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 "libyuv/basic_types.h" 14 15 #ifdef __cplusplus 16 namespace libyuv { 17 extern "C" { 18 #endif 19 20 // TODO(fbarchard): Consider 'any' functions handling any quantity of pixels. 21 // TODO(fbarchard): Consider 'any' functions handling odd alignment. 22 // YUV to RGB does multiple of 8 with SIMD and remainder with C. 23 #define YANY(NAMEANY, I420TORGB_SIMD, I420TORGB_C, UV_SHIFT, BPP, MASK) \ 24 void NAMEANY(const uint8* y_buf, \ 25 const uint8* u_buf, \ 26 const uint8* v_buf, \ 27 uint8* rgb_buf, \ 28 int width) { \ 29 int n = width & ~MASK; \ 30 I420TORGB_SIMD(y_buf, u_buf, v_buf, rgb_buf, n); \ 31 I420TORGB_C(y_buf + n, \ 32 u_buf + (n >> UV_SHIFT), \ 33 v_buf + (n >> UV_SHIFT), \ 34 rgb_buf + n * BPP, width & MASK); \ 35 } 36 37 #ifdef HAS_I422TOARGBROW_SSSE3 38 YANY(I444ToARGBRow_Any_SSSE3, I444ToARGBRow_Unaligned_SSSE3, I444ToARGBRow_C, 39 0, 4, 7) 40 YANY(I422ToARGBRow_Any_SSSE3, I422ToARGBRow_Unaligned_SSSE3, I422ToARGBRow_C, 41 1, 4, 7) 42 YANY(I411ToARGBRow_Any_SSSE3, I411ToARGBRow_Unaligned_SSSE3, I411ToARGBRow_C, 43 2, 4, 7) 44 YANY(I422ToBGRARow_Any_SSSE3, I422ToBGRARow_Unaligned_SSSE3, I422ToBGRARow_C, 45 1, 4, 7) 46 YANY(I422ToABGRRow_Any_SSSE3, I422ToABGRRow_Unaligned_SSSE3, I422ToABGRRow_C, 47 1, 4, 7) 48 YANY(I422ToRGBARow_Any_SSSE3, I422ToRGBARow_Unaligned_SSSE3, I422ToRGBARow_C, 49 1, 4, 7) 50 // I422ToRGB565Row_SSSE3 is unaligned. 51 YANY(I422ToARGB4444Row_Any_SSSE3, I422ToARGB4444Row_SSSE3, I422ToARGB4444Row_C, 52 1, 2, 7) 53 YANY(I422ToARGB1555Row_Any_SSSE3, I422ToARGB1555Row_SSSE3, I422ToARGB1555Row_C, 54 1, 2, 7) 55 YANY(I422ToRGB565Row_Any_SSSE3, I422ToRGB565Row_SSSE3, I422ToRGB565Row_C, 56 1, 2, 7) 57 // I422ToRGB24Row_SSSE3 is unaligned. 58 YANY(I422ToRGB24Row_Any_SSSE3, I422ToRGB24Row_SSSE3, I422ToRGB24Row_C, 1, 3, 7) 59 YANY(I422ToRAWRow_Any_SSSE3, I422ToRAWRow_SSSE3, I422ToRAWRow_C, 1, 3, 7) 60 YANY(I422ToYUY2Row_Any_SSE2, I422ToYUY2Row_SSE2, I422ToYUY2Row_C, 1, 2, 15) 61 YANY(I422ToUYVYRow_Any_SSE2, I422ToUYVYRow_SSE2, I422ToUYVYRow_C, 1, 2, 15) 62 #endif // HAS_I422TOARGBROW_SSSE3 63 #ifdef HAS_I422TOARGBROW_AVX2 64 YANY(I422ToARGBRow_Any_AVX2, I422ToARGBRow_AVX2, I422ToARGBRow_C, 1, 4, 15) 65 #endif // HAS_I422TOARGBROW_AVX2 66 #ifdef HAS_I422TOARGBROW_NEON 67 YANY(I444ToARGBRow_Any_NEON, I444ToARGBRow_NEON, I444ToARGBRow_C, 0, 4, 7) 68 YANY(I422ToARGBRow_Any_NEON, I422ToARGBRow_NEON, I422ToARGBRow_C, 1, 4, 7) 69 YANY(I411ToARGBRow_Any_NEON, I411ToARGBRow_NEON, I411ToARGBRow_C, 2, 4, 7) 70 YANY(I422ToBGRARow_Any_NEON, I422ToBGRARow_NEON, I422ToBGRARow_C, 1, 4, 7) 71 YANY(I422ToABGRRow_Any_NEON, I422ToABGRRow_NEON, I422ToABGRRow_C, 1, 4, 7) 