_m1) + (sv_minor_m1) + 8u) >> 4u);
*dp++ = (uint8_t)((sv + (3u * sv_major_p1) + (sv_minor_p1) + 7u) >> 4u);
*dp++ = (uint8_t)((sv + (3u * sv_major_p1) + (sv_minor_p1) + oc) >> 4u);
}
} else {
@@ -562,8 +564,8 @@ wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle_x86_avx2(
// the u16 right value by 4. On the right (p1), shift right by 4 and then
// shift left by 8 so that, when still in the u16x16 little-endian
// interpretation, we have:
// - m1_element = (etcetera + 8) >> 4
// - p1_element = ((etcetera + 7) >> 4) << 8
// - m1_element = (etcetera + 8) >> 4
// - p1_element = ((etcetera + oc) >> 4) << 8
//
// step4_m1_lo = [0x00?? 0x00?? ... 0x00?? 0x00??]
// step4_p1_lo = [0x??00 0x??00 ... 0x??00 0x??00]
@@ -572,14 +574,14 @@ wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle_x86_avx2(
__m256i step4_m1_lo = _mm256_srli_epi16(
_mm256_add_epi16(step3_m1_lo, _mm256_set1_epi16(8)), 4);
__m256i step4_p1_lo = _mm256_slli_epi16(
_mm256_srli_epi16(_mm256_add_epi16(step3_p1_lo, _mm256_set1_epi16(7)),
4),
_mm256_srli_epi16(
_mm256_add_epi16(step3_p1_lo, _mm256_set1_epi16(oc)), 4),
8);
__m256i step4_m1_hi = _mm256_srli_epi16(
_mm256_add_epi16(step3_m1_hi, _mm256_set1_epi16(8)), 4);
__m256i step4_p1_hi = _mm256_slli_epi16(
_mm256_srli_epi16(_mm256_add_epi16(step3_p1_hi, _mm256_set1_epi16(7)),
4),
_mm256_srli_epi16(
_mm256_add_epi16(step3_p1_hi, _mm256_set1_epi16(oc)), 4),
8);
// Bitwise-or two "0x00"-rich u16x16 vectors to get a u8x32 vector. Do
@@ -630,7 +632,7 @@ wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle_x86_avx2(
uint32_t sv = (9u * ((uint32_t)(*sp_major++))) + //
(3u * ((uint32_t)(*sp_minor++)));
*dp++ = (uint8_t)((sv + (3u * sv_major_m1) + (sv_minor_m1) + 8u) >> 4u);
*dp++ = (uint8_t)((sv + (3u * sv_major_p1) + (sv_minor_p1) + 7u) >> 4u);
*dp++ = (uint8_t)((sv + (3u * sv_major_p1) + (sv_minor_p1) + oc) >> 4u);
}
return dst_ptr;
+77 -13
View File
@@ -45,7 +45,36 @@ wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle_x86_avx2(
size_t src_len,
uint32_t h1v2_bias_ignored,
bool first_column,
bool last_column);
bool last_column,
uint8_t odd_column_rounding);
static const uint8_t* //
wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle_libjpeg_x86_avx2(
uint8_t* dst_ptr,
const uint8_t* src_ptr_major,
const uint8_t* src_ptr_minor,
size_t src_len,
uint32_t h1v2_bias_ignored,
bool first_column,
bool last_column) {
return wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle_x86_avx2(
dst_ptr, src_ptr_major, src_ptr_minor, src_len, h1v2_bias_ignored,
first_column, last_column, 7);
}
static const uint8_t* //
wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle_libwebp_x86_avx2(
uint8_t* dst_ptr,
const uint8_t* src_ptr_major,
const uint8_t* src_ptr_minor,
size_t src_len,
uint32_t h1v2_bias_ignored,
bool first_column,
bool last_column) {
return wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle_x86_avx2(
dst_ptr, src_ptr_major, src_ptr_minor, src_len, h1v2_bias_ignored,
first_column, last_column, 8);
}
#endif
#endif // defined(WUFFS_PRIVATE_IMPL__CPU_ARCH__X86_64_V3)
@@ -404,8 +433,8 @@ wuffs_private_impl__swizzle_ycc__upsample_inv_h1v2_triangle(
