1/* 2 * Copyright 2012 The LibYuv Project Authors. All rights reserved. 3 * 4 * Use of this source code is governed by a BSD-style license 5 * that can be found in the LICENSE file in the root of the source 6 * tree. An additional intellectual property rights grant can be found 7 * in the file PATENTS. All contributing project authors may 8 * be found in the AUTHORS file in the root of the source tree. 9 */ 10 11#include "libyuv/row.h" 12 13#include <string.h> // For memset. 14 15#include "libyuv/basic_types.h" 16 17#ifdef __cplusplus 18namespace libyuv { 19extern "C" { 20#endif 21 22// Subsampled source needs to be increase by 1 of not even. 23#define SS(width, shift) (((width) + (1 << (shift)) - 1) >> (shift)) 24 25// Any 3 planes to 1. 26#define ANY31(NAMEANY, ANY_SIMD, UVSHIFT, DUVSHIFT, BPP, MASK) \ 27 void NAMEANY(const uint8* y_buf, const uint8* u_buf, const uint8* v_buf, \ 28 uint8* dst_ptr, int width) { \ 29 SIMD_ALIGNED(uint8 temp[64 * 4]); \ 30 memset(temp, 0, 64 * 3); /* for YUY2 and msan */ \ 31 int r = width & MASK; \ 32 int n = width & ~MASK; \ 33 if (n > 0) { \ 34 ANY_SIMD(y_buf, u_buf, v_buf, dst_ptr, n); \ 35 } \ 36 memcpy(temp, y_buf + n, r); \ 37 memcpy(temp + 64, u_buf + (n >> UVSHIFT), SS(r, UVSHIFT)); \ 38 memcpy(temp + 128, v_buf + (n >> UVSHIFT), SS(r, UVSHIFT)); \ 39 ANY_SIMD(temp, temp + 64, temp + 128, temp + 192, MASK + 1); \ 40 memcpy(dst_ptr + (n >> DUVSHIFT) * BPP, temp + 192, \ 41 SS(r, DUVSHIFT) * BPP); \ 42 } 43 44#ifdef HAS_I422TOARGBROW_SSSE3 45ANY31(I422ToARGBRow_Any_SSSE3, I422ToARGBRow_SSSE3, 1, 0, 4, 7) 46#endif 47#ifdef HAS_I444TOARGBROW_SSSE3 48ANY31(I444ToARGBRow_Any_SSSE3, I444ToARGBRow_SSSE3, 0, 0, 4, 7) 49ANY31(I411ToARGBRow_Any_SSSE3, I411ToARGBRow_SSSE3, 2, 0, 4, 7) 50ANY31(I422ToBGRARow_Any_SSSE3, I422ToBGRARow_SSSE3, 1, 0, 4, 7) 51ANY31(I422ToABGRRow_Any_SSSE3, I422ToABGRRow_SSSE3, 1, 0, 4, 7) 52ANY31(I422ToRGBARow_Any_SSSE3, I422ToRGBARow_SSSE3, 1, 0, 4, 7) 53ANY31(I422ToARGB4444Row_Any_SSSE3, I422ToARGB4444Row_SSSE3, 1, 0, 2, 7) 54ANY31(I422ToARGB1555Row_Any_SSSE3, I422ToARGB1555Row_SSSE3, 1, 0, 2, 7) 55ANY31(I422ToRGB565Row_Any_SSSE3, I422ToRGB565Row_SSSE3, 1, 0, 2, 7) 56ANY31(I422ToRGB24Row_Any_SSSE3, I422ToRGB24Row_SSSE3, 1, 0, 3, 7) 57ANY31(I422ToRAWRow_Any_SSSE3, I422ToRAWRow_SSSE3, 1, 0, 3, 7) 58ANY31(I422ToYUY2Row_Any_SSE2, I422ToYUY2Row_SSE2, 1, 1, 4, 15) 59ANY31(I422ToUYVYRow_Any_SSE2, I422ToUYVYRow_SSE2, 1, 1, 4, 15) 60#endif // HAS_I444TOARGBROW_SSSE3 61#ifdef HAS_I422TORGB24ROW_AVX2 62ANY31(I422ToRGB24Row_Any_AVX2, I422ToRGB24Row_AVX2, 1, 0, 3, 15) 63#endif 64#ifdef HAS_I422TORAWROW_AVX2 65ANY31(I422ToRAWRow_Any_AVX2, I422ToRAWRow_AVX2, 1, 0, 3, 15) 66#endif 67#ifdef HAS_J422TOARGBROW_SSSE3 68ANY31(J422ToARGBRow_Any_SSSE3, J422ToARGBRow_SSSE3, 1, 0, 4, 7) 69#endif 70#ifdef HAS_J422TOARGBROW_AVX2 71ANY31(J422ToARGBRow_Any_AVX2, J422ToARGBRow_AVX2, 1, 0, 4, 15) 72#endif 73#ifdef HAS_I422TOARGBROW_AVX2 74ANY31(I422ToARGBRow_Any_AVX2, I422ToARGBRow_AVX2, 1, 0, 4, 15) 75#endif 76#ifdef HAS_I422TOBGRAROW_AVX2 77ANY31(I422ToBGRARow_Any_AVX2, I422ToBGRARow_AVX2, 1, 0, 4, 15) 78#endif 79#ifdef HAS_I422TORGBAROW_AVX2 80ANY31(I422ToRGBARow_Any_AVX2, I422ToRGBARow_AVX2, 1, 0, 4, 15) 81#endif 82#ifdef HAS_I422TOABGRROW_AVX2 83ANY31(I422ToABGRRow_Any_AVX2, I422ToABGRRow_AVX2, 