mirror of
https://chromium.googlesource.com/libyuv/libyuv
synced 2026-11-09 14:49:57 +08:00
d2a6496353
Follow up on the LoongArch *_la.cc consolidation with runtime CPU detection, build/test disable flags, full LASX parity for all LSX functions, and 38 new LSX/LASX convert and planar row functions: - Update LoongArchCpuCaps() in cpu_id.cc to detect LSX and LASX via the unprivileged cpucfg instruction combined with getauxval(AT_HWCAP) OS support, and wire GetCpuFlags() for __loongarch__. - Support LIBYUV_DISABLE_LSX, LIBYUV_DISABLE_LASX, and LIBYUV_DISABLE_ASM at build time in cpu_support.h and CMakeLists.txt, and fix the __loongarch__ environment variable check in unit_test.cc. - Port all 24 LSX-only Rotate and Scale kernels to 256-bit LASX so every LSX function has a matching LASX implementation. - Move ScaleMaxSamples, ScaleSumSamples, and ScaleSamples from row_* to scale_* across all platforms and add LSX and LASX implementations in scale_la.cc. - Add 38 LSX and LASX row implementations (and _Any wrappers): SwapUVRow, ARGBCopyYToAlphaRow, HalfMergeUVRow, Convert16To8Row, Convert8To16Row, MultiplyRow_16, DivideRow_16, MergeUVRow_16, SplitUVRow_16, CopyRow_16, HalfRow, HalfRow_16, ARGBCopyAlphaRow, YUY2ToNVUVRow, AR64ToARGBRow, AB64ToARGBRow, ARGBToAR64Row, ARGBToAB64Row, Convert8To8Row, HalfRow_16To8, AR30ToAB30Row, ARGBToAR30Row, ABGRToAR30Row, AR30ToARGBRow, AR30ToABGRRow, AYUVToYRow, AYUVToUVRow, AYUVToVURow, DetileRow, DetileRow_16, MergeARGBRow, MergeXRGBRow, SplitARGBRow, SplitXRGBRow, RAWToRGBARow, RGB24MirrorRow, BlendPlaneRow, and InterpolateRow_16. Compare SIMD ISA per *_C: ------------------------------------------------------------------------ C Function x86 LoongArch ------------------------------------------------------------------------ ARGBDetectRow_C AVX2 LSX+LASX HammingDistance_C SSSE3+AVX2+AVX512BW+AVX512VNNI LSX+LASX HashDjb2_C SSE41+AVX2+AVX512BW LSX+LASX Ssim4x4Row_C SSE2+AVX2+AVX512BW LSX+LASX SsimWindowRow_C SSE2+AVX2+AVX512BW LSX+LASX SumSquareError_C SSE2+AVX2+AVX512BW+AVX512VNNI LSX+LASX ------------------------------------------------------------------------ Compare: 6 (100%) 6 (100%) 6 (100%) ------------------------------------------------------------------------ Convert SIMD ISA per *_C: ------------------------------------------------------------------------ C Function x86 LoongArch ------------------------------------------------------------------------ AB64ToARGBRow_C SSSE3+AVX2 LSX+LASX ABGRToAR30Row_C SSSE3+AVX2+AVX512BW LSX+LASX AR30ToAB30Row_C AVX2+AVX512BW LSX+LASX AR30ToABGRRow_C AVX2+AVX512BW LSX+LASX AR30ToARGBRow_C AVX2+AVX512BW LSX+LASX AR64ShuffleRow_C SSSE3+AVX2+AVX512BW LSX+LASX AR64ToAB64Row_C SSSE3+AVX2+AVX512BW LSX+LASX AR64ToARGBRow_C SSSE3+AVX2 LSX+LASX ARGB1555ToARGBRow_C AVX2+AVX512BW LSX+LASX ARGB1555ToUVMatrixRow_C AVX2 ARGB1555ToUVRow_C AVX2 LSX+LASX ARGB1555ToYMatrixRow_C AVX2 ARGB1555ToYRow_C AVX2 LSX+LASX ARGB4444ToARGBRow_C AVX2 LSX+LASX ARGB4444ToUVMatrixRow_C AVX2 ARGB4444ToUVRow_C AVX2 ARGB4444ToYMatrixRow_C AVX2 ARGB4444ToYRow_C AVX2 ARGBAddRow_C SSE2+AVX2 LSX+LASX ARGBAffineRow_C AVX2 ARGBAlphaBlendRow_C SSSE3+AVX2+AVX512BW LSX+LASX ARGBAttenuateRow_C SSSE3+AVX2+AVX512BW LSX+LASX ARGBBlendRow_C AVX2+AVX512BW LSX+LASX ARGBColorMatrixRow_C AVX2+AVX512VNNI LSX+LASX ARGBColorTableRow_C AVX2 ARGBCopyAlphaRow_C SSE2+AVX2 LSX+LASX ARGBCopyYToAlphaRow_C SSE2+AVX2+AVX512BW LSX+LASX ARGBExtractAlphaRow_C SSE2+AVX2+AVX512BW LSX+LASX ARGBGrayRow_C AVX2+AVX512BW+AVX512VNNI LSX+LASX ARGBLumaColorTableRow_C AVX2 ARGBMirrorRow_C SSE2+AVX2+AVX512BW LSX+LASX ARGBMultiplyRow_C SSE2+AVX2+AVX512BW LSX+LASX ARGBOrRow_C SSE2+AVX2+AVX512BW LSX+LASX ARGBPolynomialRow_C SSE2+AVX2 ARGBQuantizeRow_C AVX2 LSX+LASX ARGBSepiaRow_C AVX2+AVX512VNNI LSX+LASX ARGBSetRow_C AVX2 LSX+LASX ARGBShadeRow_C AVX2 LSX+LASX ARGBShuffleRow_C SSSE3+AVX2+AVX512BW LSX+LASX ARGBSubtractRow_C SSE2+AVX2 LSX+LASX ARGBToAB64Row_C SSSE3+AVX2 LSX+LASX ARGBToABGRRow_C SSSE3+AVX2+AVX512BW LSX+LASX ARGBToAR30Row_C SSSE3+AVX2+AVX512BW LSX+LASX ARGBToAR64Row_C SSSE3+AVX2 LSX+LASX ARGBToARGB1555Row_C AVX2 LSX+LASX ARGBToARGB4444Row_C AVX2 LSX+LASX ARGBToBGRARow_C SSSE3+AVX2+AVX512BW LSX+LASX ARGBToRAWRow_C SSSE3+AVX2+AVX512BW LSX+LASX ARGBToRGB24Row_C SSSE3+AVX2+AVX512BW LSX+LASX ARGBToRGB565DitherRow_C AVX2+AVX512BW LSX+LASX ARGBToRGB565Row_C AVX2+AVX512BW LSX+LASX ARGBToRGBARow_C SSSE3+AVX2+AVX512BW LSX+LASX ARGBToUV444MatrixRow_C SSSE3+AVX2+AVX512BW+AVX512VNNI+AVX512VNNIVBMI ARGBToUVMatrixRow_C SSSE3+AVX2+AVX512BW+AVX512VNNI+AVX512VNNIVBMI ARGBToYMatrixRow_C