const uint8_t* src_ptr_minor,
size_t src_len,
uint32_t h1v2_bias,
bool first_column,
bool last_column) {
bool first_column_ignored,
bool last_column_ignored) {
uint8_t* dp = dst_ptr;
const uint8_t* sp_major = src_ptr_major;
const uint8_t* sp_minor = src_ptr_minor;
@@ -477,10 +506,12 @@ wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle(
size_t src_len,
uint32_t h1v2_bias_ignored,
bool first_column,
bool last_column) {
bool last_column,
uint8_t odd_column_rounding) {
uint8_t* dp = dst_ptr;
const uint8_t* sp_major = src_ptr_major;
const uint8_t* sp_minor = src_ptr_minor;
const uint8_t oc = odd_column_rounding;
if (first_column) {
src_len--;
@@ -488,7 +519,7 @@ wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle(
uint32_t sv = (12u * ((uint32_t)(*sp_major++))) + //
(4u * ((uint32_t)(*sp_minor++)));
*dp++ = (uint8_t)((sv + 8u) >> 4u);
*dp++ = (uint8_t)((sv + 7u) >> 4u);
*dp++ = (uint8_t)((sv + oc) >> 4u);
return dst_ptr;
}
@@ -500,7 +531,7 @@ wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle(
uint32_t sv = (9u * ((uint32_t)(*sp_major++))) + //
(3u * ((uint32_t)(*sp_minor++)));
*dp++ = (uint8_t)((sv + (3u * sv_major_m1) + (sv_minor_m1) + 8u) >> 4u);
*dp++ = (uint8_t)((sv + (3u * sv_major_p1) + (sv_minor_p1) + 7u) >> 4u);
*dp++ = (uint8_t)((sv + (3u * sv_major_p1) + (sv_minor_p1) + oc) >> 4u);
if (src_len <= 0u) {
return dst_ptr;
}
@@ -519,7 +550,7 @@ wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle(
uint32_t sv = (9u * ((uint32_t)(*sp_major++))) + //
(3u * ((uint32_t)(*sp_minor++)));
*dp++ = (uint8_t)((sv + (3u * sv_major_m1) + (sv_minor_m1) + 8u) >> 4u);
*dp++ = (uint8_t)((sv + (3u * sv_major_p1) + (sv_minor_p1) + 7u) >> 4u);
*dp++ = (uint8_t)((sv + (3u * sv_major_p1) + (sv_minor_p1) + oc) >> 4u);
}
if (last_column) {
@@ -531,12 +562,40 @@ wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle(
uint32_t sv = (9u * ((uint32_t)(*sp_major++))) + //
(3u * ((uint32_t)(*sp_minor++)));
*dp++ = (uint8_t)((sv + (3u * sv_major_m1) + (sv_minor_m1) + 8u) >> 4u);
*dp++ = (uint8_t)((sv + (3u * sv_major_p1) + (sv_minor_p1) + 7u) >> 4u);
*dp++ = (uint8_t)((sv + (3u * sv_major_p1) + (sv_minor_p1) + oc) >> 4u);
}
return dst_ptr;
}
static const uint8_t* //
wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle_libjpeg(
uint8_t* dst_ptr,
const uint8_t* src_ptr_major,
const uint8_t* src_ptr_minor,
size_t src_len,
uint32_t h1v2_bias_ignored,
bool first_column,
bool last_column) {
return wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle(
dst_ptr, src_ptr_major, src_ptr_minor, src_len, h1v2_bias_ignored,
first_column, last_column, 7);
}
static const uint8_t* //
wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle_libwebp(
uint8_t* dst_ptr,
const uint8_t* src_ptr_major,
const uint8_t* src_ptr_minor,
size_t src_len,
uint32_t h1v2_bias_ignored,
bool first_column,
bool last_column) {
return wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle(
dst_ptr, src_ptr_major, src_ptr_minor, src_len, h1v2_bias_ignored,
first_column, last_column, 8);
}
// wuffs_private_impl__swizzle_ycc__upsample_funcs is indexed by inv_h and then
// inv_v.
