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 18 namespace libyuv { 19 extern "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 4 planes to 1 with yuvconstants 26 #define ANY41C(NAMEANY, ANY_SIMD, UVSHIFT, DUVSHIFT, BPP, MASK) \ 27 void NAMEANY(const uint8* y_buf, const uint8* u_buf, const uint8* v_buf, \ 28 const uint8* a_buf, uint8* dst_ptr, \ 29 const struct YuvConstants* yuvconstants, int width) { \ 30 SIMD_ALIGNED(uint8 temp[64 * 5]); \ 31 memset(temp, 0, 64 * 4); /* for msan */ \ 32 int r = width & MASK; \ 33 int n = width & ~MASK; \ 34 if (n > 0) { \ 35 ANY_SIMD(y_buf, u_buf, v_buf, a_buf, dst_ptr, yuvconstants, n); \ 36 } \ 37 memcpy(temp, y_buf + n, r); \ 38 memcpy(temp + 64, u_buf + (n >> UVSHIFT), SS(r, UVSHIFT)); \ 39 memcpy(temp + 128, v_buf + (n >> UVSHIFT), SS(r, UVSHIFT)); \ 40 memcpy(temp + 192, a_buf + n, r); \ 41 ANY_SIMD(temp, temp + 64, temp + 128, temp + 192, temp + 256, \ 42 yuvconstants, MASK + 1); \ 43 memcpy(dst_ptr + (n >> DUVSHIFT) * BPP, temp + 256, \ 44 SS(r, DUVSHIFT) * BPP); \ 45 } 46 47 #ifdef HAS_I422ALPHATOARGBROW_SSSE3 48 ANY41C(I422AlphaToARGBRow_Any_SSSE3, I422AlphaToARGBRow_SSSE3, 1, 0, 4, 7) 49 #endif 50 #ifdef HAS_I422ALPHATOARGBROW_AVX2 51 ANY41C(I422AlphaToARGBRow_Any_AVX2, I422AlphaToARGBRow_AVX2, 1, 0, 4, 15) 52 #endif 53 #ifdef HAS_I422ALPHATOARGBROW_NEON 54 ANY41C(I422AlphaToARGBRow_Any_NEON, I422AlphaToARGBRow_NEON, 1, 0, 4, 7) 55 #endif 56 #ifdef HAS_I422ALPHATOARGBROW_MSA 57 ANY41C(I422AlphaToARGBRow_Any_MSA, I422AlphaToARGBRow_MSA, 1, 0, 4, 7) 58 #endif 59 #undef ANY41C 60 61 // Any 3 planes to 1. 62 #define ANY31(NAMEANY, ANY_SIMD, UVSHIFT, DUVSHIFT, BPP, MASK) \ 63 void NAMEANY(const uint8* y_buf, const uint8* u_buf, const uint8* v_buf, \ 64 uint8* dst_ptr, int width) { \ 65 SIMD_ALIGNED(uint8 temp[64 * 4]); \ 66 memset(temp, 0, 64 * 3); /* for YUY2 and msan */ \ 67 int r = width & MASK; \ 68 int n = width & ~MASK; \ 69 if (n > 0) { \ 70 ANY_SIMD(y_buf, u_buf, v_buf, dst_ptr, n); \ 71 } \ 72 memcpy(temp, y_buf + n, r); \ 73 memcpy(temp + 64, u_buf + (n >> UVSHIFT), SS(r, UVSHIFT)); \ 74 memcpy(temp + 128, v_buf + (n >> UVSHIFT), SS(r, UVSHIFT)); \ 75 ANY_SIMD(temp, temp + 64, temp + 128, temp + 192, MASK + 1); \ 76 memcpy(dst_ptr + (n >> DUVSHIFT) * BPP, temp + 192, \ 77 SS(r, DUVSHIFT) * BPP); \ 78 } 79 #ifdef HAS_I422TOYUY2ROW_SSE2 80 ANY31(I422ToYUY2Row_Any_SSE2, I422ToYUY2Row_SSE2, 1, 1, 4, 15) 81 ANY31(I422ToUYVYRow_Any_SSE2, I422ToUYVYRow_SSE2, 1, 1, 4, 15) 82 #endif 83 #ifdef HAS_I422TOYUY2ROW_NEON 84 ANY31(I422ToYUY2Row_Any_NEON, I422ToYUY2Row_NEON, 1, 1, 4, 15) 85 #endif 86 #ifdef HAS_I422TOYUY2ROW_MSA 87 ANY31(I422ToYUY2Row_Any_MSA, I422ToYUY2Row_MSA, 1, 1, 4, 31) 88 #endif 89 #ifdef HAS_I422TOUYVYROW_NEON 90 ANY31(I422ToUYVYRow_Any_NEON, I422ToUYVYRow_NEON, 1, 1, 4, 15) 91 #endif 92 #ifdef HAS_I422TOUYVYROW_MSA 93 ANY31(I422ToUYVYRow_Any_MSA, I422ToUYVYRow_MSA, 1, 1, 4, 31) 94 #endif 95 #ifdef HAS_BLENDPLANEROW_AVX2 96 ANY31(BlendPlaneRow_Any_AVX2, BlendPlaneRow_AVX2, 0, 0, 1, 31) 97 #endif 98 #ifdef HAS_BLENDPLANEROW_SSSE3 99 ANY31(BlendPlaneRow_Any_SSSE3, BlendPlaneRow_SSSE3, 0, 0, 1, 7) 100 #endif 101 #undef ANY31 102 103 // Note that odd width replication includes 444 due to implementation 104 // on arm that subsamples 444 to 422 internally. 105 // Any 3 planes to 1 with yuvconstants 106 #define ANY31C(NAMEANY, ANY_SIMD, UVSHIFT, DUVSHIFT, BPP, MASK) \ 107 void NAMEANY(const uint8* y_buf, const uint8* u_buf, const uint8* v_buf, \ 108 uint8* dst_ptr, const struct YuvConstants* yuvconstants, \ 109 int width) { \ 110 SIMD_ALIGNED(uint8 temp[64 * 4]); \ 111 memset(temp, 0, 64 * 3); /* for YUY2 and msan */ \ 112 int r = width & MASK; \ 113 int n = width & ~MASK; \ 114 if (n > 0) { \ 115 ANY_SIMD(y_buf, u_buf, v_buf, dst_ptr, yuvconstants, n); \ 116 } \ 117 memcpy(temp, y_buf + n, r); \ 118 memcpy(temp + 64, u_buf + (n >> UVSHIFT), SS(r, UVSHIFT)); \ 119 memcpy(temp + 128, v_buf + (n >> UVSHIFT), SS(r, UVSHIFT)); \ 120 if (width & 1) { \ 121 temp[64 + SS(r, UVSHIFT)] = temp[64 + SS(r, UVSHIFT) - 1]; \ 122 temp[128 + SS(r, UVSHIFT)] = temp[128 + SS(r, UVSHIFT) - 1]; \ 123 } \ 124 ANY_SIMD(temp, temp + 64, temp + 128, temp + 192, yuvconstants, MASK + 1); \ 125 memcpy(dst_ptr + (n >> DUVSHIFT) * BPP, temp + 192, \ 126 SS(r, DUVSHIFT) * BPP); \ 127 } 128 129 #ifdef HAS_I422TOARGBROW_SSSE3 130 ANY31C(I422ToARGBRow_Any_SSSE3, I422ToARGBRow_SSSE3, 1, 0, 4, 7) 131 #endif 132 #ifdef HAS_I444TOARGBROW_SSSE3 133 ANY31C(I444ToARGBRow_Any_SSSE3, I444ToARGBRow_SSSE3, 0, 0, 4, 7) 134 ANY31C(I422ToRGBARow_Any_SSSE3, I422ToRGBARow_SSSE3, 1, 0, 4, 7) 135 ANY31C(I422ToARGB4444Row_Any_SSSE3, I422ToARGB4444Row_SSSE3, 1, 0, 2, 7) 136 ANY31C(I422ToARGB1555Row_Any_SSSE3, I422ToARGB1555Row_SSSE3, 1, 0, 2, 7) 137 ANY31C(I422ToRGB565Row_Any_SSSE3, I422ToRGB565Row_SSSE3, 1, 0, 2, 7) 138 ANY31C(I422ToRGB24Row_Any_SSSE3, I422ToRGB24Row_SSSE3, 1, 0, 3, 7) 139 #endif // HAS_I444TOARGBROW_SSSE3 140 #ifdef HAS_I422TORGB24ROW_AVX2 141 