SSSE3+AVX2+AVX512BW+AVX512VNNI+AVX512VNNIVBMI LSX+LASX ARGBUnattenuateRow_C SSE2+AVX2 AYUVToUVRow_C AVX2+AVX512VBMI LSX+LASX AYUVToVURow_C AVX2+AVX512VBMI LSX+LASX AYUVToYRow_C AVX2+AVX512VBMI LSX+LASX BlendPlaneRow_C SSSE3+AVX2+AVX512BW LSX+LASX ByteToFloatRow_C AVX2 ComputeCumulativeSumRow_C AVX2 Convert16To8Row_C SSSE3+AVX2+AVX512BW LSX+LASX Convert8To16Row_C SSE2+AVX2+AVX512BW LSX+LASX Convert8To8Row_C AVX2+AVX512BW LSX+LASX CopyRow_16_C AVX2 LSX+LASX CopyRow_C ERMS+SSE2+AVX+AVX512BW LSX+LASX CumulativeSumToAverageRow_C AVX2 DetileRow_16_C AVX2 LSX+LASX DetileRow_C AVX2 LSX+LASX DetileSplitUVRow_C AVX2 DetileToYUY2_C AVX2 DivideRow_16_C AVX2+AVX512BW LSX+LASX GaussCol_C AVX2 GaussCol_F32_C AVX2 GaussRow_C AVX2 GaussRow_F32_C AVX2 HalfFloatRow_C F16C+AVX2+AVX512BW LSX+LASX HalfMergeUVRow_C SSSE3+AVX2+AVX512BW LSX+LASX HalfRow_16To8_C SSSE3+AVX2+AVX512BW LSX+LASX HalfRow_16_C AVX2 LSX+LASX HalfRow_C AVX2 LSX+LASX HalfWidthRow_16To8_C SSSE3+AVX2+AVX512BW HalfWidthRow_16To8_Odd_C AVX2 I210AlphaToARGBRow_C SSSE3+AVX2+AVX512BW I210ToAR30Row_C SSSE3+AVX2+AVX512BW I210ToARGBRow_C SSSE3+AVX2+AVX512BW I212ToAR30Row_C SSSE3+AVX2+AVX512BW I212ToARGBRow_C SSSE3+AVX2+AVX512BW I400ToARGBRow_C SSE2+AVX2+AVX512BW LSX+LASX I410AlphaToARGBRow_C SSSE3+AVX2+AVX512BW I410ToAR30Row_C SSSE3+AVX2+AVX512BW I410ToARGBRow_C SSSE3+AVX2+AVX512BW I422AlphaToARGBRow_C SSSE3+AVX2+AVX512BW LSX+LASX I422ToAR30Row_C SSSE3+AVX2+AVX512BW I422ToARGB1555Row_C SSSE3+AVX2+AVX512BW LSX+LASX I422ToARGB4444Row_C SSSE3+AVX2+AVX512BW LSX+LASX I422ToARGBRow_C SSSE3+AVX2+AVX512BW LSX+LASX I422ToRGB24Row_C SSSE3+AVX2+AVX512BW+AVX512VBMI LSX+LASX I422ToRGB565Row_C SSSE3+AVX2+AVX512BW LSX+LASX I422ToRGBARow_C SSSE3+AVX2+AVX512BW LSX+LASX I422ToUYVYRow_C SSE2+AVX2 LSX+LASX I422ToYUY2Row_C SSE2+AVX2 LSX+LASX I444AlphaToARGBRow_C SSSE3+AVX2+AVX512BW I444ToARGBRow_C SSSE3+AVX2+AVX512BW LSX+LASX I444ToRGB24Row_C SSSE3+AVX2+AVX512VBMI InterpolateRow_16_C AVX2+AVX512BW LSX+LASX InterpolateRow_C AVX2+AVX512BW LSX+LASX J400ToARGBRow_C AVX2+AVX512BW LSX+LASX MergeAR64Row_C AVX2 MergeARGB16To8Row_C AVX2 MergeARGBRow_C SSE2+AVX2 LSX+LASX MergeRGBRow_C AVX2+AVX512VBMI MergeUVRow_16_C AVX2+AVX512BW LSX+LASX MergeUVRow_C SSE2+AVX2+AVX512BW LSX+LASX MergeXR30Row_C AVX2 MergeXR64Row_C AVX2 MergeXRGB16To8Row_C AVX2 MergeXRGBRow_C SSE2+AVX2 LSX+LASX MirrorRow_16_C SSSE3+AVX2+AVX512BW LSX+LASX MirrorRow_C SSSE3+AVX2+AVX512BW LSX+LASX MirrorSplitUVRow_C SSSE3+AVX2+AVX512BW LSX+LASX MirrorUVRow_C SSSE3+AVX2+AVX512BW LSX+LASX MultiplyRow_16_C AVX2+AVX512BW LSX+LASX NV12ToARGBRow_C SSSE3+AVX2+AVX512BW LSX+LASX NV12ToRGB24Row_C SSSE3+AVX2+AVX512VBMI NV12ToRGB565Row_C SSSE3+AVX2 LSX+LASX NV21ToARGBRow_C SSSE3+AVX2+AVX512BW LSX+LASX NV21ToRGB24Row_C SSSE3+AVX2+AVX512VBMI NV21ToYUV24Row_C SSSE3+AVX2+AVX512VBMI P210ToAR30Row_C SSSE3+AVX2+AVX512BW P210ToARGBRow_C SSSE3+AVX2+AVX512BW P410ToAR30Row_C SSSE3+AVX2+AVX512BW P410ToARGBRow_C SSSE3+AVX2+AVX512BW RAWToARGBRow_C SSSE3+AVX2+AVX512BW LSX+LASX RAWToRGB24Row_C SSSE3+AVX2+AVX512BW+AVX512VBMI LSX+LASX RAWToRGBARow_C AVX2+AVX512VBMI LSX+LASX RAWToUVRow LSX+LASX RGB24MirrorRow_C SSSE3+AVX2+AVX512VBMI LSX+LASX RGB24ToARGBRow_C SSSE3+AVX2+AVX512BW LSX+LASX RGB24ToUVRow LSX+LASX RGB565ToARGBRow_C AVX2 LSX+LASX RGB565ToUVMatrixRow_C AVX2 RGB565ToUVRow_C AVX2 LSX+LASX RGB565ToYMatrixRow_C AVX2 RGB565ToYRow_C AVX2 LSX+LASX RGBAToARGBRow_C SSSE3+AVX2+AVX512BW LSX+LASX RGBColorTableRow_C AVX2 RGBToUV444MatrixRow_C SSSE3+AVX2+AVX512BW RGBToUVMatrixRow_C SSSE3+AVX2+AVX512BW RGBToYMatrixRow_C SSSE3+AVX2+AVX512BW LSX+LASX SetRow_C ERMS+AVX2 LSX+LASX SobelRow_C AVX2 LSX+LASX SobelToPlaneRow_C AVX2 LSX+LASX SobelXRow_C AVX2 SobelXYRow_C AVX2 LSX+LASX SobelYRow_C AVX2 SplitARGBRow_C SSE2+SSSE3+AVX2 LSX+LASX SplitRGBRow_C SSSE3+SSE41+AVX2 SplitUVRow_16_C AVX2 LSX+LASX SplitUVRow_C SSE2+AVX2+AVX512BW LSX+LASX SplitXRGBRow_C SSE2+SSSE3+AVX2 LSX+LASX SwapUVRow_C SSSE3+AVX2+AVX512BW LSX+LASX UYVYToARGBRow_C SSSE3+AVX2+AVX512BW LSX+LASX UYVYToUV422Row_C SSE2+AVX2+AVX512BW LSX+LASX UYVYToUVRow_C SSE2+AVX2+AVX512BW LSX+LASX UYVYToYRow_C SSE2+AVX2+AVX512BW LSX+LASX UnpackMT2T_C AVX2 YUY2ToARGBRow_C SSSE3+AVX2+AVX512BW LSX+LASX YUY2ToNVUVRow_C SSE2+AVX2 LSX+LASX YUY2ToUV422Row_C SSE2+AVX2+AVX512BW LSX+LASX YUY2ToUVRow_C SSE2+AVX2+AVX512BW LSX+LASX YUY2ToYRow_C SSE2+AVX2+AVX512BW LSX+LASX ------------------------------------------------------------------------ Convert: 173 (100%) 171 (99%) 116 (67%) ------------------------------------------------------------------------ Rotate SIMD ISA per *_C: ------------------------------------------------------------------------ C Function x86 LoongArch ------------------------------------------------------------------------ TransposeUVWx16_C AVX2+AVX512BW LSX+LASX TransposeUVWx8_C AVX2 LSX+LASX TransposeWx16_C AVX2+AVX512BW LSX+LASX TransposeWx4_32_C AVX2 LSX+LASX TransposeWx8_16_C AVX2 LSX+LASX TransposeWx8_C AVX2 LSX+LASX TransposeWxH_32_C AVX2 ------------------------------------------------------------------------ Rotate: 7 (100%) 7 (100%) 6 (86%) ------------------------------------------------------------------------ Scale SIMD ISA per *_C: ------------------------------------------------------------------------ C Function x86 LoongArch ------------------------------------------------------------------------ FixedDiv1_C X86 FixedDiv_C X86 ScaleARGBCols64_C ScaleARGBColsUp2_C AVX2 ScaleARGBCols_C AVX2 LSX+LASX ScaleARGBFilterCols64_C ScaleARGBFilterCols_C AVX2 LSX+LASX ScaleARGBRowDown2Box_C AVX2 LSX+LASX ScaleARGBRowDown2Linear_C AVX2 LSX+LASX ScaleARGBRowDown2_C AVX2 LSX+LASX ScaleARGBRowDownEvenBox_C AVX2 LSX+LASX ScaleARGBRowDownEven_C AVX2 LSX+LASX ScaleAddCols0_C ScaleAddCols1_16_C ScaleAddCols1_C AVX512BW ScaleAddCols2_16_C ScaleAddCols2_C AVX512BW ScaleAddRow_16_C ScaleAddRow_C SSE2+AVX2+AVX512BW LSX+LASX ScaleColsUp2_16_C ScaleColsUp2_C AVX2 ScaleCols_16_C ScaleCols_C ScaleFilterCols64_16_C ScaleFilterCols64_C ScaleFilterCols_16_C ScaleFilterCols_C AVX2 LSX+LASX ScaleMaxSamples_C AVX2 LSX+LASX ScaleRowDown2Box_16_C AVX2 ScaleRowDown2Box_C SSSE3+AVX2+AVX512BW LSX+LASX ScaleRowDown2Box_Odd_C SSSE3+AVX2+AVX512BW ScaleRowDown2Linear_16_C AVX2 ScaleRowDown2Linear_C SSSE3+AVX2+AVX512BW LSX+LASX ScaleRowDown2_16_C AVX2 ScaleRowDown2_C SSSE3+AVX2+AVX512BW LSX+LASX ScaleRowDown34_0_Box_16_C ScaleRowDown34_0_Box_C AVX2 LSX+LASX ScaleRowDown34_16_C ScaleRowDown34_1_Box_16_C ScaleRowDown34_1_Box_C AVX2 LSX+LASX ScaleRowDown34_C AVX2 LSX+LASX ScaleRowDown38_16_C ScaleRowDown38_2_Box_16_C ScaleRowDown38_2_Box_C AVX2 LSX+LASX ScaleRowDown38_3_Box_16_C ScaleRowDown38_3_Box_C AVX2 LSX+LASX ScaleRowDown38_C AVX2 LSX+LASX ScaleRowDown4Box_16_C ScaleRowDown4Box_C SSSE3+AVX2 LSX+LASX ScaleRowDown4_16_C ScaleRowDown4_C SSSE3+AVX2 LSX+LASX ScaleRowUp2_Bilinear_16_Any_C SSE2+AVX2 ScaleRowUp2_Bilinear_16_C SSE2+AVX2 ScaleRowUp2_Bilinear_Any_C SSE2+SSSE3+AVX2 ScaleRowUp2_Bilinear_C SSE2+SSSE3+AVX2 ScaleRowUp2_Linear_16_Any_C SSE2+AVX2 ScaleRowUp2_Linear_16_C SSE2+AVX2 ScaleRowUp2_Linear_Any_C SSE2+SSSE3+AVX2 ScaleRowUp2_Linear_C SSE2+SSSE3+AVX2 ScaleSamples_C AVX2 LSX+LASX ScaleSumSamples_C AVX2 LSX+LASX ScaleUVAddCols1_16_C ScaleUVAddCols1_C ScaleUVAddCols2_16_C ScaleUVAddCols2_C ScaleUVAddCols3_16_C ScaleUVAddCols3_C AVX2+AVX512VBMI LSX+LASX ScaleUVCols64_16_C ScaleUVCols64_C ScaleUVColsUp2_16_C ScaleUVColsUp2_C ScaleUVCols_16_C ScaleUVCols_C ScaleUVFilterCols64_16_C ScaleUVFilterCols64_C ScaleUVFilterCols_16_C ScaleUVFilterCols_C ScaleUVRowDown2Box_16_C ScaleUVRowDown2Box_C AVX2 ScaleUVRowDown2Linear_16_C ScaleUVRowDown2Linear_C AVX2 ScaleUVRowDown2_16_C ScaleUVRowDown2_C AVX2 ScaleUVRowDownEvenBox_16_C ScaleUVRowDownEvenBox_C ScaleUVRowDownEven_16_C ScaleUVRowDownEven_C AVX2 ScaleUVRowUp2_Bilinear_16_Any_C SSE41+AVX2 ScaleUVRowUp2_Bilinear_16_C SSE41+AVX2 ScaleUVRowUp2_Bilinear_Any_C SSSE3+AVX2 ScaleUVRowUp2_Bilinear_C SSSE3+AVX2 ScaleUVRowUp2_Linear_16_Any_C SSE41+AVX2 ScaleUVRowUp2_Linear_16_C SSE41+AVX2 ScaleUVRowUp2_Linear_Any_C SSSE3+AVX2 ScaleUVRowUp2_Linear_C SSSE3+AVX2 ------------------------------------------------------------------------ Scale: 95 (100%) 54 (57%) 24 (25%) ------------------------------------------------------------------------ Bug: 42280942 Change-Id: I663f277db5e8a6a73319df7e3153f9f78a147a7a Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/8718861 Reviewed-by: richard winterton <rrwinterton@gmail.com> Commit-Queue: Frank Barchard <fbarchard@google.com>
3519 lines
145 KiB
C++
3519 lines
145 KiB
C++
/*
|
|
* Copyright 2012 The LibYuv Project Authors. All rights reserved.
|
|
*
|
|
* Use of this source code is governed by a BSD-style license
|
|
* that can be found in the LICENSE file in the root of the source
|
|
* tree. An additional intellectual property rights grant can be found
|
|
* in the file PATENTS. All contributing project authors may
|
|
* be found in the AUTHORS file in the root of the source tree.
|
|
*/
|
|
|
|
#include "libyuv/row.h"
|
|
|
|
#include <assert.h>
|
|
#include <string.h> // For memset.
|
|
|
|
#ifdef __cplusplus
|
|
namespace libyuv {
|
|
extern "C" {
|
|
#endif
|
|
|
|
// Subsampled source needs to be increase by 1 if not even.