72 YANY(I422ToRGBARow_Any_NEON, I422ToRGBARow_NEON, I422ToRGBARow_C, 1, 4, 7) 73 YANY(I422ToRGB24Row_Any_NEON, I422ToRGB24Row_NEON, I422ToRGB24Row_C, 1, 3, 7) 74 YANY(I422ToRAWRow_Any_NEON, I422ToRAWRow_NEON, I422ToRAWRow_C, 1, 3, 7) 75 YANY(I422ToARGB4444Row_Any_NEON, I422ToARGB4444Row_NEON, I422ToARGB4444Row_C, 76 1, 2, 7) 77 YANY(I422ToARGB1555Row_Any_NEON, I422ToARGB1555Row_NEON, I422ToARGB1555Row_C, 78 1, 2, 7) 79 YANY(I422ToRGB565Row_Any_NEON, I422ToRGB565Row_NEON, I422ToRGB565Row_C, 1, 2, 7) 80 YANY(I422ToYUY2Row_Any_NEON, I422ToYUY2Row_NEON, I422ToYUY2Row_C, 1, 2, 15) 81 YANY(I422ToUYVYRow_Any_NEON, I422ToUYVYRow_NEON, I422ToUYVYRow_C, 1, 2, 15) 82 #endif // HAS_I422TOARGBROW_NEON 83 #undef YANY 84 85 // Wrappers to handle odd width 86 #define NV2NY(NAMEANY, NV12TORGB_SIMD, NV12TORGB_C, UV_SHIFT, BPP) \ 87 void NAMEANY(const uint8* y_buf, \ 88 const uint8* uv_buf, \ 89 uint8* rgb_buf, \ 90 int width) { \ 91 int n = width & ~7; \ 92 NV12TORGB_SIMD(y_buf, uv_buf, rgb_buf, n); \ 93 NV12TORGB_C(y_buf + n, \ 94 uv_buf + (n >> UV_SHIFT), \ 95 rgb_buf + n * BPP, width & 7); \ 96 } 97 98 #ifdef HAS_NV12TOARGBROW_SSSE3 99 NV2NY(NV12ToARGBRow_Any_SSSE3, NV12ToARGBRow_Unaligned_SSSE3, NV12ToARGBRow_C, 100 0, 4) 101 NV2NY(NV21ToARGBRow_Any_SSSE3, NV21ToARGBRow_Unaligned_SSSE3, NV21ToARGBRow_C, 102 0, 4) 103 #endif // HAS_NV12TOARGBROW_SSSE3 104 #ifdef HAS_NV12TOARGBROW_NEON 105 NV2NY(NV12ToARGBRow_Any_NEON, NV12ToARGBRow_NEON, NV12ToARGBRow_C, 0, 4) 106 NV2NY(NV21ToARGBRow_Any_NEON, NV21ToARGBRow_NEON, NV21ToARGBRow_C, 0, 4) 107 #endif // HAS_NV12TOARGBROW_NEON 108 #ifdef HAS_NV12TORGB565ROW_SSSE3 109 NV2NY(NV12ToRGB565Row_Any_SSSE3, NV12ToRGB565Row_SSSE3, NV12ToRGB565Row_C, 110 0, 2) 111 NV2NY(NV21ToRGB565Row_Any_SSSE3, NV21ToRGB565Row_SSSE3, NV21ToRGB565Row_C, 112 0, 2) 113 #endif // HAS_NV12TORGB565ROW_SSSE3 114 #ifdef HAS_NV12TORGB565ROW_NEON 115 NV2NY(NV12ToRGB565Row_Any_NEON, NV12ToRGB565Row_NEON, NV12ToRGB565Row_C, 0, 2) 116 NV2NY(NV21ToRGB565Row_Any_NEON, NV21ToRGB565Row_NEON, NV21ToRGB565Row_C, 0, 2) 117 #endif // HAS_NV12TORGB565ROW_NEON 118 #undef NVANY 119 120 #define RGBANY(NAMEANY, ARGBTORGB_SIMD, ARGBTORGB_C, MASK, SBPP, BPP) \ 121 void NAMEANY(const uint8* src, \ 122 uint8* dst, \ 123 int width) { \ 124 int n = width & ~MASK; \ 125 ARGBTORGB_SIMD(src, dst, n); \ 126 ARGBTORGB_C(src + n * SBPP, dst + n * BPP, width & MASK); \ 127 } 128 129 #if defined(HAS_ARGBTORGB24ROW_SSSE3) 130 RGBANY(ARGBToRGB24Row_Any_SSSE3, ARGBToRGB24Row_SSSE3, ARGBToRGB24Row_C, 131 15, 4, 3) 132 RGBANY(ARGBToRAWRow_Any_SSSE3, ARGBToRAWRow_SSSE3, ARGBToRAWRow_C, 133 15, 4, 3) 134 RGBANY(ARGBToRGB565Row_Any_SSE2, ARGBToRGB565Row_SSE2, ARGBToRGB565Row_C, 