dany/libwebp
1
0
Fork 0
mirror of https://github.com/webmproject/libwebp.git synced 2025-08-29 23:32:05 +02:00
Code Issues Actions Packages Projects Releases Wiki Activity

apply clang-format

Browse Source 
(Debian clang-format version 19.1.7 (3+build4)) with `--style=Google`.

Manual changes:
* clang-format disabled around macros with stringification (mostly
  assembly)
* some inline assembly strings were adjusted to avoid awkward line
  breaks
* trailing commas, `//` or suffixes (`ull`) added to help array
  formatting
* thread_utils.c: parameter comments were changed to the more common
  /*...=*/ style to improve formatting

The automatically generated code under swig/ was skipped.

Bug: 433996651
Change-Id: Iea3f24160d78d2a2653971cdf13fa932e47ff1b3
This commit is contained in:
clang-format
2025-07-28 18:23:12 -07:00
committed by James Zern
parent b569988d3f
commit 44257cb826
224 changed files with 16312 additions and 16734 deletions
Download Patch File Download Diff File Expand all files Collapse all files

250
src/dsp/lossless_sse2.c
View File

@@ -129,8 +129,8 @@ static WEBP_INLINE uint32_t Average3_SSE2(uint32_t a0, uint32_t a1,
return (uint32_t)_mm_cvtsi128_si32(A2);
}
static WEBP_INLINE uint32_t Average4_SSE2(uint32_t a0, uint32_t a1,
uint32_t a2, uint32_t a3) {
static WEBP_INLINE uint32_t Average4_SSE2(uint32_t a0, uint32_t a1, uint32_t a2,
uint32_t a3) {
const __m128i avg1 = Average2_uint32_16_SSE2(a0, a1);
const __m128i avg2 = Average2_uint32_16_SSE2(a2, a3);
const __m128i sum = _mm_add_epi16(avg2, avg1);
@@ -233,21 +233,21 @@ static void PredictorAdd1_SSE2(const uint32_t* in, const uint32_t* upper,
// Macro that adds 32-bit integers from IN using mod 256 arithmetic
// per 8 bit channel.
#define GENERATE_PREDICTOR_1(X, IN) \
static void PredictorAdd##X##_SSE2(const uint32_t* in, const uint32_t* upper, \
int num_pixels, \
uint32_t* WEBP_RESTRICT out) { \
int i; \
for (i = 0; i + 4 <= num_pixels; i += 4) { \
const __m128i src = _mm_loadu_si128((const __m128i*)&in[i]); \
const __m128i other = _mm_loadu_si128((const __m128i*)&(IN)); \
const __m128i res = _mm_add_epi8(src, other); \
_mm_storeu_si128((__m128i*)&out[i], res); \
} \
if (i != num_pixels) { \
VP8LPredictorsAdd_C[(X)](in + i, upper + i, num_pixels - i, out + i); \
} \
}
#define GENERATE_PREDICTOR_1(X, IN) \
static void PredictorAdd##X##_SSE2(const uint32_t* in, \
const uint32_t* upper, int num_pixels, \
uint32_t* WEBP_RESTRICT out) { \
int i; \
for (i = 0; i + 4 <= num_pixels; i += 4) { \
const __m128i src = _mm_loadu_si128((const __m128i*)&in[i]); \
const __m128i other = _mm_loadu_si128((const __m128i*)&(IN)); \
const __m128i res = _mm_add_epi8(src, other); \
_mm_storeu_si128((__m128i*)&out[i], res); \
} \
if (i != num_pixels) { \
VP8LPredictorsAdd_C[(X)](in + i, upper + i, num_pixels - i, out + i); \
} \
}
// Predictor2: Top.