|
|
#define SS(width, shift) (((width) + (1 << (shift)) - 1) >> (shift))
|
|
|
|
#ifdef LIBYUV_MSAN
|
|
#define MSAN_MEMSET(a, b, c) memset(a, b, c)
|
|
#else
|
|
#define MSAN_MEMSET(a, b, c)
|
|
#endif
|
|
|
|
// Any 4 planes to 1
|
|
#define ANY41(NAMEANY, ANY_SIMD, UVSHIFT, DUVSHIFT, BPP, MASK) \
|
|
void NAMEANY(const uint8_t* y_buf, const uint8_t* u_buf, \
|
|
const uint8_t* v_buf, const uint8_t* a_buf, uint8_t* dst_ptr, \
|
|
int width) { \
|
|
SIMD_ALIGNED(uint8_t vin[64 * 4]); \
|
|
static_assert(MASK + 1 <= sizeof(vin) / 4, "vin buffer too small"); \
|
|
SIMD_ALIGNED(uint8_t vout[64]); \
|
|
static_assert(SS(MASK + 1, DUVSHIFT) * BPP <= sizeof(vout), \
|
|
"vout buffer too small"); \
|
|
MSAN_MEMSET(vin, 0, sizeof(vin)); /* for msan */ \
|
|
MSAN_MEMSET(vout, 0, sizeof(vout)); /* for msan */ \
|
|
int r = width & MASK; \
|
|
int n = width & ~MASK; \
|
|
if (n > 0) { \
|
|
ANY_SIMD(y_buf, u_buf, v_buf, a_buf, dst_ptr, n); \
|
|
} \
|
|
ptrdiff_t np = n; \
|
|
memcpy(vin, y_buf + np, r); \
|
|
memcpy(vin + 64, u_buf + (np >> UVSHIFT), SS(r, UVSHIFT)); \
|
|
memcpy(vin + 128, v_buf + (np >> UVSHIFT), SS(r, UVSHIFT)); \
|
|
memcpy(vin + 192, a_buf + np, r); \
|
|
ANY_SIMD(vin, vin + 64, vin + 128, vin + 192, vout, MASK + 1); \
|
|
memcpy(dst_ptr + (np >> DUVSHIFT) * BPP, vout, SS(r, DUVSHIFT) * BPP); \
|
|
}
|
|
|
|
#ifdef HAS_MERGEARGBROW_SSE2
|
|
ANY41(MergeARGBRow_Any_SSE2, MergeARGBRow_SSE2, 0, 0, 4, 7)
|
|
#endif
|
|
#ifdef HAS_MERGEARGBROW_AVX2
|
|
ANY41(MergeARGBRow_Any_AVX2, MergeARGBRow_AVX2, 0, 0, 4, 15)
|
|
#endif
|
|
#ifdef HAS_MERGEARGBROW_NEON
|
|
ANY41(MergeARGBRow_Any_NEON, MergeARGBRow_NEON, 0, 0, 4, 15)
|
|
#endif
|
|
#ifdef HAS_MERGEARGBROW_LSX
|
|
ANY41(MergeARGBRow_Any_LSX, MergeARGBRow_LSX, 0, 0, 4, 15)
|
|
#endif
|
|
#ifdef HAS_MERGEARGBROW_LASX
|
|
ANY41(MergeARGBRow_Any_LASX, MergeARGBRow_LASX, 0, 0, 4, 15)
|
|
#endif
|
|
|
|
// Note that odd width replication includes 444 due to implementation
|
|
// on arm that subsamples 444 to 422 internally.
|
|
// Any 4 planes to 1 with yuvconstants
|
|
#define ANY41C(NAMEANY, ANY_SIMD, UVSHIFT, DUVSHIFT, BPP, MASK) \
|
|
void NAMEANY(const uint8_t* y_buf, const uint8_t* u_buf, \
|
|
const uint8_t* v_buf, const uint8_t* a_buf, uint8_t* dst_ptr, \
|
|
const struct YuvConstants* yuvconstants, int width) { \
|
|
SIMD_ALIGNED(uint8_t vin[64 * 4]); \
|
|
static_assert(MASK + 1 <= sizeof(vin) / 4, "vin buffer too small"); \
|
|
SIMD_ALIGNED(uint8_t vout[64]); \
|
|
static_assert(SS(MASK + 1, DUVSHIFT) * BPP <= sizeof(vout), \
|
|
"vout buffer too small"); \
|
|
MSAN_MEMSET(vin, 0, sizeof(vin)); /* for msan */ \
|
|
MSAN_MEMSET(vout, 0, sizeof(vout)); /* for msan */ \
|
|
int r = width & MASK; \
|
|
int n = width & ~MASK; \
|
|
if (n > 0) { \
|
|
ANY_SIMD(y_buf, u_buf, v_buf, a_buf, dst_ptr, yuvconstants, n); \
|
|
} \
|
|
ptrdiff_t np = n; \
|
|
memcpy(vin, y_buf + np, r); \
|
|
memcpy(vin + 64, u_buf + (np >> UVSHIFT), SS(r, UVSHIFT)); \
|
|
memcpy(vin + 128, v_buf + (np >> UVSHIFT), SS(r, UVSHIFT)); \
|
|
memcpy(vin + 192, a_buf + np, r); \
|
|
if (width & 1) { \
|
|
vin[64 + SS(r, UVSHIFT)] = vin[64 + SS(r, UVSHIFT) - 1]; \
|
|
vin[128 + SS(r, UVSHIFT)] = vin[128 + SS(r, UVSHIFT) - 1]; \
|
|
} \
|
|
ANY_SIMD(vin, vin + 64, vin + 128, vin + 192, vout, yuvconstants, \
|
|
MASK + 1); \
|
|
memcpy(dst_ptr + (np >> DUVSHIFT) * BPP, vout, SS(r, DUVSHIFT) * BPP); \
|
|
}
|
|
|
|
#ifdef HAS_I444ALPHATOARGBROW_SSSE3
|
|
ANY41C(I444AlphaToARGBRow_Any_SSSE3, I444AlphaToARGBRow_SSSE3, 0, 0, 4, 7)
|
|
#endif
|
|
#ifdef HAS_I444ALPHATOARGBROW_AVX2
|
|
ANY41C(I444AlphaToARGBRow_Any_AVX2, I444AlphaToARGBRow_AVX2, 0, 0, 4, 15)
|
|
#endif
|
|
#ifdef HAS_I422ALPHATOARGBROW_SSSE3
|
|
ANY41C(I422AlphaToARGBRow_Any_SSSE3, I422AlphaToARGBRow_SSSE3, 1, 0, 4, 7)
|
|
#endif
|
|
#ifdef HAS_I422ALPHATOARGBROW_AVX2
|
|
ANY41C(I422AlphaToARGBRow_Any_AVX2, I422AlphaToARGBRow_AVX2, 1, 0, 4, 15)
|
|
#endif
|
|
#ifdef HAS_I444ALPHATOARGBROW_NEON
|
|
ANY41C(I444AlphaToARGBRow_Any_NEON, I444AlphaToARGBRow_NEON, 0, 0, 4, 7)
|
|
#endif
|
|
#ifdef HAS_I422ALPHATOARGBROW_NEON
|
|
ANY41C(I422AlphaToARGBRow_Any_NEON, I422AlphaToARGBRow_NEON, 1, 0, 4, 7)
|
|
#endif
|
|
#ifdef HAS_I422ALPHATOARGBROW_LSX
|
|
ANY41C(I422AlphaToARGBRow_Any_LSX, I422AlphaToARGBRow_LSX, 1, 0, 4, 7)
|
|
#endif
|
|
#ifdef HAS_I422ALPHATOARGBROW_LASX
|
|
ANY41C(I422AlphaToARGBRow_Any_LASX, I422AlphaToARGBRow_LASX, 1, 0, 4, 15)
|
|
#endif
|
|
#undef ANY41C
|
|
|
|
// Any 4 planes to 1 plane of 8 bit with yuvconstants
|
|
#define ANY41CT(NAMEANY, ANY_SIMD, UVSHIFT, DUVSHIFT, T, SBPP, BPP, MASK) \
|
|
void NAMEANY(const T* y_buf, const T* u_buf, const T* v_buf, const T* a_buf, \
|
|
uint8_t* dst_ptr, const struct YuvConstants* yuvconstants, \
|
|
int width) { \
|
|
SIMD_ALIGNED(T vin[16 * 4]); \
|
|