static const wuffs_private_impl__swizzle_ycc__upsample_func
@@ -1288,7 +1347,7 @@ wuffs_base__pixel_swizzler__swizzle_ycck(
uint8_t v2,
uint8_t v3,
uint8_t ycc_model,
bool triangle_filter_for_2to1,
uint8_t ycc_upsampling,
wuffs_base__slice_u8 scratch_buffer_2k) {
if (!p) {
return wuffs_base__make_status(wuffs_base__error__bad_receiver);
@@ -1303,7 +1362,7 @@ wuffs_base__pixel_swizzler__swizzle_ycck(
(4u <= ((unsigned int)v0 - 1u)) || //
(4u <= ((unsigned int)v1 - 1u)) || //
(4u <= ((unsigned int)v2 - 1u)) || //
(triangle_filter_for_2to1 && ((x_min_incl | y_min_incl) > 0u)) ||
((ycc_upsampling != 0) && ((x_min_incl | y_min_incl) > 0u)) ||
(scratch_buffer_2k.len < 2048u)) {
return wuffs_base__make_status(wuffs_base__error__bad_argument);
}
@@ -1611,7 +1670,7 @@ wuffs_base__pixel_swizzler__swizzle_ycck(
memcpy(&upfuncs, &wuffs_private_impl__swizzle_ycc__upsample_funcs,
sizeof upfuncs);
if (triangle_filter_for_2to1 &&
if ((ycc_upsampling != 0) &&
(wuffs_private_impl__swizzle_has_triangle_upsampler(inv_h0, inv_v0) ||
wuffs_private_impl__swizzle_has_triangle_upsampler(inv_h1, inv_v1) ||
wuffs_private_impl__swizzle_has_triangle_upsampler(inv_h2, inv_v2) ||
@@ -1621,7 +1680,10 @@ wuffs_base__pixel_swizzler__swizzle_ycck(
upfuncs[0][1] = wuffs_private_impl__swizzle_ycc__upsample_inv_h1v2_triangle;
upfuncs[1][0] = wuffs_private_impl__swizzle_ycc__upsample_inv_h2v1_triangle;
upfuncs[1][1] = wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle;
upfuncs[1][1] =
(ycc_upsampling == 1)
? wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle_libjpeg
: wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle_libwebp;
#if defined(WUFFS_PRIVATE_IMPL__CPU_ARCH__X86_64_V3)
#if defined(__GNUC__) && !defined(__clang__)
@@ -1639,7 +1701,9 @@ wuffs_base__pixel_swizzler__swizzle_ycck(
#else
if (wuffs_base__cpu_arch__have_x86_avx2()) {
upfuncs[1][1] =
wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle_x86_avx2;
(ycc_upsampling == 1)
? wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle_libjpeg_x86_avx2
: wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle_libwebp_x86_avx2;
}
#endif
#endif
+7 -1
View File
@@ -248,6 +248,12 @@ var Consts = [...]struct {
{t.IDU32, "0x01", "YCC_MODEL__BT_601_STUDIO_RANGE"},
{t.IDU32, "0x40", "YCC_MODEL__RGB"},
{t.IDU32, "0x80", "YCC_MODEL__CMY"},
// ----
{t.IDU32, "0x00", "YCC_UPSAMPLING__BASIC"},
{t.IDU32, "0x01", "YCC_UPSAMPLING__FANCY_LIKE_LIBJPEG"},
{t.IDU32, "0x02", "YCC_UPSAMPLING__FANCY_LIKE_LIBWEBP"},
}
var Statuses = [...]string{
@@ -742,7 +748,7 @@ var funcsOther = [...]string{
"v2: u8[..= 4]," +
"v3: u8[..= 4]," +
"ycc_model: u8," +
"triangle_filter_for_2to1: bool," +
"ycc_upsampling: u8," +
"scratch_buffer_2k: slice u8) status",
// ---- arm_crc32_utility
+98 -27
View File
@@ -19173,7 +19173,7 @@ wuffs_base__pixel_swizzler__swizzle_ycck(
uint8_t v2,
uint8_t v3,
uint8_t ycc_model,
bool triangle_filter_for_2to1,
uint8_t ycc_upsampling,
wuffs_base__slice_u8 scratch_buffer_2k);
// ---------------- Images (Utility)
@@ -32370,7 +32370,36 @@ wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle_x86_avx2(
size_t src_len,
uint32_t h1v2_bias_ignored,
bool first_column,
bool last_column);
bool last_column,
uint8_t odd_column_rounding);
static const uint8_t* //
wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle_libjpeg_x86_avx2(
uint8_t* dst_ptr,
const uint8_t* src_ptr_major,
const uint8_t* src_ptr_minor,
size_t src_len,
uint32_t h1v2_bias_ignored,
bool first_column,
bool last_column) {
return wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle_x86_avx2(
dst_ptr, src_ptr_major, src_ptr_minor, src_len, h1v2_bias_ignored,
first_column, last_column, 7);
}
static const uint8_t* //
wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle_libwebp_x86_avx2(
uint8_t* dst_ptr,
const uint8_t* src_ptr_major,
const uint8_t* src_ptr_minor,
size_t src_len,
uint32_t h1v2_bias_ignored,
bool first_column,
bool last_column) {
return wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle_x86_avx2(
dst_ptr, src_ptr_major, src_ptr_minor, src_len, h1v2_bias_ignored,
first_column, last_column, 8);
}
#endif
#endif // defined(WUFFS_PRIVATE_IMPL__CPU_ARCH__X86_64_V3)
@@ -32729,8 +32758,8 @@ wuffs_private_impl__swizzle_ycc__upsample_inv_h1v2_triangle(
const uint8_t* src_ptr_minor,
size_t src_len,
uint32_t h1v2_bias,
bool first_column,
bool last_column) {
bool first_column_ignored,
bool last_column_ignored) {
uint8_t* dp = dst_ptr;
const uint8_t* sp_major = src_ptr_major;
const uint8_t* sp_minor = src_ptr_minor;
@@ -32802,10 +32831,12 @@ wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle(
size_t src_len,
uint32_t h1v2_bias_ignored,
bool first_column,
bool last_column) {
bool last_column,
uint8_t odd_column_rounding) {
uint8_t* dp = dst_ptr;
const uint8_t* sp_major = src_ptr_major;
const uint8_t* sp_minor = src_ptr_minor;
const uint8_t oc = odd_column_rounding;
if (first_column) {
src_len--;
@@ -32813,7 +32844,7 @@ wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle(
uint32_t sv = (12u * ((uint32_t)(*sp_major++))) + //
(4u * ((uint32_t)(*sp_minor++)));
*dp++ = (uint8_t)((sv + 8u) >> 4u);
*dp++ = (uint8_t)((sv + 7u) >> 4u);
*dp++ = (uint8_t)((sv + oc) >> 4u);
return dst_ptr;
}
@@ -32825,7 +32856,7 @@ wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle(
uint32_t sv = (9u * ((uint32_t)(*sp_major++))) + //
(3u * ((uint32_t)(*sp_minor++)));
*dp++ = (uint8_t)((sv + (3u * sv_major_m1) + (sv_minor_m1) + 8u) >> 4u);
*dp++ = (uint8_t)((sv + (3u * sv_major_p1) + (sv_minor_p1) + 7u) >> 4u);
*dp++ = (uint8_t)((sv + (3u * sv_major_p1) + (sv_minor_p1) + oc) >> 4u);
if (src_len <= 0u) {
return dst_ptr;
}
@@ -32844,7 +32875,7 @@ wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle(
uint32_t sv = (9u * ((uint32_t)(*sp_major++))) + //
(3u * ((uint32_t)(*sp_minor++)));
*dp++ = (uint8_t)((sv + (3u * sv_major_m1) + (sv_minor_m1) + 8u) >> 4u);
*dp++ = (uint8_t)((sv + (3u * sv_major_p1) + (sv_minor_p1) + 7u) >> 4u);
*dp++ = (uint8_t)((sv + (3u * sv_major_p1) + (sv_minor_p1) + oc) >> 4u);
}
if (last_column) {
@@ -32856,12 +32887,40 @@ wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle(
uint32_t sv = (9u * ((uint32_t)(*sp_major++))) + //
(3u * ((uint32_t)(*sp_minor++)));
*dp++ = (uint8_t)((sv + (3u * sv_major_m1) + (sv_minor_m1) + 8u) >> 4u);
*dp++ = (uint8_t)((sv + (3u * sv_major_p1) + (sv_minor_p1) + 7u) >> 4u);
*dp++ = (uint8_t)((sv + (3u * sv_major_p1) + (sv_minor_p1) + oc) >> 4u);
}
return dst_ptr;
}
static const uint8_t* //
wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle_libjpeg(
uint8_t* dst_ptr,
const uint8_t* src_ptr_major,
const uint8_t* src_ptr_minor,
size_t src_len,
uint32_t h1v2_bias_ignored,
bool first_column,
bool last_column) {
return wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle(
dst_ptr, src_ptr_major, src_ptr_minor, src_len, h1v2_bias_ignored,
first_column, last_column, 7);
}
static const uint8_t* //
wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle_libwebp(
uint8_t* dst_ptr,
const uint8_t* src_ptr_major,
const uint8_t* src_ptr_minor,
size_t src_len,
uint32_t h1v2_bias_ignored,
bool first_column,
bool last_column) {
return wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle(
dst_ptr, src_ptr_major, src_ptr_minor, src_len, h1v2_bias_ignored,
first_column, last_column, 8);
}
// wuffs_private_impl__swizzle_ycc__upsample_funcs is indexed by inv_h and then
// inv_v.