1, 0, 4, 15) 84#endif 85#ifdef HAS_I444TOARGBROW_AVX2 86ANY31(I444ToARGBRow_Any_AVX2, I444ToARGBRow_AVX2, 0, 0, 4, 15) 87#endif 88#ifdef HAS_I411TOARGBROW_AVX2 89ANY31(I411ToARGBRow_Any_AVX2, I411ToARGBRow_AVX2, 2, 0, 4, 15) 90#endif 91#ifdef HAS_I422TOARGB4444ROW_AVX2 92ANY31(I422ToARGB4444Row_Any_AVX2, I422ToARGB4444Row_AVX2, 1, 0, 2, 7) 93#endif 94#ifdef HAS_I422TOARGB1555ROW_AVX2 95ANY31(I422ToARGB1555Row_Any_AVX2, I422ToARGB1555Row_AVX2, 1, 0, 2, 7) 96#endif 97#ifdef HAS_I422TORGB565ROW_AVX2 98ANY31(I422ToRGB565Row_Any_AVX2, I422ToRGB565Row_AVX2, 1, 0, 2, 7) 99#endif 100#ifdef HAS_I422TOARGBROW_NEON 101ANY31(I444ToARGBRow_Any_NEON, I444ToARGBRow_NEON, 0, 0, 4, 7) 102ANY31(I422ToARGBRow_Any_NEON, I422ToARGBRow_NEON, 1, 0, 4, 7) 103ANY31(I411ToARGBRow_Any_NEON, I411ToARGBRow_NEON, 2, 0, 4, 7) 104ANY31(I422ToBGRARow_Any_NEON, I422ToBGRARow_NEON, 1, 0, 4, 7) 105ANY31(I422ToABGRRow_Any_NEON, I422ToABGRRow_NEON, 1, 0, 4, 7) 106ANY31(I422ToRGBARow_Any_NEON, I422ToRGBARow_NEON, 1, 0, 4, 7) 107ANY31(I422ToRGB24Row_Any_NEON, I422ToRGB24Row_NEON, 1, 0, 3, 7) 108ANY31(I422ToRAWRow_Any_NEON, I422ToRAWRow_NEON, 1, 0, 3, 7) 109ANY31(I422ToARGB4444Row_Any_NEON, I422ToARGB4444Row_NEON, 1, 0, 2, 7) 110ANY31(I422ToARGB1555Row_Any_NEON, I422ToARGB1555Row_NEON, 1, 0, 2, 7) 111ANY31(I422ToRGB565Row_Any_NEON, I422ToRGB565Row_NEON, 1, 0, 2, 7) 112#endif 113#ifdef HAS_I422TOYUY2ROW_NEON 114ANY31(I422ToYUY2Row_Any_NEON, I422ToYUY2Row_NEON, 1, 1, 4, 15) 115#endif 116#ifdef HAS_I422TOUYVYROW_NEON 117ANY31(I422ToUYVYRow_Any_NEON, I422ToUYVYRow_NEON, 1, 1, 4, 15) 118#endif 119#undef ANY31 120 121// Any 2 planes to 1. 122#define ANY21(NAMEANY, ANY_SIMD, UVSHIFT, SBPP, SBPP2, BPP, MASK) \ 123 void NAMEANY(const uint8* y_buf, const uint8* uv_buf, \ 124 uint8* dst_ptr, int width) { \ 125 SIMD_ALIGNED(uint8 temp[64 * 3]); \ 126 memset(temp, 0, 64 * 2); /* for msan */ \ 127 int r = width & MASK; \ 128 int n = width & ~MASK; \ 129 if (n > 0) { \ 130 ANY_SIMD(y_buf, uv_buf, dst_ptr, n); \ 131 } \ 132 memcpy(temp, y_buf + n * SBPP, r * SBPP); \ 133 memcpy(temp + 64, uv_buf + (n >> UVSHIFT) * SBPP2, \ 134 SS(r, UVSHIFT) * SBPP2); \ 135 ANY_SIMD(temp, temp + 64, temp + 128, MASK + 1); \ 136 memcpy(dst_ptr + n * BPP, temp + 128, r * BPP); \ 137 } 138 139// Biplanar to RGB. 140#ifdef HAS_NV12TOARGBROW_SSSE3 141ANY21(NV12ToARGBRow_Any_SSSE3, NV12ToARGBRow_SSSE3, 1, 1, 2, 4, 7) 142ANY21(NV21ToARGBRow_Any_SSSE3, NV21ToARGBRow_SSSE3, 1, 1, 2, 4, 7) 143#endif 144#ifdef HAS_NV12TOARGBROW_AVX2 145ANY21(NV12ToARGBRow_Any_AVX2, NV12ToARGBRow_AVX2, 1, 1, 2, 4, 15) 146ANY21(NV21ToARGBRow_Any_AVX2, NV21ToARGBRow_AVX2, 1, 1, 2, 4, 15) 147#endif 148#ifdef HAS_NV12TOARGBROW_NEON 149ANY21(NV12ToARGBRow_Any_NEON, NV12ToARGBRow_NEON, 1, 1, 2, 4, 7) 150ANY21(NV21ToARGBRow_Any_NEON, NV21ToARGBRow_NEON, 1, 1, 2, 4, 7) 151#endif 152#ifdef HAS_NV12TORGB565ROW_SSSE3 153ANY21(NV12ToRGB565Row_Any_SSSE3, NV12ToRGB565Row_SSSE3, 1, 1, 2, 2, 7) 154ANY21(NV21ToRGB565Row_Any_SSSE3, NV21ToRGB565Row_SSSE3, 1, 1, 2, 2, 7) 155#endif 156#ifdef HAS_NV12TORGB565ROW_AVX2 157ANY21(NV12ToRGB565Row_Any_AVX2, NV12ToRGB565Row_AVX2, 1, 1, 2, 2, 15) 158ANY21(NV21ToRGB565Row_Any_AVX2, NV21ToRGB565Row_AVX2, 1, 1, 2, 2, 15) 159#endif 160#ifdef HAS_NV12TORGB565ROW_NEON 161ANY21(NV12ToRGB565Row_Any_NEON, NV12ToRGB565Row_NEON, 1, 1, 2, 2, 7) 162ANY21(NV21ToRGB565Row_Any_NEON, NV21ToRGB565Row_NEON, 1, 1, 2, 2, 7) 163#endif 164 165// Merge functions. 