static_assert((MASK + 1) * SBPP <= sizeof(vin) / 4, \
|
|
"vin buffer too small"); \
|
|
SIMD_ALIGNED(uint8_t vout[64]); \
|
|
static_assert(SS(MASK + 1, DUVSHIFT) * BPP <= sizeof(vout), \
|
|
"vout buffer too small"); \
|
|
MSAN_MEMSET(vin, 0, sizeof(vin)); /* for YUY2 and msan */ \
|
|
MSAN_MEMSET(vout, 0, sizeof(vout)); /* for msan */ \
|
|
int r = width & MASK; \
|
|
int n = width & ~MASK; \
|
|
if (n > 0) { \
|
|
ANY_SIMD(y_buf, u_buf, v_buf, a_buf, dst_ptr, yuvconstants, n); \
|
|
} \
|
|
ptrdiff_t np = n; \
|
|
memcpy(vin, y_buf + np, r * SBPP); \
|
|
memcpy(vin + 16, u_buf + (np >> UVSHIFT), SS(r, UVSHIFT) * SBPP); \
|
|
memcpy(vin + 32, v_buf + (np >> UVSHIFT), SS(r, UVSHIFT) * SBPP); \
|
|
memcpy(vin + 48, a_buf + np, r * SBPP); \
|
|
ANY_SIMD(vin, vin + 16, vin + 32, vin + 48, vout, yuvconstants, MASK + 1); \
|
|
memcpy(dst_ptr + (np >> DUVSHIFT) * BPP, vout, SS(r, DUVSHIFT) * BPP); \
|
|
}
|
|
|
|
#ifdef HAS_I210ALPHATOARGBROW_NEON
|
|
ANY41CT(I210AlphaToARGBRow_Any_NEON,
|
|
I210AlphaToARGBRow_NEON,
|
|
1,
|
|
0,
|
|
uint16_t,
|
|
2,
|
|
4,
|
|
7)
|
|
#endif
|
|
#ifdef HAS_I410ALPHATOARGBROW_NEON
|
|
ANY41CT(I410AlphaToARGBRow_Any_NEON,
|
|
I410AlphaToARGBRow_NEON,
|
|
0,
|
|
0,
|
|
uint16_t,
|
|
2,
|
|
4,
|
|
7)
|
|
#endif
|
|
|
|
#ifdef HAS_I210ALPHATOARGBROW_SSSE3
|
|
ANY41CT(I210AlphaToARGBRow_Any_SSSE3,
|
|
I210AlphaToARGBRow_SSSE3,
|
|
1,
|
|
0,
|
|
uint16_t,
|
|
2,
|
|
4,
|
|
7)
|
|
#endif
|
|
|
|
#ifdef HAS_I210ALPHATOARGBROW_AVX2
|
|
ANY41CT(I210AlphaToARGBRow_Any_AVX2,
|
|
I210AlphaToARGBRow_AVX2,
|
|
1,
|
|
0,
|
|
uint16_t,
|
|
2,
|
|
4,
|
|
15)
|
|
#endif
|
|
|
|
#ifdef HAS_I410ALPHATOARGBROW_SSSE3
|
|
ANY41CT(I410AlphaToARGBRow_Any_SSSE3,
|
|
I410AlphaToARGBRow_SSSE3,
|
|
0,
|
|
0,
|
|
uint16_t,
|
|
2,
|
|
4,
|
|
7)
|
|
#endif
|
|
|
|
#ifdef HAS_I410ALPHATOARGBROW_AVX2
|
|
ANY41CT(I410AlphaToARGBRow_Any_AVX2,
|
|
I410AlphaToARGBRow_AVX2,
|
|
0,
|
|
0,
|
|
uint16_t,
|
|
2,
|
|
4,
|
|
15)
|
|
#endif
|
|
|
|
#undef ANY41CT
|
|
|
|
// Any 4 planes to 1 plane with parameter
|
|
#define ANY41PT(NAMEANY, ANY_SIMD, STYPE, SBPP, DTYPE, BPP, MASK) \
|
|
void NAMEANY(const STYPE* r_buf, const STYPE* g_buf, const STYPE* b_buf, \
|
|
const STYPE* a_buf, DTYPE* dst_ptr, int depth, int width) { \
|
|
SIMD_ALIGNED(STYPE vin[16 * 4]); \
|
|
static_assert((MASK + 1) * SBPP <= sizeof(vin) / 4, \
|
|
"vin buffer too small"); \
|
|
SIMD_ALIGNED(DTYPE vout[64]); \
|
|
static_assert((MASK + 1) * BPP <= sizeof(vout), "vout buffer too small"); \
|
|
MSAN_MEMSET(vin, 0, sizeof(vin)); /* for msan */ \
|
|
MSAN_MEMSET(vout, 0, sizeof(vout)); /* for msan */ \
|
|
int r = width & MASK; \
|
|
int n = width & ~MASK; \
|
|
if (n > 0) { \
|
|
ANY_SIMD(r_buf, g_buf, b_buf, a_buf, dst_ptr, depth, n); \
|
|
} \
|
|
ptrdiff_t np = n; \
|
|
memcpy(vin, r_buf + np, r * SBPP); \
|
|
memcpy(vin + 16, g_buf + np, r * SBPP); \
|
|
memcpy(vin + 32, b_buf + np, r * SBPP); \
|
|
memcpy(vin + 48, a_buf + np, r * SBPP); \
|
|
ANY_SIMD(vin, vin + 16, vin + 32, vin + 48, vout, depth, MASK + 1); \
|
|
memcpy((uint8_t*)dst_ptr + np * BPP, vout, r * BPP); \
|
|
}
|
|
|
|
#ifdef HAS_MERGEAR64ROW_AVX2
|
|
ANY41PT(MergeAR64Row_Any_AVX2, MergeAR64Row_AVX2, uint16_t, 2, uint16_t, 8, 15)
|
|
#endif
|
|
|
|
#ifdef HAS_MERGEAR64ROW_NEON
|
|
ANY41PT(MergeAR64Row_Any_NEON, MergeAR64Row_NEON, uint16_t, 2, uint16_t, 8, 7)
|
|
#endif
|
|
|
|
#ifdef HAS_MERGEARGB16TO8ROW_AVX2
|
|
ANY41PT(MergeARGB16To8Row_Any_AVX2,
|
|
MergeARGB16To8Row_AVX2,
|
|
uint16_t,
|
|
2,
|
|
uint8_t,
|
|
4,
|
|
15)
|
|
#endif
|
|
|
|
#ifdef HAS_MERGEARGB16TO8ROW_NEON
|
|
ANY41PT(MergeARGB16To8Row_Any_NEON,
|
|
MergeARGB16To8Row_NEON,
|
|
uint16_t,
|
|
2,
|
|
uint8_t,
|
|
4,
|
|
7)
|
|
#endif
|
|
|
|
#undef ANY41PT
|
|
|
|
// Any 3 planes to 1.
|
|
#define ANY31(NAMEANY, ANY_SIMD, UVSHIFT, DUVSHIFT, BPP, MASK) \
|
|
void NAMEANY(const uint8_t* y_buf, const uint8_t* u_buf, \
|
|
const uint8_t* v_buf, uint8_t* dst_ptr, int width) { \
|
|
SIMD_ALIGNED(uint8_t vin[64 * 3]); \
|
|
static_assert(MASK + 1 <= sizeof(vin) / 3, "vin buffer too small"); \
|
|
SIMD_ALIGNED(uint8_t vout[64 * 4]); \
|
|
static_assert(SS(MASK + 1, DUVSHIFT) * BPP <= sizeof(vout), \
|
|
"vout buffer too small"); \
|
|
memset(vin, 0, sizeof(vin)); /* for YUY2 and msan */ \
|
|
MSAN_MEMSET(vout, 0, sizeof(vout)); /* for msan */ \
|
|
int r = width & MASK; \
|
|
int n = width & ~MASK; \
|
|
if (n > 0) { \
|
|
ANY_SIMD(y_buf, u_buf, v_buf, dst_ptr, n); \
|
|
} \
|
|
ptrdiff_t np = n; \
|
|
memcpy(vin, y_buf + np, r); \
|
|
memcpy(vin + 64, u_buf + (np >> UVSHIFT), SS(r, UVSHIFT)); \
|
|
memcpy(vin + 128, v_buf + (np >> UVSHIFT), SS(r, UVSHIFT)); \
|
|
ANY_SIMD(vin, vin + 64, vin + 128, vout, MASK + 1); \
|
|
memcpy(dst_ptr + (np >> DUVSHIFT) * BPP, vout, SS(r, DUVSHIFT) * BPP); \
|
|
}
|
|
|
|
// Merge functions.
