Chromium Code Reviews
chromiumcodereview-hr@appspot.gserviceaccount.com (chromiumcodereview-hr) | Please choose your nickname with Settings | Help | Chromium Project | Gerrit Changes | Sign out
(15)

Issues Search
    My Issues | Starred     Open | Closed | All

Unified Diff: source/libvpx/third_party/libyuv/source/row_any.cc

Issue 1302353004: libvpx: Pull from upstream (Closed) Base URL: https://chromium.googlesource.com/chromium/deps/libvpx.git@master
Patch Set: Created 5 years, 4 months ago
Use n/p to move between diff chunks; N/P to move between comments. Draft comments are only viewable by you.
Jump to:
View side-by-side diff with in-line comments
Download patch
« no previous file with comments | « source/libvpx/third_party/libyuv/source/rotate_win.cc ('k') | source/libvpx/third_party/libyuv/source/row_common.cc » ('j') | no next file with comments »
Expand Comments ('e') | Collapse Comments ('c') | Show Comments Hide Comments ('s')
Index: source/libvpx/third_party/libyuv/source/row_any.cc
diff --git a/source/libvpx/third_party/libyuv/source/row_any.cc b/source/libvpx/third_party/libyuv/source/row_any.cc
index 19340b3b74cc01902facfe59e966f524113c8612..1cb1f6b9309993d9913d847c7a3a28b306ccd023 100644
--- a/source/libvpx/third_party/libyuv/source/row_any.cc
+++ b/source/libvpx/third_party/libyuv/source/row_any.cc
@@ -10,6 +10,8 @@
#include "libyuv/row.h"
+#include <string.h> // For memset.
+
#include "libyuv/basic_types.h"
#ifdef __cplusplus
@@ -17,699 +19,660 @@ namespace libyuv {
extern "C" {
#endif
-// YUV to RGB does multiple of 8 with SIMD and remainder with C.
-#define YANY(NAMEANY, I420TORGB_SIMD, I420TORGB_C, UV_SHIFT, BPP, MASK) \
+// Subsampled source needs to be increase by 1 of not even.
+#define SS(width, shift) (((width) + (1 << (shift)) - 1) >> (shift))
+
+// Any 3 planes to 1.
+#define ANY31(NAMEANY, ANY_SIMD, UVSHIFT, DUVSHIFT, BPP, MASK) \
void NAMEANY(const uint8* y_buf, const uint8* u_buf, const uint8* v_buf, \
- uint8* rgb_buf, int width) { \
+ uint8* dst_ptr, int width) { \
+ SIMD_ALIGNED(uint8 temp[64 * 4]); \
+ memset(temp, 0, 64 * 3); /* for YUY2 and msan */ \
+ int r = width & MASK; \
int n = width & ~MASK; \
if (n > 0) { \
- I420TORGB_SIMD(y_buf, u_buf, v_buf, rgb_buf, n); \
+ ANY_SIMD(y_buf, u_buf, v_buf, dst_ptr, n); \
} \
- I420TORGB_C(y_buf + n, \
- u_buf + (n >> UV_SHIFT), \
- v_buf + (n >> UV_SHIFT), \
- rgb_buf + n * BPP, width & MASK); \
+ memcpy(temp, y_buf + n, r); \
+ memcpy(temp + 64, u_buf + (n >> UVSHIFT), SS(r, UVSHIFT)); \
+ memcpy(temp + 128, v_buf + (n >> UVSHIFT), SS(r, UVSHIFT)); \
+ ANY_SIMD(temp, temp + 64, temp + 128, temp + 192, MASK + 1); \
+ memcpy(dst_ptr + (n >> DUVSHIFT) * BPP, temp + 192, \
+ SS(r, DUVSHIFT) * BPP); \
}
#ifdef HAS_I422TOARGBROW_SSSE3
-YANY(I422ToARGBRow_Any_SSSE3, I422ToARGBRow_SSSE3, I422ToARGBRow_C,
- 1, 4, 7)
+ANY31(I422ToARGBRow_Any_SSSE3, I422ToARGBRow_SSSE3, 1, 0, 4, 7)
#endif
#ifdef HAS_I444TOARGBROW_SSSE3
-YANY(I444ToARGBRow_Any_SSSE3, I444ToARGBRow_SSSE3, I444ToARGBRow_C,
- 0, 4, 7)
-YANY(I411ToARGBRow_Any_SSSE3, I411ToARGBRow_SSSE3, I411ToARGBRow_C,
- 2, 4, 7)
-YANY(I422ToBGRARow_Any_SSSE3, I422ToBGRARow_SSSE3, I422ToBGRARow_C,
- 1, 4, 7)
-YANY(I422ToABGRRow_Any_SSSE3, I422ToABGRRow_SSSE3, I422ToABGRRow_C,
- 1, 4, 7)
-YANY(I422ToRGBARow_Any_SSSE3, I422ToRGBARow_SSSE3, I422ToRGBARow_C,
- 1, 4, 7)
-YANY(I422ToARGB4444Row_Any_SSSE3, I422ToARGB4444Row_SSSE3, I422ToARGB4444Row_C,
- 1, 2, 7)
-YANY(I422ToARGB1555Row_Any_SSSE3, I422ToARGB1555Row_SSSE3, I422ToARGB1555Row_C,
- 1, 2, 7)
-YANY(I422ToRGB565Row_Any_SSSE3, I422ToRGB565Row_SSSE3, I422ToRGB565Row_C,
- 1, 2, 7)
-YANY(I422ToRGB24Row_Any_SSSE3, I422ToRGB24Row_SSSE3, I422ToRGB24Row_C, 1, 3, 7)
-YANY(I422ToRAWRow_Any_SSSE3, I422ToRAWRow_SSSE3, I422ToRAWRow_C, 1, 3, 7)
-YANY(I422ToYUY2Row_Any_SSE2, I422ToYUY2Row_SSE2, I422ToYUY2Row_C, 1, 2, 15)
-YANY(I422ToUYVYRow_Any_SSE2, I422ToUYVYRow_SSE2, I422ToUYVYRow_C, 1, 2, 15)
+ANY31(I444ToARGBRow_Any_SSSE3, I444ToARGBRow_SSSE3, 0, 0, 4, 7)
+ANY31(I411ToARGBRow_Any_SSSE3, I411ToARGBRow_SSSE3, 2, 0, 4, 7)
+ANY31(I422ToBGRARow_Any_SSSE3, I422ToBGRARow_SSSE3, 1, 0, 4, 7)
+ANY31(I422ToABGRRow_Any_SSSE3, I422ToABGRRow_SSSE3, 1, 0, 4, 7)
+ANY31(I422ToRGBARow_Any_SSSE3, I422ToRGBARow_SSSE3, 1, 0, 4, 7)
+ANY31(I422ToARGB4444Row_Any_SSSE3, I422ToARGB4444Row_SSSE3, 1, 0, 2, 7)
+ANY31(I422ToARGB1555Row_Any_SSSE3, I422ToARGB1555Row_SSSE3, 1, 0, 2, 7)
+ANY31(I422ToRGB565Row_Any_SSSE3, I422ToRGB565Row_SSSE3, 1, 0, 2, 7)
+ANY31(I422ToRGB24Row_Any_SSSE3, I422ToRGB24Row_SSSE3, 1, 0, 3, 7)
+ANY31(I422ToRAWRow_Any_SSSE3, I422ToRAWRow_SSSE3, 1, 0, 3, 7)
+ANY31(I422ToYUY2Row_Any_SSE2, I422ToYUY2Row_SSE2, 1, 1, 4, 15)
+ANY31(I422ToUYVYRow_Any_SSE2, I422ToUYVYRow_SSE2, 1, 1, 4, 15)
#endif // HAS_I444TOARGBROW_SSSE3
+#ifdef HAS_I422TORGB24ROW_AVX2
+ANY31(I422ToRGB24Row_Any_AVX2, I422ToRGB24Row_AVX2, 1, 0, 3, 15)
+#endif
+#ifdef HAS_I422TORAWROW_AVX2
+ANY31(I422ToRAWRow_Any_AVX2, I422ToRAWRow_AVX2, 1, 0, 3, 15)
+#endif
#ifdef HAS_J422TOARGBROW_SSSE3
-YANY(J422ToARGBRow_Any_SSSE3, J422ToARGBRow_SSSE3, J422ToARGBRow_C,
- 1, 4, 7)
+ANY31(J422ToARGBRow_Any_SSSE3, J422ToARGBRow_SSSE3, 1, 0, 4, 7)
+#endif
+#ifdef HAS_J422TOARGBROW_AVX2
+ANY31(J422ToARGBRow_Any_AVX2, J422ToARGBRow_AVX2, 1, 0, 4, 15)
#endif
#ifdef HAS_I422TOARGBROW_AVX2
-YANY(I422ToARGBRow_Any_AVX2, I422ToARGBRow_AVX2, I422ToARGBRow_C, 1, 4, 15)
+ANY31(I422ToARGBRow_Any_AVX2, I422ToARGBRow_AVX2, 1, 0, 4, 15)
#endif
#ifdef HAS_I422TOBGRAROW_AVX2
-YANY(I422ToBGRARow_Any_AVX2, I422ToBGRARow_AVX2, I422ToBGRARow_C, 1, 4, 15)
+ANY31(I422ToBGRARow_Any_AVX2, I422ToBGRARow_AVX2, 1, 0, 4, 15)
#endif
#ifdef HAS_I422TORGBAROW_AVX2
-YANY(I422ToRGBARow_Any_AVX2, I422ToRGBARow_AVX2, I422ToRGBARow_C, 1, 4, 15)
+ANY31(I422ToRGBARow_Any_AVX2, I422ToRGBARow_AVX2, 1, 0, 4, 15)
#endif
#ifdef HAS_I422TOABGRROW_AVX2
-YANY(I422ToABGRRow_Any_AVX2, I422ToABGRRow_AVX2, I422ToABGRRow_C, 1, 4, 15)
+ANY31(I422ToABGRRow_Any_AVX2, I422ToABGRRow_AVX2, 1, 0, 4, 15)
+#endif
+#ifdef HAS_I444TOARGBROW_AVX2
+ANY31(I444ToARGBRow_Any_AVX2, I444ToARGBRow_AVX2, 0, 0, 4, 15)
+#endif
+#ifdef HAS_I411TOARGBROW_AVX2