static const wuffs_private_impl__swizzle_ycc__upsample_func
@@ -33613,7 +33672,7 @@ wuffs_base__pixel_swizzler__swizzle_ycck(
uint8_t v2,
uint8_t v3,
uint8_t ycc_model,
bool triangle_filter_for_2to1,
uint8_t ycc_upsampling,
wuffs_base__slice_u8 scratch_buffer_2k) {
if (!p) {
return wuffs_base__make_status(wuffs_base__error__bad_receiver);
@@ -33628,7 +33687,7 @@ wuffs_base__pixel_swizzler__swizzle_ycck(
(4u <= ((unsigned int)v0 - 1u)) || //
(4u <= ((unsigned int)v1 - 1u)) || //
(4u <= ((unsigned int)v2 - 1u)) || //
(triangle_filter_for_2to1 && ((x_min_incl | y_min_incl) > 0u)) ||
((ycc_upsampling != 0) && ((x_min_incl | y_min_incl) > 0u)) ||
(scratch_buffer_2k.len < 2048u)) {
return wuffs_base__make_status(wuffs_base__error__bad_argument);
}
@@ -33936,7 +33995,7 @@ wuffs_base__pixel_swizzler__swizzle_ycck(
memcpy(&upfuncs, &wuffs_private_impl__swizzle_ycc__upsample_funcs,
sizeof upfuncs);
if (triangle_filter_for_2to1 &&
if ((ycc_upsampling != 0) &&
(wuffs_private_impl__swizzle_has_triangle_upsampler(inv_h0, inv_v0) ||
wuffs_private_impl__swizzle_has_triangle_upsampler(inv_h1, inv_v1) ||
wuffs_private_impl__swizzle_has_triangle_upsampler(inv_h2, inv_v2) ||
@@ -33946,7 +34005,10 @@ wuffs_base__pixel_swizzler__swizzle_ycck(
upfuncs[0][1] = wuffs_private_impl__swizzle_ycc__upsample_inv_h1v2_triangle;
upfuncs[1][0] = wuffs_private_impl__swizzle_ycc__upsample_inv_h2v1_triangle;
upfuncs[1][1] = wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle;
upfuncs[1][1] =
(ycc_upsampling == 1)
? wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle_libjpeg
: wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle_libwebp;
#if defined(WUFFS_PRIVATE_IMPL__CPU_ARCH__X86_64_V3)
#if defined(__GNUC__) && !defined(__clang__)
@@ -33964,7 +34026,9 @@ wuffs_base__pixel_swizzler__swizzle_ycck(
#else
if (wuffs_base__cpu_arch__have_x86_avx2()) {
upfuncs[1][1] =
wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle_x86_avx2;
(ycc_upsampling == 1)
? wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle_libjpeg_x86_avx2
: wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle_libwebp_x86_avx2;
}
#endif
#endif
@@ -34439,10 +34503,12 @@ wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle_x86_avx2(
size_t src_len,
uint32_t h1v2_bias_ignored,
bool first_column,
bool last_column) {
bool last_column,
uint8_t odd_column_rounding) {
uint8_t* dp = dst_ptr;
const uint8_t* sp_major = src_ptr_major;
const uint8_t* sp_minor = src_ptr_minor;
const uint8_t oc = odd_column_rounding;
if (first_column) {
src_len--;
@@ -34450,7 +34516,7 @@ wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle_x86_avx2(
uint32_t sv = (12u * ((uint32_t)(*sp_major++))) + //
(4u * ((uint32_t)(*sp_minor++)));
*dp++ = (uint8_t)((sv + 8u) >> 4u);