GENERATE_PREDICTOR_1(2, upper[i])
@@ -263,24 +263,24 @@ GENERATE_PREDICTOR_ADD(Predictor5_SSE2, PredictorAdd5_SSE2)
GENERATE_PREDICTOR_ADD(Predictor6_SSE2, PredictorAdd6_SSE2)
GENERATE_PREDICTOR_ADD(Predictor7_SSE2, PredictorAdd7_SSE2)
#define GENERATE_PREDICTOR_2(X, IN) \
static void PredictorAdd##X##_SSE2(const uint32_t* in, const uint32_t* upper, \
int num_pixels, \
uint32_t* WEBP_RESTRICT out) { \
int i; \
for (i = 0; i + 4 <= num_pixels; i += 4) { \
const __m128i Tother = _mm_loadu_si128((const __m128i*)&(IN)); \
const __m128i T = _mm_loadu_si128((const __m128i*)&upper[i]); \
const __m128i src = _mm_loadu_si128((const __m128i*)&in[i]); \
__m128i avg, res; \
Average2_m128i(&T, &Tother, &avg); \
res = _mm_add_epi8(avg, src); \
_mm_storeu_si128((__m128i*)&out[i], res); \
} \
if (i != num_pixels) { \
VP8LPredictorsAdd_C[(X)](in + i, upper + i, num_pixels - i, out + i); \
} \
}
#define GENERATE_PREDICTOR_2(X, IN) \
static void PredictorAdd##X##_SSE2(const uint32_t* in, \
const uint32_t* upper, int num_pixels, \
uint32_t* WEBP_RESTRICT out) { \
int i; \
for (i = 0; i + 4 <= num_pixels; i += 4) { \
const __m128i Tother = _mm_loadu_si128((const __m128i*)&(IN)); \
const __m128i T = _mm_loadu_si128((const __m128i*)&upper[i]); \
const __m128i src = _mm_loadu_si128((const __m128i*)&in[i]); \
__m128i avg, res; \
Average2_m128i(&T, &Tother, &avg); \
res = _mm_add_epi8(avg, src); \
_mm_storeu_si128((__m128i*)&out[i], res); \
} \
if (i != num_pixels) { \
VP8LPredictorsAdd_C[(X)](in + i, upper + i, num_pixels - i, out + i); \
} \
}
// Predictor8: average TL T.
GENERATE_PREDICTOR_2(8, upper[i - 1])
// Predictor9: average T TR.
@@ -288,20 +288,22 @@ GENERATE_PREDICTOR_2(9, upper[i + 1])
#undef GENERATE_PREDICTOR_2
// Predictor10: average of (average of (L,TL), average of (T, TR)).
#define DO_PRED10(OUT) do { \
__m128i avgLTL, avg; \
Average2_m128i(&L, &TL, &avgLTL); \
Average2_m128i(&avgTTR, &avgLTL, &avg); \
L = _mm_add_epi8(avg, src); \
out[i + (OUT)] = (uint32_t)_mm_cvtsi128_si32(L); \
} while (0)
#define DO_PRED10(OUT) \
do { \
__m128i avgLTL, avg; \
Average2_m128i(&L, &TL, &avgLTL); \
Average2_m128i(&avgTTR, &avgLTL, &avg); \
L = _mm_add_epi8(avg, src); \
out[i + (OUT)] = (uint32_t)_mm_cvtsi128_si32(L); \
} while (0)
#define DO_PRED10_SHIFT do { \
/* Rotate the pre-computed values for the next iteration.*/ \
avgTTR = _mm_srli_si128(avgTTR, 4); \
TL = _mm_srli_si128(TL, 4); \
src = _mm_srli_si128(src, 4); \
} while (0)
#define DO_PRED10_SHIFT \
do { \
/* Rotate the pre-computed values for the next iteration.*/ \
avgTTR = _mm_srli_si128(avgTTR, 4); \
TL = _mm_srli_si128(TL, 4); \
src = _mm_srli_si128(src, 4); \
} while (0)
static void PredictorAdd10_SSE2(const uint32_t* in, const uint32_t* upper,
int num_pixels, uint32_t* WEBP_RESTRICT out) {
@@ -330,25 +332,27 @@ static void PredictorAdd10_SSE2(const uint32_t* in, const uint32_t* upper,
#undef DO_PRED10_SHIFT
// Predictor11: select.