+ANY31(I411ToARGBRow_Any_AVX2, I411ToARGBRow_AVX2, 2, 0, 4, 15)
#endif
#ifdef HAS_I422TOARGB4444ROW_AVX2
-YANY(I422ToARGB4444Row_Any_AVX2, I422ToARGB4444Row_AVX2, I422ToARGB4444Row_C,
- 1, 2, 7)
+ANY31(I422ToARGB4444Row_Any_AVX2, I422ToARGB4444Row_AVX2, 1, 0, 2, 7)
#endif
#ifdef HAS_I422TOARGB1555ROW_AVX2
-YANY(I422ToARGB1555Row_Any_AVX2, I422ToARGB1555Row_AVX2, I422ToARGB1555Row_C,
- 1, 2, 7)
+ANY31(I422ToARGB1555Row_Any_AVX2, I422ToARGB1555Row_AVX2, 1, 0, 2, 7)
#endif
#ifdef HAS_I422TORGB565ROW_AVX2
-YANY(I422ToRGB565Row_Any_AVX2, I422ToRGB565Row_AVX2, I422ToRGB565Row_C,
- 1, 2, 7)
+ANY31(I422ToRGB565Row_Any_AVX2, I422ToRGB565Row_AVX2, 1, 0, 2, 7)
#endif
#ifdef HAS_I422TOARGBROW_NEON
-YANY(I444ToARGBRow_Any_NEON, I444ToARGBRow_NEON, I444ToARGBRow_C, 0, 4, 7)
-YANY(I422ToARGBRow_Any_NEON, I422ToARGBRow_NEON, I422ToARGBRow_C, 1, 4, 7)
-YANY(I411ToARGBRow_Any_NEON, I411ToARGBRow_NEON, I411ToARGBRow_C, 2, 4, 7)
-YANY(I422ToBGRARow_Any_NEON, I422ToBGRARow_NEON, I422ToBGRARow_C, 1, 4, 7)
-YANY(I422ToABGRRow_Any_NEON, I422ToABGRRow_NEON, I422ToABGRRow_C, 1, 4, 7)
-YANY(I422ToRGBARow_Any_NEON, I422ToRGBARow_NEON, I422ToRGBARow_C, 1, 4, 7)
-YANY(I422ToRGB24Row_Any_NEON, I422ToRGB24Row_NEON, I422ToRGB24Row_C, 1, 3, 7)
-YANY(I422ToRAWRow_Any_NEON, I422ToRAWRow_NEON, I422ToRAWRow_C, 1, 3, 7)
-YANY(I422ToARGB4444Row_Any_NEON, I422ToARGB4444Row_NEON, I422ToARGB4444Row_C,
- 1, 2, 7)
-YANY(I422ToARGB1555Row_Any_NEON, I422ToARGB1555Row_NEON, I422ToARGB1555Row_C,
- 1, 2, 7)
-YANY(I422ToRGB565Row_Any_NEON, I422ToRGB565Row_NEON, I422ToRGB565Row_C, 1, 2, 7)
+ANY31(I444ToARGBRow_Any_NEON, I444ToARGBRow_NEON, 0, 0, 4, 7)
+ANY31(I422ToARGBRow_Any_NEON, I422ToARGBRow_NEON, 1, 0, 4, 7)
+ANY31(I411ToARGBRow_Any_NEON, I411ToARGBRow_NEON, 2, 0, 4, 7)
+ANY31(I422ToBGRARow_Any_NEON, I422ToBGRARow_NEON, 1, 0, 4, 7)
+ANY31(I422ToABGRRow_Any_NEON, I422ToABGRRow_NEON, 1, 0, 4, 7)
+ANY31(I422ToRGBARow_Any_NEON, I422ToRGBARow_NEON, 1, 0, 4, 7)
+ANY31(I422ToRGB24Row_Any_NEON, I422ToRGB24Row_NEON, 1, 0, 3, 7)
+ANY31(I422ToRAWRow_Any_NEON, I422ToRAWRow_NEON, 1, 0, 3, 7)
+ANY31(I422ToARGB4444Row_Any_NEON, I422ToARGB4444Row_NEON, 1, 0, 2, 7)
+ANY31(I422ToARGB1555Row_Any_NEON, I422ToARGB1555Row_NEON, 1, 0, 2, 7)
+ANY31(I422ToRGB565Row_Any_NEON, I422ToRGB565Row_NEON, 1, 0, 2, 7)
#endif
#ifdef HAS_I422TOYUY2ROW_NEON
-YANY(I422ToYUY2Row_Any_NEON, I422ToYUY2Row_NEON, I422ToYUY2Row_C, 1, 2, 15)
+ANY31(I422ToYUY2Row_Any_NEON, I422ToYUY2Row_NEON, 1, 1, 4, 15)
#endif
#ifdef HAS_I422TOUYVYROW_NEON
-YANY(I422ToUYVYRow_Any_NEON, I422ToUYVYRow_NEON, I422ToUYVYRow_C, 1, 2, 15)
+ANY31(I422ToUYVYRow_Any_NEON, I422ToUYVYRow_NEON, 1, 1, 4, 15)
#endif
-#undef YANY
+#undef ANY31
-// Wrappers to handle odd width
-#define NV2NY(NAMEANY, NV12TORGB_SIMD, NV12TORGB_C, UV_SHIFT, BPP, MASK) \
+// Any 2 planes to 1.
+#define ANY21(NAMEANY, ANY_SIMD, UVSHIFT, SBPP, SBPP2, BPP, MASK) \
void NAMEANY(const uint8* y_buf, const uint8* uv_buf, \
- uint8* rgb_buf, int width) { \
+ uint8* dst_ptr, int width) { \
+ SIMD_ALIGNED(uint8 temp[64 * 3]); \
+ memset(temp, 0, 64 * 2); /* for msan */ \
+ int r = width & MASK; \
int n = width & ~MASK; \
if (n > 0) { \
- NV12TORGB_SIMD(y_buf, uv_buf, rgb_buf, n); \
+ ANY_SIMD(y_buf, uv_buf, dst_ptr, n); \
} \
- NV12TORGB_C(y_buf + n, \
- uv_buf + (n >> UV_SHIFT), \
- rgb_buf + n * BPP, width & MASK); \
+ memcpy(temp, y_buf + n * SBPP, r * SBPP); \
+ memcpy(temp + 64, uv_buf + (n >> UVSHIFT) * SBPP2, \
+ SS(r, UVSHIFT) * SBPP2); \
+ ANY_SIMD(temp, temp + 64, temp + 128, MASK + 1); \
+ memcpy(dst_ptr + n * BPP, temp + 128, r * BPP); \
}
+// Biplanar to RGB.
#ifdef HAS_NV12TOARGBROW_SSSE3
-NV2NY(NV12ToARGBRow_Any_SSSE3, NV12ToARGBRow_SSSE3, NV12ToARGBRow_C, 0, 4, 7)
-NV2NY(NV21ToARGBRow_Any_SSSE3, NV21ToARGBRow_SSSE3, NV21ToARGBRow_C, 0, 4, 7)
+ANY21(NV12ToARGBRow_Any_SSSE3, NV12ToARGBRow_SSSE3, 1, 1, 2, 4, 7)
+ANY21(NV21ToARGBRow_Any_SSSE3, NV21ToARGBRow_SSSE3, 1, 1, 2, 4, 7)
#endif
#ifdef HAS_NV12TOARGBROW_AVX2
-NV2NY(NV12ToARGBRow_Any_AVX2, NV12ToARGBRow_AVX2, NV12ToARGBRow_C, 0, 4, 15)
-NV2NY(NV21ToARGBRow_Any_AVX2, NV21ToARGBRow_AVX2, NV21ToARGBRow_C, 0, 4, 15)
+ANY21(NV12ToARGBRow_Any_AVX2, NV12ToARGBRow_AVX2, 1, 1, 2, 4, 15)
+ANY21(NV21ToARGBRow_Any_AVX2, NV21ToARGBRow_AVX2, 1, 1, 2, 4, 15)
#endif
#ifdef HAS_NV12TOARGBROW_NEON
-NV2NY(NV12ToARGBRow_Any_NEON, NV12ToARGBRow_NEON, NV12ToARGBRow_C, 0, 4, 7)
-NV2NY(NV21ToARGBRow_Any_NEON, NV21ToARGBRow_NEON, NV21ToARGBRow_C, 0, 4, 7)
+ANY21(NV12ToARGBRow_Any_NEON, NV12ToARGBRow_NEON, 1, 1, 2, 4, 7)
+ANY21(NV21ToARGBRow_Any_NEON, NV21ToARGBRow_NEON, 1, 1, 2, 4, 7)
#endif
#ifdef HAS_NV12TORGB565ROW_SSSE3
-NV2NY(NV12ToRGB565Row_Any_SSSE3, NV12ToRGB565Row_SSSE3, NV12ToRGB565Row_C,
- 0, 2, 7)
-NV2NY(NV21ToRGB565Row_Any_SSSE3, NV21ToRGB565Row_SSSE3, NV21ToRGB565Row_C,
- 0, 2, 7)
+ANY21(NV12ToRGB565Row_Any_SSSE3, NV12ToRGB565Row_SSSE3, 1, 1, 2, 2, 7)
+ANY21(NV21ToRGB565Row_Any_SSSE3, NV21ToRGB565Row_SSSE3, 1, 1, 2, 2, 7)
#endif
#ifdef HAS_NV12TORGB565ROW_AVX2
-NV2NY(NV12ToRGB565Row_Any_AVX2, NV12ToRGB565Row_AVX2, NV12ToRGB565Row_C,
- 0, 2, 15)
-NV2NY(NV21ToRGB565Row_Any_AVX2, NV21ToRGB565Row_AVX2, NV21ToRGB565Row_C,
- 0, 2, 15)
+ANY21(NV12ToRGB565Row_Any_AVX2, NV12ToRGB565Row_AVX2, 1, 1, 2, 2, 15)
+ANY21(NV21ToRGB565Row_Any_AVX2, NV21ToRGB565Row_AVX2, 1, 1, 2, 2, 15)
#endif
#ifdef HAS_NV12TORGB565ROW_NEON
-NV2NY(NV12ToRGB565Row_Any_NEON, NV12ToRGB565Row_NEON, NV12ToRGB565Row_C,
- 0, 2, 7)
-NV2NY(NV21ToRGB565Row_Any_NEON, NV21ToRGB565Row_NEON, NV21ToRGB565Row_C,
- 0, 2, 7)
+ANY21(NV12ToRGB565Row_Any_NEON, NV12ToRGB565Row_NEON, 1, 1, 2, 2, 7)
+ANY21(NV21ToRGB565Row_Any_NEON, NV21ToRGB565Row_NEON, 1, 1, 2, 2, 7)
+#endif
+
+// Merge functions.
+#ifdef HAS_MERGEUVROW_SSE2
+ANY21(MergeUVRow_Any_SSE2, MergeUVRow_SSE2, 0, 1, 1, 2, 15)
+#endif
+#ifdef HAS_MERGEUVROW_AVX2
+ANY21(MergeUVRow_Any_AVX2, MergeUVRow_AVX2, 0, 1, 1, 2, 31)
+#endif
+#ifdef HAS_MERGEUVROW_NEON
+ANY21(MergeUVRow_Any_NEON, MergeUVRow_NEON, 0, 1, 1, 2, 15)
#endif
-#undef NVANY
-#define RGBANY(NAMEANY, ARGBTORGB_SIMD, ARGBTORGB_C, SBPP, BPP, MASK) \
- void NAMEANY(const uint8* src, uint8* dst, int width) { \
+// Math functions.