135 3, 4, 2) 136 RGBANY(ARGBToARGB1555Row_Any_SSE2, ARGBToARGB1555Row_SSE2, ARGBToARGB1555Row_C, 137 3, 4, 2) 138 RGBANY(ARGBToARGB4444Row_Any_SSE2, ARGBToARGB4444Row_SSE2, ARGBToARGB4444Row_C, 139 3, 4, 2) 140 #endif 141 #if defined(HAS_I400TOARGBROW_SSE2) 142 RGBANY(I400ToARGBRow_Any_SSE2, I400ToARGBRow_Unaligned_SSE2, I400ToARGBRow_C, 143 7, 1, 4) 144 #endif 145 #if defined(HAS_YTOARGBROW_SSE2) 146 RGBANY(YToARGBRow_Any_SSE2, YToARGBRow_SSE2, YToARGBRow_C, 147 7, 1, 4) 148 RGBANY(YUY2ToARGBRow_Any_SSSE3, YUY2ToARGBRow_Unaligned_SSSE3, YUY2ToARGBRow_C, 149 15, 2, 4) 150 RGBANY(UYVYToARGBRow_Any_SSSE3, UYVYToARGBRow_Unaligned_SSSE3, UYVYToARGBRow_C, 151 15, 2, 4) 152 // These require alignment on ARGB, so C is used for remainder. 153 RGBANY(RGB24ToARGBRow_Any_SSSE3, RGB24ToARGBRow_SSSE3, RGB24ToARGBRow_C, 154 15, 3, 4) 155 RGBANY(RAWToARGBRow_Any_SSSE3, RAWToARGBRow_SSSE3, RAWToARGBRow_C, 156 15, 3, 4) 157 RGBANY(RGB565ToARGBRow_Any_SSE2, RGB565ToARGBRow_SSE2, RGB565ToARGBRow_C, 158 7, 2, 4) 159 RGBANY(ARGB1555ToARGBRow_Any_SSE2, ARGB1555ToARGBRow_SSE2, ARGB1555ToARGBRow_C, 160 7, 2, 4) 161 RGBANY(ARGB4444ToARGBRow_Any_SSE2, ARGB4444ToARGBRow_SSE2, ARGB4444ToARGBRow_C, 162 7, 2, 4) 163 #endif 164 #if defined(HAS_ARGBTORGB24ROW_NEON) 165 RGBANY(ARGBToRGB24Row_Any_NEON, ARGBToRGB24Row_NEON, ARGBToRGB24Row_C, 7, 4, 3) 166 RGBANY(ARGBToRAWRow_Any_NEON, ARGBToRAWRow_NEON, ARGBToRAWRow_C, 7, 4, 3) 167 RGBANY(ARGBToRGB565Row_Any_NEON, ARGBToRGB565Row_NEON, ARGBToRGB565Row_C, 168 7, 4, 2) 169 RGBANY(ARGBToARGB1555Row_Any_NEON, ARGBToARGB1555Row_NEON, ARGBToARGB1555Row_C, 170 7, 4, 2) 171 RGBANY(ARGBToARGB4444Row_Any_NEON, ARGBToARGB4444Row_NEON, ARGBToARGB4444Row_C, 172 7, 4, 2) 173 RGBANY(I400ToARGBRow_Any_NEON, I400ToARGBRow_NEON, I400ToARGBRow_C, 174 7, 1, 4) 175 RGBANY(YToARGBRow_Any_NEON, YToARGBRow_NEON, YToARGBRow_C, 176 7, 1, 4) 177 RGBANY(YUY2ToARGBRow_Any_NEON, YUY2ToARGBRow_NEON, YUY2ToARGBRow_C, 178 7, 2, 4) 179 RGBANY(UYVYToARGBRow_Any_NEON, UYVYToARGBRow_NEON, UYVYToARGBRow_C, 180 7, 2, 4) 181 #endif 182 #undef RGBANY 183 184 // ARGB to Bayer does multiple of 4 pixels, SSSE3 aligned src, unaligned dst. 185 #define BAYERANY(NAMEANY, ARGBTORGB_SIMD, ARGBTORGB_C, MASK, SBPP, BPP) \ 186 void NAMEANY(const uint8* src, \ 187 uint8* dst, uint32 selector, \ 188 int width) { \ 189 int n = width & ~MASK; \ 190 ARGBTORGB_SIMD(src, dst, selector, n); \ 191 ARGBTORGB_C(src + n * SBPP, dst + n * BPP, selector, width & MASK); \ 192 } 193 194 #if defined(HAS_ARGBTOBAYERROW_SSSE3) 195 BAYERANY(ARGBToBayerRow_Any_SSSE3, ARGBToBayerRow_SSSE3, ARGBToBayerRow_C, 196 7, 4, 1) 197 #endif 198 #if defined(HAS_ARGBTOBAYERROW_NEON) 199 