#define DO_PRED11(OUT) do { \
const __m128i L_lo = _mm_unpacklo_epi32(L, T); \
const __m128i TL_lo = _mm_unpacklo_epi32(TL, T); \
const __m128i pb = _mm_sad_epu8(L_lo, TL_lo); /* pb = sum |L-TL|*/ \
const __m128i mask = _mm_cmpgt_epi32(pb, pa); \
const __m128i A = _mm_and_si128(mask, L); \
const __m128i B = _mm_andnot_si128(mask, T); \
const __m128i pred = _mm_or_si128(A, B); /* pred = (pa > b)? L : T*/ \
L = _mm_add_epi8(src, pred); \
out[i + (OUT)] = (uint32_t)_mm_cvtsi128_si32(L); \
} while (0)
#define DO_PRED11(OUT) \
do { \
const __m128i L_lo = _mm_unpacklo_epi32(L, T); \
const __m128i TL_lo = _mm_unpacklo_epi32(TL, T); \
const __m128i pb = _mm_sad_epu8(L_lo, TL_lo); /* pb = sum |L-TL|*/ \
const __m128i mask = _mm_cmpgt_epi32(pb, pa); \
const __m128i A = _mm_and_si128(mask, L); \
const __m128i B = _mm_andnot_si128(mask, T); \
const __m128i pred = _mm_or_si128(A, B); /* pred = (pa > b)? L : T*/ \
L = _mm_add_epi8(src, pred); \
out[i + (OUT)] = (uint32_t)_mm_cvtsi128_si32(L); \
} while (0)
#define DO_PRED11_SHIFT do { \
/* Shift the pre-computed value for the next iteration.*/ \
T = _mm_srli_si128(T, 4); \
TL = _mm_srli_si128(TL, 4); \
src = _mm_srli_si128(src, 4); \
pa = _mm_srli_si128(pa, 4); \
} while (0)
#define DO_PRED11_SHIFT \
do { \
/* Shift the pre-computed value for the next iteration.*/ \
T = _mm_srli_si128(T, 4); \
TL = _mm_srli_si128(TL, 4); \
src = _mm_srli_si128(src, 4); \
pa = _mm_srli_si128(pa, 4); \
} while (0)
static void PredictorAdd11_SSE2(const uint32_t* in, const uint32_t* upper,
int num_pixels, uint32_t* WEBP_RESTRICT out) {
@@ -387,19 +391,21 @@ static void PredictorAdd11_SSE2(const uint32_t* in, const uint32_t* upper,
#undef DO_PRED11_SHIFT
// Predictor12: ClampedAddSubtractFull.
#define DO_PRED12(DIFF, LANE, OUT) do { \
const __m128i all = _mm_add_epi16(L, (DIFF)); \
const __m128i alls = _mm_packus_epi16(all, all); \
const __m128i res = _mm_add_epi8(src, alls); \
out[i + (OUT)] = (uint32_t)_mm_cvtsi128_si32(res); \
L = _mm_unpacklo_epi8(res, zero); \
} while (0)
#define DO_PRED12(DIFF, LANE, OUT) \
do { \
const __m128i all = _mm_add_epi16(L, (DIFF)); \
const __m128i alls = _mm_packus_epi16(all, all); \
const __m128i res = _mm_add_epi8(src, alls); \
out[i + (OUT)] = (uint32_t)_mm_cvtsi128_si32(res); \
L = _mm_unpacklo_epi8(res, zero); \
} while (0)
#define DO_PRED12_SHIFT(DIFF, LANE) do { \
/* Shift the pre-computed value for the next iteration.*/ \
if ((LANE) == 0) (DIFF) = _mm_srli_si128((DIFF), 8); \
src = _mm_srli_si128(src, 4); \
} while (0)
#define DO_PRED12_SHIFT(DIFF, LANE) \
do { \
/* Shift the pre-computed value for the next iteration.*/ \
if ((LANE) == 0) (DIFF) = _mm_srli_si128((DIFF), 8); \
src = _mm_srli_si128(src, 4); \
} while (0)
static void PredictorAdd12_SSE2(const uint32_t* in, const uint32_t* upper,
int num_pixels, uint32_t* WEBP_RESTRICT out) {
@@ -444,8 +450,8 @@ static void AddGreenToBlueAndRed_SSE2(const uint32_t* const src, int num_pixels,
uint32_t* dst) {
int i;
for (i = 0; i + 4 <= num_pixels; i += 4) {
const __m128i in = _mm_loadu_si128((const __m128i*)&src[i]); // argb
const __m128i A = _mm_srli_epi16(in, 8); // 0 a 0 g
const __m128i in = _mm_loadu_si128((const __m128i*)&src[i]); // argb
const __m128i A = _mm_srli_epi16(in, 8); // 0 a 0 g
const __m128i B = _mm_shufflelo_epi16(A, _MM_SHUFFLE(2, 2, 0, 0));
const __m128i C = _mm_shufflehi_epi16(B, _MM_SHUFFLE(2, 2, 0, 0)); // 0g0g
const __m128i out = _mm_add_epi8(in, C);
@@ -464,7 +470,7 @@ static void TransformColorInverse_SSE2(const VP8LMultipliers* const m,
const uint32_t* const src,
int num_pixels, uint32_t* dst) {
// sign-extended multiplying constants, pre-shifted by 5.
