1.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000
1.2 +++ b/media/libyuv/source/row_any.cc Wed Dec 31 06:09:35 2014 +0100
1.3 @@ -0,0 +1,542 @@
1.4 +/*
1.5 + * Copyright 2012 The LibYuv Project Authors. All rights reserved.
1.6 + *
1.7 + * Use of this source code is governed by a BSD-style license
1.8 + * that can be found in the LICENSE file in the root of the source
1.9 + * tree. An additional intellectual property rights grant can be found
1.10 + * in the file PATENTS. All contributing project authors may
1.11 + * be found in the AUTHORS file in the root of the source tree.
1.12 + */
1.13 +
1.14 +#include "libyuv/row.h"
1.15 +
1.16 +#include "libyuv/basic_types.h"
1.17 +
1.18 +#ifdef __cplusplus
1.19 +namespace libyuv {
1.20 +extern "C" {
1.21 +#endif
1.22 +
1.23 +// TODO(fbarchard): Consider 'any' functions handling any quantity of pixels.
1.24 +// TODO(fbarchard): Consider 'any' functions handling odd alignment.
1.25 +// YUV to RGB does multiple of 8 with SIMD and remainder with C.
1.26 +#define YANY(NAMEANY, I420TORGB_SIMD, I420TORGB_C, UV_SHIFT, BPP, MASK) \
1.27 + void NAMEANY(const uint8* y_buf, \
1.28 + const uint8* u_buf, \
1.29 + const uint8* v_buf, \
1.30 + uint8* rgb_buf, \
1.31 + int width) { \
1.32 + int n = width & ~MASK; \
1.33 + I420TORGB_SIMD(y_buf, u_buf, v_buf, rgb_buf, n); \
1.34 + I420TORGB_C(y_buf + n, \
1.35 + u_buf + (n >> UV_SHIFT), \
1.36 + v_buf + (n >> UV_SHIFT), \
1.37 + rgb_buf + n * BPP, width & MASK); \
1.38 + }
1.39 +
1.40 +#ifdef HAS_I422TOARGBROW_SSSE3
1.41 +YANY(I444ToARGBRow_Any_SSSE3, I444ToARGBRow_Unaligned_SSSE3, I444ToARGBRow_C,
1.42 + 0, 4, 7)
1.43 +YANY(I422ToARGBRow_Any_SSSE3, I422ToARGBRow_Unaligned_SSSE3, I422ToARGBRow_C,
1.44 + 1, 4, 7)
1.45 +YANY(I411ToARGBRow_Any_SSSE3, I411ToARGBRow_Unaligned_SSSE3, I411ToARGBRow_C,
1.46 + 2, 4, 7)
1.47 +YANY(I422ToBGRARow_Any_SSSE3, I422ToBGRARow_Unaligned_SSSE3, I422ToBGRARow_C,
1.48 + 1, 4, 7)
1.49 +YANY(I422ToABGRRow_Any_SSSE3, I422ToABGRRow_Unaligned_SSSE3, I422ToABGRRow_C,
1.50 + 1, 4, 7)
1.51 +YANY(I422ToRGBARow_Any_SSSE3, I422ToRGBARow_Unaligned_SSSE3, I422ToRGBARow_C,
1.52 + 1, 4, 7)
1.53 +// I422ToRGB565Row_SSSE3 is unaligned.
1.54 +YANY(I422ToARGB4444Row_Any_SSSE3, I422ToARGB4444Row_SSSE3, I422ToARGB4444Row_C,
1.55 + 1, 2, 7)
1.56 +YANY(I422ToARGB1555Row_Any_SSSE3, I422ToARGB1555Row_SSSE3, I422ToARGB1555Row_C,
1.57 + 1, 2, 7)
1.58 +YANY(I422ToRGB565Row_Any_SSSE3, I422ToRGB565Row_SSSE3, I422ToRGB565Row_C,
1.59 + 1, 2, 7)
1.60 +// I422ToRGB24Row_SSSE3 is unaligned.