ANY31C(I422ToRGB24Row_Any_AVX2, I422ToRGB24Row_AVX2, 1, 0, 3, 15) 142 #endif 143 #ifdef HAS_I422TOARGBROW_AVX2 144 ANY31C(I422ToARGBRow_Any_AVX2, I422ToARGBRow_AVX2, 1, 0, 4, 15) 145 #endif 146 #ifdef HAS_I422TORGBAROW_AVX2 147 ANY31C(I422ToRGBARow_Any_AVX2, I422ToRGBARow_AVX2, 1, 0, 4, 15) 148 #endif 149 #ifdef HAS_I444TOARGBROW_AVX2 150 ANY31C(I444ToARGBRow_Any_AVX2, I444ToARGBRow_AVX2, 0, 0, 4, 15) 151 #endif 152 #ifdef HAS_I422TOARGB4444ROW_AVX2 153 ANY31C(I422ToARGB4444Row_Any_AVX2, I422ToARGB4444Row_AVX2, 1, 0, 2, 7) 154 #endif 155 #ifdef HAS_I422TOARGB1555ROW_AVX2 156 ANY31C(I422ToARGB1555Row_Any_AVX2, I422ToARGB1555Row_AVX2, 1, 0, 2, 7) 157 #endif 158 #ifdef HAS_I422TORGB565ROW_AVX2 159 ANY31C(I422ToRGB565Row_Any_AVX2, I422ToRGB565Row_AVX2, 1, 0, 2, 7) 160 #endif 161 #ifdef HAS_I422TOARGBROW_NEON 162 ANY31C(I444ToARGBRow_Any_NEON, I444ToARGBRow_NEON, 0, 0, 4, 7) 163 ANY31C(I422ToARGBRow_Any_NEON, I422ToARGBRow_NEON, 1, 0, 4, 7) 164 ANY31C(I422ToRGBARow_Any_NEON, I422ToRGBARow_NEON, 1, 0, 4, 7) 165 ANY31C(I422ToRGB24Row_Any_NEON, I422ToRGB24Row_NEON, 1, 0, 3, 7) 166 ANY31C(I422ToARGB4444Row_Any_NEON, I422ToARGB4444Row_NEON, 1, 0, 2, 7) 167 ANY31C(I422ToARGB1555Row_Any_NEON, I422ToARGB1555Row_NEON, 1, 0, 2, 7) 168 ANY31C(I422ToRGB565Row_Any_NEON, I422ToRGB565Row_NEON, 1, 0, 2, 7) 169 #endif 170 #ifdef HAS_I422TOARGBROW_DSPR2 171 ANY31C(I444ToARGBRow_Any_DSPR2, I444ToARGBRow_DSPR2, 0, 0, 4, 7) 172 ANY31C(I422ToARGBRow_Any_DSPR2, I422ToARGBRow_DSPR2, 1, 0, 4, 7) 173 ANY31C(I422ToARGB4444Row_Any_DSPR2, I422ToARGB4444Row_DSPR2, 1, 0, 2, 7) 174 ANY31C(I422ToARGB1555Row_Any_DSPR2, I422ToARGB1555Row_DSPR2, 1, 0, 2, 7) 175 #endif 176 #ifdef HAS_I422TOARGBROW_MSA 177 ANY31C(I444ToARGBRow_Any_MSA, I444ToARGBRow_MSA, 0, 0, 4, 7) 178 ANY31C(I422ToARGBRow_Any_MSA, I422ToARGBRow_MSA, 1, 0, 4, 7) 179 ANY31C(I422ToRGBARow_Any_MSA, I422ToRGBARow_MSA, 1, 0, 4, 7) 180 ANY31C(I422ToRGB24Row_Any_MSA, I422ToRGB24Row_MSA, 1, 0, 3, 15) 181 ANY31C(I422ToARGB4444Row_Any_MSA, I422ToARGB4444Row_MSA, 1, 0, 2, 7) 182 ANY31C(I422ToARGB1555Row_Any_MSA, I422ToARGB1555Row_MSA, 1, 0, 2, 7) 183 ANY31C(I422ToRGB565Row_Any_MSA, I422ToRGB565Row_MSA, 1, 0, 2, 7) 184 #endif 185 #undef ANY31C 186 187 // Any 2 planes to 1. 188 #define ANY21(NAMEANY, ANY_SIMD, UVSHIFT, SBPP, SBPP2, BPP, MASK) \ 189 void NAMEANY(const uint8* y_buf, const uint8* uv_buf, uint8* dst_ptr, \ 190 int width) { \ 191 SIMD_ALIGNED(uint8 temp[64 * 3]); \ 192 memset(temp, 0, 64 * 2); /* for msan */ \ 193 int r = width & MASK; \ 194 int n = width & ~MASK; \ 195 if (n > 0) { \ 196 ANY_SIMD(y_buf, uv_buf, dst_ptr, n); \ 197 } \ 198 memcpy(temp, y_buf + n * SBPP, r * SBPP); \ 199 memcpy(temp + 64, uv_buf + (n >> UVSHIFT) * SBPP2, \ 200 SS(r, UVSHIFT) * SBPP2); \ 201 ANY_SIMD(temp, temp + 64, temp + 128, MASK + 1); \ 202 memcpy(dst_ptr + n * BPP, temp + 128, r * BPP); \ 203 } 204 205 // Merge functions. 206 #ifdef HAS_MERGEUVROW_SSE2 207 ANY21(MergeUVRow_Any_SSE2, MergeUVRow_SSE2, 0, 1, 1, 2, 15) 208 #endif 209 #ifdef HAS_MERGEUVROW_AVX2 210 ANY21(MergeUVRow_Any_AVX2, MergeUVRow_AVX2, 0, 1, 1, 2, 31) 211 #endif 212 #ifdef HAS_MERGEUVROW_NEON 213 ANY21(MergeUVRow_Any_NEON, MergeUVRow_NEON, 0, 1, 1, 2, 15) 214 #endif 215 216 // Math functions. 217 #ifdef HAS_ARGBMULTIPLYROW_SSE2 218 ANY21(ARGBMultiplyRow_Any_SSE2, ARGBMultiplyRow_SSE2, 0, 4, 4, 4, 3) 219 #endif 220 #ifdef HAS_ARGBADDROW_SSE2 221 ANY21(ARGBAddRow_Any_SSE2, ARGBAddRow_SSE2, 0, 4, 4, 4, 3) 222 #endif 223 #ifdef HAS_ARGBSUBTRACTROW_SSE2 224 ANY21(ARGBSubtractRow_Any_SSE2, ARGBSubtractRow_SSE2, 0, 4, 4, 4, 3) 225 #endif 226 #ifdef HAS_ARGBMULTIPLYROW_AVX2 227 ANY21(ARGBMultiplyRow_Any_AVX2, ARGBMultiplyRow_AVX2, 0, 4, 4, 4, 7) 228 #endif 229 #ifdef HAS_ARGBADDROW_AVX2 230 ANY21(ARGBAddRow_Any_AVX2, ARGBAddRow_AVX2, 0, 4, 4, 4, 7) 231 #endif 232 #ifdef HAS_ARGBSUBTRACTROW_AVX2 233 ANY21(ARGBSubtractRow_Any_AVX2, ARGBSubtractRow_AVX2, 0, 4, 4, 4, 7) 234 #endif 235 #ifdef HAS_ARGBMULTIPLYROW_NEON 236 ANY21(ARGBMultiplyRow_Any_NEON, ARGBMultiplyRow_NEON, 0, 4, 4, 4, 7) 237 #endif 238 #ifdef HAS_ARGBADDROW_NEON 239 ANY21(ARGBAddRow_Any_NEON, ARGBAddRow_NEON, 0, 4, 4, 4, 7) 240 #endif 241 #ifdef HAS_ARGBSUBTRACTROW_NEON 242 ANY21(ARGBSubtractRow_Any_NEON, ARGBSubtractRow_NEON, 0, 4, 4, 4, 7) 243 #endif 244 #ifdef HAS_ARGBMULTIPLYROW_MSA 245 ANY21(ARGBMultiplyRow_Any_MSA, ARGBMultiplyRow_MSA, 0, 4, 4, 4, 3) 246 #endif 247 #ifdef HAS_ARGBADDROW_MSA 248 ANY21(ARGBAddRow_Any_MSA, ARGBAddRow_MSA, 0, 4, 4, 4, 7) 249 #endif 250 #ifdef HAS_ARGBSUBTRACTROW_MSA 251 ANY21(ARGBSubtractRow_Any_MSA, ARGBSubtractRow_MSA, 0, 4, 4, 4, 7) 252 #endif 253 #ifdef HAS_SOBELROW_SSE2 254 ANY21(SobelRow_Any_SSE2, SobelRow_SSE2, 0, 1, 1, 4, 15) 255 #endif 256 #ifdef HAS_SOBELROW_NEON 257 ANY21(SobelRow_Any_NEON, SobelRow_NEON, 0, 1, 1, 4, 7) 258 #endif 259 #ifdef HAS_SOBELROW_MSA 260 ANY21(SobelRow_Any_MSA, SobelRow_MSA, 0, 1, 1, 4, 15) 261 #endif 262 #ifdef HAS_SOBELTOPLANEROW_SSE2 263 ANY21(SobelToPlaneRow_Any_SSE2, SobelToPlaneRow_SSE2, 