166#ifdef HAS_MERGEUVROW_SSE2 167ANY21(MergeUVRow_Any_SSE2, MergeUVRow_SSE2, 0, 1, 1, 2, 15) 168#endif 169#ifdef HAS_MERGEUVROW_AVX2 170ANY21(MergeUVRow_Any_AVX2, MergeUVRow_AVX2, 0, 1, 1, 2, 31) 171#endif 172#ifdef HAS_MERGEUVROW_NEON 173ANY21(MergeUVRow_Any_NEON, MergeUVRow_NEON, 0, 1, 1, 2, 15) 174#endif 175 176// Math functions. 177#ifdef HAS_ARGBMULTIPLYROW_SSE2 178ANY21(ARGBMultiplyRow_Any_SSE2, ARGBMultiplyRow_SSE2, 0, 4, 4, 4, 3) 179#endif 180#ifdef HAS_ARGBADDROW_SSE2 181ANY21(ARGBAddRow_Any_SSE2, ARGBAddRow_SSE2, 0, 4, 4, 4, 3) 182#endif 183#ifdef HAS_ARGBSUBTRACTROW_SSE2 184ANY21(ARGBSubtractRow_Any_SSE2, ARGBSubtractRow_SSE2, 0, 4, 4, 4, 3) 185#endif 186#ifdef HAS_ARGBMULTIPLYROW_AVX2 187ANY21(ARGBMultiplyRow_Any_AVX2, ARGBMultiplyRow_AVX2, 0, 4, 4, 4, 7) 188#endif 189#ifdef HAS_ARGBADDROW_AVX2 190ANY21(ARGBAddRow_Any_AVX2, ARGBAddRow_AVX2, 0, 4, 4, 4, 7) 191#endif 192#ifdef HAS_ARGBSUBTRACTROW_AVX2 193ANY21(ARGBSubtractRow_Any_AVX2, ARGBSubtractRow_AVX2, 0, 4, 4, 4, 7) 194#endif 195#ifdef HAS_ARGBMULTIPLYROW_NEON 196ANY21(ARGBMultiplyRow_Any_NEON, ARGBMultiplyRow_NEON, 0, 4, 4, 4, 7) 197#endif 198#ifdef HAS_ARGBADDROW_NEON 199ANY21(ARGBAddRow_Any_NEON, ARGBAddRow_NEON, 0, 4, 4, 4, 7) 200#endif 201#ifdef HAS_ARGBSUBTRACTROW_NEON 202ANY21(ARGBSubtractRow_Any_NEON, ARGBSubtractRow_NEON, 0, 4, 4, 4, 7) 203#endif 204#ifdef HAS_SOBELROW_SSE2 205ANY21(SobelRow_Any_SSE2, SobelRow_SSE2, 0, 1, 1, 4, 15) 206#endif 207#ifdef HAS_SOBELROW_NEON 208ANY21(SobelRow_Any_NEON, SobelRow_NEON, 0, 1, 1, 4, 7) 209#endif 210#ifdef HAS_SOBELTOPLANEROW_SSE2 211ANY21(SobelToPlaneRow_Any_SSE2, SobelToPlaneRow_SSE2, 0, 1, 1, 1, 15) 212#endif 213#ifdef HAS_SOBELTOPLANEROW_NEON 214ANY21(SobelToPlaneRow_Any_NEON, SobelToPlaneRow_NEON, 0, 1, 1, 1, 15) 215#endif 216#ifdef HAS_SOBELXYROW_SSE2 217ANY21(SobelXYRow_Any_SSE2, SobelXYRow_SSE2, 0, 1, 1, 4, 15) 218#endif 219#ifdef HAS_SOBELXYROW_NEON 220ANY21(SobelXYRow_Any_NEON, SobelXYRow_NEON, 0, 1, 1, 4, 7) 221#endif 222#undef ANY21 223 224// Any 1 to 1. 225#define ANY11(NAMEANY, ANY_SIMD, UVSHIFT, SBPP, BPP, MASK) \ 226 void NAMEANY(const uint8* src_ptr, uint8* dst_ptr, int width) { \ 227 SIMD_ALIGNED(uint8 temp[128 * 2]); \ 228 memset(temp, 0, 128); /* for YUY2 and msan */ \ 229 int r = width & MASK; \ 230 int n = width & ~MASK; \ 231 if (n > 0) { \ 232 ANY_SIMD(src_ptr, dst_ptr, n); \ 233 } \ 234 memcpy(temp, src_ptr + (n >> UVSHIFT) * SBPP, SS(r, UVSHIFT) * SBPP); \ 235 ANY_SIMD(temp, temp + 128, MASK + 1); \ 236 memcpy(dst_ptr + n * BPP, temp + 128, r * BPP); \ 237 } 238 239#ifdef HAS_COPYROW_AVX 240ANY11(CopyRow_Any_AVX, CopyRow_AVX, 0, 1, 1, 63) 241#endif 242#ifdef HAS_COPYROW_SSE2 243ANY11(CopyRow_Any_SSE2, CopyRow_SSE2, 0, 1, 1, 31) 244#endif 245#ifdef HAS_COPYROW_NEON 246ANY11(CopyRow_Any_NEON, CopyRow_NEON, 0, 1, 1, 31) 247#endif 248#if defined(HAS_ARGBTORGB24ROW_SSSE3) 249ANY11(ARGBToRGB24Row_Any_SSSE3, ARGBToRGB24Row_SSSE3, 0, 4, 3, 15) 250ANY11(ARGBToRAWRow_Any_SSSE3, ARGBToRAWRow_SSSE3, 0, 4, 3, 15) 251ANY11(ARGBToRGB565Row_Any_SSE2, ARGBToRGB565Row_SSE2, 0, 4, 2, 3) 252ANY11(ARGBToARGB1555Row_Any_SSE2, ARGBToARGB1555Row_SSE2, 