1.61 +YANY(I422ToRGB24Row_Any_SSSE3, I422ToRGB24Row_SSSE3, I422ToRGB24Row_C, 1, 3, 7)
1.62 +YANY(I422ToRAWRow_Any_SSSE3, I422ToRAWRow_SSSE3, I422ToRAWRow_C, 1, 3, 7)
1.63 +YANY(I422ToYUY2Row_Any_SSE2, I422ToYUY2Row_SSE2, I422ToYUY2Row_C, 1, 2, 15)
1.64 +YANY(I422ToUYVYRow_Any_SSE2, I422ToUYVYRow_SSE2, I422ToUYVYRow_C, 1, 2, 15)
1.65 +#endif // HAS_I422TOARGBROW_SSSE3
1.66 +#ifdef HAS_I422TOARGBROW_AVX2
1.67 +YANY(I422ToARGBRow_Any_AVX2, I422ToARGBRow_AVX2, I422ToARGBRow_C, 1, 4, 15)
1.68 +#endif // HAS_I422TOARGBROW_AVX2
1.69 +#ifdef HAS_I422TOARGBROW_NEON
1.70 +YANY(I444ToARGBRow_Any_NEON, I444ToARGBRow_NEON, I444ToARGBRow_C, 0, 4, 7)
1.71 +YANY(I422ToARGBRow_Any_NEON, I422ToARGBRow_NEON, I422ToARGBRow_C, 1, 4, 7)
1.72 +YANY(I411ToARGBRow_Any_NEON, I411ToARGBRow_NEON, I411ToARGBRow_C, 2, 4, 7)
1.73 +YANY(I422ToBGRARow_Any_NEON, I422ToBGRARow_NEON, I422ToBGRARow_C, 1, 4, 7)
1.74 +YANY(I422ToABGRRow_Any_NEON, I422ToABGRRow_NEON, I422ToABGRRow_C, 1, 4, 7)
1.75 +YANY(I422ToRGBARow_Any_NEON, I422ToRGBARow_NEON, I422ToRGBARow_C, 1, 4, 7)
1.76 +YANY(I422ToRGB24Row_Any_NEON, I422ToRGB24Row_NEON, I422ToRGB24Row_C, 1, 3, 7)
1.77 +YANY(I422ToRAWRow_Any_NEON, I422ToRAWRow_NEON, I422ToRAWRow_C, 1, 3, 7)
1.78 +YANY(I422ToARGB4444Row_Any_NEON, I422ToARGB4444Row_NEON, I422ToARGB4444Row_C,
1.79 + 1, 2, 7)
1.80 +YANY(I422ToARGB1555Row_Any_NEON, I422ToARGB1555Row_NEON, I422ToARGB1555Row_C,
1.81 + 1, 2, 7)
1.82 +YANY(I422ToRGB565Row_Any_NEON, I422ToRGB565Row_NEON, I422ToRGB565Row_C, 1, 2, 7)
1.83 +YANY(I422ToYUY2Row_Any_NEON, I422ToYUY2Row_NEON, I422ToYUY2Row_C, 1, 2, 15)
1.84 +YANY(I422ToUYVYRow_Any_NEON, I422ToUYVYRow_NEON, I422ToUYVYRow_C, 1, 2, 15)
1.85 +#endif // HAS_I422TOARGBROW_NEON
1.86 +#undef YANY
1.87 +
1.88 +// Wrappers to handle odd width
1.89 +#define NV2NY(NAMEANY, NV12TORGB_SIMD, NV12TORGB_C, UV_SHIFT, BPP) \
1.90 + void NAMEANY(const uint8* y_buf, \
1.91 + const uint8* uv_buf, \
1.92 + uint8* rgb_buf, \
1.93 + int width) { \
1.94 + int n = width & ~7; \
1.95 + NV12TORGB_SIMD(y_buf, uv_buf, rgb_buf, n); \
1.96 + NV12TORGB_C(y_buf + n, \
1.97 + uv_buf + (n >> UV_SHIFT), \
1.98 + rgb_buf + n * BPP, width & 7); \
1.99 + }
1.100 +
1.101 +#ifdef HAS_NV12TOARGBROW_SSSE3
1.102 +NV2NY(NV12ToARGBRow_Any_SSSE3, NV12ToARGBRow_Unaligned_SSSE3, NV12ToARGBRow_C,
1.103 + 0, 4)
1.104 +NV2NY(NV21ToARGBRow_Any_SSSE3, NV21ToARGBRow_Unaligned_SSSE3, NV21ToARGBRow_C,
1.105 + 0, 4)
1.106 +#endif // HAS_NV12TOARGBROW_SSSE3
1.107 +#ifdef HAS_NV12TOARGBROW_NEON
1.108 +NV2NY(NV12ToARGBRow_Any_NEON, NV12ToARGBRow_NEON, NV12ToARGBRow_C, 0, 4)
1.109 +NV2NY(NV21ToARGBRow_Any_NEON, NV21ToARGBRow_NEON, NV21ToARGBRow_C, 0, 4)
1.110 +#endif // HAS_NV12TOARGBROW_NEON
1.111 +#ifdef HAS_NV12TORGB565ROW_SSSE3
1.112 +NV2NY(NV12ToRGB565Row_Any_SSSE3, NV12ToRGB565Row_SSSE3, NV12ToRGB565Row_C,
1.113 + 0, 2)
1.114 +NV2NY(NV21ToRGB565Row_Any_SSSE3, NV21ToRGB565Row_SSSE3, NV21ToRGB565Row_C,
1.115 + 0, 2)
1.116 +#endif // HAS_NV12TORGB565ROW_SSSE3
1.117 +#ifdef HAS_NV12TORGB565ROW_NEON
1.118 +NV2NY(NV12ToRGB565Row_Any_NEON, NV12ToRGB565Row_NEON, NV12ToRGB565Row_C, 0, 2)