0, 1, 1, 1, 15) 264 #endif 265 #ifdef HAS_SOBELTOPLANEROW_NEON 266 ANY21(SobelToPlaneRow_Any_NEON, SobelToPlaneRow_NEON, 0, 1, 1, 1, 15) 267 #endif 268 #ifdef HAS_SOBELTOPLANEROW_MSA 269 ANY21(SobelToPlaneRow_Any_MSA, SobelToPlaneRow_MSA, 0, 1, 1, 1, 31) 270 #endif 271 #ifdef HAS_SOBELXYROW_SSE2 272 ANY21(SobelXYRow_Any_SSE2, SobelXYRow_SSE2, 0, 1, 1, 4, 15) 273 #endif 274 #ifdef HAS_SOBELXYROW_NEON 275 ANY21(SobelXYRow_Any_NEON, SobelXYRow_NEON, 0, 1, 1, 4, 7) 276 #endif 277 #ifdef HAS_SOBELXYROW_MSA 278 ANY21(SobelXYRow_Any_MSA, SobelXYRow_MSA, 0, 1, 1, 4, 15) 279 #endif 280 #undef ANY21 281 282 // Any 2 planes to 1 with yuvconstants 283 #define ANY21C(NAMEANY, ANY_SIMD, UVSHIFT, SBPP, SBPP2, BPP, MASK) \ 284 void NAMEANY(const uint8* y_buf, const uint8* uv_buf, uint8* dst_ptr, \ 285 const struct YuvConstants* yuvconstants, int width) { \ 286 SIMD_ALIGNED(uint8 temp[64 * 3]); \ 287 memset(temp, 0, 64 * 2); /* for msan */ \ 288 int r = width & MASK; \ 289 int n = width & ~MASK; \ 290 if (n > 0) { \ 291 ANY_SIMD(y_buf, uv_buf, dst_ptr, yuvconstants, n); \ 292 } \ 293 memcpy(temp, y_buf + n * SBPP, r * SBPP); \ 294 memcpy(temp + 64, uv_buf + (n >> UVSHIFT) * SBPP2, \ 295 SS(r, UVSHIFT) * SBPP2); \ 296 ANY_SIMD(temp, temp + 64, temp + 128, yuvconstants, MASK + 1); \ 297 memcpy(dst_ptr + n * BPP, temp + 128, r * BPP); \ 298 } 299 300 // Biplanar to RGB. 301 #ifdef HAS_NV12TOARGBROW_SSSE3 302 ANY21C(NV12ToARGBRow_Any_SSSE3, NV12ToARGBRow_SSSE3, 1, 1, 2, 4, 7) 303 #endif 304 #ifdef HAS_NV12TOARGBROW_AVX2 305 ANY21C(NV12ToARGBRow_Any_AVX2, NV12ToARGBRow_AVX2, 1, 1, 2, 4, 15) 306 #endif 307 #ifdef HAS_NV12TOARGBROW_NEON 308 ANY21C(NV12ToARGBRow_Any_NEON, NV12ToARGBRow_NEON, 1, 1, 2, 4, 7) 309 #endif 310 #ifdef HAS_NV12TOARGBROW_DSPR2 311 ANY21C(NV12ToARGBRow_Any_DSPR2, NV12ToARGBRow_DSPR2, 1, 1, 2, 4, 7) 312 #endif 313 #ifdef HAS_NV12TOARGBROW_MSA 314 ANY21C(NV12ToARGBRow_Any_MSA, NV12ToARGBRow_MSA, 1, 1, 2, 4, 7) 315 #endif 316 #ifdef HAS_NV21TOARGBROW_SSSE3 317 ANY21C(NV21ToARGBRow_Any_SSSE3, NV21ToARGBRow_SSSE3, 1, 1, 2, 4, 7) 318 #endif 319 #ifdef HAS_NV21TOARGBROW_AVX2 320 ANY21C(NV21ToARGBRow_Any_AVX2, NV21ToARGBRow_AVX2, 1, 1, 2, 4, 15) 321 #endif 322 #ifdef HAS_NV21TOARGBROW_NEON 323 ANY21C(NV21ToARGBRow_Any_NEON, NV21ToARGBRow_NEON, 1, 1, 2, 4, 7) 324 #endif 325 #ifdef HAS_NV21TOARGBROW_MSA 326 ANY21C(NV21ToARGBRow_Any_MSA, NV21ToARGBRow_MSA, 1, 1, 2, 4, 7) 327 #endif 328 #ifdef HAS_NV12TORGB565ROW_SSSE3 329 ANY21C(NV12ToRGB565Row_Any_SSSE3, NV12ToRGB565Row_SSSE3, 1, 1, 2, 2, 7) 330 #endif 331 #ifdef HAS_NV12TORGB565ROW_AVX2 332 ANY21C(NV12ToRGB565Row_Any_AVX2, NV12ToRGB565Row_AVX2, 1, 1, 2, 2, 15) 333 #endif 334 #ifdef HAS_NV12TORGB565ROW_NEON 335 ANY21C(NV12ToRGB565Row_Any_NEON, NV12ToRGB565Row_NEON, 1, 1, 2, 2, 7) 336 #endif 337 #ifdef HAS_NV12TORGB565ROW_MSA 338 ANY21C(NV12ToRGB565Row_Any_MSA, NV12ToRGB565Row_MSA, 1, 1, 2, 2, 7) 339 #endif 340 #undef ANY21C 341 342 // Any 1 to 1. 343 #define ANY11(NAMEANY, ANY_SIMD, UVSHIFT, SBPP, BPP, MASK) \ 344 void NAMEANY(const uint8* src_ptr, uint8* dst_ptr, int width) { \ 345 SIMD_ALIGNED(uint8 temp[128 * 2]); \ 346 memset(temp, 0, 128); /* for YUY2 and msan */ \ 347 int r = width & MASK; \ 348 int n = width & ~MASK; \ 349 if (n > 0) { \ 350 ANY_SIMD(src_ptr, dst_ptr, n); \ 351 } \ 352 memcpy(temp, src_ptr + (n >> UVSHIFT) * SBPP, SS(r, UVSHIFT) * SBPP); \ 353 ANY_SIMD(temp, temp + 128, MASK + 1); \ 354 memcpy(dst_ptr + n * BPP, temp + 128, r * BPP); \ 355 } 356 357 #ifdef HAS_COPYROW_AVX 358 ANY11(CopyRow_Any_AVX, CopyRow_AVX, 0, 1, 1, 63) 359 #endif 360 #ifdef HAS_COPYROW_SSE2 361 ANY11(CopyRow_Any_SSE2, CopyRow_SSE2, 0, 1, 1, 31) 362 #endif 363 #ifdef HAS_COPYROW_NEON 364 ANY11(CopyRow_Any_NEON, CopyRow_NEON, 0, 1, 1, 31) 365 #endif 366 #if defined(HAS_ARGBTORGB24ROW_SSSE3) 367 ANY11(ARGBToRGB24Row_Any_SSSE3, ARGBToRGB24Row_SSSE3, 0, 4, 3, 15) 368 ANY11(ARGBToRAWRow_Any_SSSE3, ARGBToRAWRow_SSSE3, 0, 4, 3, 15) 369 ANY11(ARGBToRGB565Row_Any_SSE2, ARGBToRGB565Row_SSE2, 0, 4, 2, 3) 370 ANY11(ARGBToARGB1555Row_Any_SSE2, ARGBToARGB1555Row_SSE2, 0, 4, 2, 3) 371 ANY11(ARGBToARGB4444Row_Any_SSE2, ARGBToARGB4444Row_SSE2, 0, 4, 2, 3) 372 #endif 373 #if defined(HAS_ARGBTORGB565ROW_AVX2) 374 ANY11(ARGBToRGB565Row_Any_AVX2, ARGBToRGB565Row_AVX2, 0, 4, 2, 7) 375 #endif 376 #if defined(HAS_ARGBTOARGB4444ROW_AVX2) 377 ANY11(ARGBToARGB1555Row_Any_AVX2, ARGBToARGB1555Row_AVX2, 0, 4, 2, 7) 378 ANY11(ARGBToARGB4444Row_Any_AVX2, ARGBToARGB4444Row_AVX2, 0, 4, 2, 7) 379 #endif 380 #if defined(HAS_J400TOARGBROW_SSE2) 381 ANY11(J400ToARGBRow_Any_SSE2, J400ToARGBRow_SSE2, 0, 1, 4, 7) 382 #endif 383 #if defined(HAS_J400TOARGBROW_AVX2) 384 ANY11(J400ToARGBRow_Any_AVX2, J400ToARGBRow_AVX2, 0, 1, 4, 15) 385 #endif 386 #if defined(HAS_I400TOARGBROW_SSE2) 387 ANY11(I400ToARGBRow_Any_SSE2, I400ToARGBRow_SSE2, 0, 1, 4, 7) 388 #endif 389 #if defined(HAS_I400TOARGBROW_AVX2) 390 ANY11(I400ToARGBRow_Any_AVX2, I400ToARGBRow_AVX2, 0, 1, 4, 15) 391 #endif 392 #if