0, 4, 2, 3) 253ANY11(ARGBToARGB4444Row_Any_SSE2, ARGBToARGB4444Row_SSE2, 0, 4, 2, 3) 254#endif 255#if defined(HAS_ARGBTOARGB4444ROW_AVX2) 256ANY11(ARGBToRGB565Row_Any_AVX2, ARGBToRGB565Row_AVX2, 0, 4, 2, 7) 257ANY11(ARGBToARGB1555Row_Any_AVX2, ARGBToARGB1555Row_AVX2, 0, 4, 2, 7) 258ANY11(ARGBToARGB4444Row_Any_AVX2, ARGBToARGB4444Row_AVX2, 0, 4, 2, 7) 259#endif 260#if defined(HAS_J400TOARGBROW_SSE2) 261ANY11(J400ToARGBRow_Any_SSE2, J400ToARGBRow_SSE2, 0, 1, 4, 7) 262#endif 263#if defined(HAS_J400TOARGBROW_AVX2) 264ANY11(J400ToARGBRow_Any_AVX2, J400ToARGBRow_AVX2, 0, 1, 4, 15) 265#endif 266#if defined(HAS_I400TOARGBROW_SSE2) 267ANY11(I400ToARGBRow_Any_SSE2, I400ToARGBRow_SSE2, 0, 1, 4, 7) 268#endif 269#if defined(HAS_I400TOARGBROW_AVX2) 270ANY11(I400ToARGBRow_Any_AVX2, I400ToARGBRow_AVX2, 0, 1, 4, 15) 271#endif 272#if defined(HAS_YUY2TOARGBROW_SSSE3) 273ANY11(YUY2ToARGBRow_Any_SSSE3, YUY2ToARGBRow_SSSE3, 1, 4, 4, 15) 274ANY11(UYVYToARGBRow_Any_SSSE3, UYVYToARGBRow_SSSE3, 1, 4, 4, 15) 275ANY11(RGB24ToARGBRow_Any_SSSE3, RGB24ToARGBRow_SSSE3, 0, 3, 4, 15) 276ANY11(RAWToARGBRow_Any_SSSE3, RAWToARGBRow_SSSE3, 0, 3, 4, 15) 277ANY11(RGB565ToARGBRow_Any_SSE2, RGB565ToARGBRow_SSE2, 0, 2, 4, 7) 278ANY11(ARGB1555ToARGBRow_Any_SSE2, ARGB1555ToARGBRow_SSE2, 0, 2, 4, 7) 279ANY11(ARGB4444ToARGBRow_Any_SSE2, ARGB4444ToARGBRow_SSE2, 0, 2, 4, 7) 280#endif 281#if defined(HAS_RGB565TOARGBROW_AVX2) 282ANY11(RGB565ToARGBRow_Any_AVX2, RGB565ToARGBRow_AVX2, 0, 2, 4, 15) 283#endif 284#if defined(HAS_ARGB1555TOARGBROW_AVX2) 285ANY11(ARGB1555ToARGBRow_Any_AVX2, ARGB1555ToARGBRow_AVX2, 0, 2, 4, 15) 286#endif 287#if defined(HAS_ARGB4444TOARGBROW_AVX2) 288ANY11(ARGB4444ToARGBRow_Any_AVX2, ARGB4444ToARGBRow_AVX2, 0, 2, 4, 15) 289#endif 290#if defined(HAS_YUY2TOARGBROW_AVX2) 291ANY11(YUY2ToARGBRow_Any_AVX2, YUY2ToARGBRow_AVX2, 1, 4, 4, 31) 292ANY11(UYVYToARGBRow_Any_AVX2, UYVYToARGBRow_AVX2, 1, 4, 4, 31) 293#endif 294#if defined(HAS_ARGBTORGB24ROW_NEON) 295ANY11(ARGBToRGB24Row_Any_NEON, ARGBToRGB24Row_NEON, 0, 4, 3, 7) 296ANY11(ARGBToRAWRow_Any_NEON, ARGBToRAWRow_NEON, 0, 4, 3, 7) 297ANY11(ARGBToRGB565Row_Any_NEON, ARGBToRGB565Row_NEON, 0, 4, 2, 7) 298ANY11(ARGBToARGB1555Row_Any_NEON, ARGBToARGB1555Row_NEON, 0, 4, 2, 7) 299ANY11(ARGBToARGB4444Row_Any_NEON, ARGBToARGB4444Row_NEON, 0, 4, 2, 7) 300ANY11(J400ToARGBRow_Any_NEON, J400ToARGBRow_NEON, 0, 1, 4, 7) 301ANY11(I400ToARGBRow_Any_NEON, I400ToARGBRow_NEON, 0, 1, 4, 7) 302ANY11(YUY2ToARGBRow_Any_NEON, YUY2ToARGBRow_NEON, 1, 4, 4, 7) 303ANY11(UYVYToARGBRow_Any_NEON, UYVYToARGBRow_NEON, 1, 4, 4, 7) 304#endif 305#ifdef HAS_ARGBTOYROW_AVX2 306ANY11(ARGBToYRow_Any_AVX2, ARGBToYRow_AVX2, 0, 4, 1, 31) 307#endif 308#ifdef HAS_ARGBTOYJROW_AVX2 309ANY11(ARGBToYJRow_Any_AVX2, ARGBToYJRow_AVX2, 0, 4, 1, 31) 310#endif 311#ifdef HAS_UYVYTOYROW_AVX2 312ANY11(UYVYToYRow_Any_AVX2, UYVYToYRow_AVX2, 0, 2, 1, 31) 313#endif 314#ifdef HAS_YUY2TOYROW_AVX2 315ANY11(YUY2ToYRow_Any_AVX2, YUY2ToYRow_AVX2, 1, 4, 1, 31) 316#endif 317#ifdef HAS_ARGBTOYROW_SSSE3 318ANY11(ARGBToYRow_Any_SSSE3, ARGBToYRow_SSSE3, 0, 4, 1, 15) 319#endif 320#ifdef HAS_BGRATOYROW_SSSE3 321ANY11(BGRAToYRow_Any_SSSE3, BGRAToYRow_SSSE3, 0, 4, 1, 15) 322ANY11(ABGRToYRow_Any_SSSE3, ABGRToYRow_SSSE3, 0, 4, 1, 15) 323ANY11(RGBAToYRow_Any_SSSE3, RGBAToYRow_SSSE3, 