1.119 +NV2NY(NV21ToRGB565Row_Any_NEON, NV21ToRGB565Row_NEON, NV21ToRGB565Row_C, 0, 2)
1.120 +#endif // HAS_NV12TORGB565ROW_NEON
1.121 +#undef NVANY
1.122 +
1.123 +#define RGBANY(NAMEANY, ARGBTORGB_SIMD, ARGBTORGB_C, MASK, SBPP, BPP) \
1.124 + void NAMEANY(const uint8* src, \
1.125 + uint8* dst, \
1.126 + int width) { \
1.127 + int n = width & ~MASK; \
1.128 + ARGBTORGB_SIMD(src, dst, n); \
1.129 + ARGBTORGB_C(src + n * SBPP, dst + n * BPP, width & MASK); \
1.130 + }
1.131 +
1.132 +#if defined(HAS_ARGBTORGB24ROW_SSSE3)
1.133 +RGBANY(ARGBToRGB24Row_Any_SSSE3, ARGBToRGB24Row_SSSE3, ARGBToRGB24Row_C,
1.134 + 15, 4, 3)
1.135 +RGBANY(ARGBToRAWRow_Any_SSSE3, ARGBToRAWRow_SSSE3, ARGBToRAWRow_C,
1.136 + 15, 4, 3)
1.137 +RGBANY(ARGBToRGB565Row_Any_SSE2, ARGBToRGB565Row_SSE2, ARGBToRGB565Row_C,
1.138 + 3, 4, 2)
1.139 +RGBANY(ARGBToARGB1555Row_Any_SSE2, ARGBToARGB1555Row_SSE2, ARGBToARGB1555Row_C,
1.140 + 3, 4, 2)
1.141 +RGBANY(ARGBToARGB4444Row_Any_SSE2, ARGBToARGB4444Row_SSE2, ARGBToARGB4444Row_C,
1.142 + 3, 4, 2)
1.143 +#endif
1.144 +#if defined(HAS_I400TOARGBROW_SSE2)
1.145 +RGBANY(I400ToARGBRow_Any_SSE2, I400ToARGBRow_Unaligned_SSE2, I400ToARGBRow_C,
1.146 + 7, 1, 4)
1.147 +#endif
1.148 +#if defined(HAS_YTOARGBROW_SSE2)
1.149 +RGBANY(YToARGBRow_Any_SSE2, YToARGBRow_SSE2, YToARGBRow_C,
1.150 + 7, 1, 4)
1.151 +RGBANY(YUY2ToARGBRow_Any_SSSE3, YUY2ToARGBRow_Unaligned_SSSE3, YUY2ToARGBRow_C,
1.152 + 15, 2, 4)
1.153 +RGBANY(UYVYToARGBRow_Any_SSSE3, UYVYToARGBRow_Unaligned_SSSE3, UYVYToARGBRow_C,
1.154 + 15, 2, 4)
1.155 +// These require alignment on ARGB, so C is used for remainder.
1.156 +RGBANY(RGB24ToARGBRow_Any_SSSE3, RGB24ToARGBRow_SSSE3, RGB24ToARGBRow_C,
1.157 + 15, 3, 4)
1.158 +RGBANY(RAWToARGBRow_Any_SSSE3, RAWToARGBRow_SSSE3, RAWToARGBRow_C,
1.159 + 15, 3, 4)
1.160 +RGBANY(RGB565ToARGBRow_Any_SSE2, RGB565ToARGBRow_SSE2, RGB565ToARGBRow_C,
1.161 + 7, 2, 4)
1.162 +RGBANY(ARGB1555ToARGBRow_Any_SSE2, ARGB1555ToARGBRow_SSE2, ARGB1555ToARGBRow_C,
1.163 + 7, 2, 4)
1.164 +RGBANY(ARGB4444ToARGBRow_Any_SSE2, ARGB4444ToARGBRow_SSE2, ARGB4444ToARGBRow_C,
1.165 + 7, 2, 4)
1.166 +#endif
1.167 +#if defined(HAS_ARGBTORGB24ROW_NEON)
1.168 +RGBANY(ARGBToRGB24Row_Any_NEON, ARGBToRGB24Row_NEON, ARGBToRGB24Row_C, 7, 4, 3)
1.169 +RGBANY(ARGBToRAWRow_Any_NEON, ARGBToRAWRow_NEON, ARGBToRAWRow_C, 7, 4, 3)
1.170 +RGBANY(ARGBToRGB565Row_Any_NEON, ARGBToRGB565Row_NEON, ARGBToRGB565Row_C,
1.171 + 7, 4, 2)
1.172 +RGBANY(ARGBToARGB1555Row_Any_NEON, ARGBToARGB1555Row_NEON, ARGBToARGB1555Row_C,
1.173 + 7, 4, 2)
1.174 +RGBANY(ARGBToARGB4444Row_Any_NEON, ARGBToARGB4444Row_NEON, ARGBToARGB4444Row_C,
1.175 + 7, 4, 2)
1.176 +RGBANY(I400ToARGBRow_Any_NEON, I400ToARGBRow_NEON, I400ToARGBRow_C,
1.177 + 7, 1, 4)
1.178 +RGBANY(YToARGBRow_Any_NEON, YToARGBRow_NEON, YToARGBRow_C,
1.179 + 7, 1, 4)
1.180 +RGBANY(YUY2ToARGBRow_Any_NEON, YUY2ToARGBRow_NEON, YUY2ToARGBRow_C,
1.181 + 7, 2, 4)
1.182 +RGBANY(UYVYToARGBRow_Any_NEON, UYVYToARGBRow_NEON, UYVYToARGBRow_C,
1.183 + 7, 2, 4)
1.184 +#endif
1.185 +#undef RGBANY
1.186 +
1.187 +// ARGB to Bayer does multiple of 4 pixels, SSSE3 aligned src, unaligned dst.