+#ifdef HAS_ARGBMULTIPLYROW_SSE2
+ANY21(ARGBMultiplyRow_Any_SSE2, ARGBMultiplyRow_SSE2, 0, 4, 4, 4, 3)
+#endif
+#ifdef HAS_ARGBADDROW_SSE2
+ANY21(ARGBAddRow_Any_SSE2, ARGBAddRow_SSE2, 0, 4, 4, 4, 3)
+#endif
+#ifdef HAS_ARGBSUBTRACTROW_SSE2
+ANY21(ARGBSubtractRow_Any_SSE2, ARGBSubtractRow_SSE2, 0, 4, 4, 4, 3)
+#endif
+#ifdef HAS_ARGBMULTIPLYROW_AVX2
+ANY21(ARGBMultiplyRow_Any_AVX2, ARGBMultiplyRow_AVX2, 0, 4, 4, 4, 7)
+#endif
+#ifdef HAS_ARGBADDROW_AVX2
+ANY21(ARGBAddRow_Any_AVX2, ARGBAddRow_AVX2, 0, 4, 4, 4, 7)
+#endif
+#ifdef HAS_ARGBSUBTRACTROW_AVX2
+ANY21(ARGBSubtractRow_Any_AVX2, ARGBSubtractRow_AVX2, 0, 4, 4, 4, 7)
+#endif
+#ifdef HAS_ARGBMULTIPLYROW_NEON
+ANY21(ARGBMultiplyRow_Any_NEON, ARGBMultiplyRow_NEON, 0, 4, 4, 4, 7)
+#endif
+#ifdef HAS_ARGBADDROW_NEON
+ANY21(ARGBAddRow_Any_NEON, ARGBAddRow_NEON, 0, 4, 4, 4, 7)
+#endif
+#ifdef HAS_ARGBSUBTRACTROW_NEON
+ANY21(ARGBSubtractRow_Any_NEON, ARGBSubtractRow_NEON, 0, 4, 4, 4, 7)
+#endif
+#ifdef HAS_SOBELROW_SSE2
+ANY21(SobelRow_Any_SSE2, SobelRow_SSE2, 0, 1, 1, 4, 15)
+#endif
+#ifdef HAS_SOBELROW_NEON
+ANY21(SobelRow_Any_NEON, SobelRow_NEON, 0, 1, 1, 4, 7)
+#endif
+#ifdef HAS_SOBELTOPLANEROW_SSE2
+ANY21(SobelToPlaneRow_Any_SSE2, SobelToPlaneRow_SSE2, 0, 1, 1, 1, 15)
+#endif
+#ifdef HAS_SOBELTOPLANEROW_NEON
+ANY21(SobelToPlaneRow_Any_NEON, SobelToPlaneRow_NEON, 0, 1, 1, 1, 15)
+#endif
+#ifdef HAS_SOBELXYROW_SSE2
+ANY21(SobelXYRow_Any_SSE2, SobelXYRow_SSE2, 0, 1, 1, 4, 15)
+#endif
+#ifdef HAS_SOBELXYROW_NEON
+ANY21(SobelXYRow_Any_NEON, SobelXYRow_NEON, 0, 1, 1, 4, 7)
+#endif
+#undef ANY21
+
+// Any 1 to 1.
+#define ANY11(NAMEANY, ANY_SIMD, UVSHIFT, SBPP, BPP, MASK) \
+ void NAMEANY(const uint8* src_ptr, uint8* dst_ptr, int width) { \
+ SIMD_ALIGNED(uint8 temp[128 * 2]); \
+ memset(temp, 0, 128); /* for YUY2 and msan */ \
+ int r = width & MASK; \
int n = width & ~MASK; \
if (n > 0) { \
- ARGBTORGB_SIMD(src, dst, n); \
+ ANY_SIMD(src_ptr, dst_ptr, n); \
} \
- ARGBTORGB_C(src + n * SBPP, dst + n * BPP, width & MASK); \
+ memcpy(temp, src_ptr + (n >> UVSHIFT) * SBPP, SS(r, UVSHIFT) * SBPP); \
+ ANY_SIMD(temp, temp + 128, MASK + 1); \
+ memcpy(dst_ptr + n * BPP, temp + 128, r * BPP); \
}
+#ifdef HAS_COPYROW_AVX
+ANY11(CopyRow_Any_AVX, CopyRow_AVX, 0, 1, 1, 63)
+#endif
+#ifdef HAS_COPYROW_SSE2
+ANY11(CopyRow_Any_SSE2, CopyRow_SSE2, 0, 1, 1, 31)
+#endif
+#ifdef HAS_COPYROW_NEON
+ANY11(CopyRow_Any_NEON, CopyRow_NEON, 0, 1, 1, 31)
+#endif
#if defined(HAS_ARGBTORGB24ROW_SSSE3)
-RGBANY(ARGBToRGB24Row_Any_SSSE3, ARGBToRGB24Row_SSSE3, ARGBToRGB24Row_C,
- 4, 3, 15)
-RGBANY(ARGBToRAWRow_Any_SSSE3, ARGBToRAWRow_SSSE3, ARGBToRAWRow_C,
- 4, 3, 15)
-RGBANY(ARGBToRGB565Row_Any_SSE2, ARGBToRGB565Row_SSE2, ARGBToRGB565Row_C,
- 4, 2, 3)
-RGBANY(ARGBToARGB1555Row_Any_SSE2, ARGBToARGB1555Row_SSE2, ARGBToARGB1555Row_C,
- 4, 2, 3)
-RGBANY(ARGBToARGB4444Row_Any_SSE2, ARGBToARGB4444Row_SSE2, ARGBToARGB4444Row_C,
- 4, 2, 3)
+ANY11(ARGBToRGB24Row_Any_SSSE3, ARGBToRGB24Row_SSSE3, 0, 4, 3, 15)
+ANY11(ARGBToRAWRow_Any_SSSE3, ARGBToRAWRow_SSSE3, 0, 4, 3, 15)
+ANY11(ARGBToRGB565Row_Any_SSE2, ARGBToRGB565Row_SSE2, 0, 4, 2, 3)
+ANY11(ARGBToARGB1555Row_Any_SSE2, ARGBToARGB1555Row_SSE2, 0, 4, 2, 3)
+ANY11(ARGBToARGB4444Row_Any_SSE2, ARGBToARGB4444Row_SSE2, 0, 4, 2, 3)
#endif
#if defined(HAS_ARGBTOARGB4444ROW_AVX2)
-RGBANY(ARGBToRGB565Row_Any_AVX2, ARGBToRGB565Row_AVX2, ARGBToRGB565Row_C,
- 4, 2, 7)
-RGBANY(ARGBToARGB1555Row_Any_AVX2, ARGBToARGB1555Row_AVX2, ARGBToARGB1555Row_C,
- 4, 2, 7)
-RGBANY(ARGBToARGB4444Row_Any_AVX2, ARGBToARGB4444Row_AVX2, ARGBToARGB4444Row_C,
- 4, 2, 7)
+ANY11(ARGBToRGB565Row_Any_AVX2, ARGBToRGB565Row_AVX2, 0, 4, 2, 7)
+ANY11(ARGBToARGB1555Row_Any_AVX2, ARGBToARGB1555Row_AVX2, 0, 4, 2, 7)
+ANY11(ARGBToARGB4444Row_Any_AVX2, ARGBToARGB4444Row_AVX2, 0, 4, 2, 7)
#endif
-
-#if defined(HAS_I400TOARGBROW_SSE2)
-RGBANY(I400ToARGBRow_Any_SSE2, I400ToARGBRow_SSE2, I400ToARGBRow_C, 1, 4, 7)
+#if defined(HAS_J400TOARGBROW_SSE2)
+ANY11(J400ToARGBRow_Any_SSE2, J400ToARGBRow_SSE2, 0, 1, 4, 7)
+#endif
+#if defined(HAS_J400TOARGBROW_AVX2)
+ANY11(J400ToARGBRow_Any_AVX2, J400ToARGBRow_AVX2, 0, 1, 4, 15)
#endif
-#if defined(HAS_YTOARGBROW_SSE2)
-RGBANY(YToARGBRow_Any_SSE2, YToARGBRow_SSE2, YToARGBRow_C, 1, 4, 7)
+#if defined(HAS_I400TOARGBROW_SSE2)
+ANY11(I400ToARGBRow_Any_SSE2, I400ToARGBRow_SSE2, 0, 1, 4, 7)
#endif
-#if defined(HAS_YTOARGBROW_AVX2)
-RGBANY(YToARGBRow_Any_AVX2, YToARGBRow_AVX2, YToARGBRow_C, 1, 4, 15)
+#if defined(HAS_I400TOARGBROW_AVX2)
+ANY11(I400ToARGBRow_Any_AVX2, I400ToARGBRow_AVX2, 0, 1, 4, 15)
#endif
#if defined(HAS_YUY2TOARGBROW_SSSE3)
-RGBANY(YUY2ToARGBRow_Any_SSSE3, YUY2ToARGBRow_SSSE3, YUY2ToARGBRow_C, 2, 4, 15)
-RGBANY(UYVYToARGBRow_Any_SSSE3, UYVYToARGBRow_SSSE3, UYVYToARGBRow_C, 2, 4, 15)
-RGBANY(RGB24ToARGBRow_Any_SSSE3, RGB24ToARGBRow_SSSE3, RGB24ToARGBRow_C,
- 3, 4, 15)
-RGBANY(RAWToARGBRow_Any_SSSE3, RAWToARGBRow_SSSE3, RAWToARGBRow_C, 3, 4, 15)
-RGBANY(RGB565ToARGBRow_Any_SSE2, RGB565ToARGBRow_SSE2, RGB565ToARGBRow_C,
- 2, 4, 7)
-RGBANY(ARGB1555ToARGBRow_Any_SSE2, ARGB1555ToARGBRow_SSE2, ARGB1555ToARGBRow_C,
- 2, 4, 7)
-RGBANY(ARGB4444ToARGBRow_Any_SSE2, ARGB4444ToARGBRow_SSE2, ARGB4444ToARGBRow_C,
- 2, 4, 7)
+ANY11(YUY2ToARGBRow_Any_SSSE3, YUY2ToARGBRow_SSSE3, 1, 4, 4, 15)
+ANY11(UYVYToARGBRow_Any_SSSE3, UYVYToARGBRow_SSSE3, 1, 4, 4, 15)
+ANY11(RGB24ToARGBRow_Any_SSSE3, RGB24ToARGBRow_SSSE3, 0, 3, 4, 15)
+ANY11(RAWToARGBRow_Any_SSSE3, RAWToARGBRow_SSSE3, 0, 3, 4, 15)
+ANY11(RGB565ToARGBRow_Any_SSE2, RGB565ToARGBRow_SSE2, 0, 2, 4, 7)
+ANY11(ARGB1555ToARGBRow_Any_SSE2, ARGB1555ToARGBRow_SSE2, 0, 2, 4, 7)
+ANY11(ARGB4444ToARGBRow_Any_SSE2, ARGB4444ToARGBRow_SSE2, 0, 2, 4, 7)
+#endif
+#if defined(HAS_RGB565TOARGBROW_AVX2)
+ANY11(RGB565ToARGBRow_Any_AVX2, RGB565ToARGBRow_AVX2, 0, 2, 4, 15)
+#endif
+#if defined(HAS_ARGB1555TOARGBROW_AVX2)
+ANY11(ARGB1555ToARGBRow_Any_AVX2, ARGB1555ToARGBRow_AVX2, 0, 2, 4, 15)
+#endif
+#if defined(HAS_ARGB4444TOARGBROW_AVX2)
+ANY11(ARGB4444ToARGBRow_Any_AVX2, ARGB4444ToARGBRow_AVX2, 0, 2, 4, 15)
#endif
#if defined(HAS_YUY2TOARGBROW_AVX2)
-RGBANY(YUY2ToARGBRow_Any_AVX2, YUY2ToARGBRow_AVX2, YUY2ToARGBRow_C, 2, 4, 31)
-RGBANY(UYVYToARGBRow_Any_AVX2, UYVYToARGBRow_AVX2, UYVYToARGBRow_C, 2, 4, 31)
+ANY11(YUY2ToARGBRow_Any_AVX2, YUY2ToARGBRow_AVX2, 1, 4, 4, 31)
+ANY11(UYVYToARGBRow_Any_AVX2, UYVYToARGBRow_AVX2, 1, 4, 4, 31)
#endif
#if defined(HAS_ARGBTORGB24ROW_NEON)
-RGBANY(ARGBToRGB24Row_Any_NEON, ARGBToRGB24Row_NEON, ARGBToRGB24Row_C, 4, 3, 7)
-RGBANY(ARGBToRAWRow_Any_NEON, ARGBToRAWRow_NEON, ARGBToRAWRow_C, 4, 3, 7)
-RGBANY(ARGBToRGB565Row_Any_NEON, ARGBToRGB565Row_NEON, ARGBToRGB565Row_C,
- 4, 2, 7)
-RGBANY(ARGBToARGB1555Row_Any_NEON, ARGBToARGB1555Row_NEON, ARGBToARGB1555Row_C,
- 4, 2, 7)
-RGBANY(ARGBToARGB4444Row_Any_NEON, ARGBToARGB4444Row_NEON, ARGBToARGB4444Row_C,
- 4, 2, 7)
-RGBANY(I400ToARGBRow_Any_NEON, I400ToARGBRow_NEON, I400ToARGBRow_C, 1, 4, 7)
-RGBANY(YToARGBRow_Any_NEON, YToARGBRow_NEON, YToARGBRow_C, 1, 4, 7)
-RGBANY(YUY2ToARGBRow_Any_NEON, YUY2ToARGBRow_NEON, YUY2ToARGBRow_C, 2, 4, 7)
-RGBANY(UYVYToARGBRow_Any_NEON, UYVYToARGBRow_NEON, UYVYToARGBRow_C, 2, 4, 7)
-#endif
-#undef RGBANY
-
-// ARGB to Bayer does multiple of 4 pixels, SSSE3 aligned src, unaligned dst.