defined(HAS_RGB24TOARGBROW_SSSE3) 393 ANY11(RGB24ToARGBRow_Any_SSSE3, RGB24ToARGBRow_SSSE3, 0, 3, 4, 15) 394 ANY11(RAWToARGBRow_Any_SSSE3, RAWToARGBRow_SSSE3, 0, 3, 4, 15) 395 ANY11(RGB565ToARGBRow_Any_SSE2, RGB565ToARGBRow_SSE2, 0, 2, 4, 7) 396 ANY11(ARGB1555ToARGBRow_Any_SSE2, ARGB1555ToARGBRow_SSE2, 0, 2, 4, 7) 397 ANY11(ARGB4444ToARGBRow_Any_SSE2, ARGB4444ToARGBRow_SSE2, 0, 2, 4, 7) 398 #endif 399 #if defined(HAS_RAWTORGB24ROW_SSSE3) 400 ANY11(RAWToRGB24Row_Any_SSSE3, RAWToRGB24Row_SSSE3, 0, 3, 3, 7) 401 #endif 402 #if defined(HAS_RGB565TOARGBROW_AVX2) 403 ANY11(RGB565ToARGBRow_Any_AVX2, RGB565ToARGBRow_AVX2, 0, 2, 4, 15) 404 #endif 405 #if defined(HAS_ARGB1555TOARGBROW_AVX2) 406 ANY11(ARGB1555ToARGBRow_Any_AVX2, ARGB1555ToARGBRow_AVX2, 0, 2, 4, 15) 407 #endif 408 #if defined(HAS_ARGB4444TOARGBROW_AVX2) 409 ANY11(ARGB4444ToARGBRow_Any_AVX2, ARGB4444ToARGBRow_AVX2, 0, 2, 4, 15) 410 #endif 411 #if defined(HAS_ARGBTORGB24ROW_NEON) 412 ANY11(ARGBToRGB24Row_Any_NEON, ARGBToRGB24Row_NEON, 0, 4, 3, 7) 413 ANY11(ARGBToRAWRow_Any_NEON, ARGBToRAWRow_NEON, 0, 4, 3, 7) 414 ANY11(ARGBToRGB565Row_Any_NEON, ARGBToRGB565Row_NEON, 0, 4, 2, 7) 415 ANY11(ARGBToARGB1555Row_Any_NEON, ARGBToARGB1555Row_NEON, 0, 4, 2, 7) 416 ANY11(ARGBToARGB4444Row_Any_NEON, ARGBToARGB4444Row_NEON, 0, 4, 2, 7) 417 ANY11(J400ToARGBRow_Any_NEON, J400ToARGBRow_NEON, 0, 1, 4, 7) 418 ANY11(I400ToARGBRow_Any_NEON, I400ToARGBRow_NEON, 0, 1, 4, 7) 419 #endif 420 #if defined(HAS_ARGBTORGB24ROW_MSA) 421 ANY11(ARGBToRGB24Row_Any_MSA, ARGBToRGB24Row_MSA, 0, 4, 3, 15) 422 ANY11(ARGBToRAWRow_Any_MSA, ARGBToRAWRow_MSA, 0, 4, 3, 15) 423 ANY11(ARGBToRGB565Row_Any_MSA, ARGBToRGB565Row_MSA, 0, 4, 2, 7) 424 ANY11(ARGBToARGB1555Row_Any_MSA, ARGBToARGB1555Row_MSA, 0, 4, 2, 7) 425 ANY11(ARGBToARGB4444Row_Any_MSA, ARGBToARGB4444Row_MSA, 0, 4, 2, 7) 426 ANY11(J400ToARGBRow_Any_MSA, J400ToARGBRow_MSA, 0, 1, 4, 15) 427 ANY11(I400ToARGBRow_Any_MSA, I400ToARGBRow_MSA, 0, 1, 4, 15) 428 #endif 429 #if defined(HAS_RAWTORGB24ROW_NEON) 430 ANY11(RAWToRGB24Row_Any_NEON, RAWToRGB24Row_NEON, 0, 3, 3, 7) 431 #endif 432 #ifdef HAS_ARGBTOYROW_AVX2 433 ANY11(ARGBToYRow_Any_AVX2, ARGBToYRow_AVX2, 0, 4, 1, 31) 434 #endif 435 #ifdef HAS_ARGBTOYJROW_AVX2 436 ANY11(ARGBToYJRow_Any_AVX2, ARGBToYJRow_AVX2, 0, 4, 1, 31) 437 #endif 438 #ifdef HAS_UYVYTOYROW_AVX2 439 ANY11(UYVYToYRow_Any_AVX2, UYVYToYRow_AVX2, 0, 2, 1, 31) 440 #endif 441 #ifdef HAS_YUY2TOYROW_AVX2 442 ANY11(YUY2ToYRow_Any_AVX2, YUY2ToYRow_AVX2, 1, 4, 1, 31) 443 #endif 444 #ifdef HAS_ARGBTOYROW_SSSE3 445 ANY11(ARGBToYRow_Any_SSSE3, ARGBToYRow_SSSE3, 0, 4, 1, 15) 446 #endif 447 #ifdef HAS_BGRATOYROW_SSSE3 448 ANY11(BGRAToYRow_Any_SSSE3, BGRAToYRow_SSSE3, 0, 4, 1, 15) 449 ANY11(ABGRToYRow_Any_SSSE3, ABGRToYRow_SSSE3, 0, 4, 1, 15) 450 ANY11(RGBAToYRow_Any_SSSE3, RGBAToYRow_SSSE3, 0, 4, 1, 15) 451 ANY11(YUY2ToYRow_Any_SSE2, YUY2ToYRow_SSE2, 1, 4, 1, 15) 452 ANY11(UYVYToYRow_Any_SSE2, UYVYToYRow_SSE2, 1, 4, 1, 15) 453 #endif 454 #ifdef HAS_ARGBTOYJROW_SSSE3 455 ANY11(ARGBToYJRow_Any_SSSE3, ARGBToYJRow_SSSE3, 0, 4, 1, 15) 456 #endif 457 #ifdef HAS_ARGBTOYROW_NEON 458 ANY11(ARGBToYRow_Any_NEON, ARGBToYRow_NEON, 0, 4, 1, 7) 459 #endif 460 #ifdef HAS_ARGBTOYROW_MSA 461 ANY11(ARGBToYRow_Any_MSA, ARGBToYRow_MSA, 0, 4, 1, 15) 462 #endif 463 #ifdef HAS_ARGBTOYJROW_NEON 464 ANY11(ARGBToYJRow_Any_NEON, ARGBToYJRow_NEON, 0, 4, 1, 7) 465 #endif 466 #ifdef HAS_ARGBTOYJROW_MSA 467 ANY11(ARGBToYJRow_Any_MSA, ARGBToYJRow_MSA, 0, 4, 1, 15) 468 #endif 469 #ifdef HAS_BGRATOYROW_NEON 470 ANY11(BGRAToYRow_Any_NEON, BGRAToYRow_NEON, 0, 4, 1, 7) 471 #endif 472 #ifdef HAS_BGRATOYROW_MSA 473 ANY11(BGRAToYRow_Any_MSA, BGRAToYRow_MSA, 0, 4, 1, 15) 474 #endif 475 #ifdef HAS_ABGRTOYROW_NEON 476 ANY11(ABGRToYRow_Any_NEON, ABGRToYRow_NEON, 0, 4, 1, 7) 477 #endif 478 #ifdef HAS_ABGRTOYROW_MSA 479 ANY11(ABGRToYRow_Any_MSA, ABGRToYRow_MSA, 0, 4, 1, 7) 480 #endif 481 #ifdef HAS_RGBATOYROW_NEON 482 ANY11(RGBAToYRow_Any_NEON, RGBAToYRow_NEON, 0, 4, 1, 7) 483 #endif 484 #ifdef HAS_RGBATOYROW_MSA 485 ANY11(RGBAToYRow_Any_MSA, RGBAToYRow_MSA, 0, 4, 1, 15) 486 #endif 487 #ifdef HAS_RGB24TOYROW_NEON 488 ANY11(RGB24ToYRow_Any_NEON, RGB24ToYRow_NEON, 0, 3, 1, 7) 489 #endif 490 #ifdef HAS_RGB24TOYROW_MSA 491 ANY11(RGB24ToYRow_Any_MSA, RGB24ToYRow_MSA, 0, 3, 1, 15) 492 #endif 493 #ifdef HAS_RAWTOYROW_NEON 494 ANY11(RAWToYRow_Any_NEON, RAWToYRow_NEON, 0, 3, 1, 7) 495 #endif 496 #ifdef HAS_RAWTOYROW_MSA 497 ANY11(RAWToYRow_Any_MSA, RAWToYRow_MSA, 0, 3, 1, 15) 498 #endif 499 #ifdef HAS_RGB565TOYROW_NEON 500 ANY11(RGB565ToYRow_Any_NEON, RGB565ToYRow_NEON, 0, 2, 1, 7) 501 #endif 502 #ifdef HAS_RGB565TOYROW_MSA 503 ANY11(RGB565ToYRow_Any_MSA, RGB565ToYRow_MSA, 0, 2, 1, 15) 504 #endif 505 #ifdef HAS_ARGB1555TOYROW_NEON 506 ANY11(ARGB1555ToYRow_Any_NEON, ARGB1555ToYRow_NEON, 0, 2, 1, 7) 507 #endif 508 #ifdef HAS_ARGB1555TOYROW_MSA 509 ANY11(ARGB1555ToYRow_Any_MSA, ARGB1555ToYRow_MSA, 0, 2, 1, 15) 510 #endif 511 #ifdef HAS_ARGB4444TOYROW_NEON 512 ANY11(ARGB4444ToYRow_Any_NEON, ARGB4444ToYRow_NEON, 