BAYERANY(ARGBToBayerRow_Any_NEON, ARGBToBayerRow_NEON, ARGBToBayerRow_C, 200 7, 4, 1) 201 #endif 202 #if defined(HAS_ARGBTOBAYERGGROW_SSE2) 203 BAYERANY(ARGBToBayerGGRow_Any_SSE2, ARGBToBayerGGRow_SSE2, ARGBToBayerGGRow_C, 204 7, 4, 1) 205 #endif 206 #if defined(HAS_ARGBTOBAYERGGROW_NEON) 207 BAYERANY(ARGBToBayerGGRow_Any_NEON, ARGBToBayerGGRow_NEON, ARGBToBayerGGRow_C, 208 7, 4, 1) 209 #endif 210 211 #undef BAYERANY 212 213 // RGB/YUV to Y does multiple of 16 with SIMD and last 16 with SIMD. 214 #define YANY(NAMEANY, ARGBTOY_SIMD, SBPP, BPP, NUM) \ 215 void NAMEANY(const uint8* src_argb, uint8* dst_y, int width) { \ 216 ARGBTOY_SIMD(src_argb, dst_y, width - NUM); \ 217 ARGBTOY_SIMD(src_argb + (width - NUM) * SBPP, \ 218 dst_y + (width - NUM) * BPP, NUM); \ 219 } 220 221 #ifdef HAS_ARGBTOYROW_AVX2 222 YANY(ARGBToYRow_Any_AVX2, ARGBToYRow_AVX2, 4, 1, 32) 223 YANY(ARGBToYJRow_Any_AVX2, ARGBToYJRow_AVX2, 4, 1, 32) 224 YANY(YUY2ToYRow_Any_AVX2, YUY2ToYRow_AVX2, 2, 1, 32) 225 YANY(UYVYToYRow_Any_AVX2, UYVYToYRow_AVX2, 2, 1, 32) 226 #endif 227 #ifdef HAS_ARGBTOYROW_SSSE3 228 YANY(ARGBToYRow_Any_SSSE3, ARGBToYRow_Unaligned_SSSE3, 4, 1, 16) 229 #endif 230 #ifdef HAS_BGRATOYROW_SSSE3 231 YANY(BGRAToYRow_Any_SSSE3, BGRAToYRow_Unaligned_SSSE3, 4, 1, 16) 232 YANY(ABGRToYRow_Any_SSSE3, ABGRToYRow_Unaligned_SSSE3, 4, 1, 16) 233 YANY(RGBAToYRow_Any_SSSE3, RGBAToYRow_Unaligned_SSSE3, 4, 1, 16) 234 YANY(YUY2ToYRow_Any_SSE2, YUY2ToYRow_Unaligned_SSE2, 2, 1, 16) 235 YANY(UYVYToYRow_Any_SSE2, UYVYToYRow_Unaligned_SSE2, 2, 1, 16) 236 #endif 237 #ifdef HAS_ARGBTOYJROW_SSSE3 238 YANY(ARGBToYJRow_Any_SSSE3, ARGBToYJRow_Unaligned_SSSE3, 4, 1, 16) 239 #endif 240 #ifdef HAS_ARGBTOYROW_NEON 241 YANY(ARGBToYRow_Any_NEON, ARGBToYRow_NEON, 4, 1, 8) 242 YANY(ARGBToYJRow_Any_NEON, ARGBToYJRow_NEON, 4, 1, 8) 243 YANY(BGRAToYRow_Any_NEON, BGRAToYRow_NEON, 4, 1, 8) 244 YANY(ABGRToYRow_Any_NEON, ABGRToYRow_NEON, 4, 1, 8) 245 YANY(RGBAToYRow_Any_NEON, RGBAToYRow_NEON, 4, 1, 8) 246 YANY(RGB24ToYRow_Any_NEON, RGB24ToYRow_NEON, 3, 1, 8) 247 YANY(RAWToYRow_Any_NEON, RAWToYRow_NEON, 3, 1, 8) 248 YANY(RGB565ToYRow_Any_NEON, RGB565ToYRow_NEON, 2, 1, 8) 249 YANY(ARGB1555ToYRow_Any_NEON, ARGB1555ToYRow_NEON, 2, 1, 8) 250 YANY(ARGB4444ToYRow_Any_NEON, ARGB4444ToYRow_NEON, 2, 1, 8) 251 YANY(YUY2ToYRow_Any_NEON, YUY2ToYRow_NEON, 2, 1, 16) 252 YANY(UYVYToYRow_Any_NEON, UYVYToYRow_NEON, 2, 1, 16) 253 YANY(RGB24ToARGBRow_Any_NEON, RGB24ToARGBRow_NEON, 3, 4, 8) 254 YANY(RAWToARGBRow_Any_NEON, RAWToARGBRow_NEON, 3, 4, 8) 255 YANY(RGB565ToARGBRow_Any_NEON, RGB565ToARGBRow_NEON, 2, 4, 8) 256 YANY(ARGB1555ToARGBRow_Any_NEON, ARGB1555ToARGBRow_NEON, 2, 4, 8) 257 YANY(ARGB4444ToARGBRow_Any_NEON, ARGB4444ToARGBRow_NEON, 