-#define BAYERANY(NAMEANY, ARGBTORGB_SIMD, ARGBTORGB_C, SBPP, BPP, MASK) \
- void NAMEANY(const uint8* src, uint8* dst, uint32 selector, int width) { \
- int n = width & ~MASK; \
- if (n > 0) { \
- ARGBTORGB_SIMD(src, dst, selector, n); \
- } \
- ARGBTORGB_C(src + n * SBPP, dst + n * BPP, selector, width & MASK); \
- }
-
-#if defined(HAS_ARGBTOBAYERGGROW_SSE2)
-BAYERANY(ARGBToBayerGGRow_Any_SSE2, ARGBToBayerGGRow_SSE2, ARGBToBayerGGRow_C,
- 4, 1, 7)
+ANY11(ARGBToRGB24Row_Any_NEON, ARGBToRGB24Row_NEON, 0, 4, 3, 7)
+ANY11(ARGBToRAWRow_Any_NEON, ARGBToRAWRow_NEON, 0, 4, 3, 7)
+ANY11(ARGBToRGB565Row_Any_NEON, ARGBToRGB565Row_NEON, 0, 4, 2, 7)
+ANY11(ARGBToARGB1555Row_Any_NEON, ARGBToARGB1555Row_NEON, 0, 4, 2, 7)
+ANY11(ARGBToARGB4444Row_Any_NEON, ARGBToARGB4444Row_NEON, 0, 4, 2, 7)
+ANY11(J400ToARGBRow_Any_NEON, J400ToARGBRow_NEON, 0, 1, 4, 7)
+ANY11(I400ToARGBRow_Any_NEON, I400ToARGBRow_NEON, 0, 1, 4, 7)
+ANY11(YUY2ToARGBRow_Any_NEON, YUY2ToARGBRow_NEON, 1, 4, 4, 7)
+ANY11(UYVYToARGBRow_Any_NEON, UYVYToARGBRow_NEON, 1, 4, 4, 7)
#endif
-#if defined(HAS_ARGBTOBAYERGGROW_NEON)
-BAYERANY(ARGBToBayerGGRow_Any_NEON, ARGBToBayerGGRow_NEON, ARGBToBayerGGRow_C,
- 4, 1, 7)
-#endif
-
-#undef BAYERANY
-
-#define YANY(NAMEANY, ARGBTOY_SIMD, ARGBTOY_C, SBPP, BPP, MASK) \
- void NAMEANY(const uint8* src_argb, uint8* dst_y, int width) { \
- int n = width & ~MASK; \
- if (n > 0) { \
- ARGBTOY_SIMD(src_argb, dst_y, n); \
- } \
- ARGBTOY_C(src_argb + n * SBPP, \
- dst_y + n * BPP, width & MASK); \
- }
#ifdef HAS_ARGBTOYROW_AVX2
-YANY(ARGBToYRow_Any_AVX2, ARGBToYRow_AVX2, ARGBToYRow_C, 4, 1, 31)
+ANY11(ARGBToYRow_Any_AVX2, ARGBToYRow_AVX2, 0, 4, 1, 31)
#endif
#ifdef HAS_ARGBTOYJROW_AVX2
-YANY(ARGBToYJRow_Any_AVX2, ARGBToYJRow_AVX2, ARGBToYJRow_C, 4, 1, 31)
+ANY11(ARGBToYJRow_Any_AVX2, ARGBToYJRow_AVX2, 0, 4, 1, 31)
#endif
#ifdef HAS_UYVYTOYROW_AVX2
-YANY(UYVYToYRow_Any_AVX2, UYVYToYRow_AVX2, UYVYToYRow_C, 2, 1, 31)
+ANY11(UYVYToYRow_Any_AVX2, UYVYToYRow_AVX2, 0, 2, 1, 31)
#endif
#ifdef HAS_YUY2TOYROW_AVX2
-YANY(YUY2ToYRow_Any_AVX2, YUY2ToYRow_AVX2, YUY2ToYRow_C, 2, 1, 31)
+ANY11(YUY2ToYRow_Any_AVX2, YUY2ToYRow_AVX2, 1, 4, 1, 31)
#endif
#ifdef HAS_ARGBTOYROW_SSSE3
-YANY(ARGBToYRow_Any_SSSE3, ARGBToYRow_SSSE3, ARGBToYRow_C, 4, 1, 15)
+ANY11(ARGBToYRow_Any_SSSE3, ARGBToYRow_SSSE3, 0, 4, 1, 15)
#endif
#ifdef HAS_BGRATOYROW_SSSE3
-YANY(BGRAToYRow_Any_SSSE3, BGRAToYRow_SSSE3, BGRAToYRow_C, 4, 1, 15)
-YANY(ABGRToYRow_Any_SSSE3, ABGRToYRow_SSSE3, ABGRToYRow_C, 4, 1, 15)
-YANY(RGBAToYRow_Any_SSSE3, RGBAToYRow_SSSE3, RGBAToYRow_C, 4, 1, 15)
-YANY(YUY2ToYRow_Any_SSE2, YUY2ToYRow_SSE2, YUY2ToYRow_C, 2, 1, 15)
-YANY(UYVYToYRow_Any_SSE2, UYVYToYRow_SSE2, UYVYToYRow_C, 2, 1, 15)
+ANY11(BGRAToYRow_Any_SSSE3, BGRAToYRow_SSSE3, 0, 4, 1, 15)
+ANY11(ABGRToYRow_Any_SSSE3, ABGRToYRow_SSSE3, 0, 4, 1, 15)
+ANY11(RGBAToYRow_Any_SSSE3, RGBAToYRow_SSSE3, 0, 4, 1, 15)
+ANY11(YUY2ToYRow_Any_SSE2, YUY2ToYRow_SSE2, 1, 4, 1, 15)
+ANY11(UYVYToYRow_Any_SSE2, UYVYToYRow_SSE2, 1, 4, 1, 15)
#endif
#ifdef HAS_ARGBTOYJROW_SSSE3
-YANY(ARGBToYJRow_Any_SSSE3, ARGBToYJRow_SSSE3, ARGBToYJRow_C, 4, 1, 15)
+ANY11(ARGBToYJRow_Any_SSSE3, ARGBToYJRow_SSSE3, 0, 4, 1, 15)
#endif
#ifdef HAS_ARGBTOYROW_NEON
-YANY(ARGBToYRow_Any_NEON, ARGBToYRow_NEON, ARGBToYRow_C, 4, 1, 7)
+ANY11(ARGBToYRow_Any_NEON, ARGBToYRow_NEON, 0, 4, 1, 7)
#endif
#ifdef HAS_ARGBTOYJROW_NEON
-YANY(ARGBToYJRow_Any_NEON, ARGBToYJRow_NEON, ARGBToYJRow_C, 4, 1, 7)
+ANY11(ARGBToYJRow_Any_NEON, ARGBToYJRow_NEON, 0, 4, 1, 7)
#endif
#ifdef HAS_BGRATOYROW_NEON
-YANY(BGRAToYRow_Any_NEON, BGRAToYRow_NEON, BGRAToYRow_C, 4, 1, 7)
+ANY11(BGRAToYRow_Any_NEON, BGRAToYRow_NEON, 0, 4, 1, 7)
#endif
#ifdef HAS_ABGRTOYROW_NEON
-YANY(ABGRToYRow_Any_NEON, ABGRToYRow_NEON, ABGRToYRow_C, 4, 1, 7)
+ANY11(ABGRToYRow_Any_NEON, ABGRToYRow_NEON, 0, 4, 1, 7)
#endif
#ifdef HAS_RGBATOYROW_NEON
-YANY(RGBAToYRow_Any_NEON, RGBAToYRow_NEON, RGBAToYRow_C, 4, 1, 7)
+ANY11(RGBAToYRow_Any_NEON, RGBAToYRow_NEON, 0, 4, 1, 7)
#endif
#ifdef HAS_RGB24TOYROW_NEON
-YANY(RGB24ToYRow_Any_NEON, RGB24ToYRow_NEON, RGB24ToYRow_C, 3, 1, 7)
+ANY11(RGB24ToYRow_Any_NEON, RGB24ToYRow_NEON, 0, 3, 1, 7)
#endif
#ifdef HAS_RAWTOYROW_NEON
-YANY(RAWToYRow_Any_NEON, RAWToYRow_NEON, RAWToYRow_C, 3, 1, 7)
+ANY11(RAWToYRow_Any_NEON, RAWToYRow_NEON, 0, 3, 1, 7)
#endif
#ifdef HAS_RGB565TOYROW_NEON
-YANY(RGB565ToYRow_Any_NEON, RGB565ToYRow_NEON, RGB565ToYRow_C, 2, 1, 7)
+ANY11(RGB565ToYRow_Any_NEON, RGB565ToYRow_NEON, 0, 2, 1, 7)
#endif
#ifdef HAS_ARGB1555TOYROW_NEON
-YANY(ARGB1555ToYRow_Any_NEON, ARGB1555ToYRow_NEON, ARGB1555ToYRow_C, 2, 1, 7)
+ANY11(ARGB1555ToYRow_Any_NEON, ARGB1555ToYRow_NEON, 0, 2, 1, 7)
#endif
#ifdef HAS_ARGB4444TOYROW_NEON
-YANY(ARGB4444ToYRow_Any_NEON, ARGB4444ToYRow_NEON, ARGB4444ToYRow_C, 2, 1, 7)