0, 2, 1, 7) 513 #endif 514 #ifdef HAS_YUY2TOYROW_NEON 515 ANY11(YUY2ToYRow_Any_NEON, YUY2ToYRow_NEON, 1, 4, 1, 15) 516 #endif 517 #ifdef HAS_UYVYTOYROW_NEON 518 ANY11(UYVYToYRow_Any_NEON, UYVYToYRow_NEON, 1, 4, 1, 15) 519 #endif 520 #ifdef HAS_YUY2TOYROW_MSA 521 ANY11(YUY2ToYRow_Any_MSA, YUY2ToYRow_MSA, 1, 4, 1, 31) 522 #endif 523 #ifdef HAS_UYVYTOYROW_MSA 524 ANY11(UYVYToYRow_Any_MSA, UYVYToYRow_MSA, 1, 4, 1, 31) 525 #endif 526 #ifdef HAS_RGB24TOARGBROW_NEON 527 ANY11(RGB24ToARGBRow_Any_NEON, RGB24ToARGBRow_NEON, 0, 3, 4, 7) 528 #endif 529 #ifdef HAS_RGB24TOARGBROW_MSA 530 ANY11(RGB24ToARGBRow_Any_MSA, RGB24ToARGBRow_MSA, 0, 3, 4, 15) 531 #endif 532 #ifdef HAS_RAWTOARGBROW_NEON 533 ANY11(RAWToARGBRow_Any_NEON, RAWToARGBRow_NEON, 0, 3, 4, 7) 534 #endif 535 #ifdef HAS_RAWTOARGBROW_MSA 536 ANY11(RAWToARGBRow_Any_MSA, RAWToARGBRow_MSA, 0, 3, 4, 15) 537 #endif 538 #ifdef HAS_RGB565TOARGBROW_NEON 539 ANY11(RGB565ToARGBRow_Any_NEON, RGB565ToARGBRow_NEON, 0, 2, 4, 7) 540 #endif 541 #ifdef HAS_RGB565TOARGBROW_MSA 542 ANY11(RGB565ToARGBRow_Any_MSA, RGB565ToARGBRow_MSA, 0, 2, 4, 15) 543 #endif 544 #ifdef HAS_ARGB1555TOARGBROW_NEON 545 ANY11(ARGB1555ToARGBRow_Any_NEON, ARGB1555ToARGBRow_NEON, 0, 2, 4, 7) 546 #endif 547 #ifdef HAS_ARGB1555TOARGBROW_MSA 548 ANY11(ARGB1555ToARGBRow_Any_MSA, ARGB1555ToARGBRow_MSA, 0, 2, 4, 15) 549 #endif 550 #ifdef HAS_ARGB4444TOARGBROW_NEON 551 ANY11(ARGB4444ToARGBRow_Any_NEON, ARGB4444ToARGBRow_NEON, 0, 2, 4, 7) 552 #endif 553 #ifdef HAS_RGB24TOARGBROW_DSPR2 554 ANY11(RGB24ToARGBRow_Any_DSPR2, RGB24ToARGBRow_DSPR2, 0, 3, 4, 7) 555 #endif 556 #ifdef HAS_RAWTOARGBROW_DSPR2 557 ANY11(RAWToARGBRow_Any_DSPR2, RAWToARGBRow_DSPR2, 0, 3, 4, 7) 558 #endif 559 #ifdef HAS_RGB565TOARGBROW_DSPR2 560 ANY11(RGB565ToARGBRow_Any_DSPR2, RGB565ToARGBRow_DSPR2, 0, 2, 4, 7) 561 #endif 562 #ifdef HAS_ARGB1555TOARGBROW_DSPR2 563 ANY11(ARGB1555ToARGBRow_Any_DSPR2, ARGB1555ToARGBRow_DSPR2, 0, 2, 4, 7) 564 #endif 565 #ifdef HAS_ARGB4444TOARGBROW_DSPR2 566 ANY11(ARGB4444ToARGBRow_Any_DSPR2, ARGB4444ToARGBRow_DSPR2, 0, 2, 4, 7) 567 #endif 568 #ifdef HAS_BGRATOYROW_DSPR2 569 ANY11(BGRAToYRow_Any_DSPR2, BGRAToYRow_DSPR2, 0, 4, 1, 7) 570 #endif 571 #ifdef HAS_ARGBTOYROW_DSPR2 572 ANY11(ARGBToYRow_Any_DSPR2, ARGBToYRow_DSPR2, 0, 4, 1, 7) 573 #endif 574 #ifdef HAS_ABGRTOYROW_DSPR2 575 ANY11(ABGRToYRow_Any_DSPR2, ABGRToYRow_DSPR2, 0, 4, 1, 7) 576 #endif 577 #ifdef HAS_RGBATOYROW_DSPR2 578 ANY11(RGBAToYRow_Any_DSPR2, RGBAToYRow_DSPR2, 0, 4, 1, 7) 579 #endif 580 #ifdef HAS_ARGB4444TOARGBROW_MSA 581 ANY11(ARGB4444ToARGBRow_Any_MSA, ARGB4444ToARGBRow_MSA, 0, 2, 4, 15) 582 #endif 583 #ifdef HAS_ARGBATTENUATEROW_SSSE3 584 ANY11(ARGBAttenuateRow_Any_SSSE3, ARGBAttenuateRow_SSSE3, 0, 4, 4, 3) 585 #endif 586 #ifdef HAS_ARGBUNATTENUATEROW_SSE2 587 ANY11(ARGBUnattenuateRow_Any_SSE2, ARGBUnattenuateRow_SSE2, 0, 4, 4, 3) 588 #endif 589 #ifdef HAS_ARGBATTENUATEROW_AVX2 590 ANY11(ARGBAttenuateRow_Any_AVX2, ARGBAttenuateRow_AVX2, 0, 4, 4, 7) 591 #endif 592 #ifdef HAS_ARGBUNATTENUATEROW_AVX2 593 ANY11(ARGBUnattenuateRow_Any_AVX2, ARGBUnattenuateRow_AVX2, 0, 4, 4, 7) 594 #endif 595 #ifdef HAS_ARGBATTENUATEROW_NEON 596 ANY11(ARGBAttenuateRow_Any_NEON, ARGBAttenuateRow_NEON, 0, 4, 4, 7) 597 #endif 598 #ifdef HAS_ARGBATTENUATEROW_MSA 599 ANY11(ARGBAttenuateRow_Any_MSA, ARGBAttenuateRow_MSA, 0, 4, 4, 7) 600 #endif 601 #ifdef HAS_ARGBEXTRACTALPHAROW_SSE2 602 ANY11(ARGBExtractAlphaRow_Any_SSE2, ARGBExtractAlphaRow_SSE2, 0, 4, 1, 7) 603 #endif 604 #ifdef HAS_ARGBEXTRACTALPHAROW_AVX2 605 ANY11(ARGBExtractAlphaRow_Any_AVX2, ARGBExtractAlphaRow_AVX2, 0, 4, 1, 32) 606 #endif 607 #ifdef HAS_ARGBEXTRACTALPHAROW_NEON 608 ANY11(ARGBExtractAlphaRow_Any_NEON, ARGBExtractAlphaRow_NEON, 0, 4, 1, 15) 609 #endif 610 #undef ANY11 611 612 // Any 1 to 1 blended. Destination is read, modify, write. 613 #define ANY11B(NAMEANY, ANY_SIMD, UVSHIFT, SBPP, BPP, MASK) \ 614 void NAMEANY(const uint8* src_ptr, uint8* dst_ptr, int width) { \ 615 SIMD_ALIGNED(uint8 temp[128 * 2]); \ 616 memset(temp, 0, 128 * 2); /* for YUY2 and msan */ \ 617 int r = width & MASK; \ 618 int n = width & ~MASK; \ 619 if (n > 0) { \ 620 ANY_SIMD(src_ptr, dst_ptr, n); \ 621 } \ 622 memcpy(temp, src_ptr + (n >> UVSHIFT) * SBPP, SS(r, UVSHIFT) * SBPP); \ 623 memcpy(temp + 128, dst_ptr + n * BPP, r * BPP); \ 624 ANY_SIMD(temp, temp + 128, MASK + 1); \ 625 memcpy(dst_ptr + n * BPP, temp + 128, r * BPP); \ 626 } 627 628 #ifdef HAS_ARGBCOPYALPHAROW_AVX2 629 ANY11B(ARGBCopyAlphaRow_Any_AVX2, ARGBCopyAlphaRow_AVX2, 0, 4, 4, 15) 630 #endif 631 #ifdef HAS_ARGBCOPYALPHAROW_SSE2 632 ANY11B(ARGBCopyAlphaRow_Any_SSE2, ARGBCopyAlphaRow_SSE2, 0, 4, 4, 7) 633 #endif 634 #ifdef HAS_ARGBCOPYYTOALPHAROW_AVX2 635 ANY11B(ARGBCopyYToAlphaRow_Any_AVX2, ARGBCopyYToAlphaRow_AVX2, 0, 1, 4, 15) 636 #endif 637 #ifdef HAS_ARGBCOPYYTOALPHAROW_SSE2 638 ANY11B(ARGBCopyYToAlphaRow_Any_SSE2, ARGBCopyYToAlphaRow_SSE2, 0, 1, 4, 7) 639 #endif 640 #undef ANY11B 641 642 // Any 1 to 1 with parameter. 