|
|
#ifdef HAS_MERGERGBROW_AVX2
|
|
ANY31(MergeRGBRow_Any_AVX2, MergeRGBRow_AVX2, 0, 0, 3, 31)
|
|
#endif
|
|
#ifdef HAS_MERGERGBROW_NEON
|
|
ANY31(MergeRGBRow_Any_NEON, MergeRGBRow_NEON, 0, 0, 3, 15)
|
|
#endif
|
|
#ifdef HAS_MERGEXRGBROW_SSE2
|
|
ANY31(MergeXRGBRow_Any_SSE2, MergeXRGBRow_SSE2, 0, 0, 4, 7)
|
|
#endif
|
|
#ifdef HAS_MERGEXRGBROW_AVX2
|
|
ANY31(MergeXRGBRow_Any_AVX2, MergeXRGBRow_AVX2, 0, 0, 4, 15)
|
|
#endif
|
|
#ifdef HAS_MERGEXRGBROW_NEON
|
|
ANY31(MergeXRGBRow_Any_NEON, MergeXRGBRow_NEON, 0, 0, 4, 15)
|
|
#endif
|
|
#ifdef HAS_MERGEXRGBROW_LSX
|
|
ANY31(MergeXRGBRow_Any_LSX, MergeXRGBRow_LSX, 0, 0, 4, 15)
|
|
#endif
|
|
#ifdef HAS_MERGEXRGBROW_LASX
|
|
ANY31(MergeXRGBRow_Any_LASX, MergeXRGBRow_LASX, 0, 0, 4, 15)
|
|
#endif
|
|
#ifdef HAS_I422TOYUY2ROW_SSE2
|
|
ANY31(I422ToYUY2Row_Any_SSE2, I422ToYUY2Row_SSE2, 1, 1, 4, 15)
|
|
ANY31(I422ToUYVYRow_Any_SSE2, I422ToUYVYRow_SSE2, 1, 1, 4, 15)
|
|
#endif
|
|
#ifdef HAS_I422TOYUY2ROW_AVX2
|
|
ANY31(I422ToYUY2Row_Any_AVX2, I422ToYUY2Row_AVX2, 1, 1, 4, 31)
|
|
ANY31(I422ToUYVYRow_Any_AVX2, I422ToUYVYRow_AVX2, 1, 1, 4, 31)
|
|
#endif
|
|
#ifdef HAS_I422TOYUY2ROW_NEON
|
|
ANY31(I422ToYUY2Row_Any_NEON, I422ToYUY2Row_NEON, 1, 1, 4, 15)
|
|
#endif
|
|
#ifdef HAS_I422TOYUY2ROW_LSX
|
|
ANY31(I422ToYUY2Row_Any_LSX, I422ToYUY2Row_LSX, 1, 1, 4, 15)
|
|
#endif
|
|
#ifdef HAS_I422TOYUY2ROW_LASX
|
|
ANY31(I422ToYUY2Row_Any_LASX, I422ToYUY2Row_LASX, 1, 1, 4, 31)
|
|
#endif
|
|
#ifdef HAS_I422TOUYVYROW_NEON
|
|
ANY31(I422ToUYVYRow_Any_NEON, I422ToUYVYRow_NEON, 1, 1, 4, 15)
|
|
#endif
|
|
#ifdef HAS_I422TOUYVYROW_LSX
|
|
ANY31(I422ToUYVYRow_Any_LSX, I422ToUYVYRow_LSX, 1, 1, 4, 15)
|
|
#endif
|
|
#ifdef HAS_I422TOUYVYROW_LASX
|
|
ANY31(I422ToUYVYRow_Any_LASX, I422ToUYVYRow_LASX, 1, 1, 4, 31)
|
|
#endif
|
|
#ifdef HAS_BLENDPLANEROW_AVX2
|
|
ANY31(BlendPlaneRow_Any_AVX2, BlendPlaneRow_AVX2, 0, 0, 1, 31)
|
|
#endif
|
|
#ifdef HAS_BLENDPLANEROW_SSSE3
|
|
ANY31(BlendPlaneRow_Any_SSSE3, BlendPlaneRow_SSSE3, 0, 0, 1, 7)
|
|
#endif
|
|
#ifdef HAS_BLENDPLANEROW_NEON
|
|
ANY31(BlendPlaneRow_Any_NEON, BlendPlaneRow_NEON, 0, 0, 1, 15)
|
|
#endif
|
|
#ifdef HAS_BLENDPLANEROW_LSX
|
|
ANY31(BlendPlaneRow_Any_LSX, BlendPlaneRow_LSX, 0, 0, 1, 15)
|
|
#endif
|
|
#ifdef HAS_BLENDPLANEROW_LASX
|
|
ANY31(BlendPlaneRow_Any_LASX, BlendPlaneRow_LASX, 0, 0, 1, 31)
|
|
#endif
|
|
#ifdef HAS_BLENDPLANEROW_WASMSIMD
|
|
ANY31(BlendPlaneRow_Any_WASMSIMD, BlendPlaneRow_WASMSIMD, 0, 0, 1, 15)
|
|
#endif
|
|
#undef ANY31
|
|
|
|
// Note that odd width replication includes 444 due to implementation
|
|
// on arm that subsamples 444 to 422 internally.
|
|
// Any 3 planes to 1 with yuvconstants
|
|
#define ANY31C(NAMEANY, ANY_SIMD, UVSHIFT, DUVSHIFT, BPP, MASK) \
|
|
void NAMEANY(const uint8_t* y_buf, const uint8_t* u_buf, \
|
|
const uint8_t* v_buf, uint8_t* dst_ptr, \
|
|
const struct YuvConstants* yuvconstants, int width) { \
|
|
SIMD_ALIGNED(uint8_t vin[128 * 3]); \
|
|
static_assert(MASK + 1 <= sizeof(vin) / 3, "vin buffer too small"); \
|
|
SIMD_ALIGNED(uint8_t vout[128]); \
|
|
static_assert(SS(MASK + 1, DUVSHIFT) * BPP <= sizeof(vout), \
|
|
"vout buffer too small"); \
|
|
MSAN_MEMSET(vin, 0, sizeof(vin)); /* for YUY2 and msan */ \
|
|
MSAN_MEMSET(vout, 0, sizeof(vout)); /* for msan */ \
|
|
int r = width & MASK; \
|
|
int n = width & ~MASK; \
|
|
if (n > 0) { \
|
|
ANY_SIMD(y_buf, u_buf, v_buf, dst_ptr, yuvconstants, n); \
|
|
} \
|
|
ptrdiff_t np = n; \
|
|
memcpy(vin, y_buf + np, r); \
|
|
memcpy(vin + 128, u_buf + (np >> UVSHIFT), SS(r, UVSHIFT)); \
|
|
memcpy(vin + 256, v_buf + (np >> UVSHIFT), SS(r, UVSHIFT)); \
|
|
if (width & 1) { \
|
|
vin[128 + SS(r, UVSHIFT)] = vin[128 + SS(r, UVSHIFT) - 1]; \
|
|
vin[256 + SS(r, UVSHIFT)] = vin[256 + SS(r, UVSHIFT) - 1]; \
|
|
} \
|
|
ANY_SIMD(vin, vin + 128, vin + 256, vout, yuvconstants, MASK + 1); \
|
|
memcpy(dst_ptr + (np >> DUVSHIFT) * BPP, vout, SS(r, DUVSHIFT) * BPP); \
|
|
}
|
|
|
|
#ifdef HAS_I422TOARGBROW_SSSE3
|
|
ANY31C(I422ToARGBRow_Any_SSSE3, I422ToARGBRow_SSSE3, 1, 0, 4, 7)
|
|
#endif
|
|
#ifdef HAS_I422TORGBAROW_SSSE3
|
|
ANY31C(I422ToRGBARow_Any_SSSE3, I422ToRGBARow_SSSE3, 1, 0, 4, 7)
|
|
#endif
|
|
#ifdef HAS_I422TOARGB4444ROW_SSSE3
|
|
ANY31C(I422ToARGB4444Row_Any_SSSE3, I422ToARGB4444Row_SSSE3, 1, 0, 2, 7)
|
|
#endif
|
|
#ifdef HAS_I422TOARGB1555ROW_SSSE3