+ANY11(ARGB4444ToYRow_Any_NEON, ARGB4444ToYRow_NEON, 0, 2, 1, 7)
#endif
#ifdef HAS_YUY2TOYROW_NEON
-YANY(YUY2ToYRow_Any_NEON, YUY2ToYRow_NEON, YUY2ToYRow_C, 2, 1, 15)
+ANY11(YUY2ToYRow_Any_NEON, YUY2ToYRow_NEON, 1, 4, 1, 15)
#endif
#ifdef HAS_UYVYTOYROW_NEON
-YANY(UYVYToYRow_Any_NEON, UYVYToYRow_NEON, UYVYToYRow_C, 2, 1, 15)
+ANY11(UYVYToYRow_Any_NEON, UYVYToYRow_NEON, 0, 2, 1, 15)
#endif
#ifdef HAS_RGB24TOARGBROW_NEON
-YANY(RGB24ToARGBRow_Any_NEON, RGB24ToARGBRow_NEON, RGB24ToARGBRow_C, 3, 4, 7)
+ANY11(RGB24ToARGBRow_Any_NEON, RGB24ToARGBRow_NEON, 0, 3, 4, 7)
#endif
#ifdef HAS_RAWTOARGBROW_NEON
-YANY(RAWToARGBRow_Any_NEON, RAWToARGBRow_NEON, RAWToARGBRow_C, 3, 4, 7)
+ANY11(RAWToARGBRow_Any_NEON, RAWToARGBRow_NEON, 0, 3, 4, 7)
#endif
#ifdef HAS_RGB565TOARGBROW_NEON
-YANY(RGB565ToARGBRow_Any_NEON, RGB565ToARGBRow_NEON, RGB565ToARGBRow_C, 2, 4, 7)
+ANY11(RGB565ToARGBRow_Any_NEON, RGB565ToARGBRow_NEON, 0, 2, 4, 7)
#endif
#ifdef HAS_ARGB1555TOARGBROW_NEON
-YANY(ARGB1555ToARGBRow_Any_NEON, ARGB1555ToARGBRow_NEON, ARGB1555ToARGBRow_C,
- 2, 4, 7)
+ANY11(ARGB1555ToARGBRow_Any_NEON, ARGB1555ToARGBRow_NEON, 0, 2, 4, 7)
#endif
#ifdef HAS_ARGB4444TOARGBROW_NEON
-YANY(ARGB4444ToARGBRow_Any_NEON, ARGB4444ToARGBRow_NEON, ARGB4444ToARGBRow_C,
- 2, 4, 7)
+ANY11(ARGB4444ToARGBRow_Any_NEON, ARGB4444ToARGBRow_NEON, 0, 2, 4, 7)
#endif
#ifdef HAS_ARGBATTENUATEROW_SSSE3
-YANY(ARGBAttenuateRow_Any_SSSE3, ARGBAttenuateRow_SSSE3, ARGBAttenuateRow_C,
- 4, 4, 3)
+ANY11(ARGBAttenuateRow_Any_SSSE3, ARGBAttenuateRow_SSSE3, 0, 4, 4, 3)
#endif
#ifdef HAS_ARGBATTENUATEROW_SSE2
-YANY(ARGBAttenuateRow_Any_SSE2, ARGBAttenuateRow_SSE2, ARGBAttenuateRow_C,
- 4, 4, 3)
+ANY11(ARGBAttenuateRow_Any_SSE2, ARGBAttenuateRow_SSE2, 0, 4, 4, 3)
#endif
#ifdef HAS_ARGBUNATTENUATEROW_SSE2
-YANY(ARGBUnattenuateRow_Any_SSE2, ARGBUnattenuateRow_SSE2, ARGBUnattenuateRow_C,
- 4, 4, 3)
+ANY11(ARGBUnattenuateRow_Any_SSE2, ARGBUnattenuateRow_SSE2, 0, 4, 4, 3)
#endif
#ifdef HAS_ARGBATTENUATEROW_AVX2
-YANY(ARGBAttenuateRow_Any_AVX2, ARGBAttenuateRow_AVX2, ARGBAttenuateRow_C,
- 4, 4, 7)
+ANY11(ARGBAttenuateRow_Any_AVX2, ARGBAttenuateRow_AVX2, 0, 4, 4, 7)
#endif
#ifdef HAS_ARGBUNATTENUATEROW_AVX2
-YANY(ARGBUnattenuateRow_Any_AVX2, ARGBUnattenuateRow_AVX2, ARGBUnattenuateRow_C,
- 4, 4, 7)
+ANY11(ARGBUnattenuateRow_Any_AVX2, ARGBUnattenuateRow_AVX2, 0, 4, 4, 7)
#endif
#ifdef HAS_ARGBATTENUATEROW_NEON
-YANY(ARGBAttenuateRow_Any_NEON, ARGBAttenuateRow_NEON, ARGBAttenuateRow_C,
- 4, 4, 7)
+ANY11(ARGBAttenuateRow_Any_NEON, ARGBAttenuateRow_NEON, 0, 4, 4, 7)
#endif
-#undef YANY
+#undef ANY11
-// RGB/YUV to UV does multiple of 16 with SIMD and remainder with C.
-#define UVANY(NAMEANY, ANYTOUV_SIMD, ANYTOUV_C, BPP, MASK) \
- void NAMEANY(const uint8* src_argb, int src_stride_argb, \
- uint8* dst_u, uint8* dst_v, int width) { \
+// Any 1 to 1 with parameter.
+#define ANY11P(NAMEANY, ANY_SIMD, T, SBPP, BPP, MASK) \
+ void NAMEANY(const uint8* src_ptr, uint8* dst_ptr, \
+ T shuffler, int width) { \
+ SIMD_ALIGNED(uint8 temp[64 * 2]); \
+ memset(temp, 0, 64); /* for msan */ \
+ int r = width & MASK; \
int n = width & ~MASK; \
if (n > 0) { \
- ANYTOUV_SIMD(src_argb, src_stride_argb, dst_u, dst_v, n); \
+ ANY_SIMD(src_ptr, dst_ptr, shuffler, n); \
} \
- ANYTOUV_C(src_argb + n * BPP, src_stride_argb, \
- dst_u + (n >> 1), \
- dst_v + (n >> 1), \
- width & MASK); \
+ memcpy(temp, src_ptr + n * SBPP, r * SBPP); \
+ ANY_SIMD(temp, temp + 64, shuffler, MASK + 1); \
+ memcpy(dst_ptr + n * BPP, temp + 64, r * BPP); \
}
-#ifdef HAS_ARGBTOUVROW_AVX2
-UVANY(ARGBToUVRow_Any_AVX2, ARGBToUVRow_AVX2, ARGBToUVRow_C, 4, 31)
-#endif
-#ifdef HAS_ARGBTOUVROW_SSSE3
-UVANY(ARGBToUVRow_Any_SSSE3, ARGBToUVRow_SSSE3, ARGBToUVRow_C, 4, 15)
-UVANY(ARGBToUVJRow_Any_SSSE3, ARGBToUVJRow_SSSE3, ARGBToUVJRow_C, 4, 15)
-UVANY(BGRAToUVRow_Any_SSSE3, BGRAToUVRow_SSSE3, BGRAToUVRow_C, 4, 15)
-UVANY(ABGRToUVRow_Any_SSSE3, ABGRToUVRow_SSSE3, ABGRToUVRow_C, 4, 15)
-UVANY(RGBAToUVRow_Any_SSSE3, RGBAToUVRow_SSSE3, RGBAToUVRow_C, 4, 15)
-#endif
-#ifdef HAS_YUY2TOUVROW_AVX2
-UVANY(YUY2ToUVRow_Any_AVX2, YUY2ToUVRow_AVX2, YUY2ToUVRow_C, 2, 31)
-UVANY(UYVYToUVRow_Any_AVX2, UYVYToUVRow_AVX2, UYVYToUVRow_C, 2, 31)
-#endif
-#ifdef HAS_YUY2TOUVROW_SSE2
-UVANY(YUY2ToUVRow_Any_SSE2, YUY2ToUVRow_SSE2, YUY2ToUVRow_C, 2, 15)
-UVANY(UYVYToUVRow_Any_SSE2, UYVYToUVRow_SSE2, UYVYToUVRow_C, 2, 15)
-#endif
-#ifdef HAS_ARGBTOUVROW_NEON
-UVANY(ARGBToUVRow_Any_NEON, ARGBToUVRow_NEON, ARGBToUVRow_C, 4, 15)
-#endif
-#ifdef HAS_ARGBTOUVJROW_NEON
-UVANY(ARGBToUVJRow_Any_NEON, ARGBToUVJRow_NEON, ARGBToUVJRow_C, 4, 15)
-#endif
-#ifdef HAS_BGRATOUVROW_NEON
-UVANY(BGRAToUVRow_Any_NEON, BGRAToUVRow_NEON, BGRAToUVRow_C, 4, 15)
-#endif