1.188 +#define BAYERANY(NAMEANY, ARGBTORGB_SIMD, ARGBTORGB_C, MASK, SBPP, BPP) \
1.189 + void NAMEANY(const uint8* src, \
1.190 + uint8* dst, uint32 selector, \
1.191 + int width) { \
1.192 + int n = width & ~MASK; \
1.193 + ARGBTORGB_SIMD(src, dst, selector, n); \
1.194 + ARGBTORGB_C(src + n * SBPP, dst + n * BPP, selector, width & MASK); \
1.195 + }
1.196 +
1.197 +#if defined(HAS_ARGBTOBAYERROW_SSSE3)
1.198 +BAYERANY(ARGBToBayerRow_Any_SSSE3, ARGBToBayerRow_SSSE3, ARGBToBayerRow_C,
1.199 + 7, 4, 1)
1.200 +#endif
1.201 +#if defined(HAS_ARGBTOBAYERROW_NEON)
1.202 +BAYERANY(ARGBToBayerRow_Any_NEON, ARGBToBayerRow_NEON, ARGBToBayerRow_C,
1.203 + 7, 4, 1)
1.204 +#endif
1.205 +#if defined(HAS_ARGBTOBAYERGGROW_SSE2)
1.206 +BAYERANY(ARGBToBayerGGRow_Any_SSE2, ARGBToBayerGGRow_SSE2, ARGBToBayerGGRow_C,
1.207 + 7, 4, 1)
1.208 +#endif
1.209 +#if defined(HAS_ARGBTOBAYERGGROW_NEON)
1.210 +BAYERANY(ARGBToBayerGGRow_Any_NEON, ARGBToBayerGGRow_NEON, ARGBToBayerGGRow_C,
1.211 + 7, 4, 1)
1.212 +#endif
1.213 +
1.214 +#undef BAYERANY
1.215 +
1.216 +// RGB/YUV to Y does multiple of 16 with SIMD and last 16 with SIMD.
1.217 +#define YANY(NAMEANY, ARGBTOY_SIMD, SBPP, BPP, NUM) \
1.218 + void NAMEANY(const uint8* src_argb, uint8* dst_y, int width) { \
1.219 + ARGBTOY_SIMD(src_argb, dst_y, width - NUM); \
1.220 + ARGBTOY_SIMD(src_argb + (width - NUM) * SBPP, \
1.221 + dst_y + (width - NUM) * BPP, NUM); \
1.222 + }
1.223 +
1.224 +#ifdef HAS_ARGBTOYROW_AVX2
1.225 +YANY(ARGBToYRow_Any_AVX2, ARGBToYRow_AVX2, 4, 1, 32)
1.226 +YANY(ARGBToYJRow_Any_AVX2, ARGBToYJRow_AVX2, 4, 1, 32)
1.227 +YANY(YUY2ToYRow_Any_AVX2, YUY2ToYRow_AVX2, 2, 1, 32)
1.228 +YANY(UYVYToYRow_Any_AVX2, UYVYToYRow_AVX2, 2, 1, 32)
1.229 +#endif
1.230 +#ifdef HAS_ARGBTOYROW_SSSE3
1.231 +YANY(ARGBToYRow_Any_SSSE3, ARGBToYRow_Unaligned_SSSE3, 4, 1, 16)
1.232 +#endif
1.233 +#ifdef HAS_BGRATOYROW_SSSE3
1.234 +YANY(BGRAToYRow_Any_SSSE3, BGRAToYRow_Unaligned_SSSE3, 4, 1, 16)
1.235 +YANY(ABGRToYRow_Any_SSSE3, ABGRToYRow_Unaligned_SSSE3, 4, 1, 16)
1.236 +YANY(RGBAToYRow_Any_SSSE3, RGBAToYRow_Unaligned_SSSE3, 4, 1, 16)
1.237 +YANY(YUY2ToYRow_Any_SSE2, YUY2ToYRow_Unaligned_SSE2, 2, 1, 16)
1.238 +YANY(UYVYToYRow_Any_SSE2, UYVYToYRow_Unaligned_SSE2, 2, 1, 16)
1.239 +#endif
1.240 +#ifdef HAS_ARGBTOYJROW_SSSE3
1.241 +YANY(ARGBToYJRow_Any_SSSE3, ARGBToYJRow_Unaligned_SSSE3, 4, 1, 16)
1.242 +#endif
1.243 +#ifdef HAS_ARGBTOYROW_NEON
1.244 +YANY(ARGBToYRow_Any_NEON, ARGBToYRow_NEON, 4, 1, 8)
1.245 +YANY(ARGBToYJRow_Any_NEON, ARGBToYJRow_NEON, 4, 1, 8)
1.246 +YANY(BGRAToYRow_Any_NEON, BGRAToYRow_NEON, 4, 1, 8)
1.247 +YANY(ABGRToYRow_Any_NEON, ABGRToYRow_NEON, 4, 1, 8)
1.248 +YANY(RGBAToYRow_Any_NEON, RGBAToYRow_NEON, 4, 1, 8)
1.249 +YANY(RGB24ToYRow_Any_NEON, RGB24ToYRow_NEON, 3, 1, 8)
1.250 +YANY(RAWToYRow_Any_NEON, RAWToYRow_NEON, 3, 1, 8)
1.251 +YANY(RGB565ToYRow_Any_NEON, RGB565ToYRow_NEON, 2, 1, 8)
1.252 +YANY(ARGB1555ToYRow_Any_NEON, ARGB1555ToYRow_NEON, 2, 1, 8)
1.253 +YANY(ARGB4444ToYRow_Any_NEON, ARGB4444ToYRow_NEON, 2, 1, 8)
1.254 +YANY(YUY2ToYRow_Any_NEON, YUY2ToYRow_NEON, 2, 1, 16)
1.255 +YANY(UYVYToYRow_Any_NEON, UYVYToYRow_NEON, 2, 1, 16)
1.256 +YANY(RGB24ToARGBRow_Any_NEON, RGB24ToARGBRow_NEON, 3, 4, 8)
1.257 +YANY(RAWToARGBRow_Any_NEON, RAWToARGBRow_NEON, 3, 4, 8)
1.258 +YANY(RGB565ToARGBRow_Any_NEON, RGB565ToARGBRow_NEON, 2, 4, 8)
1.259 +YANY(ARGB1555ToARGBRow_Any_NEON, ARGB1555ToARGBRow_NEON, 2, 4, 8)
1.260 +YANY(ARGB4444ToARGBRow_Any_NEON, ARGB4444ToARGBRow_NEON, 2, 4, 8)
1.261 +#endif
1.262 +#undef YANY
1.263 +
1.264 +#define YANY(NAMEANY, ARGBTOY_SIMD, ARGBTOY_C, SBPP, BPP, MASK) \
1.265 + void NAMEANY(const uint8* src_argb, uint8* dst_y, int width) { \
1.266 + int n = width & ~MASK; \
1.267 + ARGBTOY_SIMD(src_argb, dst_y, n); \
1.268 + ARGBTOY_C(src_argb + n * SBPP, \
1.269 + dst_y + n * BPP, width & MASK); \
1.270 + }
1.271 +
1.272 +// Attenuate is destructive so last16 method can not be used due to overlap.