|
|
ANY31C(I422ToARGB1555Row_Any_SSSE3, I422ToARGB1555Row_SSSE3, 1, 0, 2, 7)
|
|
#endif
|
|
#ifdef HAS_I422TORGB565ROW_SSSE3
|
|
ANY31C(I422ToRGB565Row_Any_SSSE3, I422ToRGB565Row_SSSE3, 1, 0, 2, 7)
|
|
#endif
|
|
#ifdef HAS_I422TORGB24ROW_SSSE3
|
|
ANY31C(I422ToRGB24Row_Any_SSSE3, I422ToRGB24Row_SSSE3, 1, 0, 3, 7)
|
|
#endif
|
|
#ifdef HAS_I422TOAR30ROW_SSSE3
|
|
ANY31C(I422ToAR30Row_Any_SSSE3, I422ToAR30Row_SSSE3, 1, 0, 4, 7)
|
|
#endif
|
|
#ifdef HAS_I422TOAR30ROW_AVX2
|
|
ANY31C(I422ToAR30Row_Any_AVX2, I422ToAR30Row_AVX2, 1, 0, 4, 15)
|
|
#endif
|
|
#ifdef HAS_I444TOARGBROW_SSSE3
|
|
ANY31C(I444ToARGBRow_Any_SSSE3, I444ToARGBRow_SSSE3, 0, 0, 4, 7)
|
|
#endif
|
|
#ifdef HAS_I444TORGB24ROW_SSSE3
|
|
ANY31C(I444ToRGB24Row_Any_SSSE3, I444ToRGB24Row_SSSE3, 0, 0, 3, 7)
|
|
#endif
|
|
#ifdef HAS_I422TORGB24ROW_AVX2
|
|
ANY31C(I422ToRGB24Row_Any_AVX2, I422ToRGB24Row_AVX2, 1, 0, 3, 15)
|
|
#endif
|
|
#ifdef HAS_I422TOARGBROW_AVX2
|
|
ANY31C(I422ToARGBRow_Any_AVX2, I422ToARGBRow_AVX2, 1, 0, 4, 15)
|
|
#endif
|
|
#ifdef HAS_I422TORGBAROW_AVX2
|
|
ANY31C(I422ToRGBARow_Any_AVX2, I422ToRGBARow_AVX2, 1, 0, 4, 15)
|
|
#endif
|
|
#ifdef HAS_I444TOARGBROW_AVX2
|
|
ANY31C(I444ToARGBRow_Any_AVX2, I444ToARGBRow_AVX2, 0, 0, 4, 15)
|
|
#endif
|
|
#ifdef HAS_I444TORGB24ROW_AVX2
|
|
ANY31C(I444ToRGB24Row_Any_AVX2, I444ToRGB24Row_AVX2, 0, 0, 3, 31)
|
|
#endif
|
|
#ifdef HAS_I422TOARGB4444ROW_AVX2
|
|
ANY31C(I422ToARGB4444Row_Any_AVX2, I422ToARGB4444Row_AVX2, 1, 0, 2, 15)
|
|
#endif
|
|
#ifdef HAS_I422TOARGB1555ROW_AVX2
|
|
ANY31C(I422ToARGB1555Row_Any_AVX2, I422ToARGB1555Row_AVX2, 1, 0, 2, 15)
|
|
#endif
|
|
#ifdef HAS_I422TORGB565ROW_AVX2
|
|
ANY31C(I422ToRGB565Row_Any_AVX2, I422ToRGB565Row_AVX2, 1, 0, 2, 15)
|
|
#endif
|
|
#ifdef HAS_I444TORGB24ROW_NEON
|
|
ANY31C(I444ToRGB24Row_Any_NEON, I444ToRGB24Row_NEON, 0, 0, 3, 7)
|
|
#endif
|
|
#ifdef HAS_I422TOAR30ROW_NEON
|
|
ANY31C(I422ToAR30Row_Any_NEON, I422ToAR30Row_NEON, 1, 0, 4, 7)
|
|
#endif
|
|
#ifdef HAS_I422TOARGBROW_NEON
|
|
ANY31C(I444ToARGBRow_Any_NEON, I444ToARGBRow_NEON, 0, 0, 4, 7)
|
|
ANY31C(I422ToARGBRow_Any_NEON, I422ToARGBRow_NEON, 1, 0, 4, 7)
|
|
ANY31C(I422ToRGBARow_Any_NEON, I422ToRGBARow_NEON, 1, 0, 4, 7)
|
|
ANY31C(I422ToRGB24Row_Any_NEON, I422ToRGB24Row_NEON, 1, 0, 3, 7)
|
|
ANY31C(I422ToARGB4444Row_Any_NEON, I422ToARGB4444Row_NEON, 1, 0, 2, 7)
|
|
ANY31C(I422ToARGB1555Row_Any_NEON, I422ToARGB1555Row_NEON, 1, 0, 2, 7)
|
|
ANY31C(I422ToRGB565Row_Any_NEON, I422ToRGB565Row_NEON, 1, 0, 2, 7)
|
|
#endif
|
|
#ifdef HAS_I422TOARGBROW_LSX
|
|
ANY31C(I422ToARGBRow_Any_LSX, I422ToARGBRow_LSX, 1, 0, 4, 7)
|
|
ANY31C(I422ToRGBARow_Any_LSX, I422ToRGBARow_LSX, 1, 0, 4, 7)
|
|
ANY31C(I422ToRGB24Row_Any_LSX, I422ToRGB24Row_LSX, 1, 0, 3, 7)
|
|
ANY31C(I422ToRGB565Row_Any_LSX, I422ToRGB565Row_LSX, 1, 0, 2, 7)
|
|
ANY31C(I422ToARGB4444Row_Any_LSX, I422ToARGB4444Row_LSX, 1, 0, 2, 7)
|
|
ANY31C(I422ToARGB1555Row_Any_LSX, I422ToARGB1555Row_LSX, 1, 0, 2, 7)
|
|
#endif
|
|
#ifdef HAS_I422TOARGBROW_LASX
|
|
ANY31C(I422ToARGBRow_Any_LASX, I422ToARGBRow_LASX, 1, 0, 4, 15)
|
|
ANY31C(I422ToRGBARow_Any_LASX, I422ToRGBARow_LASX, 1, 0, 4, 15)
|
|
ANY31C(I422ToRGB24Row_Any_LASX, I422ToRGB24Row_LASX, 1, 0, 3, 15)
|
|
ANY31C(I422ToRGB565Row_Any_LASX, I422ToRGB565Row_LASX, 1, 0, 2, 15)
|
|
ANY31C(I422ToARGB4444Row_Any_LASX, I422ToARGB4444Row_LASX, 1, 0, 2, 15)
|
|
ANY31C(I422ToARGB1555Row_Any_LASX, I422ToARGB1555Row_LASX, 1, 0, 2, 15)
|
|
#endif
|
|
#ifdef HAS_I444TOARGBROW_LSX
|
|
ANY31C(I444ToARGBRow_Any_LSX, I444ToARGBRow_LSX, 0, 0, 4, 7)
|
|
#endif
|
|
#ifdef HAS_I444TOARGBROW_LASX
|
|
ANY31C(I444ToARGBRow_Any_LASX, I444ToARGBRow_LASX, 0, 0, 4, 15)
|
|
#endif
|
|
#ifdef HAS_I422TOARGBROW_WASMSIMD
|
|
ANY31C(I422ToARGBRow_Any_WASMSIMD, I422ToARGBRow_WASMSIMD, 1, 0, 4, 7)
|
|
#endif
|
|
#ifdef HAS_I422TOAR30ROW_WASMSIMD
|
|
ANY31C(I422ToAR30Row_Any_WASMSIMD, I422ToAR30Row_WASMSIMD, 1, 0, 4, 7)
|
|
#endif
|
|
#ifdef HAS_I422TORGB24ROW_WASMSIMD
|
|
ANY31C(I422ToRGB24Row_Any_WASMSIMD, I422ToRGB24Row_WASMSIMD, 1, 0, 3, 15)
|
|
#endif
|
|
#ifdef HAS_I422TORGB565ROW_WASMSIMD
|
|
ANY31C(I422ToRGB565Row_Any_WASMSIMD, I422ToRGB565Row_WASMSIMD, 1, 0, 2, 7)
|
|
#endif
|
|
#ifdef HAS_I422TOARGB1555ROW_WASMSIMD
|
|
ANY31C(I422ToARGB1555Row_Any_WASMSIMD, I422ToARGB1555Row_WASMSIMD, 1, 0, 2, 7)
|
|
#endif
|
|