-#ifdef HAS_ABGRTOUVROW_NEON
-UVANY(ABGRToUVRow_Any_NEON, ABGRToUVRow_NEON, ABGRToUVRow_C, 4, 15)
-#endif
-#ifdef HAS_RGBATOUVROW_NEON
-UVANY(RGBAToUVRow_Any_NEON, RGBAToUVRow_NEON, RGBAToUVRow_C, 4, 15)
-#endif
-#ifdef HAS_RGB24TOUVROW_NEON
-UVANY(RGB24ToUVRow_Any_NEON, RGB24ToUVRow_NEON, RGB24ToUVRow_C, 3, 15)
+#if defined(HAS_ARGBTORGB565DITHERROW_SSE2)
+ANY11P(ARGBToRGB565DitherRow_Any_SSE2, ARGBToRGB565DitherRow_SSE2,
+ const uint32, 4, 2, 3)
#endif
-#ifdef HAS_RAWTOUVROW_NEON
-UVANY(RAWToUVRow_Any_NEON, RAWToUVRow_NEON, RAWToUVRow_C, 3, 15)
+#if defined(HAS_ARGBTORGB565DITHERROW_AVX2)
+ANY11P(ARGBToRGB565DitherRow_Any_AVX2, ARGBToRGB565DitherRow_AVX2,
+ const uint32, 4, 2, 7)
#endif
-#ifdef HAS_RGB565TOUVROW_NEON
-UVANY(RGB565ToUVRow_Any_NEON, RGB565ToUVRow_NEON, RGB565ToUVRow_C, 2, 15)
+#if defined(HAS_ARGBTORGB565DITHERROW_NEON)
+ANY11P(ARGBToRGB565DitherRow_Any_NEON, ARGBToRGB565DitherRow_NEON,
+ const uint32, 4, 2, 7)
#endif
-#ifdef HAS_ARGB1555TOUVROW_NEON
-UVANY(ARGB1555ToUVRow_Any_NEON, ARGB1555ToUVRow_NEON, ARGB1555ToUVRow_C, 2, 15)
+#ifdef HAS_ARGBSHUFFLEROW_SSE2
+ANY11P(ARGBShuffleRow_Any_SSE2, ARGBShuffleRow_SSE2, const uint8*, 4, 4, 3)
#endif
-#ifdef HAS_ARGB4444TOUVROW_NEON
-UVANY(ARGB4444ToUVRow_Any_NEON, ARGB4444ToUVRow_NEON, ARGB4444ToUVRow_C, 2, 15)
+#ifdef HAS_ARGBSHUFFLEROW_SSSE3
+ANY11P(ARGBShuffleRow_Any_SSSE3, ARGBShuffleRow_SSSE3, const uint8*, 4, 4, 7)
#endif
-#ifdef HAS_YUY2TOUVROW_NEON
-UVANY(YUY2ToUVRow_Any_NEON, YUY2ToUVRow_NEON, YUY2ToUVRow_C, 2, 15)
+#ifdef HAS_ARGBSHUFFLEROW_AVX2
+ANY11P(ARGBShuffleRow_Any_AVX2, ARGBShuffleRow_AVX2, const uint8*, 4, 4, 15)
#endif
-#ifdef HAS_UYVYTOUVROW_NEON
-UVANY(UYVYToUVRow_Any_NEON, UYVYToUVRow_NEON, UYVYToUVRow_C, 2, 15)
+#ifdef HAS_ARGBSHUFFLEROW_NEON
+ANY11P(ARGBShuffleRow_Any_NEON, ARGBShuffleRow_NEON, const uint8*, 4, 4, 3)
#endif
-#undef UVANY
+#undef ANY11P
-#define UV422ANY(NAMEANY, ANYTOUV_SIMD, ANYTOUV_C, BPP, SHIFT, MASK) \
- void NAMEANY(const uint8* src_uv, uint8* dst_u, uint8* dst_v, int width) { \
+// Any 1 to 1 interpolate. Takes 2 rows of source via stride.
+#define ANY11T(NAMEANY, ANY_SIMD, SBPP, BPP, MASK) \
+ void NAMEANY(uint8* dst_ptr, const uint8* src_ptr, \
+ ptrdiff_t src_stride_ptr, int width, \
+ int source_y_fraction) { \
+ SIMD_ALIGNED(uint8 temp[64 * 3]); \
+ memset(temp, 0, 64 * 2); /* for msan */ \
+ int r = width & MASK; \
int n = width & ~MASK; \
if (n > 0) { \
- ANYTOUV_SIMD(src_uv, dst_u, dst_v, n); \
+ ANY_SIMD(dst_ptr, src_ptr, src_stride_ptr, n, source_y_fraction); \
} \
- ANYTOUV_C(src_uv + n * BPP, \
- dst_u + (n >> SHIFT), \
- dst_v + (n >> SHIFT), \
- width & MASK); \
+ memcpy(temp, src_ptr + n * SBPP, r * SBPP); \
+ memcpy(temp + 64, src_ptr + src_stride_ptr + n * SBPP, r * SBPP); \
+ ANY_SIMD(temp + 128, temp, 64, MASK + 1, source_y_fraction); \
+ memcpy(dst_ptr + n * BPP, temp + 128, r * BPP); \
}
-#ifdef HAS_ARGBTOUV444ROW_SSSE3
-UV422ANY(ARGBToUV444Row_Any_SSSE3, ARGBToUV444Row_SSSE3,
- ARGBToUV444Row_C, 4, 0, 15)
+#ifdef HAS_INTERPOLATEROW_AVX2
+ANY11T(InterpolateRow_Any_AVX2, InterpolateRow_AVX2, 1, 1, 31)
#endif
-#ifdef HAS_YUY2TOUV422ROW_AVX2
-UV422ANY(YUY2ToUV422Row_Any_AVX2, YUY2ToUV422Row_AVX2,
- YUY2ToUV422Row_C, 2, 1, 31)
-UV422ANY(UYVYToUV422Row_Any_AVX2, UYVYToUV422Row_AVX2,
- UYVYToUV422Row_C, 2, 1, 31)
+#ifdef HAS_INTERPOLATEROW_SSSE3
+ANY11T(InterpolateRow_Any_SSSE3, InterpolateRow_SSSE3, 1, 1, 15)
#endif
-#ifdef HAS_ARGBTOUV422ROW_SSSE3
-UV422ANY(ARGBToUV422Row_Any_SSSE3, ARGBToUV422Row_SSSE3,
- ARGBToUV422Row_C, 4, 1, 15)
+#ifdef HAS_INTERPOLATEROW_SSE2
+ANY11T(InterpolateRow_Any_SSE2, InterpolateRow_SSE2, 1, 1, 15)
#endif
-#ifdef HAS_YUY2TOUV422ROW_SSE2
-UV422ANY(YUY2ToUV422Row_Any_SSE2, YUY2ToUV422Row_SSE2,
- YUY2ToUV422Row_C, 2, 1, 15)
-UV422ANY(UYVYToUV422Row_Any_SSE2, UYVYToUV422Row_SSE2,
- UYVYToUV422Row_C, 2, 1, 15)
+#ifdef HAS_INTERPOLATEROW_NEON
+ANY11T(InterpolateRow_Any_NEON, InterpolateRow_NEON, 1, 1, 15)
#endif
-#ifdef HAS_YUY2TOUV422ROW_NEON
-UV422ANY(ARGBToUV444Row_Any_NEON, ARGBToUV444Row_NEON,
- ARGBToUV444Row_C, 4, 0, 7)
-UV422ANY(ARGBToUV422Row_Any_NEON, ARGBToUV422Row_NEON,
- ARGBToUV422Row_C, 4, 1, 15)
-UV422ANY(ARGBToUV411Row_Any_NEON, ARGBToUV411Row_NEON,
- ARGBToUV411Row_C, 4, 2, 31)
-UV422ANY(YUY2ToUV422Row_Any_NEON, YUY2ToUV422Row_NEON,
- YUY2ToUV422Row_C, 2, 1, 15)
-UV422ANY(UYVYToUV422Row_Any_NEON, UYVYToUV422Row_NEON,
- UYVYToUV422Row_C, 2, 1, 15)
-#endif
-#undef UV422ANY
+#ifdef HAS_INTERPOLATEROW_MIPS_DSPR2
+ANY11T(InterpolateRow_Any_MIPS_DSPR2, InterpolateRow_MIPS_DSPR2, 1, 1, 3)
+#endif
+#undef ANY11T
-#define SPLITUVROWANY(NAMEANY, ANYTOUV_SIMD, ANYTOUV_C, MASK) \
- void NAMEANY(const uint8* src_uv, uint8* dst_u, uint8* dst_v, int width) { \
+// Any 1 to 1 mirror.