2, 4, 8) 258 #endif 259 #undef YANY 260 261 #define YANY(NAMEANY, ARGBTOY_SIMD, ARGBTOY_C, SBPP, BPP, MASK) \ 262 void NAMEANY(const uint8* src_argb, uint8* dst_y, int width) { \ 263 int n = width & ~MASK; \ 264 ARGBTOY_SIMD(src_argb, dst_y, n); \ 265 ARGBTOY_C(src_argb + n * SBPP, \ 266 dst_y + n * BPP, width & MASK); \ 267 } 268 269 // Attenuate is destructive so last16 method can not be used due to overlap. 270 #ifdef HAS_ARGBATTENUATEROW_SSSE3 271 YANY(ARGBAttenuateRow_Any_SSSE3, ARGBAttenuateRow_SSSE3, ARGBAttenuateRow_C, 272 4, 4, 3) 273 #endif 274 #ifdef HAS_ARGBATTENUATEROW_SSE2 275 YANY(ARGBAttenuateRow_Any_SSE2, ARGBAttenuateRow_SSE2, ARGBAttenuateRow_C, 276 4, 4, 3) 277 #endif 278 #ifdef HAS_ARGBUNATTENUATEROW_SSE2 279 YANY(ARGBUnattenuateRow_Any_SSE2, ARGBUnattenuateRow_SSE2, ARGBUnattenuateRow_C, 280 4, 4, 3) 281 #endif 282 #ifdef HAS_ARGBATTENUATEROW_AVX2 283 YANY(ARGBAttenuateRow_Any_AVX2, ARGBAttenuateRow_AVX2, ARGBAttenuateRow_C, 284 4, 4, 7) 285 #endif 286 #ifdef HAS_ARGBUNATTENUATEROW_AVX2 287 YANY(ARGBUnattenuateRow_Any_AVX2, ARGBUnattenuateRow_AVX2, ARGBUnattenuateRow_C, 288 4, 4, 7) 289 #endif 290 #ifdef HAS_ARGBATTENUATEROW_NEON 291 YANY(ARGBAttenuateRow_Any_NEON, ARGBAttenuateRow_NEON, ARGBAttenuateRow_C, 292 4, 4, 7) 293 #endif 294 #undef YANY 295 296 // RGB/YUV to UV does multiple of 16 with SIMD and remainder with C. 297 #define UVANY(NAMEANY, ANYTOUV_SIMD, ANYTOUV_C, BPP, MASK) \ 298 void NAMEANY(const uint8* src_argb, int src_stride_argb, \ 299 uint8* dst_u, uint8* dst_v, int width) { \ 300 int n = width & ~MASK; \ 301 ANYTOUV_SIMD(src_argb, src_stride_argb, dst_u, dst_v, n); \ 302 ANYTOUV_C(src_argb + n * BPP, src_stride_argb, \ 303 dst_u + (n >> 1), \ 304 dst_v + (n >> 1), \ 305 width & MASK); \ 306 } 307 308 #ifdef HAS_ARGBTOUVROW_AVX2 309 UVANY(ARGBToUVRow_Any_AVX2, ARGBToUVRow_AVX2, ARGBToUVRow_C, 4, 31) 310 UVANY(YUY2ToUVRow_Any_AVX2, YUY2ToUVRow_AVX2, YUY2ToUVRow_C, 2, 31) 311 UVANY(UYVYToUVRow_Any_AVX2, UYVYToUVRow_AVX2, UYVYToUVRow_C, 2, 31) 312 #endif 313 #ifdef HAS_ARGBTOUVROW_SSSE3 314 UVANY(ARGBToUVRow_Any_SSSE3, ARGBToUVRow_Unaligned_SSSE3, ARGBToUVRow_C, 4, 15) 315 UVANY(ARGBToUVJRow_Any_SSSE3, ARGBToUVJRow_Unaligned_SSSE3, ARGBToUVJRow_C, 316 4, 15) 317 UVANY(BGRAToUVRow_Any_SSSE3, BGRAToUVRow_Unaligned_SSSE3, BGRAToUVRow_C, 4, 15) 318 UVANY(ABGRToUVRow_Any_SSSE3, ABGRToUVRow_Unaligned_SSSE3, ABGRToUVRow_C, 4, 15) 319 UVANY(RGBAToUVRow_Any_SSSE3, RGBAToUVRow_Unaligned_SSSE3, RGBAToUVRow_C, 4, 15) 320 UVANY(YUY2ToUVRow_Any_SSE2, YUY2ToUVRow_Unaligned_SSE2, YUY2ToUVRow_C, 2, 15) 321 UVANY(UYVYToUVRow_Any_SSE2, UYVYToUVRow_Unaligned_SSE2, UYVYToUVRow_C, 2, 15) 