*dp++ = (uint8_t)((sv + 7u) >> 4u);
*dp++ = (uint8_t)((sv + oc) >> 4u);
return dst_ptr;
}
@@ -34462,7 +34528,7 @@ wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle_x86_avx2(
uint32_t sv = (9u * ((uint32_t)(*sp_major++))) + //
(3u * ((uint32_t)(*sp_minor++)));
*dp++ = (uint8_t)((sv + (3u * sv_major_m1) + (sv_minor_m1) + 8u) >> 4u);
*dp++ = (uint8_t)((sv + (3u * sv_major_p1) + (sv_minor_p1) + 7u) >> 4u);
*dp++ = (uint8_t)((sv + (3u * sv_major_p1) + (sv_minor_p1) + oc) >> 4u);
if (src_len <= 0u) {
return dst_ptr;
}
@@ -34483,7 +34549,7 @@ wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle_x86_avx2(
uint32_t sv = (9u * ((uint32_t)(*sp_major++))) + //
(3u * ((uint32_t)(*sp_minor++)));
*dp++ = (uint8_t)((sv + (3u * sv_major_m1) + (sv_minor_m1) + 8u) >> 4u);
*dp++ = (uint8_t)((sv + (3u * sv_major_p1) + (sv_minor_p1) + 7u) >> 4u);
*dp++ = (uint8_t)((sv + (3u * sv_major_p1) + (sv_minor_p1) + oc) >> 4u);
}
} else {
@@ -34556,8 +34622,8 @@ wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle_x86_avx2(
// the u16 right value by 4. On the right (p1), shift right by 4 and then
// shift left by 8 so that, when still in the u16x16 little-endian
// interpretation, we have:
// - m1_element = (etcetera + 8) >> 4
// - p1_element = ((etcetera + 7) >> 4) << 8
// - m1_element = (etcetera + 8) >> 4
// - p1_element = ((etcetera + oc) >> 4) << 8
//
// step4_m1_lo = [0x00?? 0x00?? ... 0x00?? 0x00??]
// step4_p1_lo = [0x??00 0x??00 ... 0x??00 0x??00]
@@ -34566,14 +34632,14 @@ wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle_x86_avx2(
__m256i step4_m1_lo = _mm256_srli_epi16(
_mm256_add_epi16(step3_m1_lo, _mm256_set1_epi16(8)), 4);
__m256i step4_p1_lo = _mm256_slli_epi16(
_mm256_srli_epi16(_mm256_add_epi16(step3_p1_lo, _mm256_set1_epi16(7)),
4),
_mm256_srli_epi16(
_mm256_add_epi16(step3_p1_lo, _mm256_set1_epi16(oc)), 4),
8);
__m256i step4_m1_hi = _mm256_srli_epi16(
_mm256_add_epi16(step3_m1_hi, _mm256_set1_epi16(8)), 4);
__m256i step4_p1_hi = _mm256_slli_epi16(
_mm256_srli_epi16(_mm256_add_epi16(step3_p1_hi, _mm256_set1_epi16(7)),
4),
_mm256_srli_epi16(
_mm256_add_epi16(step3_p1_hi, _mm256_set1_epi16(oc)), 4),
8);
// Bitwise-or two "0x00"-rich u16x16 vectors to get a u8x32 vector. Do
@@ -34624,7 +34690,7 @@ wuffs_private_impl__swizzle_ycc__upsample_inv_h2v2_triangle_x86_avx2(
uint32_t sv = (9u * ((uint32_t)(*sp_major++))) + //
(3u * ((uint32_t)(*sp_minor++)));
*dp++ = (uint8_t)((sv + (3u * sv_major_m1) + (sv_minor_m1) + 8u) >> 4u);
*dp++ = (uint8_t)((sv + (3u * sv_major_p1) + (sv_minor_p1) + 7u) >> 4u);
*dp++ = (uint8_t)((sv + (3u * sv_major_p1) + (sv_minor_p1) + oc) >> 4u);
}
return dst_ptr;
@@ -57760,6 +57826,7 @@ wuffs_jpeg__decoder__swizzle_colorful(
uint32_t v_height1 = 0;
uint32_t v_height2 = 0;