+#define ANY11M(NAMEANY, ANY_SIMD, BPP, MASK) \
+ void NAMEANY(const uint8* src_ptr, uint8* dst_ptr, int width) { \
+ SIMD_ALIGNED(uint8 temp[64 * 2]); \
+ memset(temp, 0, 64); /* for msan */ \
+ int r = width & MASK; \
int n = width & ~MASK; \
if (n > 0) { \
- ANYTOUV_SIMD(src_uv, dst_u, dst_v, n); \
+ ANY_SIMD(src_ptr + r * BPP, dst_ptr, n); \
} \
- ANYTOUV_C(src_uv + n * 2, \
- dst_u + n, \
- dst_v + n, \
- width & MASK); \
+ memcpy(temp, src_ptr, r * BPP); \
+ ANY_SIMD(temp, temp + 64, MASK + 1); \
+ memcpy(dst_ptr + n * BPP, temp + 64 + (MASK + 1 - r) * BPP, r * BPP); \
}
-#ifdef HAS_SPLITUVROW_SSE2
-SPLITUVROWANY(SplitUVRow_Any_SSE2, SplitUVRow_SSE2, SplitUVRow_C, 15)
+#ifdef HAS_MIRRORROW_AVX2
+ANY11M(MirrorRow_Any_AVX2, MirrorRow_AVX2, 1, 31)
#endif
-#ifdef HAS_SPLITUVROW_AVX2
-SPLITUVROWANY(SplitUVRow_Any_AVX2, SplitUVRow_AVX2, SplitUVRow_C, 31)
+#ifdef HAS_MIRRORROW_SSSE3
+ANY11M(MirrorRow_Any_SSSE3, MirrorRow_SSSE3, 1, 15)
#endif
-#ifdef HAS_SPLITUVROW_NEON
-SPLITUVROWANY(SplitUVRow_Any_NEON, SplitUVRow_NEON, SplitUVRow_C, 15)
+#ifdef HAS_MIRRORROW_SSE2
+ANY11M(MirrorRow_Any_SSE2, MirrorRow_SSE2, 1, 15)
#endif
-#ifdef HAS_SPLITUVROW_MIPS_DSPR2
-SPLITUVROWANY(SplitUVRow_Any_MIPS_DSPR2, SplitUVRow_MIPS_DSPR2,
- SplitUVRow_C, 15)
+#ifdef HAS_MIRRORROW_NEON
+ANY11M(MirrorRow_Any_NEON, MirrorRow_NEON, 1, 15)
+#endif
+#ifdef HAS_ARGBMIRRORROW_AVX2
+ANY11M(ARGBMirrorRow_Any_AVX2, ARGBMirrorRow_AVX2, 4, 7)
#endif
-#undef SPLITUVROWANY
+#ifdef HAS_ARGBMIRRORROW_SSE2
+ANY11M(ARGBMirrorRow_Any_SSE2, ARGBMirrorRow_SSE2, 4, 3)
+#endif
+#ifdef HAS_ARGBMIRRORROW_NEON
+ANY11M(ARGBMirrorRow_Any_NEON, ARGBMirrorRow_NEON, 4, 3)
+#endif
+#undef ANY11M
-#define MERGEUVROW_ANY(NAMEANY, ANYTOUV_SIMD, ANYTOUV_C, MASK) \
- void NAMEANY(const uint8* src_u, const uint8* src_v, \
- uint8* dst_uv, int width) { \
+// Any 1 plane. (memset)
+#define ANY1(NAMEANY, ANY_SIMD, T, BPP, MASK) \
+ void NAMEANY(uint8* dst_ptr, T v32, int width) { \
+ SIMD_ALIGNED(uint8 temp[64]); \
+ int r = width & MASK; \
int n = width & ~MASK; \
if (n > 0) { \
- ANYTOUV_SIMD(src_u, src_v, dst_uv, n); \
+ ANY_SIMD(dst_ptr, v32, n); \
} \
- ANYTOUV_C(src_u + n, \
- src_v + n, \
- dst_uv + n * 2, \
- width & MASK); \
+ ANY_SIMD(temp, v32, MASK + 1); \
+ memcpy(dst_ptr + n * BPP, temp, r * BPP); \
}
-#ifdef HAS_MERGEUVROW_SSE2
-MERGEUVROW_ANY(MergeUVRow_Any_SSE2, MergeUVRow_SSE2, MergeUVRow_C, 15)
+#ifdef HAS_SETROW_X86
+ANY1(SetRow_Any_X86, SetRow_X86, uint8, 1, 3)
#endif
-#ifdef HAS_MERGEUVROW_AVX2
-MERGEUVROW_ANY(MergeUVRow_Any_AVX2, MergeUVRow_AVX2, MergeUVRow_C, 31)
+#ifdef HAS_SETROW_NEON
+ANY1(SetRow_Any_NEON, SetRow_NEON, uint8, 1, 15)
#endif
-#ifdef HAS_MERGEUVROW_NEON
-MERGEUVROW_ANY(MergeUVRow_Any_NEON, MergeUVRow_NEON, MergeUVRow_C, 15)
+#ifdef HAS_ARGBSETROW_NEON
+ANY1(ARGBSetRow_Any_NEON, ARGBSetRow_NEON, uint32, 4, 3)
#endif
-#undef MERGEUVROW_ANY
+#undef ANY1
-#define MATHROW_ANY(NAMEANY, ARGBMATH_SIMD, ARGBMATH_C, MASK) \
- void NAMEANY(const uint8* src_argb0, const uint8* src_argb1, \
- uint8* dst_argb, int width) { \
+// Any 1 to 2. Outputs UV planes.
+#define ANY12(NAMEANY, ANY_SIMD, UVSHIFT, BPP, DUVSHIFT, MASK) \
+ void NAMEANY(const uint8* src_ptr, uint8* dst_u, uint8* dst_v, int width) {\
+ SIMD_ALIGNED(uint8 temp[128 * 3]); \
+ memset(temp, 0, 128); /* for msan */ \
+ int r = width & MASK; \
int n = width & ~MASK; \
if (n > 0) { \
- ARGBMATH_SIMD(src_argb0, src_argb1, dst_argb, n); \
+ ANY_SIMD(src_ptr, dst_u, dst_v, n); \
+ } \
+ memcpy(temp, src_ptr + (n >> UVSHIFT) * BPP, SS(r, UVSHIFT) * BPP); \
+ if ((width & 1) && BPP == 4) { /* repeat last 4 bytes for subsampler */ \
+ memcpy(temp + SS(r, UVSHIFT) * BPP, \
+ temp + SS(r, UVSHIFT) * BPP - BPP, 4); \
} \
- ARGBMATH_C(src_argb0 + n * 4, \
- src_argb1 + n * 4, \
- dst_argb + n * 4, \
- width & MASK); \
+ ANY_SIMD(temp, temp + 128, temp + 256, MASK + 1); \
+ memcpy(dst_u + (n >> DUVSHIFT), temp + 128, SS(r, DUVSHIFT)); \
+ memcpy(dst_v + (n >> DUVSHIFT), temp + 256, SS(r, DUVSHIFT)); \
}
-#ifdef HAS_ARGBMULTIPLYROW_SSE2
-MATHROW_ANY(ARGBMultiplyRow_Any_SSE2, ARGBMultiplyRow_SSE2, ARGBMultiplyRow_C,
- 3)
-#endif
-#ifdef HAS_ARGBADDROW_SSE2
-MATHROW_ANY(ARGBAddRow_Any_SSE2, ARGBAddRow_SSE2, ARGBAddRow_C, 3)
+#ifdef HAS_SPLITUVROW_SSE2
+ANY12(SplitUVRow_Any_SSE2, SplitUVRow_SSE2, 0, 2, 0, 15)
#endif
-#ifdef HAS_ARGBSUBTRACTROW_SSE2
-MATHROW_ANY(ARGBSubtractRow_Any_SSE2, ARGBSubtractRow_SSE2, ARGBSubtractRow_C,
- 3)
+#ifdef HAS_SPLITUVROW_AVX2
+ANY12(SplitUVRow_Any_AVX2, SplitUVRow_AVX2, 0, 2, 0, 31)
#endif
-#ifdef HAS_ARGBMULTIPLYROW_AVX2
-MATHROW_ANY(ARGBMultiplyRow_Any_AVX2, ARGBMultiplyRow_AVX2, ARGBMultiplyRow_C,
- 7)
+#ifdef HAS_SPLITUVROW_NEON
+ANY12(SplitUVRow_Any_NEON, SplitUVRow_NEON, 0, 2, 0, 15)
#endif
-#ifdef HAS_ARGBADDROW_AVX2
-MATHROW_ANY(ARGBAddRow_Any_AVX2, ARGBAddRow_AVX2, ARGBAddRow_C, 7)
+#ifdef HAS_SPLITUVROW_MIPS_DSPR2
+ANY12(SplitUVRow_Any_MIPS_DSPR2, SplitUVRow_MIPS_DSPR2, 0, 2, 0, 15)
#endif
-#ifdef HAS_ARGBSUBTRACTROW_AVX2
-MATHROW_ANY(ARGBSubtractRow_Any_AVX2, ARGBSubtractRow_AVX2, ARGBSubtractRow_C,
- 7)
+#ifdef HAS_ARGBTOUV444ROW_SSSE3
+ANY12(ARGBToUV444Row_Any_SSSE3, ARGBToUV444Row_SSSE3, 0, 4, 0, 15)
#endif
-#ifdef HAS_ARGBMULTIPLYROW_NEON
-MATHROW_ANY(ARGBMultiplyRow_Any_NEON, ARGBMultiplyRow_NEON, ARGBMultiplyRow_C,
- 7)
+#ifdef HAS_YUY2TOUV422ROW_AVX2
+ANY12(YUY2ToUV422Row_Any_AVX2, YUY2ToUV422Row_AVX2, 1, 4, 1, 31)
+ANY12(UYVYToUV422Row_Any_AVX2, UYVYToUV422Row_AVX2, 1, 4, 1, 31)
#endif
-#ifdef HAS_ARGBADDROW_NEON
-MATHROW_ANY(ARGBAddRow_Any_NEON, ARGBAddRow_NEON, ARGBAddRow_C, 7)
+#ifdef HAS_ARGBTOUV422ROW_SSSE3
+ANY12(ARGBToUV422Row_Any_SSSE3, ARGBToUV422Row_SSSE3, 0, 4, 1, 15)
#endif
-#ifdef HAS_ARGBSUBTRACTROW_NEON
-MATHROW_ANY(ARGBSubtractRow_Any_NEON, ARGBSubtractRow_NEON, ARGBSubtractRow_C,
- 7)
+#ifdef HAS_YUY2TOUV422ROW_SSE2
+ANY12(YUY2ToUV422Row_Any_SSE2, YUY2ToUV422Row_SSE2, 1, 4, 1, 15)
+ANY12(UYVYToUV422Row_Any_SSE2, UYVYToUV422Row_SSE2, 1, 4, 1, 15)
#endif
-#undef MATHROW_ANY
-
-// Shuffle may want to work in place, so last16 method can not be used.
-#define YANY(NAMEANY, ARGBTOY_SIMD, ARGBTOY_C, SBPP, BPP, MASK) \
- void NAMEANY(const uint8* src_argb, uint8* dst_argb, \
- const uint8* shuffler, int width) { \
+#ifdef HAS_YUY2TOUV422ROW_NEON
+ANY12(ARGBToUV444Row_Any_NEON, ARGBToUV444Row_NEON, 0, 4, 0, 7)
+ANY12(ARGBToUV422Row_Any_NEON, ARGBToUV422Row_NEON, 0, 4, 1, 15)
+ANY12(ARGBToUV411Row_Any_NEON, ARGBToUV411Row_NEON, 0, 4, 2, 31)
+ANY12(YUY2ToUV422Row_Any_NEON, YUY2ToUV422Row_NEON, 1, 4, 1, 15)
+ANY12(UYVYToUV422Row_Any_NEON, UYVYToUV422Row_NEON, 1, 4, 1, 15)
+#endif
+#undef ANY12
+
+// Any 1 to 2 with source stride (2 rows of source). Outputs UV planes.
+// 128 byte row allows for 32 avx ARGB pixels.
+#define ANY12S(NAMEANY, ANY_SIMD, UVSHIFT, BPP, MASK) \
+ void NAMEANY(const uint8* src_ptr, int src_stride_ptr, \
+ uint8* dst_u, uint8* dst_v, int width) { \
+ SIMD_ALIGNED(uint8 temp[128 * 4]); \
+ memset(temp, 0, 128 * 2); /* for msan */ \
+ int r = width & MASK; \
int n = width & ~MASK; \
if (n > 0) { \
- ARGBTOY_SIMD(src_argb, dst_argb, shuffler, n); \
+ ANY_SIMD(src_ptr, src_stride_ptr, dst_u, dst_v, n); \
} \
- ARGBTOY_C(src_argb + n * SBPP, \
- dst_argb + n * BPP, shuffler, width & MASK); \
- }
-
-#ifdef HAS_ARGBSHUFFLEROW_SSE2
-YANY(ARGBShuffleRow_Any_SSE2, ARGBShuffleRow_SSE2,
- ARGBShuffleRow_C, 4, 4, 3)
-#endif
-#ifdef HAS_ARGBSHUFFLEROW_SSSE3
-YANY(ARGBShuffleRow_Any_SSSE3, ARGBShuffleRow_SSSE3,
- ARGBShuffleRow_C, 4, 4, 7)
-#endif
-#ifdef HAS_ARGBSHUFFLEROW_AVX2
-YANY(ARGBShuffleRow_Any_AVX2, ARGBShuffleRow_AVX2,
- ARGBShuffleRow_C, 4, 4, 15)
-#endif
-#ifdef HAS_ARGBSHUFFLEROW_NEON
-YANY(ARGBShuffleRow_Any_NEON, ARGBShuffleRow_NEON,
- ARGBShuffleRow_C, 4, 4, 3)
-#endif
-#undef YANY
-
-// Interpolate may want to work in place, so last16 method can not be used.