322 #endif 323 #ifdef HAS_ARGBTOUVROW_NEON 324 UVANY(ARGBToUVRow_Any_NEON, ARGBToUVRow_NEON, ARGBToUVRow_C, 4, 15) 325 UVANY(ARGBToUVJRow_Any_NEON, ARGBToUVJRow_NEON, ARGBToUVJRow_C, 4, 15) 326 UVANY(BGRAToUVRow_Any_NEON, BGRAToUVRow_NEON, BGRAToUVRow_C, 4, 15) 327 UVANY(ABGRToUVRow_Any_NEON, ABGRToUVRow_NEON, ABGRToUVRow_C, 4, 15) 328 UVANY(RGBAToUVRow_Any_NEON, RGBAToUVRow_NEON, RGBAToUVRow_C, 4, 15) 329 UVANY(RGB24ToUVRow_Any_NEON, RGB24ToUVRow_NEON, RGB24ToUVRow_C, 3, 15) 330 UVANY(RAWToUVRow_Any_NEON, RAWToUVRow_NEON, RAWToUVRow_C, 3, 15) 331 UVANY(RGB565ToUVRow_Any_NEON, RGB565ToUVRow_NEON, RGB565ToUVRow_C, 2, 15) 332 UVANY(ARGB1555ToUVRow_Any_NEON, ARGB1555ToUVRow_NEON, ARGB1555ToUVRow_C, 2, 15) 333 UVANY(ARGB4444ToUVRow_Any_NEON, ARGB4444ToUVRow_NEON, ARGB4444ToUVRow_C, 2, 15) 334 UVANY(YUY2ToUVRow_Any_NEON, YUY2ToUVRow_NEON, YUY2ToUVRow_C, 2, 15) 335 UVANY(UYVYToUVRow_Any_NEON, UYVYToUVRow_NEON, UYVYToUVRow_C, 2, 15) 336 #endif 337 #undef UVANY 338 339 #define UV422ANY(NAMEANY, ANYTOUV_SIMD, ANYTOUV_C, BPP, MASK, SHIFT) \ 340 void NAMEANY(const uint8* src_uv, \ 341 uint8* dst_u, uint8* dst_v, int width) { \ 342 int n = width & ~MASK; \ 343 ANYTOUV_SIMD(src_uv, dst_u, dst_v, n); \ 344 ANYTOUV_C(src_uv + n * BPP, \ 345 dst_u + (n >> SHIFT), \ 346 dst_v + (n >> SHIFT), \ 347 width & MASK); \ 348 } 349 350 #ifdef HAS_ARGBTOUV444ROW_SSSE3 351 UV422ANY(ARGBToUV444Row_Any_SSSE3, ARGBToUV444Row_Unaligned_SSSE3, 352 ARGBToUV444Row_C, 4, 15, 0) 353 #endif 354 #ifdef HAS_YUY2TOUV422ROW_AVX2 355 UV422ANY(YUY2ToUV422Row_Any_AVX2, YUY2ToUV422Row_AVX2, 356 YUY2ToUV422Row_C, 2, 31, 1) 357 UV422ANY(UYVYToUV422Row_Any_AVX2, UYVYToUV422Row_AVX2, 358 UYVYToUV422Row_C, 2, 31, 1) 359 #endif 360 #ifdef HAS_ARGBTOUVROW_SSSE3 361 UV422ANY(ARGBToUV422Row_Any_SSSE3, ARGBToUV422Row_Unaligned_SSSE3, 362 ARGBToUV422Row_C, 4, 15, 1) 363 UV422ANY(YUY2ToUV422Row_Any_SSE2, YUY2ToUV422Row_Unaligned_SSE2, 364 YUY2ToUV422Row_C, 2, 15, 1) 365 UV422ANY(UYVYToUV422Row_Any_SSE2, UYVYToUV422Row_Unaligned_SSE2, 366 UYVYToUV422Row_C, 2, 15, 1) 367 #endif 368 #ifdef HAS_YUY2TOUV422ROW_NEON 369 UV422ANY(ARGBToUV444Row_Any_NEON, ARGBToUV444Row_NEON, 370 ARGBToUV444Row_C, 4, 7, 0) 371 UV422ANY(ARGBToUV422Row_Any_NEON, ARGBToUV422Row_NEON, 372 ARGBToUV422Row_C, 4, 15, 1) 373 UV422ANY(ARGBToUV411Row_Any_NEON, ARGBToUV411Row_NEON, 374 ARGBToUV411Row_C, 4, 31, 2) 375 UV422ANY(YUY2ToUV422Row_Any_NEON, YUY2ToUV422Row_NEON, 376 YUY2ToUV422Row_C, 2, 15, 1) 377 UV422ANY(UYVYToUV422Row_Any_NEON, UYVYToUV422Row_NEON, 378 UYVYToUV422Row_C, 2, 15, 1) 379 #endif 380 #undef UV422ANY 381 382 #define SPLITUVROWANY(NAMEANY, ANYTOUV_SIMD, ANYTOUV_C, MASK) \ 383 void NAMEANY(const