uint32_t v_height3 = 0;
uint8_t v_ycc_upsampling = 0;
wuffs_base__status v_status = wuffs_base__make_status(NULL);
if (self->private_impl.f_swizzle_immediately) {
@@ -57821,6 +57888,10 @@ wuffs_jpeg__decoder__swizzle_colorful(
v_height3 = self->private_impl.f_components_workbuf_heights[3u];
}
}
v_ycc_upsampling = ((uint8_t)(0u));
if ( ! self->private_impl.f_use_lower_quality) {
v_ycc_upsampling = ((uint8_t)(1u));
}
v_status = wuffs_base__pixel_swizzler__swizzle_ycck(&self->private_impl.f_swizzler,
a_dst,
wuffs_base__pixel_buffer__palette_or_else(a_dst, wuffs_base__make_slice_u8(self->private_data.f_dst_palette, 1024)),
@@ -57853,7 +57924,7 @@ wuffs_jpeg__decoder__swizzle_colorful(
self->private_impl.f_components_v[2u],
self->private_impl.f_components_v[3u],
self->private_impl.f_ycc_model,
! self->private_impl.f_use_lower_quality,
v_ycc_upsampling,
wuffs_base__make_slice_u8(self->private_data.f_swizzle_ycck_scratch_buffer_2k, 2048));
return wuffs_private_impl__status__ensure_not_a_suspension(v_status);
}
@@ -84265,7 +84336,7 @@ wuffs_vp8__decoder__swizzle_one_row_of_macroblocks(
1u,
0u,
((uint8_t)(1u)),
false,
((uint8_t)(0u)),
wuffs_base__make_slice_u8(self->private_data.f_swizzle_ycck_scratch_buffer_2k, 2048));
return wuffs_private_impl__status__ensure_not_a_suspension(v_status);
}
+9 -2
View File
@@ -2072,7 +2072,9 @@ pri func decoder.swizzle_colorful!(dst: ptr base.pixel_buffer, workbuf: slice ba
var height1 : base.u32[..= 0x1_0008]
var height2 : base.u32[..= 0x1_0008]
var height3 : base.u32[..= 0x1_0008]
var status : base.status
var ycc_upsampling : base.u8
var status : base.status
if this.swizzle_immediately {
i = this.swizzle_immediately_c_offsets[0] as base.u64
@@ -2137,6 +2139,11 @@ pri func decoder.swizzle_colorful!(dst: ptr base.pixel_buffer, workbuf: slice ba
}
}
ycc_upsampling = base.YCC_UPSAMPLING__BASIC as base.u8
if not this.use_lower_quality {
ycc_upsampling = base.YCC_UPSAMPLING__FANCY_LIKE_LIBJPEG as base.u8
}
status = this.swizzler.swizzle_ycck!(
dst: args.dst,
dst_palette: args.dst.palette_or_else(fallback: this.dst_palette[..]),
@@ -2169,7 +2176,7 @@ pri func decoder.swizzle_colorful!(dst: ptr base.pixel_buffer, workbuf: slice ba
v2: this.components_v[2],
v3: this.components_v[3],
ycc_model: this.ycc_model,
triangle_filter_for_2to1: not this.use_lower_quality,
ycc_upsampling: ycc_upsampling,
scratch_buffer_2k: this.swizzle_ycck_scratch_buffer_2k[..])
return status
}
+1 -1
View File
@@ -425,7 +425,7 @@ pri func decoder.swizzle_one_row_of_macroblocks!(
v2: 1,
v3: 0,
ycc_model: base.YCC_MODEL__BT_601_STUDIO_RANGE as base.u8,
triangle_filter_for_2to1: false,
ycc_upsampling: base.YCC_UPSAMPLING__BASIC as base.u8,
scratch_buffer_2k: this.swizzle_ycck_scratch_buffer_2k[..])
return status
}