0, 4, 1, 15) 324ANY11(YUY2ToYRow_Any_SSE2, YUY2ToYRow_SSE2, 1, 4, 1, 15) 325ANY11(UYVYToYRow_Any_SSE2, UYVYToYRow_SSE2, 1, 4, 1, 15) 326#endif 327#ifdef HAS_ARGBTOYJROW_SSSE3 328ANY11(ARGBToYJRow_Any_SSSE3, ARGBToYJRow_SSSE3, 0, 4, 1, 15) 329#endif 330#ifdef HAS_ARGBTOYROW_NEON 331ANY11(ARGBToYRow_Any_NEON, ARGBToYRow_NEON, 0, 4, 1, 7) 332#endif 333#ifdef HAS_ARGBTOYJROW_NEON 334ANY11(ARGBToYJRow_Any_NEON, ARGBToYJRow_NEON, 0, 4, 1, 7) 335#endif 336#ifdef HAS_BGRATOYROW_NEON 337ANY11(BGRAToYRow_Any_NEON, BGRAToYRow_NEON, 0, 4, 1, 7) 338#endif 339#ifdef HAS_ABGRTOYROW_NEON 340ANY11(ABGRToYRow_Any_NEON, ABGRToYRow_NEON, 0, 4, 1, 7) 341#endif 342#ifdef HAS_RGBATOYROW_NEON 343ANY11(RGBAToYRow_Any_NEON, RGBAToYRow_NEON, 0, 4, 1, 7) 344#endif 345#ifdef HAS_RGB24TOYROW_NEON 346ANY11(RGB24ToYRow_Any_NEON, RGB24ToYRow_NEON, 0, 3, 1, 7) 347#endif 348#ifdef HAS_RAWTOYROW_NEON 349ANY11(RAWToYRow_Any_NEON, RAWToYRow_NEON, 0, 3, 1, 7) 350#endif 351#ifdef HAS_RGB565TOYROW_NEON 352ANY11(RGB565ToYRow_Any_NEON, RGB565ToYRow_NEON, 0, 2, 1, 7) 353#endif 354#ifdef HAS_ARGB1555TOYROW_NEON 355ANY11(ARGB1555ToYRow_Any_NEON, ARGB1555ToYRow_NEON, 0, 2, 1, 7) 356#endif 357#ifdef HAS_ARGB4444TOYROW_NEON 358ANY11(ARGB4444ToYRow_Any_NEON, ARGB4444ToYRow_NEON, 0, 2, 1, 7) 359#endif 360#ifdef HAS_YUY2TOYROW_NEON 361ANY11(YUY2ToYRow_Any_NEON, YUY2ToYRow_NEON, 1, 4, 1, 15) 362#endif 363#ifdef HAS_UYVYTOYROW_NEON 364ANY11(UYVYToYRow_Any_NEON, UYVYToYRow_NEON, 0, 2, 1, 15) 365#endif 366#ifdef HAS_RGB24TOARGBROW_NEON 367ANY11(RGB24ToARGBRow_Any_NEON, RGB24ToARGBRow_NEON, 0, 3, 4, 7) 368#endif 369#ifdef HAS_RAWTOARGBROW_NEON 370ANY11(RAWToARGBRow_Any_NEON, RAWToARGBRow_NEON, 0, 3, 4, 7) 371#endif 372#ifdef HAS_RGB565TOARGBROW_NEON 373ANY11(RGB565ToARGBRow_Any_NEON, RGB565ToARGBRow_NEON, 0, 2, 4, 7) 374#endif 375#ifdef HAS_ARGB1555TOARGBROW_NEON 376ANY11(ARGB1555ToARGBRow_Any_NEON, ARGB1555ToARGBRow_NEON, 0, 2, 4, 7) 377#endif 378#ifdef HAS_ARGB4444TOARGBROW_NEON 379ANY11(ARGB4444ToARGBRow_Any_NEON, ARGB4444ToARGBRow_NEON, 0, 2, 4, 7) 380#endif 381#ifdef HAS_ARGBATTENUATEROW_SSSE3 382ANY11(ARGBAttenuateRow_Any_SSSE3, ARGBAttenuateRow_SSSE3, 0, 4, 4, 3) 383#endif 384#ifdef HAS_ARGBATTENUATEROW_SSE2 385ANY11(ARGBAttenuateRow_Any_SSE2, ARGBAttenuateRow_SSE2, 0, 4, 4, 3) 386#endif 387#ifdef HAS_ARGBUNATTENUATEROW_SSE2 388ANY11(ARGBUnattenuateRow_Any_SSE2, ARGBUnattenuateRow_SSE2, 0, 4, 4, 3) 389#endif 390#ifdef HAS_ARGBATTENUATEROW_AVX2 391ANY11(ARGBAttenuateRow_Any_AVX2, ARGBAttenuateRow_AVX2, 0, 4, 4, 7) 392#endif 393#ifdef HAS_ARGBUNATTENUATEROW_AVX2 394ANY11(ARGBUnattenuateRow_Any_AVX2, ARGBUnattenuateRow_AVX2, 0, 4, 4, 7) 395#endif 396#ifdef HAS_ARGBATTENUATEROW_NEON 397ANY11(ARGBAttenuateRow_Any_NEON, ARGBAttenuateRow_NEON, 0, 4, 4, 7) 398#endif 399#undef ANY11 400 401// Any 1 to 1 with parameter. 