1.273 +#ifdef HAS_ARGBATTENUATEROW_SSSE3
1.274 +YANY(ARGBAttenuateRow_Any_SSSE3, ARGBAttenuateRow_SSSE3, ARGBAttenuateRow_C,
1.275 + 4, 4, 3)
1.276 +#endif
1.277 +#ifdef HAS_ARGBATTENUATEROW_SSE2
1.278 +YANY(ARGBAttenuateRow_Any_SSE2, ARGBAttenuateRow_SSE2, ARGBAttenuateRow_C,
1.279 + 4, 4, 3)
1.280 +#endif
1.281 +#ifdef HAS_ARGBUNATTENUATEROW_SSE2
1.282 +YANY(ARGBUnattenuateRow_Any_SSE2, ARGBUnattenuateRow_SSE2, ARGBUnattenuateRow_C,
1.283 + 4, 4, 3)
1.284 +#endif
1.285 +#ifdef HAS_ARGBATTENUATEROW_AVX2
1.286 +YANY(ARGBAttenuateRow_Any_AVX2, ARGBAttenuateRow_AVX2, ARGBAttenuateRow_C,
1.287 + 4, 4, 7)
1.288 +#endif
1.289 +#ifdef HAS_ARGBUNATTENUATEROW_AVX2
1.290 +YANY(ARGBUnattenuateRow_Any_AVX2, ARGBUnattenuateRow_AVX2, ARGBUnattenuateRow_C,
1.291 + 4, 4, 7)
1.292 +#endif
1.293 +#ifdef HAS_ARGBATTENUATEROW_NEON
1.294 +YANY(ARGBAttenuateRow_Any_NEON, ARGBAttenuateRow_NEON, ARGBAttenuateRow_C,
1.295 + 4, 4, 7)
1.296 +#endif
1.297 +#undef YANY
1.298 +
1.299 +// RGB/YUV to UV does multiple of 16 with SIMD and remainder with C.
1.300 +#define UVANY(NAMEANY, ANYTOUV_SIMD, ANYTOUV_C, BPP, MASK) \
1.301 + void NAMEANY(const uint8* src_argb, int src_stride_argb, \
1.302 + uint8* dst_u, uint8* dst_v, int width) { \
1.303 + int n = width & ~MASK; \
1.304 + ANYTOUV_SIMD(src_argb, src_stride_argb, dst_u, dst_v, n); \
1.305 + ANYTOUV_C(src_argb + n * BPP, src_stride_argb, \
1.306 + dst_u + (n >> 1), \
1.307 + dst_v + (n >> 1), \
1.308 + width & MASK); \
1.309 + }
1.310 +
1.311 +#ifdef HAS_ARGBTOUVROW_AVX2
1.312 +UVANY(ARGBToUVRow_Any_AVX2, ARGBToUVRow_AVX2, ARGBToUVRow_C, 4, 31)
1.313 +UVANY(YUY2ToUVRow_Any_AVX2, YUY2ToUVRow_AVX2, YUY2ToUVRow_C, 2, 31)
1.314 +UVANY(UYVYToUVRow_Any_AVX2, UYVYToUVRow_AVX2, UYVYToUVRow_C, 2, 31)
1.315 +#endif
1.316 +#ifdef HAS_ARGBTOUVROW_SSSE3
1.317 +UVANY(ARGBToUVRow_Any_SSSE3, ARGBToUVRow_Unaligned_SSSE3, ARGBToUVRow_C, 4, 15)
1.318 +UVANY(ARGBToUVJRow_Any_SSSE3, ARGBToUVJRow_Unaligned_SSSE3, ARGBToUVJRow_C,
1.319 + 4, 15)
1.320 +UVANY(BGRAToUVRow_Any_SSSE3, BGRAToUVRow_Unaligned_SSSE3, BGRAToUVRow_C, 4, 15)
1.321 +UVANY(ABGRToUVRow_Any_SSSE3, ABGRToUVRow_Unaligned_SSSE3, ABGRToUVRow_C, 4, 15)
1.322 +UVANY(RGBAToUVRow_Any_SSSE3, RGBAToUVRow_Unaligned_SSSE3, RGBAToUVRow_C, 4, 15)
1.323 +UVANY(YUY2ToUVRow_Any_SSE2, YUY2ToUVRow_Unaligned_SSE2, YUY2ToUVRow_C, 2, 15)
1.324 +UVANY(UYVYToUVRow_Any_SSE2, UYVYToUVRow_Unaligned_SSE2, UYVYToUVRow_C, 2, 15)
1.325 +#endif
1.326 +#ifdef HAS_ARGBTOUVROW_NEON
1.327 +UVANY(ARGBToUVRow_Any_NEON, ARGBToUVRow_NEON, ARGBToUVRow_C, 4, 15)
1.328 +UVANY(ARGBToUVJRow_Any_NEON, ARGBToUVJRow_NEON, ARGBToUVJRow_C, 4, 15)
1.329 +UVANY(BGRAToUVRow_Any_NEON, BGRAToUVRow_NEON, BGRAToUVRow_C, 4, 15)
1.330 +UVANY(ABGRToUVRow_Any_NEON, ABGRToUVRow_NEON, ABGRToUVRow_C, 4, 15)
1.331 +UVANY(RGBAToUVRow_Any_NEON, RGBAToUVRow_NEON, RGBAToUVRow_C, 4, 15)
1.332 +UVANY(RGB24ToUVRow_Any_NEON, RGB24ToUVRow_NEON, RGB24ToUVRow_C, 3, 15)