643 #define ANY11P(NAMEANY, ANY_SIMD, T, SBPP, BPP, MASK) \ 644 void NAMEANY(const uint8* src_ptr, uint8* dst_ptr, T shuffler, int width) { \ 645 SIMD_ALIGNED(uint8 temp[64 * 2]); \ 646 memset(temp, 0, 64); /* for msan */ \ 647 int r = width & MASK; \ 648 int n = width & ~MASK; \ 649 if (n > 0) { \ 650 ANY_SIMD(src_ptr, dst_ptr, shuffler, n); \ 651 } \ 652 memcpy(temp, src_ptr + n * SBPP, r * SBPP); \ 653 ANY_SIMD(temp, temp + 64, shuffler, MASK + 1); \ 654 memcpy(dst_ptr + n * BPP, temp + 64, r * BPP); \ 655 } 656 657 #if defined(HAS_ARGBTORGB565DITHERROW_SSE2) 658 ANY11P(ARGBToRGB565DitherRow_Any_SSE2, 659 ARGBToRGB565DitherRow_SSE2, 660 const uint32, 661 4, 662 2, 663 3) 664 #endif 665 #if defined(HAS_ARGBTORGB565DITHERROW_AVX2) 666 ANY11P(ARGBToRGB565DitherRow_Any_AVX2, 667 ARGBToRGB565DitherRow_AVX2, 668 const uint32, 669 4, 670 2, 671 7) 672 #endif 673 #if defined(HAS_ARGBTORGB565DITHERROW_NEON) 674 ANY11P(ARGBToRGB565DitherRow_Any_NEON, 675 ARGBToRGB565DitherRow_NEON, 676 const uint32, 677 4, 678 2, 679 7) 680 #endif 681 #if defined(HAS_ARGBTORGB565DITHERROW_MSA) 682 ANY11P(ARGBToRGB565DitherRow_Any_MSA, 683 ARGBToRGB565DitherRow_MSA, 684 const uint32, 685 4, 686 2, 687 7) 688 #endif 689 #ifdef HAS_ARGBSHUFFLEROW_SSE2 690 ANY11P(ARGBShuffleRow_Any_SSE2, ARGBShuffleRow_SSE2, const uint8*, 4, 4, 3) 691 #endif 692 #ifdef HAS_ARGBSHUFFLEROW_SSSE3 693 ANY11P(ARGBShuffleRow_Any_SSSE3, ARGBShuffleRow_SSSE3, const uint8*, 4, 4, 7) 694 #endif 695 #ifdef HAS_ARGBSHUFFLEROW_AVX2 696 ANY11P(ARGBShuffleRow_Any_AVX2, ARGBShuffleRow_AVX2, const uint8*, 4, 4, 15) 697 #endif 698 #ifdef HAS_ARGBSHUFFLEROW_NEON 699 ANY11P(ARGBShuffleRow_Any_NEON, ARGBShuffleRow_NEON, const uint8*, 4, 4, 3) 700 #endif 701 #ifdef HAS_ARGBSHUFFLEROW_MSA 702 ANY11P(ARGBShuffleRow_Any_MSA, ARGBShuffleRow_MSA, const uint8*, 4, 4, 7) 703 #endif 704 #undef ANY11P 705 706 // Any 1 to 1 with parameter and shorts. BPP measures in shorts. 707 #define ANY11P16(NAMEANY, ANY_SIMD, T, SBPP, BPP, MASK) \ 708 void NAMEANY(const uint16* src_ptr, uint16* dst_ptr, T shuffler, \ 709 int width) { \ 710 SIMD_ALIGNED(uint16 temp[64 * 2]); \ 711 memset(temp, 0, 64*sizeof(uint16)); /* for msan */ \ 712 int r = width & MASK; \ 713 int n = width & ~MASK; \ 714 if (n > 0) { \ 715 ANY_SIMD(src_ptr, dst_ptr, shuffler, n); \ 716 } \ 717 memcpy(temp, src_ptr + n * SBPP, r * SBPP); \ 718 ANY_SIMD(temp, temp + 64, shuffler, MASK + 1); \ 719 memcpy(dst_ptr + n * BPP, temp + 64, r * BPP); \ 720 } 721 722 #ifdef HAS_HALFFLOATROW_SSE2 723 ANY11P16(HalfFloatRow_Any_SSE2, HalfFloatRow_SSE2, float, 1, 1, 7) 724 #endif 725 #ifdef HAS_HALFFLOATROW_AVX2 726 ANY11P16(HalfFloatRow_Any_AVX2, HalfFloatRow_AVX2, float, 1, 1, 15) 727 #endif 728 #ifdef HAS_HALFFLOATROW_F16C 729 ANY11P16(HalfFloatRow_Any_F16C, HalfFloatRow_F16C, float, 1, 1, 15) 730 ANY11P16(HalfFloat1Row_Any_F16C, HalfFloat1Row_F16C, float, 1, 1, 15) 731 #endif 732 #ifdef HAS_HALFFLOATROW_NEON 733 ANY11P16(HalfFloatRow_Any_NEON, HalfFloatRow_NEON, float, 1, 1, 7) 734 ANY11P16(HalfFloat1Row_Any_NEON, HalfFloat1Row_NEON, float, 1, 1, 7) 735 #endif 736 #undef ANY11P16 737 738 // Any 1 to 1 with yuvconstants 739 #define ANY11C(NAMEANY, ANY_SIMD, UVSHIFT, SBPP, BPP, MASK) \ 740 void NAMEANY(const uint8* src_ptr, uint8* dst_ptr, \ 741 const struct YuvConstants* yuvconstants, int width) { \ 742 SIMD_ALIGNED(uint8 temp[128 * 2]); \ 743 memset(temp, 0, 128); /* for YUY2 and msan */ \ 744 int r = width & MASK; \ 745 int n = width & ~MASK; \ 746 if (n > 0) { \ 747 ANY_SIMD(src_ptr, dst_ptr, yuvconstants, n); \ 748 } \ 749 memcpy(temp, src_ptr + (n >> UVSHIFT) * SBPP, SS(r, UVSHIFT) * SBPP); \ 750 ANY_SIMD(temp, temp + 128, yuvconstants, MASK + 1); \ 751 memcpy(dst_ptr + n * BPP, temp + 128, r * BPP); \ 752 } 753 #if defined(HAS_YUY2TOARGBROW_SSSE3) 754 ANY11C(YUY2ToARGBRow_Any_SSSE3, YUY2ToARGBRow_SSSE3, 1, 4, 4, 15) 755 ANY11C(UYVYToARGBRow_Any_SSSE3, UYVYToARGBRow_SSSE3, 1, 4, 4, 15) 756 #endif 757 #if defined(HAS_YUY2TOARGBROW_AVX2) 758 ANY11C(YUY2ToARGBRow_Any_AVX2, YUY2ToARGBRow_AVX2, 1, 4, 4, 31) 759 ANY11C(UYVYToARGBRow_Any_AVX2, UYVYToARGBRow_AVX2, 1, 4, 4, 31) 760 #endif 761 #if defined(HAS_YUY2TOARGBROW_NEON) 762 ANY11C(YUY2ToARGBRow_Any_NEON, YUY2ToARGBRow_NEON, 1, 4, 4, 7) 763 ANY11C(UYVYToARGBRow_Any_NEON, UYVYToARGBRow_NEON, 1, 4, 4, 7) 764 #endif 765 #if defined(HAS_YUY2TOARGBROW_MSA) 766 ANY11C(YUY2ToARGBRow_Any_MSA, YUY2ToARGBRow_MSA, 1, 4, 4, 7) 767 ANY11C(UYVYToARGBRow_Any_MSA, UYVYToARGBRow_MSA, 1, 4, 4, 7) 768 #endif 769 #undef ANY11C 770 771 // Any 1 to 1 interpolate. Takes 2 rows of source via stride. 