402#define ANY11P(NAMEANY, ANY_SIMD, T, SBPP, BPP, MASK) \ 403 void NAMEANY(const uint8* src_ptr, uint8* dst_ptr, \ 404 T shuffler, int width) { \ 405 SIMD_ALIGNED(uint8 temp[64 * 2]); \ 406 memset(temp, 0, 64); /* for msan */ \ 407 int r = width & MASK; \ 408 int n = width & ~MASK; \ 409 if (n > 0) { \ 410 ANY_SIMD(src_ptr, dst_ptr, shuffler, n); \ 411 } \ 412 memcpy(temp, src_ptr + n * SBPP, r * SBPP); \ 413 ANY_SIMD(temp, temp + 64, shuffler, MASK + 1); \ 414 memcpy(dst_ptr + n * BPP, temp + 64, r * BPP); \ 415 } 416 417#if defined(HAS_ARGBTORGB565DITHERROW_SSE2) 418ANY11P(ARGBToRGB565DitherRow_Any_SSE2, ARGBToRGB565DitherRow_SSE2, 419 const uint32, 4, 2, 3) 420#endif 421#if defined(HAS_ARGBTORGB565DITHERROW_AVX2) 422ANY11P(ARGBToRGB565DitherRow_Any_AVX2, ARGBToRGB565DitherRow_AVX2, 423 const uint32, 4, 2, 7) 424#endif 425#if defined(HAS_ARGBTORGB565DITHERROW_NEON) 426ANY11P(ARGBToRGB565DitherRow_Any_NEON, ARGBToRGB565DitherRow_NEON, 427 const uint32, 4, 2, 7) 428#endif 429#ifdef HAS_ARGBSHUFFLEROW_SSE2 430ANY11P(ARGBShuffleRow_Any_SSE2, ARGBShuffleRow_SSE2, const uint8*, 4, 4, 3) 431#endif 432#ifdef HAS_ARGBSHUFFLEROW_SSSE3 433ANY11P(ARGBShuffleRow_Any_SSSE3, ARGBShuffleRow_SSSE3, const uint8*, 4, 4, 7) 434#endif 435#ifdef HAS_ARGBSHUFFLEROW_AVX2 436ANY11P(ARGBShuffleRow_Any_AVX2, ARGBShuffleRow_AVX2, const uint8*, 4, 4, 15) 437#endif 438#ifdef HAS_ARGBSHUFFLEROW_NEON 439ANY11P(ARGBShuffleRow_Any_NEON, ARGBShuffleRow_NEON, const uint8*, 4, 4, 3) 440#endif 441#undef ANY11P 442 443// Any 1 to 1 interpolate. Takes 2 rows of source via stride. 444#define ANY11T(NAMEANY, ANY_SIMD, SBPP, BPP, MASK) \ 445 void NAMEANY(uint8* dst_ptr, const uint8* src_ptr, \ 446 ptrdiff_t src_stride_ptr, int width, \ 447 int source_y_fraction) { \ 448 SIMD_ALIGNED(uint8 temp[64 * 3]); \ 449 memset(temp, 0, 64 * 2); /* for msan */ \ 450 int r = width & MASK; \ 451 int n = width & ~MASK; \ 452 if (n > 0) { \ 453 ANY_SIMD(dst_ptr, src_ptr, src_stride_ptr, n, source_y_fraction); \ 454 } \ 455 memcpy(temp, src_ptr + n * SBPP, r * SBPP); \ 456 memcpy(temp + 64, src_ptr + src_stride_ptr + n * SBPP, r * SBPP); \ 457 ANY_SIMD(temp + 128, temp, 64, MASK + 1, source_y_fraction); \ 458 memcpy(dst_ptr + n * BPP, temp + 128, r * BPP); \ 459 } 460 461#ifdef HAS_INTERPOLATEROW_AVX2 462ANY11T(InterpolateRow_Any_AVX2, InterpolateRow_AVX2, 1, 1, 31) 463#endif 464#ifdef HAS_INTERPOLATEROW_SSSE3 465ANY11T(InterpolateRow_Any_SSSE3, InterpolateRow_SSSE3, 1, 1, 15) 466#endif 467#ifdef HAS_INTERPOLATEROW_SSE2 468ANY11T(InterpolateRow_Any_SSE2, InterpolateRow_SSE2, 1, 1, 15) 469#endif 470#ifdef HAS_INTERPOLATEROW_NEON 471ANY11T(InterpolateRow_Any_NEON, InterpolateRow_NEON, 1, 1, 15) 472#endif 473#ifdef HAS_INTERPOLATEROW_MIPS_DSPR2 474ANY11T(InterpolateRow_Any_MIPS_DSPR2, InterpolateRow_MIPS_DSPR2, 1, 1, 3) 475#endif 476#undef ANY11T 477 478// Any 1 to 1 mirror. 479#define ANY11M(NAMEANY, ANY_SIMD, BPP, MASK) \ 480 void NAMEANY(const uint8* src_ptr, uint8* dst_ptr, int width) { \ 481 SIMD_ALIGNED(uint8 temp[64 * 2]); \ 482 memset(temp, 0, 64); /* for msan */ \ 483 int r = width & MASK; \ 484 int n = width & ~MASK; \ 485 if (n > 0) { \ 486 ANY_SIMD(src_ptr + r * BPP, dst_ptr, n); \ 487 } \ 488 memcpy(temp, src_ptr, r * BPP); \ 489 ANY_SIMD(temp, temp + 64, MASK + 1); \ 490 memcpy(dst_ptr + n * BPP, temp + 64 + (MASK + 1 - r) * BPP, r * BPP); \ 491 } 492 493#ifdef HAS_MIRRORROW_AVX2 494ANY11M(MirrorRow_Any_AVX2, MirrorRow_AVX2, 1, 31) 495#endif 496#ifdef HAS_MIRRORROW_SSSE3 497ANY11M(MirrorRow_Any_SSSE3, MirrorRow_SSSE3, 1, 15) 498#endif 499#ifdef HAS_MIRRORROW_SSE2 500ANY11M(MirrorRow_Any_SSE2, MirrorRow_SSE2, 1, 15) 501#endif 502#ifdef HAS_MIRRORROW_NEON 503ANY11M(MirrorRow_Any_NEON, MirrorRow_NEON, 