#define CST(X) (((int16_t)(m->X << 8)) >> 5) // sign-extend
#define CST(X) (((int16_t)(m->X << 8)) >> 5) // sign-extend
#define MK_CST_16(HI, LO) \
_mm_set1_epi32((int)(((uint32_t)(HI) << 16) | ((LO) & 0xffff)))
const __m128i mults_rb = MK_CST_16(CST(green_to_red), CST(green_to_blue));
@@ -474,17 +480,17 @@ static void TransformColorInverse_SSE2(const VP8LMultipliers* const m,
const __m128i mask_ag = _mm_set1_epi32((int)0xff00ff00); // alpha-green masks
int i;
for (i = 0; i + 4 <= num_pixels; i += 4) {
const __m128i in = _mm_loadu_si128((const __m128i*)&src[i]); // argb
const __m128i A = _mm_and_si128(in, mask_ag); // a 0 g 0
const __m128i in = _mm_loadu_si128((const __m128i*)&src[i]); // argb
const __m128i A = _mm_and_si128(in, mask_ag); // a 0 g 0
const __m128i B = _mm_shufflelo_epi16(A, _MM_SHUFFLE(2, 2, 0, 0));
const __m128i C = _mm_shufflehi_epi16(B, _MM_SHUFFLE(2, 2, 0, 0)); // g0g0
const __m128i D = _mm_mulhi_epi16(C, mults_rb); // x dr x db1
const __m128i E = _mm_add_epi8(in, D); // x r' x b'
const __m128i F = _mm_slli_epi16(E, 8); // r' 0 b' 0
const __m128i G = _mm_mulhi_epi16(F, mults_b2); // x db2 0 0
const __m128i H = _mm_srli_epi32(G, 8); // 0 x db2 0
const __m128i I = _mm_add_epi8(H, F); // r' x b'' 0
const __m128i J = _mm_srli_epi16(I, 8); // 0 r' 0 b''
const __m128i D = _mm_mulhi_epi16(C, mults_rb); // x dr x db1
const __m128i E = _mm_add_epi8(in, D); // x r' x b'
const __m128i F = _mm_slli_epi16(E, 8); // r' 0 b' 0
const __m128i G = _mm_mulhi_epi16(F, mults_b2); // x db2 0 0
const __m128i H = _mm_srli_epi32(G, 8); // 0 x db2 0
const __m128i I = _mm_add_epi8(H, F); // r' x b'' 0
const __m128i J = _mm_srli_epi16(I, 8); // 0 r' 0 b''
const __m128i out = _mm_or_si128(J, A);
_mm_storeu_si128((__m128i*)&dst[i], out);
}
@@ -569,21 +575,21 @@ static void ConvertBGRAToRGBA4444_SSE2(const uint32_t* WEBP_RESTRICT src,
const __m128i* in = (const __m128i*)src;
__m128i* out = (__m128i*)dst;
while (num_pixels >= 8) {
const __m128i bgra0 = _mm_loadu_si128(in++); // bgra0|bgra1|bgra2|bgra3
const __m128i bgra4 = _mm_loadu_si128(in++); // bgra4|bgra5|bgra6|bgra7
const __m128i bgra0 = _mm_loadu_si128(in++); // bgra0|bgra1|bgra2|bgra3
const __m128i bgra4 = _mm_loadu_si128(in++); // bgra4|bgra5|bgra6|bgra7
const __m128i v0l = _mm_unpacklo_epi8(bgra0, bgra4); // b0b4g0g4r0r4a0a4...