1.333 +UVANY(RAWToUVRow_Any_NEON, RAWToUVRow_NEON, RAWToUVRow_C, 3, 15)
1.334 +UVANY(RGB565ToUVRow_Any_NEON, RGB565ToUVRow_NEON, RGB565ToUVRow_C, 2, 15)
1.335 +UVANY(ARGB1555ToUVRow_Any_NEON, ARGB1555ToUVRow_NEON, ARGB1555ToUVRow_C, 2, 15)
1.336 +UVANY(ARGB4444ToUVRow_Any_NEON, ARGB4444ToUVRow_NEON, ARGB4444ToUVRow_C, 2, 15)
1.337 +UVANY(YUY2ToUVRow_Any_NEON, YUY2ToUVRow_NEON, YUY2ToUVRow_C, 2, 15)
1.338 +UVANY(UYVYToUVRow_Any_NEON, UYVYToUVRow_NEON, UYVYToUVRow_C, 2, 15)
1.339 +#endif
1.340 +#undef UVANY
1.341 +
1.342 +#define UV422ANY(NAMEANY, ANYTOUV_SIMD, ANYTOUV_C, BPP, MASK, SHIFT) \
1.343 + void NAMEANY(const uint8* src_uv, \
1.344 + uint8* dst_u, uint8* dst_v, int width) { \
1.345 + int n = width & ~MASK; \
1.346 + ANYTOUV_SIMD(src_uv, dst_u, dst_v, n); \
1.347 + ANYTOUV_C(src_uv + n * BPP, \
1.348 + dst_u + (n >> SHIFT), \
1.349 + dst_v + (n >> SHIFT), \
1.350 + width & MASK); \
1.351 + }
1.352 +
1.353 +#ifdef HAS_ARGBTOUV444ROW_SSSE3
1.354 +UV422ANY(ARGBToUV444Row_Any_SSSE3, ARGBToUV444Row_Unaligned_SSSE3,
1.355 + ARGBToUV444Row_C, 4, 15, 0)
1.356 +#endif
1.357 +#ifdef HAS_YUY2TOUV422ROW_AVX2
1.358 +UV422ANY(YUY2ToUV422Row_Any_AVX2, YUY2ToUV422Row_AVX2,
1.359 + YUY2ToUV422Row_C, 2, 31, 1)
1.360 +UV422ANY(UYVYToUV422Row_Any_AVX2, UYVYToUV422Row_AVX2,
1.361 + UYVYToUV422Row_C, 2, 31, 1)
1.362 +#endif
1.363 +#ifdef HAS_ARGBTOUVROW_SSSE3
1.364 +UV422ANY(ARGBToUV422Row_Any_SSSE3, ARGBToUV422Row_Unaligned_SSSE3,
1.365 + ARGBToUV422Row_C, 4, 15, 1)
1.366 +UV422ANY(YUY2ToUV422Row_Any_SSE2, YUY2ToUV422Row_Unaligned_SSE2,
1.367 + YUY2ToUV422Row_C, 2, 15, 1)
1.368 +UV422ANY(UYVYToUV422Row_Any_SSE2, UYVYToUV422Row_Unaligned_SSE2,
1.369 + UYVYToUV422Row_C, 2, 15, 1)
1.370 +#endif
1.371 +#ifdef HAS_YUY2TOUV422ROW_NEON
1.372 +UV422ANY(ARGBToUV444Row_Any_NEON, ARGBToUV444Row_NEON,
1.373 + ARGBToUV444Row_C, 4, 7, 0)
1.374 +UV422ANY(ARGBToUV422Row_Any_NEON, ARGBToUV422Row_NEON,
1.375 + ARGBToUV422Row_C, 4, 15, 1)
1.376 +UV422ANY(ARGBToUV411Row_Any_NEON, ARGBToUV411Row_NEON,
1.377 + ARGBToUV411Row_C, 4, 31, 2)
1.378 +UV422ANY(YUY2ToUV422Row_Any_NEON, YUY2ToUV422Row_NEON,
1.379 + YUY2ToUV422Row_C, 2, 15, 1)
1.380 +UV422ANY(UYVYToUV422Row_Any_NEON, UYVYToUV422Row_NEON,
1.381 + UYVYToUV422Row_C, 2, 15, 1)
1.382 +#endif
1.383 +#undef UV422ANY
1.384 +
1.385 +#define SPLITUVROWANY(NAMEANY, ANYTOUV_SIMD, ANYTOUV_C, MASK) \
1.386 + void NAMEANY(const uint8* src_uv, \
1.387 + uint8* dst_u, uint8* dst_v, int width) { \
1.388 + int n = width & ~MASK; \
1.389 + ANYTOUV_SIMD(src_uv, dst_u, dst_v, n); \
1.390 + ANYTOUV_C(src_uv + n * 2, \
1.391 + dst_u + n, \
1.392 + dst_v + n, \
1.393 + width & MASK); \
1.394 + }
1.395 +
1.396 +#ifdef HAS_SPLITUVROW_SSE2
1.397 +SPLITUVROWANY(SplitUVRow_Any_SSE2, SplitUVRow_Unaligned_SSE2, SplitUVRow_C, 15)
1.398 +#endif
1.399 +#ifdef HAS_SPLITUVROW_AVX2
1.400 +SPLITUVROWANY(SplitUVRow_Any_AVX2, SplitUVRow_AVX2, SplitUVRow_C, 31)
1.401 +#endif