1, 15) 504#endif 505#ifdef HAS_ARGBMIRRORROW_AVX2 506ANY11M(ARGBMirrorRow_Any_AVX2, ARGBMirrorRow_AVX2, 4, 7) 507#endif 508#ifdef HAS_ARGBMIRRORROW_SSE2 509ANY11M(ARGBMirrorRow_Any_SSE2, ARGBMirrorRow_SSE2, 4, 3) 510#endif 511#ifdef HAS_ARGBMIRRORROW_NEON 512ANY11M(ARGBMirrorRow_Any_NEON, ARGBMirrorRow_NEON, 4, 3) 513#endif 514#undef ANY11M 515 516// Any 1 plane. (memset) 517#define ANY1(NAMEANY, ANY_SIMD, T, BPP, MASK) \ 518 void NAMEANY(uint8* dst_ptr, T v32, int width) { \ 519 SIMD_ALIGNED(uint8 temp[64]); \ 520 int r = width & MASK; \ 521 int n = width & ~MASK; \ 522 if (n > 0) { \ 523 ANY_SIMD(dst_ptr, v32, n); \ 524 } \ 525 ANY_SIMD(temp, v32, MASK + 1); \ 526 memcpy(dst_ptr + n * BPP, temp, r * BPP); \ 527 } 528 529#ifdef HAS_SETROW_X86 530ANY1(SetRow_Any_X86, SetRow_X86, uint8, 1, 3) 531#endif 532#ifdef HAS_SETROW_NEON 533ANY1(SetRow_Any_NEON, SetRow_NEON, uint8, 1, 15) 534#endif 535#ifdef HAS_ARGBSETROW_NEON 536ANY1(ARGBSetRow_Any_NEON, ARGBSetRow_NEON, uint32, 4, 3) 537#endif 538#undef ANY1 539 540// Any 1 to 2. Outputs UV planes. 541#define ANY12(NAMEANY, ANY_SIMD, UVSHIFT, BPP, DUVSHIFT, MASK) \ 542 void NAMEANY(const uint8* src_ptr, uint8* dst_u, uint8* dst_v, int width) {\ 543 SIMD_ALIGNED(uint8 temp[128 * 3]); \ 544 memset(temp, 0, 128); /* for msan */ \ 545 int r = width & MASK; \ 546 int n = width & ~MASK; \ 547 if (n > 0) { \ 548 ANY_SIMD(src_ptr, dst_u, dst_v, n); \ 549 } \ 550 memcpy(temp, src_ptr + (n >> UVSHIFT) * BPP, SS(r, UVSHIFT) * BPP); \ 551 if ((width & 1) && BPP == 4) { /* repeat last 4 bytes for subsampler */ \ 552 memcpy(temp + SS(r, UVSHIFT) * BPP, \ 553 temp + SS(r, UVSHIFT) * BPP - BPP, 4); \ 554 } \ 555 ANY_SIMD(temp, temp + 128, temp + 256, MASK + 1); \ 556 memcpy(dst_u + (n >> DUVSHIFT), temp + 128, SS(r, DUVSHIFT)); \ 557 memcpy(dst_v + (n >> DUVSHIFT), temp + 256, SS(r, DUVSHIFT)); \ 558 } 559 560#ifdef HAS_SPLITUVROW_SSE2 561ANY12(SplitUVRow_Any_SSE2, SplitUVRow_SSE2, 0, 2, 0, 15) 562#endif 563#ifdef HAS_SPLITUVROW_AVX2 564ANY12(SplitUVRow_Any_AVX2, SplitUVRow_AVX2, 0, 2, 0, 31) 565#endif 566#ifdef HAS_SPLITUVROW_NEON 567ANY12(SplitUVRow_Any_NEON, SplitUVRow_NEON, 0, 2, 0, 15) 568#endif 569#ifdef HAS_SPLITUVROW_MIPS_DSPR2 570ANY12(SplitUVRow_Any_MIPS_DSPR2, SplitUVRow_MIPS_DSPR2, 0, 2, 0, 15) 571#endif 572#ifdef HAS_ARGBTOUV444ROW_SSSE3 573ANY12(ARGBToUV444Row_Any_SSSE3, ARGBToUV444Row_SSSE3, 0, 4, 0, 15) 574#endif 575#ifdef HAS_YUY2TOUV422ROW_AVX2 576ANY12(YUY2ToUV422Row_Any_AVX2, YUY2ToUV422Row_AVX2, 1, 4, 1, 31) 577ANY12(UYVYToUV422Row_Any_AVX2, UYVYToUV422Row_AVX2, 1, 4, 1, 31) 578#endif 579#ifdef HAS_ARGBTOUV422ROW_SSSE3 580ANY12(ARGBToUV422Row_Any_SSSE3, ARGBToUV422Row_SSSE3, 0, 4, 1, 15) 581#endif 582#ifdef HAS_YUY2TOUV422ROW_SSE2 583ANY12(YUY2ToUV422Row_Any_SSE2, YUY2ToUV422Row_SSE2, 1, 4, 1, 15) 584ANY12(UYVYToUV422Row_Any_SSE2, UYVYToUV422Row_SSE2, 1, 4, 1, 15) 585#endif 586#ifdef HAS_YUY2TOUV422ROW_NEON 587ANY12(ARGBToUV444Row_Any_NEON, ARGBToUV444Row_NEON, 0, 4, 0, 7) 588ANY12(ARGBToUV422Row_Any_NEON, ARGBToUV422Row_NEON, 0, 4, 1, 15) 589ANY12(ARGBToUV411Row_Any_NEON, ARGBToUV411Row_NEON, 0, 4, 2, 31) 590ANY12(YUY2ToUV422Row_Any_NEON, YUY2ToUV422Row_NEON, 1, 4, 1, 15) 591ANY12(UYVYToUV422Row_Any_NEON, UYVYToUV422Row_NEON, 1, 4, 1, 15) 592#endif 593#undef ANY12 594 595// Any 1 to 2 with source stride (2 rows of source). Outputs UV planes. 