const __m128i v0h = _mm_unpackhi_epi8(bgra0, bgra4); // b2b6g2g6r2r6a2a6...
const __m128i v1l = _mm_unpacklo_epi8(v0l, v0h); // b0b2b4b6g0g2g4g6...
const __m128i v1h = _mm_unpackhi_epi8(v0l, v0h); // b1b3b5b7g1g3g5g7...
const __m128i v2l = _mm_unpacklo_epi8(v1l, v1h); // b0...b7 | g0...g7
const __m128i v2h = _mm_unpackhi_epi8(v1l, v1h); // r0...r7 | a0...a7
const __m128i ga0 = _mm_unpackhi_epi64(v2l, v2h); // g0...g7 | a0...a7
const __m128i rb0 = _mm_unpacklo_epi64(v2h, v2l); // r0...r7 | b0...b7
const __m128i ga1 = _mm_srli_epi16(ga0, 4); // g0-|g1-|...|a6-|a7-
const __m128i rb1 = _mm_and_si128(rb0, mask_0xf0); // -r0|-r1|...|-b6|-a7
const __m128i ga2 = _mm_and_si128(ga1, mask_0x0f); // g0-|g1-|...|a6-|a7-
const __m128i rgba0 = _mm_or_si128(ga2, rb1); // rg0..rg7 | ba0..ba7
const __m128i rgba1 = _mm_srli_si128(rgba0, 8); // ba0..ba7 | 0
const __m128i v1l = _mm_unpacklo_epi8(v0l, v0h); // b0b2b4b6g0g2g4g6...
const __m128i v1h = _mm_unpackhi_epi8(v0l, v0h); // b1b3b5b7g1g3g5g7...
const __m128i v2l = _mm_unpacklo_epi8(v1l, v1h); // b0...b7 | g0...g7
const __m128i v2h = _mm_unpackhi_epi8(v1l, v1h); // r0...r7 | a0...a7
const __m128i ga0 = _mm_unpackhi_epi64(v2l, v2h); // g0...g7 | a0...a7
const __m128i rb0 = _mm_unpacklo_epi64(v2h, v2l); // r0...r7 | b0...b7
const __m128i ga1 = _mm_srli_epi16(ga0, 4); // g0-|g1-|...|a6-|a7-
const __m128i rb1 = _mm_and_si128(rb0, mask_0xf0); // -r0|-r1|...|-b6|-a7
const __m128i ga2 = _mm_and_si128(ga1, mask_0x0f); // g0-|g1-|...|a6-|a7-
const __m128i rgba0 = _mm_or_si128(ga2, rb1); // rg0..rg7 | ba0..ba7
const __m128i rgba1 = _mm_srli_si128(rgba0, 8); // ba0..ba7 | 0
#if (WEBP_SWAP_16BIT_CSP == 1)
const __m128i rgba = _mm_unpacklo_epi8(rgba1, rgba0); // barg0...barg7
#else
@@ -607,8 +613,8 @@ static void ConvertBGRAToRGB565_SSE2(const uint32_t* WEBP_RESTRICT src,
const __m128i* in = (const __m128i*)src;
__m128i* out = (__m128i*)dst;
while (num_pixels >= 8) {
const __m128i bgra0 = _mm_loadu_si128(in++); // bgra0|bgra1|bgra2|bgra3
const __m128i bgra4 = _mm_loadu_si128(in++); // bgra4|bgra5|bgra6|bgra7
const __m128i bgra0 = _mm_loadu_si128(in++); // bgra0|bgra1|bgra2|bgra3
const __m128i bgra4 = _mm_loadu_si128(in++); // bgra4|bgra5|bgra6|bgra7
const __m128i v0l = _mm_unpacklo_epi8(bgra0, bgra4); // b0b4g0g4r0r4a0a4...
const __m128i v0h = _mm_unpackhi_epi8(bgra0, bgra4); // b2b6g2g6r2r6a2a6...
const __m128i v1l = _mm_unpacklo_epi8(v0l, v0h); // b0b2b4b6g0g2g4g6...