1.402 +#ifdef HAS_SPLITUVROW_NEON
1.403 +SPLITUVROWANY(SplitUVRow_Any_NEON, SplitUVRow_NEON, SplitUVRow_C, 15)
1.404 +#endif
1.405 +#ifdef HAS_SPLITUVROW_MIPS_DSPR2
1.406 +SPLITUVROWANY(SplitUVRow_Any_MIPS_DSPR2, SplitUVRow_Unaligned_MIPS_DSPR2,
1.407 + SplitUVRow_C, 15)
1.408 +#endif
1.409 +#undef SPLITUVROWANY
1.410 +
1.411 +#define MERGEUVROW_ANY(NAMEANY, ANYTOUV_SIMD, ANYTOUV_C, MASK) \
1.412 + void NAMEANY(const uint8* src_u, const uint8* src_v, \
1.413 + uint8* dst_uv, int width) { \
1.414 + int n = width & ~MASK; \
1.415 + ANYTOUV_SIMD(src_u, src_v, dst_uv, n); \
1.416 + ANYTOUV_C(src_u + n, \
1.417 + src_v + n, \
1.418 + dst_uv + n * 2, \
1.419 + width & MASK); \
1.420 + }
1.421 +
1.422 +#ifdef HAS_MERGEUVROW_SSE2
1.423 +MERGEUVROW_ANY(MergeUVRow_Any_SSE2, MergeUVRow_Unaligned_SSE2, MergeUVRow_C, 15)
1.424 +#endif
1.425 +#ifdef HAS_MERGEUVROW_AVX2
1.426 +MERGEUVROW_ANY(MergeUVRow_Any_AVX2, MergeUVRow_AVX2, MergeUVRow_C, 31)
1.427 +#endif
1.428 +#ifdef HAS_MERGEUVROW_NEON
1.429 +MERGEUVROW_ANY(MergeUVRow_Any_NEON, MergeUVRow_NEON, MergeUVRow_C, 15)
1.430 +#endif
1.431 +#undef MERGEUVROW_ANY
1.432 +
1.433 +#define MATHROW_ANY(NAMEANY, ARGBMATH_SIMD, ARGBMATH_C, MASK) \
1.434 + void NAMEANY(const uint8* src_argb0, const uint8* src_argb1, \
1.435 + uint8* dst_argb, int width) { \
1.436 + int n = width & ~MASK; \
1.437 + ARGBMATH_SIMD(src_argb0, src_argb1, dst_argb, n); \
1.438 + ARGBMATH_C(src_argb0 + n * 4, \
1.439 + src_argb1 + n * 4, \
1.440 + dst_argb + n * 4, \
1.441 + width & MASK); \
1.442 + }
1.443 +
1.444 +#ifdef HAS_ARGBMULTIPLYROW_SSE2
1.445 +MATHROW_ANY(ARGBMultiplyRow_Any_SSE2, ARGBMultiplyRow_SSE2, ARGBMultiplyRow_C,
1.446 + 3)
1.447 +#endif
1.448 +#ifdef HAS_ARGBADDROW_SSE2
1.449 +MATHROW_ANY(ARGBAddRow_Any_SSE2, ARGBAddRow_SSE2, ARGBAddRow_C, 3)
1.450 +#endif
1.451 +#ifdef HAS_ARGBSUBTRACTROW_SSE2
1.452 +MATHROW_ANY(ARGBSubtractRow_Any_SSE2, ARGBSubtractRow_SSE2, ARGBSubtractRow_C,
1.453 + 3)
1.454 +#endif
1.455 +#ifdef HAS_ARGBMULTIPLYROW_AVX2
1.456 +MATHROW_ANY(ARGBMultiplyRow_Any_AVX2, ARGBMultiplyRow_AVX2, ARGBMultiplyRow_C,
1.457 + 7)
1.458 +#endif
1.459 +#ifdef HAS_ARGBADDROW_AVX2
1.460 +MATHROW_ANY(ARGBAddRow_Any_AVX2, ARGBAddRow_AVX2, ARGBAddRow_C, 7)
1.461 +#endif
1.462 +#ifdef HAS_ARGBSUBTRACTROW_AVX2
1.463 +MATHROW_ANY(ARGBSubtractRow_Any_AVX2, ARGBSubtractRow_AVX2, ARGBSubtractRow_C,
1.464 + 7)
1.465 +#endif
1.466 +#ifdef HAS_ARGBMULTIPLYROW_NEON
1.467 +MATHROW_ANY(ARGBMultiplyRow_Any_NEON, ARGBMultiplyRow_NEON, ARGBMultiplyRow_C,
1.468 + 7)
1.469 +#endif
1.470 +#ifdef HAS_ARGBADDROW_NEON
1.471 +MATHROW_ANY(ARGBAddRow_Any_NEON, ARGBAddRow_NEON, ARGBAddRow_C, 7)
1.472 +#endif
1.473 +#ifdef HAS_ARGBSUBTRACTROW_NEON
1.474 +MATHROW_ANY(ARGBSubtractRow_Any_NEON, ARGBSubtractRow_NEON, ARGBSubtractRow_C,
1.475 + 7)
1.476 +#endif
1.477 +#undef MATHROW_ANY
1.478 +
1.479 +// Shuffle may want to work in place, so last16 method can not be used.