596// 128 byte row allows for 32 avx ARGB pixels. 597#define ANY12S(NAMEANY, ANY_SIMD, UVSHIFT, BPP, MASK) \ 598 void NAMEANY(const uint8* src_ptr, int src_stride_ptr, \ 599 uint8* dst_u, uint8* dst_v, int width) { \ 600 SIMD_ALIGNED(uint8 temp[128 * 4]); \ 601 memset(temp, 0, 128 * 2); /* for msan */ \ 602 int r = width & MASK; \ 603 int n = width & ~MASK; \ 604 if (n > 0) { \ 605 ANY_SIMD(src_ptr, src_stride_ptr, dst_u, dst_v, n); \ 606 } \ 607 memcpy(temp, src_ptr + (n >> UVSHIFT) * BPP, SS(r, UVSHIFT) * BPP); \ 608 memcpy(temp + 128, src_ptr + src_stride_ptr + (n >> UVSHIFT) * BPP, \ 609 SS(r, UVSHIFT) * BPP); \ 610 if ((width & 1) && BPP == 4) { /* repeat last 4 bytes for subsampler */ \ 611 memcpy(temp + SS(r, UVSHIFT) * BPP, \ 612 temp + SS(r, UVSHIFT) * BPP - BPP, 4); \ 613 memcpy(temp + 128 + SS(r, UVSHIFT) * BPP, \ 614 temp + 128 + SS(r, UVSHIFT) * BPP - BPP, 4); \ 615 } \ 616 ANY_SIMD(temp, 128, temp + 256, temp + 384, MASK + 1); \ 617 memcpy(dst_u + (n >> 1), temp + 256, SS(r, 1)); \ 618 memcpy(dst_v + (n >> 1), temp + 384, SS(r, 1)); \ 619 } 620 621#ifdef HAS_ARGBTOUVROW_AVX2 622ANY12S(ARGBToUVRow_Any_AVX2, ARGBToUVRow_AVX2, 0, 4, 31) 623#endif 624#ifdef HAS_ARGBTOUVROW_SSSE3 625ANY12S(ARGBToUVRow_Any_SSSE3, ARGBToUVRow_SSSE3, 0, 4, 15) 626ANY12S(ARGBToUVJRow_Any_SSSE3, ARGBToUVJRow_SSSE3, 0, 4, 15) 627ANY12S(BGRAToUVRow_Any_SSSE3, BGRAToUVRow_SSSE3, 0, 4, 15) 628ANY12S(ABGRToUVRow_Any_SSSE3, ABGRToUVRow_SSSE3, 0, 4, 15) 629ANY12S(RGBAToUVRow_Any_SSSE3, RGBAToUVRow_SSSE3, 0, 4, 15) 630#endif 631#ifdef HAS_YUY2TOUVROW_AVX2 632ANY12S(YUY2ToUVRow_Any_AVX2, YUY2ToUVRow_AVX2, 1, 4, 31) 633ANY12S(UYVYToUVRow_Any_AVX2, UYVYToUVRow_AVX2, 1, 4, 31) 634#endif 635#ifdef HAS_YUY2TOUVROW_SSE2 636ANY12S(YUY2ToUVRow_Any_SSE2, YUY2ToUVRow_SSE2, 1, 4, 15) 637ANY12S(UYVYToUVRow_Any_SSE2, UYVYToUVRow_SSE2, 1, 4, 15) 638#endif 639#ifdef HAS_ARGBTOUVROW_NEON 640ANY12S(ARGBToUVRow_Any_NEON, ARGBToUVRow_NEON, 0, 4, 15) 641#endif 642#ifdef HAS_ARGBTOUVJROW_NEON 643ANY12S(ARGBToUVJRow_Any_NEON, ARGBToUVJRow_NEON, 0, 4, 15) 644#endif 645#ifdef HAS_BGRATOUVROW_NEON 646ANY12S(BGRAToUVRow_Any_NEON, BGRAToUVRow_NEON, 0, 4, 15) 647#endif 648#ifdef HAS_ABGRTOUVROW_NEON 649ANY12S(ABGRToUVRow_Any_NEON, ABGRToUVRow_NEON, 0, 4, 15) 650#endif 651#ifdef HAS_RGBATOUVROW_NEON 652ANY12S(RGBAToUVRow_Any_NEON, RGBAToUVRow_NEON, 0, 4, 15) 653#endif 654#ifdef HAS_RGB24TOUVROW_NEON 655ANY12S(RGB24ToUVRow_Any_NEON, RGB24ToUVRow_NEON, 0, 3, 15) 656#endif 657#ifdef HAS_RAWTOUVROW_NEON 658ANY12S(RAWToUVRow_Any_NEON, RAWToUVRow_NEON, 0, 3, 15) 659#endif 660#ifdef HAS_RGB565TOUVROW_NEON 661ANY12S(RGB565ToUVRow_Any_NEON, RGB565ToUVRow_NEON, 0, 2, 15) 662#endif 663#ifdef HAS_ARGB1555TOUVROW_NEON 664ANY12S(ARGB1555ToUVRow_Any_NEON, ARGB1555ToUVRow_NEON, 0, 2, 15) 665#endif 666#ifdef HAS_ARGB4444TOUVROW_NEON 667ANY12S(ARGB4444ToUVRow_Any_NEON, ARGB4444ToUVRow_NEON, 0, 2, 15) 668#endif 669#ifdef HAS_YUY2TOUVROW_NEON 670ANY12S(YUY2ToUVRow_Any_NEON, YUY2ToUVRow_NEON, 1, 4, 15) 671#endif 672#ifdef HAS_UYVYTOUVROW_NEON 673ANY12S(UYVYToUVRow_Any_NEON, UYVYToUVRow_NEON, 1, 4, 15) 674#endif 675#undef ANY12S 676 677#ifdef __cplusplus 678} // extern "C" 679} // namespace libyuv 680#endif 681