-#define NANY(NAMEANY, TERP_SIMD, TERP_C, SBPP, BPP, MASK) \
- void NAMEANY(uint8* dst_ptr, const uint8* src_ptr, \
- ptrdiff_t src_stride_ptr, int width, \
- int source_y_fraction) { \
- int n = width & ~MASK; \
- if (n > 0) { \
- TERP_SIMD(dst_ptr, src_ptr, src_stride_ptr, n, source_y_fraction); \
+ memcpy(temp, src_ptr + (n >> UVSHIFT) * BPP, SS(r, UVSHIFT) * BPP); \
+ memcpy(temp + 128, src_ptr + src_stride_ptr + (n >> UVSHIFT) * BPP, \
+ SS(r, UVSHIFT) * BPP); \
+ if ((width & 1) && BPP == 4) { /* repeat last 4 bytes for subsampler */ \
+ memcpy(temp + SS(r, UVSHIFT) * BPP, \
+ temp + SS(r, UVSHIFT) * BPP - BPP, 4); \
+ memcpy(temp + 128 + SS(r, UVSHIFT) * BPP, \
+ temp + 128 + SS(r, UVSHIFT) * BPP - BPP, 4); \
} \
- TERP_C(dst_ptr + n * BPP, \
- src_ptr + n * SBPP, src_stride_ptr, \
- width & MASK, source_y_fraction); \
+ ANY_SIMD(temp, 128, temp + 256, temp + 384, MASK + 1); \
+ memcpy(dst_u + (n >> 1), temp + 256, SS(r, 1)); \
+ memcpy(dst_v + (n >> 1), temp + 384, SS(r, 1)); \
}
-#ifdef HAS_INTERPOLATEROW_AVX2
-NANY(InterpolateRow_Any_AVX2, InterpolateRow_AVX2, InterpolateRow_C, 1, 1, 31)
-#endif
-#ifdef HAS_INTERPOLATEROW_SSSE3
-NANY(InterpolateRow_Any_SSSE3, InterpolateRow_SSSE3, InterpolateRow_C, 1, 1, 15)
-#endif
-#ifdef HAS_INTERPOLATEROW_SSE2
-NANY(InterpolateRow_Any_SSE2, InterpolateRow_SSE2, InterpolateRow_C, 1, 1, 15)
+#ifdef HAS_ARGBTOUVROW_AVX2
+ANY12S(ARGBToUVRow_Any_AVX2, ARGBToUVRow_AVX2, 0, 4, 31)
#endif
-#ifdef HAS_INTERPOLATEROW_NEON
-NANY(InterpolateRow_Any_NEON, InterpolateRow_NEON, InterpolateRow_C, 1, 1, 15)
+#ifdef HAS_ARGBTOUVROW_SSSE3
+ANY12S(ARGBToUVRow_Any_SSSE3, ARGBToUVRow_SSSE3, 0, 4, 15)
+ANY12S(ARGBToUVJRow_Any_SSSE3, ARGBToUVJRow_SSSE3, 0, 4, 15)
+ANY12S(BGRAToUVRow_Any_SSSE3, BGRAToUVRow_SSSE3, 0, 4, 15)
+ANY12S(ABGRToUVRow_Any_SSSE3, ABGRToUVRow_SSSE3, 0, 4, 15)
+ANY12S(RGBAToUVRow_Any_SSSE3, RGBAToUVRow_SSSE3, 0, 4, 15)
#endif
-#ifdef HAS_INTERPOLATEROW_MIPS_DSPR2
-NANY(InterpolateRow_Any_MIPS_DSPR2, InterpolateRow_MIPS_DSPR2, InterpolateRow_C,
- 1, 1, 3)
+#ifdef HAS_YUY2TOUVROW_AVX2
+ANY12S(YUY2ToUVRow_Any_AVX2, YUY2ToUVRow_AVX2, 1, 4, 31)
+ANY12S(UYVYToUVRow_Any_AVX2, UYVYToUVRow_AVX2, 1, 4, 31)
#endif
-#undef NANY
-
-#define MANY(NAMEANY, MIRROR_SIMD, MIRROR_C, BPP, MASK) \
- void NAMEANY(const uint8* src_y, uint8* dst_y, int width) { \
- int n = width & ~MASK; \
- int r = width & MASK; \
- if (n > 0) { \
- MIRROR_SIMD(src_y, dst_y + r * BPP, n); \
- } \
- MIRROR_C(src_y + n * BPP, dst_y, r); \
- }
-
-#ifdef HAS_MIRRORROW_AVX2
-MANY(MirrorRow_Any_AVX2, MirrorRow_AVX2, MirrorRow_C, 1, 31)
+#ifdef HAS_YUY2TOUVROW_SSE2
+ANY12S(YUY2ToUVRow_Any_SSE2, YUY2ToUVRow_SSE2, 1, 4, 15)
+ANY12S(UYVYToUVRow_Any_SSE2, UYVYToUVRow_SSE2, 1, 4, 15)
#endif
-#ifdef HAS_MIRRORROW_SSSE3
-MANY(MirrorRow_Any_SSSE3, MirrorRow_SSSE3, MirrorRow_C, 1, 15)
+#ifdef HAS_ARGBTOUVROW_NEON
+ANY12S(ARGBToUVRow_Any_NEON, ARGBToUVRow_NEON, 0, 4, 15)
#endif
-#ifdef HAS_MIRRORROW_SSE2
-MANY(MirrorRow_Any_SSE2, MirrorRow_SSE2, MirrorRow_C, 1, 15)
+#ifdef HAS_ARGBTOUVJROW_NEON
+ANY12S(ARGBToUVJRow_Any_NEON, ARGBToUVJRow_NEON, 0, 4, 15)
#endif
-#ifdef HAS_MIRRORROW_NEON
-MANY(MirrorRow_Any_NEON, MirrorRow_NEON, MirrorRow_C, 1, 15)
+#ifdef HAS_BGRATOUVROW_NEON
+ANY12S(BGRAToUVRow_Any_NEON, BGRAToUVRow_NEON, 0, 4, 15)
#endif
-#ifdef HAS_ARGBMIRRORROW_AVX2
-MANY(ARGBMirrorRow_Any_AVX2, ARGBMirrorRow_AVX2, ARGBMirrorRow_C, 4, 7)
+#ifdef HAS_ABGRTOUVROW_NEON
+ANY12S(ABGRToUVRow_Any_NEON, ABGRToUVRow_NEON, 0, 4, 15)
#endif
-#ifdef HAS_ARGBMIRRORROW_SSE2
-MANY(ARGBMirrorRow_Any_SSE2, ARGBMirrorRow_SSE2, ARGBMirrorRow_C, 4, 3)
+#ifdef HAS_RGBATOUVROW_NEON
+ANY12S(RGBAToUVRow_Any_NEON, RGBAToUVRow_NEON, 0, 4, 15)
#endif
-#ifdef HAS_ARGBMIRRORROW_NEON
-MANY(ARGBMirrorRow_Any_NEON, ARGBMirrorRow_NEON, ARGBMirrorRow_C, 4, 3)
+#ifdef HAS_RGB24TOUVROW_NEON
+ANY12S(RGB24ToUVRow_Any_NEON, RGB24ToUVRow_NEON, 0, 3, 15)
#endif
-#undef MANY
-
-#define MANY(NAMEANY, COPY_SIMD, COPY_C, BPP, MASK) \
- void NAMEANY(const uint8* src_y, uint8* dst_y, int width) { \
- int n = width & ~MASK; \
- int r = width & MASK; \
- if (n > 0) { \
- COPY_SIMD(src_y, dst_y, n); \
- } \
- COPY_C(src_y + n * BPP, dst_y + n * BPP, r); \
- }
-
-#ifdef HAS_COPYROW_AVX
-MANY(CopyRow_Any_AVX, CopyRow_AVX, CopyRow_C, 1, 63)
+#ifdef HAS_RAWTOUVROW_NEON
+ANY12S(RAWToUVRow_Any_NEON, RAWToUVRow_NEON, 0, 3, 15)
#endif
-#ifdef HAS_COPYROW_SSE2
-MANY(CopyRow_Any_SSE2, CopyRow_SSE2, CopyRow_C, 1, 31)
+#ifdef HAS_RGB565TOUVROW_NEON
+ANY12S(RGB565ToUVRow_Any_NEON, RGB565ToUVRow_NEON, 0, 2, 15)
#endif
-#ifdef HAS_COPYROW_NEON
-MANY(CopyRow_Any_NEON, CopyRow_NEON, CopyRow_C, 1, 31)
+#ifdef HAS_ARGB1555TOUVROW_NEON
+ANY12S(ARGB1555ToUVRow_Any_NEON, ARGB1555ToUVRow_NEON, 0, 2, 15)
#endif
-#undef MANY
-
-#define SETANY(NAMEANY, SET_SIMD, SET_C, T, BPP, MASK) \
- void NAMEANY(uint8* dst_y, T v8, int width) { \
- int n = width & ~MASK; \
- int r = width & MASK; \
- if (n > 0) { \
- SET_SIMD(dst_y, v8, n); \
- } \
- SET_C(dst_y + n * BPP, v8, r); \
- }
-
-#ifdef HAS_SETROW_X86
-SETANY(SetRow_Any_X86, SetRow_X86, SetRow_ERMS, uint8, 1, 3)
+#ifdef HAS_ARGB4444TOUVROW_NEON
+ANY12S(ARGB4444ToUVRow_Any_NEON, ARGB4444ToUVRow_NEON, 0, 2, 15)
#endif
-#ifdef HAS_SETROW_NEON
-SETANY(SetRow_Any_NEON, SetRow_NEON, SetRow_C, uint8, 1, 15)
+#ifdef HAS_YUY2TOUVROW_NEON
+ANY12S(YUY2ToUVRow_Any_NEON, YUY2ToUVRow_NEON, 1, 4, 15)
#endif
-#ifdef HAS_ARGBSETROW_NEON
-SETANY(ARGBSetRow_Any_NEON, ARGBSetRow_NEON, ARGBSetRow_C, uint32, 4, 3)
+#ifdef HAS_UYVYTOUVROW_NEON
+ANY12S(UYVYToUVRow_Any_NEON, UYVYToUVRow_NEON, 1, 4, 15)
#endif
-#undef SETANY
+#undef ANY12S
#ifdef __cplusplus
} // extern "C"
« no previous file with comments | « source/libvpx/third_party/libyuv/source/rotate_win.cc ('k') | source/libvpx/third_party/libyuv/source/row_common.cc » ('j') | no next file with comments »

Powered by Google App Engine
This is Rietveld 408576698