1.480 +#define YANY(NAMEANY, ARGBTOY_SIMD, ARGBTOY_C, SBPP, BPP, MASK) \
1.481 + void NAMEANY(const uint8* src_argb, uint8* dst_argb, \
1.482 + const uint8* shuffler, int width) { \
1.483 + int n = width & ~MASK; \
1.484 + ARGBTOY_SIMD(src_argb, dst_argb, shuffler, n); \
1.485 + ARGBTOY_C(src_argb + n * SBPP, \
1.486 + dst_argb + n * BPP, shuffler, width & MASK); \
1.487 + }
1.488 +
1.489 +#ifdef HAS_ARGBSHUFFLEROW_SSE2
1.490 +YANY(ARGBShuffleRow_Any_SSE2, ARGBShuffleRow_SSE2,
1.491 + ARGBShuffleRow_C, 4, 4, 3)
1.492 +#endif
1.493 +#ifdef HAS_ARGBSHUFFLEROW_SSSE3
1.494 +YANY(ARGBShuffleRow_Any_SSSE3, ARGBShuffleRow_Unaligned_SSSE3,
1.495 + ARGBShuffleRow_C, 4, 4, 7)
1.496 +#endif
1.497 +#ifdef HAS_ARGBSHUFFLEROW_AVX2
1.498 +YANY(ARGBShuffleRow_Any_AVX2, ARGBShuffleRow_AVX2,
1.499 + ARGBShuffleRow_C, 4, 4, 15)
1.500 +#endif
1.501 +#ifdef HAS_ARGBSHUFFLEROW_NEON
1.502 +YANY(ARGBShuffleRow_Any_NEON, ARGBShuffleRow_NEON,
1.503 + ARGBShuffleRow_C, 4, 4, 3)
1.504 +#endif
1.505 +#undef YANY
1.506 +
1.507 +// Interpolate may want to work in place, so last16 method can not be used.
1.508 +#define NANY(NAMEANY, TERP_SIMD, TERP_C, SBPP, BPP, MASK) \
1.509 + void NAMEANY(uint8* dst_ptr, const uint8* src_ptr, \
1.510 + ptrdiff_t src_stride_ptr, int width, \
1.511 + int source_y_fraction) { \
1.512 + int n = width & ~MASK; \
1.513 + TERP_SIMD(dst_ptr, src_ptr, src_stride_ptr, \
1.514 + n, source_y_fraction); \
1.515 + TERP_C(dst_ptr + n * BPP, \
1.516 + src_ptr + n * SBPP, src_stride_ptr, \
1.517 + width & MASK, source_y_fraction); \
1.518 + }
1.519 +
1.520 +#ifdef HAS_INTERPOLATEROW_AVX2
1.521 +NANY(InterpolateRow_Any_AVX2, InterpolateRow_AVX2,
1.522 + InterpolateRow_C, 1, 1, 32)
1.523 +#endif
1.524 +#ifdef HAS_INTERPOLATEROW_SSSE3
1.525 +NANY(InterpolateRow_Any_SSSE3, InterpolateRow_Unaligned_SSSE3,
1.526 + InterpolateRow_C, 1, 1, 15)
1.527 +#endif
1.528 +#ifdef HAS_INTERPOLATEROW_SSE2
1.529 +NANY(InterpolateRow_Any_SSE2, InterpolateRow_Unaligned_SSE2,
1.530 + InterpolateRow_C, 1, 1, 15)
1.531 +#endif
1.532 +#ifdef HAS_INTERPOLATEROW_NEON
1.533 +NANY(InterpolateRow_Any_NEON, InterpolateRow_NEON,
1.534 + InterpolateRow_C, 1, 1, 15)
1.535 +#endif
1.536 +#ifdef HAS_INTERPOLATEROW_MIPS_DSPR2
1.537 +NANY(InterpolateRow_Any_MIPS_DSPR2, InterpolateRow_MIPS_DSPR2,
1.538 + InterpolateRow_C, 1, 1, 3)
1.539 +#endif
1.540 +#undef NANY
1.541 +
1.542 +#ifdef __cplusplus
1.543 +} // extern "C"
1.544 +} // namespace libyuv
1.545 +#endif