@@ -625,11 +631,11 @@ static void ConvertBGRAToRGB565_SSE2(const uint32_t* WEBP_RESTRICT src,
const __m128i b0 = _mm_srli_si128(rb1, 8); // -b0...-b7|0
const __m128i rg1 = _mm_or_si128(rb1, g_lo2); // gr0...gr7|xx
const __m128i b1 = _mm_srli_epi16(b0, 3);
const __m128i gb1 = _mm_or_si128(b1, g_hi2); // bg0...bg7|xx
const __m128i gb1 = _mm_or_si128(b1, g_hi2); // bg0...bg7|xx
#if (WEBP_SWAP_16BIT_CSP == 1)
const __m128i rgba = _mm_unpacklo_epi8(gb1, rg1); // rggb0...rggb7
const __m128i rgba = _mm_unpacklo_epi8(gb1, rg1); // rggb0...rggb7
#else
const __m128i rgba = _mm_unpacklo_epi8(rg1, gb1); // bgrb0...bgrb7
const __m128i rgba = _mm_unpacklo_epi8(rg1, gb1); // bgrb0...bgrb7
#endif
_mm_storeu_si128(out++, rgba);
num_pixels -= 8;
@@ -648,22 +654,22 @@ static void ConvertBGRAToBGR_SSE2(const uint32_t* WEBP_RESTRICT src,
const uint8_t* const end = dst + num_pixels * 3;
// the last storel_epi64 below writes 8 bytes starting at offset 18
while (dst + 26 <= end) {
const __m128i bgra0 = _mm_loadu_si128(in++); // bgra0|bgra1|bgra2|bgra3
const __m128i bgra4 = _mm_loadu_si128(in++); // bgra4|bgra5|bgra6|bgra7
const __m128i a0l = _mm_and_si128(bgra0, mask_l); // bgr0|0|bgr0|0
const __m128i a4l = _mm_and_si128(bgra4, mask_l); // bgr0|0|bgr0|0
const __m128i a0h = _mm_and_si128(bgra0, mask_h); // 0|bgr0|0|bgr0
const __m128i a4h = _mm_and_si128(bgra4, mask_h); // 0|bgr0|0|bgr0
const __m128i b0h = _mm_srli_epi64(a0h, 8); // 000b|gr00|000b|gr00
const __m128i b4h = _mm_srli_epi64(a4h, 8); // 000b|gr00|000b|gr00
const __m128i c0 = _mm_or_si128(a0l, b0h); // rgbrgb00|rgbrgb00
const __m128i c4 = _mm_or_si128(a4l, b4h); // rgbrgb00|rgbrgb00
const __m128i bgra0 = _mm_loadu_si128(in++); // bgra0|bgra1|bgra2|bgra3
const __m128i bgra4 = _mm_loadu_si128(in++); // bgra4|bgra5|bgra6|bgra7
const __m128i a0l = _mm_and_si128(bgra0, mask_l); // bgr0|0|bgr0|0
const __m128i a4l = _mm_and_si128(bgra4, mask_l); // bgr0|0|bgr0|0
const __m128i a0h = _mm_and_si128(bgra0, mask_h); // 0|bgr0|0|bgr0
const __m128i a4h = _mm_and_si128(bgra4, mask_h); // 0|bgr0|0|bgr0
const __m128i b0h = _mm_srli_epi64(a0h, 8); // 000b|gr00|000b|gr00
const __m128i b4h = _mm_srli_epi64(a4h, 8); // 000b|gr00|000b|gr00
const __m128i c0 = _mm_or_si128(a0l, b0h); // rgbrgb00|rgbrgb00
const __m128i c4 = _mm_or_si128(a4l, b4h); // rgbrgb00|rgbrgb00
const __m128i c2 = _mm_srli_si128(c0, 8);
const __m128i c6 = _mm_srli_si128(c4, 8);
_mm_storel_epi64((__m128i*)(dst + 0), c0);
_mm_storel_epi64((__m128i*)(dst + 6), c2);
_mm_storel_epi64((__m128i*)(dst + 12), c4);
_mm_storel_epi64((__m128i*)(dst + 18), c6);
_mm_storel_epi64((__m128i*)(dst + 0), c0);
_mm_storel_epi64((__m128i*)(dst + 6), c2);
_mm_storel_epi64((__m128i*)(dst + 12), c4);
_mm_storel_epi64((__m128i*)(dst + 18), c6);
dst += 24;
num_pixels -= 8;
}