uint8* src_uv, \ 384 uint8* dst_u, uint8* dst_v, int width) { \ 385 int n = width & ~MASK; \ 386 ANYTOUV_SIMD(src_uv, dst_u, dst_v, n); \ 387 ANYTOUV_C(src_uv + n * 2, \ 388 dst_u + n, \ 389 dst_v + n, \ 390 width & MASK); \ 391 } 392 393 #ifdef HAS_SPLITUVROW_SSE2 394 SPLITUVROWANY(SplitUVRow_Any_SSE2, SplitUVRow_Unaligned_SSE2, SplitUVRow_C, 15) 395 #endif 396 #ifdef HAS_SPLITUVROW_AVX2 397 SPLITUVROWANY(SplitUVRow_Any_AVX2, SplitUVRow_AVX2, SplitUVRow_C, 31) 398 #endif 399 #ifdef HAS_SPLITUVROW_NEON 400 SPLITUVROWANY(SplitUVRow_Any_NEON, SplitUVRow_NEON, SplitUVRow_C, 15) 401 #endif 402 #ifdef HAS_SPLITUVROW_MIPS_DSPR2 403 SPLITUVROWANY(SplitUVRow_Any_MIPS_DSPR2, SplitUVRow_Unaligned_MIPS_DSPR2, 404 SplitUVRow_C, 15) 405 #endif 406 #undef SPLITUVROWANY 407 408 #define MERGEUVROW_ANY(NAMEANY, ANYTOUV_SIMD, ANYTOUV_C, MASK) \ 409 void NAMEANY(const uint8* src_u, const uint8* src_v, \ 410 uint8* dst_uv, int width) { \ 411 int n = width & ~MASK; \ 412 ANYTOUV_SIMD(src_u, src_v, dst_uv, n); \ 413 ANYTOUV_C(src_u + n, \ 414 src_v + n, \ 415 dst_uv + n * 2, \ 416 width & MASK); \ 417 } 418 419 #ifdef HAS_MERGEUVROW_SSE2 420 MERGEUVROW_ANY(MergeUVRow_Any_SSE2, MergeUVRow_Unaligned_SSE2, MergeUVRow_C, 15) 421 #endif 422 #ifdef HAS_MERGEUVROW_AVX2 423 MERGEUVROW_ANY(MergeUVRow_Any_AVX2, MergeUVRow_AVX2, MergeUVRow_C, 31) 424 #endif 425 #ifdef HAS_MERGEUVROW_NEON 426 MERGEUVROW_ANY(MergeUVRow_Any_NEON, MergeUVRow_NEON, MergeUVRow_C, 15) 427 #endif 428 #undef MERGEUVROW_ANY 429 430 #define MATHROW_ANY(NAMEANY, ARGBMATH_SIMD, ARGBMATH_C, MASK) \ 431 void NAMEANY(const uint8* src_argb0, const uint8* src_argb1, \ 432 uint8* dst_argb, int width) { \ 433 int n = width & ~MASK; \ 434 ARGBMATH_SIMD(src_argb0, src_argb1, dst_argb, n); \ 435 ARGBMATH_C(src_argb0 + n * 4, \ 436 src_argb1 + n * 4, \ 437 dst_argb + n * 4, \ 438 width & MASK); \ 439 } 440 441 #ifdef HAS_ARGBMULTIPLYROW_SSE2 442 MATHROW_ANY(ARGBMultiplyRow_Any_SSE2, ARGBMultiplyRow_SSE2, ARGBMultiplyRow_C, 443 3) 444 #endif 445 #ifdef HAS_ARGBADDROW_SSE2 446 MATHROW_ANY(ARGBAddRow_Any_SSE2, ARGBAddRow_SSE2, ARGBAddRow_C, 3) 447 #endif 448 #ifdef HAS_ARGBSUBTRACTROW_SSE2 449 MATHROW_ANY(ARGBSubtractRow_Any_SSE2, ARGBSubtractRow_SSE2, ARGBSubtractRow_C, 450 3) 451 #endif 452 #ifdef HAS_ARGBMULTIPLYROW_AVX2 453 MATHROW_ANY(ARGBMultiplyRow_Any_AVX2, ARGBMultiplyRow_AVX2, ARGBMultiplyRow_C, 454 7) 455 #endif 456 #ifdef HAS_ARGBADDROW_AVX2 457 MATHROW_ANY(ARGBAddRow_Any_AVX2, ARGBAddRow_AVX2, ARGBAddRow_C, 7) 458 #endif 459 #ifdef HAS_ARGBSUBTRACTROW_AVX2 460 MATHROW_ANY(ARGBSubtractRow_Any_AVX2, ARGBSubtractRow_AVX2, ARGBSubtractRow_C, 461 7) 462 #endif 463 #ifdef HAS_ARGBMULTIPLYROW_NEON 464 