772 #define ANY11T(NAMEANY, ANY_SIMD, SBPP, BPP, MASK) \ 773 void NAMEANY(uint8* dst_ptr, const uint8* src_ptr, ptrdiff_t src_stride_ptr, \ 774 int width, int source_y_fraction) { \ 775 SIMD_ALIGNED(uint8 temp[64 * 3]); \ 776 memset(temp, 0, 64 * 2); /* for msan */ \ 777 int r = width & MASK; \ 778 int n = width & ~MASK; \ 779 if (n > 0) { \ 780 ANY_SIMD(dst_ptr, src_ptr, src_stride_ptr, n, source_y_fraction); \ 781 } \ 782 memcpy(temp, src_ptr + n * SBPP, r * SBPP); \ 783 memcpy(temp + 64, src_ptr + src_stride_ptr + n * SBPP, r * SBPP); \ 784 ANY_SIMD(temp + 128, temp, 64, MASK + 1, source_y_fraction); \ 785 memcpy(dst_ptr + n * BPP, temp + 128, r * BPP); \ 786 } 787 788 #ifdef HAS_INTERPOLATEROW_AVX2 789 ANY11T(InterpolateRow_Any_AVX2, InterpolateRow_AVX2, 1, 1, 31) 790 #endif 791 #ifdef HAS_INTERPOLATEROW_SSSE3 792 ANY11T(InterpolateRow_Any_SSSE3, InterpolateRow_SSSE3, 1, 1, 15) 793 #endif 794 #ifdef HAS_INTERPOLATEROW_NEON 795 ANY11T(InterpolateRow_Any_NEON, InterpolateRow_NEON, 1, 1, 15) 796 #endif 797 #ifdef HAS_INTERPOLATEROW_DSPR2 798 ANY11T(InterpolateRow_Any_DSPR2, InterpolateRow_DSPR2, 1, 1, 3) 799 #endif 800 #undef ANY11T 801 802 // Any 1 to 1 mirror. 803 #define ANY11M(NAMEANY, ANY_SIMD, BPP, MASK) \ 804 void NAMEANY(const uint8* src_ptr, uint8* dst_ptr, int width) { \ 805 SIMD_ALIGNED(uint8 temp[64 * 2]); \ 806 memset(temp, 0, 64); /* for msan */ \ 807 int r = width & MASK; \ 808 int n = width & ~MASK; \ 809 if (n > 0) { \ 810 ANY_SIMD(src_ptr + r * BPP, dst_ptr, n); \ 811 } \ 812 memcpy(temp, src_ptr, r* BPP); \ 813 ANY_SIMD(temp, temp + 64, MASK + 1); \ 814 memcpy(dst_ptr + n * BPP, temp + 64 + (MASK + 1 - r) * BPP, r * BPP); \ 815 } 816 817 #ifdef HAS_MIRRORROW_AVX2 818 ANY11M(MirrorRow_Any_AVX2, MirrorRow_AVX2, 1, 31) 819 #endif 820 #ifdef HAS_MIRRORROW_SSSE3 821 ANY11M(MirrorRow_Any_SSSE3, MirrorRow_SSSE3, 1, 15) 822 #endif 823 #ifdef HAS_MIRRORROW_NEON 824 ANY11M(MirrorRow_Any_NEON, MirrorRow_NEON, 1, 15) 825 #endif 826 #ifdef HAS_MIRRORROW_MSA 827 ANY11M(MirrorRow_Any_MSA, MirrorRow_MSA, 1, 63) 828 #endif 829 #ifdef HAS_ARGBMIRRORROW_AVX2 830 ANY11M(ARGBMirrorRow_Any_AVX2, ARGBMirrorRow_AVX2, 4, 7) 831 #endif 832 #ifdef HAS_ARGBMIRRORROW_SSE2 833 ANY11M(ARGBMirrorRow_Any_SSE2, ARGBMirrorRow_SSE2, 4, 3) 834 #endif 835 #ifdef HAS_ARGBMIRRORROW_NEON 836 ANY11M(ARGBMirrorRow_Any_NEON, ARGBMirrorRow_NEON, 4, 3) 837 #endif 838 #ifdef HAS_ARGBMIRRORROW_MSA 839 ANY11M(ARGBMirrorRow_Any_MSA, ARGBMirrorRow_MSA, 4, 15) 840 #endif 841 #undef ANY11M 842 843 // Any 1 plane. (memset) 844 #define ANY1(NAMEANY, ANY_SIMD, T, BPP, MASK) \ 845 void NAMEANY(uint8* dst_ptr, T v32, int width) { \ 846 SIMD_ALIGNED(uint8 temp[64]); \ 847 int r = width & MASK; \ 848 int n = width & ~MASK; \ 849 if (n > 0) { \ 850 ANY_SIMD(dst_ptr, v32, n); \ 851 } \ 852 ANY_SIMD(temp, v32, MASK + 1); \ 853 memcpy(dst_ptr + n * BPP, temp, r * BPP); \ 854 } 855 856 #ifdef HAS_SETROW_X86 857 ANY1(SetRow_Any_X86, SetRow_X86, uint8, 1, 3) 858 #endif 859 #ifdef HAS_SETROW_NEON 860 ANY1(SetRow_Any_NEON, SetRow_NEON, uint8, 1, 15) 861 #endif 862 #ifdef HAS_ARGBSETROW_NEON 863 ANY1(ARGBSetRow_Any_NEON, ARGBSetRow_NEON, uint32, 4, 3) 864 #endif 865 #undef ANY1 866 867 // Any 1 to 2. Outputs UV planes. 868 #define ANY12(NAMEANY, ANY_SIMD, UVSHIFT, BPP, DUVSHIFT, MASK) \ 869 void NAMEANY(const uint8* src_ptr, uint8* dst_u, uint8* dst_v, int width) { \ 870 SIMD_ALIGNED(uint8 temp[128 * 3]); \ 871 memset(temp, 0, 128); /* for msan */ \ 872 int r = width & MASK; \ 873 int n = width & ~MASK; \ 874 if (n > 0) { \ 875 ANY_SIMD(src_ptr, dst_u, dst_v, n); \ 876 } \ 877 memcpy(temp, src_ptr + (n >> UVSHIFT) * BPP, SS(r, UVSHIFT) * BPP); \ 878 ANY_SIMD(temp, temp + 128, temp + 256, MASK + 1); \ 879 memcpy(dst_u + (n >> DUVSHIFT), temp + 128, SS(r, DUVSHIFT)); \ 880 memcpy(dst_v + (n >> DUVSHIFT), temp + 256, SS(r, DUVSHIFT)); \ 881 } 882 883 #ifdef HAS_SPLITUVROW_SSE2 884 ANY12(SplitUVRow_Any_SSE2, SplitUVRow_SSE2, 0, 2, 0, 15) 885 #endif 886 #ifdef HAS_SPLITUVROW_AVX2 887 ANY12(SplitUVRow_Any_AVX2, SplitUVRow_AVX2, 0, 2, 0, 31) 888 #endif 889 #ifdef HAS_SPLITUVROW_NEON 890 ANY12(SplitUVRow_Any_NEON, SplitUVRow_NEON, 0, 2, 0, 15) 891 #endif 892 #ifdef HAS_SPLITUVROW_DSPR2 893 ANY12(SplitUVRow_Any_DSPR2, SplitUVRow_DSPR2, 0, 2, 0, 15) 894 #endif 895 #ifdef HAS_ARGBTOUV444ROW_SSSE3 896 ANY12(ARGBToUV444Row_Any_SSSE3, ARGBToUV444Row_SSSE3, 0, 4, 0, 15) 897 #endif 898 #ifdef HAS_YUY2TOUV422ROW_AVX2 899 ANY12(YUY2ToUV422Row_Any_AVX2, YUY2ToUV422Row_AVX2, 1, 4, 1, 31) 900 ANY12(UYVYToUV422Row_Any_AVX2, UYVYToUV422Row_AVX2, 1, 4, 1, 31) 901 #endif 902 #ifdef HAS_YUY2TOUV422ROW_SSE2 903 ANY12(YUY2ToUV422Row_Any_SSE2, YUY2ToUV422Row_SSE2, 1, 4, 1, 15) 904 ANY12(UYVYToUV422Row_Any_SSE2, UYVYToUV422Row_SSE2, 1, 4, 1, 15) 905 #endif 906 #ifdef HAS_YUY2TOUV422ROW_NEON 907 ANY12(ARGBToUV444Row_Any_NEON, ARGBToUV444Row_NEON, 0, 4, 0, 7) 908 ANY12(YUY2ToUV422Row_Any_NEON, YUY2ToUV422Row_NEON, 1, 4, 1, 15) 909 ANY12(UYVYToUV422Row_Any_NEON, UYVYToUV422Row_NEON, 1, 4, 1, 15) 910 #endif 911 #ifdef HAS_YUY2TOUV422ROW_MSA 912 ANY12(ARGBToUV444Row_Any_MSA, ARGBToUV444Row_MSA, 0, 4, 0, 15) 913 ANY12(YUY2ToUV422Row_Any_MSA, YUY2ToUV422Row_MSA, 1, 4, 1, 31) 914 ANY12(UYVYToUV422Row_Any_MSA, UYVYToUV422Row_MSA, 1, 4, 1, 31) 915 #endif 916 #undef ANY12 917 918 // Any 1 to 2 with source stride (2 rows of source). Outputs UV planes. 