MATHROW_ANY(ARGBMultiplyRow_Any_NEON, ARGBMultiplyRow_NEON, ARGBMultiplyRow_C, 465 7) 466 #endif 467 #ifdef HAS_ARGBADDROW_NEON 468 MATHROW_ANY(ARGBAddRow_Any_NEON, ARGBAddRow_NEON, ARGBAddRow_C, 7) 469 #endif 470 #ifdef HAS_ARGBSUBTRACTROW_NEON 471 MATHROW_ANY(ARGBSubtractRow_Any_NEON, ARGBSubtractRow_NEON, ARGBSubtractRow_C, 472 7) 473 #endif 474 #undef MATHROW_ANY 475 476 // Shuffle may want to work in place, so last16 method can not be used. 477 #define YANY(NAMEANY, ARGBTOY_SIMD, ARGBTOY_C, SBPP, BPP, MASK) \ 478 void NAMEANY(const uint8* src_argb, uint8* dst_argb, \ 479 const uint8* shuffler, int width) { \ 480 int n = width & ~MASK; \ 481 ARGBTOY_SIMD(src_argb, dst_argb, shuffler, n); \ 482 ARGBTOY_C(src_argb + n * SBPP, \ 483 dst_argb + n * BPP, shuffler, width & MASK); \ 484 } 485 486 #ifdef HAS_ARGBSHUFFLEROW_SSE2 487 YANY(ARGBShuffleRow_Any_SSE2, ARGBShuffleRow_SSE2, 488 ARGBShuffleRow_C, 4, 4, 3) 489 #endif 490 #ifdef HAS_ARGBSHUFFLEROW_SSSE3 491 YANY(ARGBShuffleRow_Any_SSSE3, ARGBShuffleRow_Unaligned_SSSE3, 492 ARGBShuffleRow_C, 4, 4, 7) 493 #endif 494 #ifdef HAS_ARGBSHUFFLEROW_AVX2 495 YANY(ARGBShuffleRow_Any_AVX2, ARGBShuffleRow_AVX2, 496 ARGBShuffleRow_C, 4, 4, 15) 497 #endif 498 #ifdef HAS_ARGBSHUFFLEROW_NEON 499 YANY(ARGBShuffleRow_Any_NEON, ARGBShuffleRow_NEON, 500 ARGBShuffleRow_C, 4, 4, 3) 501 #endif 502 #undef YANY 503 504 // Interpolate may want to work in place, so last16 method can not be used. 505 #define NANY(NAMEANY, TERP_SIMD, TERP_C, SBPP, BPP, MASK) \ 506 void NAMEANY(uint8* dst_ptr, const uint8* src_ptr, \ 507 ptrdiff_t src_stride_ptr, int width, \ 508 int source_y_fraction) { \ 509 int n = width & ~MASK; \ 510 TERP_SIMD(dst_ptr, src_ptr, src_stride_ptr, \ 511 n, source_y_fraction); \ 512 TERP_C(dst_ptr + n * BPP, \ 513 src_ptr + n * SBPP, src_stride_ptr, \ 514 width & MASK, source_y_fraction); \ 515 } 516 517 #ifdef HAS_INTERPOLATEROW_AVX2 518 NANY(InterpolateRow_Any_AVX2, InterpolateRow_AVX2, 519 InterpolateRow_C, 1, 1, 32) 520 #endif 521 #ifdef HAS_INTERPOLATEROW_SSSE3 522 NANY(InterpolateRow_Any_SSSE3, InterpolateRow_Unaligned_SSSE3, 523 InterpolateRow_C, 1, 1, 15) 524 #endif 525 #ifdef HAS_INTERPOLATEROW_SSE2 526 NANY(InterpolateRow_Any_SSE2, InterpolateRow_Unaligned_SSE2, 527 InterpolateRow_C, 1, 1, 15) 528 #endif 529 #ifdef HAS_INTERPOLATEROW_NEON 530 NANY(InterpolateRow_Any_NEON, InterpolateRow_NEON, 531 InterpolateRow_C, 1, 1, 15) 532 #endif 533 #ifdef HAS_INTERPOLATEROW_MIPS_DSPR2 534 NANY(InterpolateRow_Any_MIPS_DSPR2, InterpolateRow_MIPS_DSPR2, 535 InterpolateRow_C, 1, 1, 3) 536 #endif 537 #undef NANY 538 539 #ifdef __cplusplus 540 } // extern "C" 541 } // namespace libyuv 542 #endif 543