919 // 128 byte row allows for 32 avx ARGB pixels. 920 #define ANY12S(NAMEANY, ANY_SIMD, UVSHIFT, BPP, MASK) \ 921 void NAMEANY(const uint8* src_ptr, int src_stride_ptr, uint8* dst_u, \ 922 uint8* dst_v, int width) { \ 923 SIMD_ALIGNED(uint8 temp[128 * 4]); \ 924 memset(temp, 0, 128 * 2); /* for msan */ \ 925 int r = width & MASK; \ 926 int n = width & ~MASK; \ 927 if (n > 0) { \ 928 ANY_SIMD(src_ptr, src_stride_ptr, dst_u, dst_v, n); \ 929 } \ 930 memcpy(temp, src_ptr + (n >> UVSHIFT) * BPP, SS(r, UVSHIFT) * BPP); \ 931 memcpy(temp + 128, src_ptr + src_stride_ptr + (n >> UVSHIFT) * BPP, \ 932 SS(r, UVSHIFT) * BPP); \ 933 if ((width & 1) && UVSHIFT == 0) { /* repeat last pixel for subsample */ \ 934 memcpy(temp + SS(r, UVSHIFT) * BPP, temp + SS(r, UVSHIFT) * BPP - BPP, \ 935 BPP); \ 936 memcpy(temp + 128 + SS(r, UVSHIFT) * BPP, \ 937 temp + 128 + SS(r, UVSHIFT) * BPP - BPP, BPP); \ 938 } \ 939 ANY_SIMD(temp, 128, temp + 256, temp + 384, MASK + 1); \ 940 memcpy(dst_u + (n >> 1), temp + 256, SS(r, 1)); \ 941 memcpy(dst_v + (n >> 1), temp + 384, SS(r, 1)); \ 942 } 943 944 #ifdef HAS_ARGBTOUVROW_AVX2 945 ANY12S(ARGBToUVRow_Any_AVX2, ARGBToUVRow_AVX2, 0, 4, 31) 946 #endif 947 #ifdef HAS_ARGBTOUVJROW_AVX2 948 ANY12S(ARGBToUVJRow_Any_AVX2, ARGBToUVJRow_AVX2, 0, 4, 31) 949 #endif 950 #ifdef HAS_ARGBTOUVROW_SSSE3 951 ANY12S(ARGBToUVRow_Any_SSSE3, ARGBToUVRow_SSSE3, 0, 4, 15) 952 ANY12S(ARGBToUVJRow_Any_SSSE3, ARGBToUVJRow_SSSE3, 0, 4, 15) 953 ANY12S(BGRAToUVRow_Any_SSSE3, BGRAToUVRow_SSSE3, 0, 4, 15) 954 ANY12S(ABGRToUVRow_Any_SSSE3, ABGRToUVRow_SSSE3, 0, 4, 15) 955 ANY12S(RGBAToUVRow_Any_SSSE3, RGBAToUVRow_SSSE3, 0, 4, 15) 956 #endif 957 #ifdef HAS_YUY2TOUVROW_AVX2 958 ANY12S(YUY2ToUVRow_Any_AVX2, YUY2ToUVRow_AVX2, 1, 4, 31) 959 ANY12S(UYVYToUVRow_Any_AVX2, UYVYToUVRow_AVX2, 1, 4, 31) 960 #endif 961 #ifdef HAS_YUY2TOUVROW_SSE2 962 ANY12S(YUY2ToUVRow_Any_SSE2, YUY2ToUVRow_SSE2, 1, 4, 15) 963 ANY12S(UYVYToUVRow_Any_SSE2, UYVYToUVRow_SSE2, 1, 4, 15) 964 #endif 965 #ifdef HAS_ARGBTOUVROW_NEON 966 ANY12S(ARGBToUVRow_Any_NEON, ARGBToUVRow_NEON, 0, 4, 15) 967 #endif 968 #ifdef HAS_ARGBTOUVROW_MSA 969 ANY12S(ARGBToUVRow_Any_MSA, ARGBToUVRow_MSA, 0, 4, 31) 970 #endif 971 #ifdef HAS_ARGBTOUVJROW_NEON 972 ANY12S(ARGBToUVJRow_Any_NEON, ARGBToUVJRow_NEON, 0, 4, 15) 973 #endif 974 #ifdef HAS_ARGBTOUVJROW_MSA 975 ANY12S(ARGBToUVJRow_Any_MSA, ARGBToUVJRow_MSA, 0, 4, 31) 976 #endif 977 #ifdef HAS_BGRATOUVROW_NEON 978 ANY12S(BGRAToUVRow_Any_NEON, BGRAToUVRow_NEON, 0, 4, 15) 979 #endif 980 #ifdef HAS_BGRATOUVROW_MSA 981 ANY12S(BGRAToUVRow_Any_MSA, BGRAToUVRow_MSA, 0, 4, 31) 982 #endif 983 #ifdef HAS_ABGRTOUVROW_NEON 984 ANY12S(ABGRToUVRow_Any_NEON, ABGRToUVRow_NEON, 0, 4, 15) 985 #endif 986 #ifdef HAS_ABGRTOUVROW_MSA 987 ANY12S(ABGRToUVRow_Any_MSA, ABGRToUVRow_MSA, 0, 4, 31) 988 #endif 989 #ifdef HAS_RGBATOUVROW_NEON 990 ANY12S(RGBAToUVRow_Any_NEON, RGBAToUVRow_NEON, 0, 4, 15) 991 #endif 992 #ifdef HAS_RGBATOUVROW_MSA 993 ANY12S(RGBAToUVRow_Any_MSA, RGBAToUVRow_MSA, 0, 4, 31) 994 #endif 995 #ifdef HAS_RGB24TOUVROW_NEON 996 ANY12S(RGB24ToUVRow_Any_NEON, RGB24ToUVRow_NEON, 0, 3, 15) 997 #endif 998 #ifdef HAS_RGB24TOUVROW_MSA 999 ANY12S(RGB24ToUVRow_Any_MSA, RGB24ToUVRow_MSA, 0, 3, 15) 1000 #endif 1001 #ifdef HAS_RAWTOUVROW_NEON 1002 ANY12S(RAWToUVRow_Any_NEON, RAWToUVRow_NEON, 0, 3, 15) 1003 #endif 1004 #ifdef HAS_RAWTOUVROW_MSA 1005 ANY12S(RAWToUVRow_Any_MSA, RAWToUVRow_MSA, 0, 3, 15) 1006 #endif 1007 #ifdef HAS_RGB565TOUVROW_NEON 1008 ANY12S(RGB565ToUVRow_Any_NEON, RGB565ToUVRow_NEON, 0, 2, 15) 1009 #endif 1010 #ifdef HAS_RGB565TOUVROW_MSA 1011 ANY12S(RGB565ToUVRow_Any_MSA, RGB565ToUVRow_MSA, 0, 2, 15) 1012 #endif 1013 #ifdef HAS_ARGB1555TOUVROW_NEON 1014 ANY12S(ARGB1555ToUVRow_Any_NEON, ARGB1555ToUVRow_NEON, 0, 2, 15) 1015 #endif 1016 #ifdef HAS_ARGB1555TOUVROW_MSA 1017 ANY12S(ARGB1555ToUVRow_Any_MSA, ARGB1555ToUVRow_MSA, 0, 2, 15) 1018 #endif 1019 #ifdef HAS_ARGB4444TOUVROW_NEON 1020 ANY12S(ARGB4444ToUVRow_Any_NEON, ARGB4444ToUVRow_NEON, 0, 2, 15) 1021 #endif 1022 #ifdef HAS_YUY2TOUVROW_NEON 1023 ANY12S(YUY2ToUVRow_Any_NEON, YUY2ToUVRow_NEON, 1, 4, 15) 1024 #endif 1025 #ifdef HAS_UYVYTOUVROW_NEON 1026 ANY12S(UYVYToUVRow_Any_NEON, UYVYToUVRow_NEON, 1, 4, 15) 1027 #endif 1028 #ifdef HAS_BGRATOUVROW_DSPR2 1029 ANY12S(BGRAToUVRow_Any_DSPR2, BGRAToUVRow_DSPR2, 0, 4, 15) 1030 #endif 1031 #ifdef HAS_ABGRTOUVROW_DSPR2 1032 ANY12S(ABGRToUVRow_Any_DSPR2, ABGRToUVRow_DSPR2, 0, 4, 15) 1033 #endif 1034 #ifdef HAS_RGBATOUVROW_DSPR2 1035 ANY12S(RGBAToUVRow_Any_DSPR2, RGBAToUVRow_DSPR2, 0, 4, 15) 1036 #endif 1037 #ifdef HAS_ARGBTOUVROW_DSPR2 1038 ANY12S(ARGBToUVRow_Any_DSPR2, ARGBToUVRow_DSPR2, 0, 4, 15) 1039 #endif 1040 #ifdef HAS_YUY2TOUVROW_MSA 1041 ANY12S(YUY2ToUVRow_Any_MSA, YUY2ToUVRow_MSA, 1, 4, 31) 1042 #endif 1043 #ifdef HAS_UYVYTOUVROW_MSA 1044 ANY12S(UYVYToUVRow_Any_MSA, UYVYToUVRow_MSA, 1, 4, 31) 1045 #endif 1046 #undef ANY12S 1047 1048 #ifdef __cplusplus 1049 } // extern "C" 1050 } // namespace libyuv 1051 #endif 1052