#ifdef HAS_I422TOARGB4444ROW_WASMSIMD
|
|
ANY31C(I422ToARGB4444Row_Any_WASMSIMD, I422ToARGB4444Row_WASMSIMD, 1, 0, 2, 7)
|
|
#endif
|
|
#undef ANY31C
|
|
|
|
// Any 3 planes of 16 bit to 1 with yuvconstants
|
|
// TODO(fbarchard): consider sharing this code with ANY31C
|
|
#define ANY31CT(NAMEANY, ANY_SIMD, UVSHIFT, DUVSHIFT, T, SBPP, BPP, MASK) \
|
|
void NAMEANY(const T* y_buf, const T* u_buf, const T* v_buf, \
|
|
uint8_t* dst_ptr, const struct YuvConstants* yuvconstants, \
|
|
int width) { \
|
|
SIMD_ALIGNED(T vin[16 * 3]); \
|
|
static_assert((MASK + 1) * SBPP <= sizeof(vin) / 3, \
|
|
"vin buffer too small"); \
|
|
SIMD_ALIGNED(uint8_t vout[64]); \
|
|
static_assert(SS(MASK + 1, DUVSHIFT) * BPP <= sizeof(vout), \
|
|
"vout buffer too small"); \
|
|
MSAN_MEMSET(vin, 0, sizeof(vin)); /* for YUY2 and msan */ \
|
|
MSAN_MEMSET(vout, 0, sizeof(vout)); /* for msan */ \
|
|
int r = width & MASK; \
|
|
int n = width & ~MASK; \
|
|
if (n > 0) { \
|
|
ANY_SIMD(y_buf, u_buf, v_buf, dst_ptr, yuvconstants, n); \
|
|
} \
|
|
ptrdiff_t np = n; \
|
|
memcpy(vin, y_buf + np, r * SBPP); \
|
|
memcpy(vin + 16, u_buf + (np >> UVSHIFT), SS(r, UVSHIFT) * SBPP); \
|
|
memcpy(vin + 32, v_buf + (np >> UVSHIFT), SS(r, UVSHIFT) * SBPP); \
|
|
ANY_SIMD(vin, vin + 16, vin + 32, vout, yuvconstants, MASK + 1); \
|
|
memcpy(dst_ptr + (np >> DUVSHIFT) * BPP, vout, SS(r, DUVSHIFT) * BPP); \
|
|
}
|
|
|
|
#ifdef HAS_I210TOAR30ROW_SSSE3
|
|
ANY31CT(I210ToAR30Row_Any_SSSE3, I210ToAR30Row_SSSE3, 1, 0, uint16_t, 2, 4, 7)
|
|
#endif
|
|
#ifdef HAS_I210TOARGBROW_SSSE3
|
|
ANY31CT(I210ToARGBRow_Any_SSSE3, I210ToARGBRow_SSSE3, 1, 0, uint16_t, 2, 4, 7)
|
|
#endif
|
|
#ifdef HAS_I210TOARGBROW_AVX2
|
|
ANY31CT(I210ToARGBRow_Any_AVX2, I210ToARGBRow_AVX2, 1, 0, uint16_t, 2, 4, 15)
|
|
#endif
|
|
#ifdef HAS_I210TOAR30ROW_AVX2
|
|
ANY31CT(I210ToAR30Row_Any_AVX2, I210ToAR30Row_AVX2, 1, 0, uint16_t, 2, 4, 15)
|
|
#endif
|
|
#ifdef HAS_I410TOAR30ROW_SSSE3
|
|
ANY31CT(I410ToAR30Row_Any_SSSE3, I410ToAR30Row_SSSE3, 0, 0, uint16_t, 2, 4, 7)
|
|
#endif
|
|
#ifdef HAS_I410TOARGBROW_SSSE3
|
|
ANY31CT(I410ToARGBRow_Any_SSSE3, I410ToARGBRow_SSSE3, 0, 0, uint16_t, 2, 4, 7)
|
|
#endif
|
|
#ifdef HAS_I410TOARGBROW_AVX2
|
|
ANY31CT(I410ToARGBRow_Any_AVX2, I410ToARGBRow_AVX2, 0, 0, uint16_t, 2, 4, 15)
|
|
#endif
|
|
#ifdef HAS_I410TOAR30ROW_AVX2
|
|
ANY31CT(I410ToAR30Row_Any_AVX2, I410ToAR30Row_AVX2, 0, 0, uint16_t, 2, 4, 15)
|
|
#endif
|
|
#ifdef HAS_I212TOAR30ROW_SSSE3
|
|
ANY31CT(I212ToAR30Row_Any_SSSE3, I212ToAR30Row_SSSE3, 1, 0, uint16_t, 2, 4, 7)
|
|
#endif
|
|
#ifdef HAS_I212TOARGBROW_SSSE3
|
|
ANY31CT(I212ToARGBRow_Any_SSSE3, I212ToARGBRow_SSSE3, 1, 0, uint16_t, 2, 4, 7)
|
|
#endif
|
|
#ifdef HAS_I212TOARGBROW_AVX2
|
|
ANY31CT(I212ToARGBRow_Any_AVX2, I212ToARGBRow_AVX2, 1, 0, uint16_t, 2, 4, 15)
|
|
#endif
|
|
#ifdef HAS_I212TOAR30ROW_AVX2
|
|
ANY31CT(I212ToAR30Row_Any_AVX2, I212ToAR30Row_AVX2, 1, 0, uint16_t, 2, 4, 15)
|
|
#endif
|
|
#ifdef HAS_I210TOARGBROW_NEON
|
|
ANY31CT(I210ToARGBRow_Any_NEON, I210ToARGBRow_NEON, 1, 0, uint16_t, 2, 4, 7)
|
|
#endif
|
|
#ifdef HAS_I410TOARGBROW_NEON
|
|
ANY31CT(I410ToARGBRow_Any_NEON, I410ToARGBRow_NEON, 0, 0, uint16_t, 2, 4, 7)
|
|
#endif
|
|
#ifdef HAS_I210TOAR30ROW_NEON
|
|
ANY31CT(I210ToAR30Row_Any_NEON, I210ToAR30Row_NEON, 1, 0, uint16_t, 2, 4, 7)
|
|
#endif
|
|
#ifdef HAS_I410TOAR30ROW_NEON
|
|
ANY31CT(I410ToAR30Row_Any_NEON, I410ToAR30Row_NEON, 0, 0, uint16_t, 2, 4, 7)
|
|
#endif
|
|
#ifdef HAS_I212TOARGBROW_NEON
|
|
ANY31CT(I212ToARGBRow_Any_NEON, I212ToARGBRow_NEON, 1, 0, uint16_t, 2, 4, 7)
|
|
#endif
|
|
#ifdef HAS_I212TOAR30ROW_NEON
|
|
ANY31CT(I212ToAR30Row_Any_NEON, I212ToAR30Row_NEON, 1, 0, uint16_t, 2, 4, 7)
|
|
#endif
|
|
#undef ANY31CT
|
|
|
|
// Any 3 planes to 1 plane with parameter
|
|
#define ANY31PT(NAMEANY, ANY_SIMD, STYPE, SBPP, DTYPE, BPP, MASK) \
|
|
void NAMEANY(const STYPE* r_buf, const STYPE* g_buf, const STYPE* b_buf, \
|
|
DTYPE* dst_ptr, int depth, int width) { \
|
|
SIMD_ALIGNED(STYPE vin[16 * 3]); \
|
|
static_assert((MASK + 1) * SBPP <= sizeof(vin) / 3, \
|
|
"vin buffer too small"); \
|
|
SIMD_ALIGNED(DTYPE vout[64]); \
|
|
static_assert((MASK + 1) * BPP <= sizeof(vout), "vout buffer too small"); \
|
|
MSAN_MEMSET(vin, 0, sizeof(vin)); /* for YUY2 and msan */ \
|
|
MSAN_MEMSET(vout, 0, sizeof(vout)); /* for msan */ \
|
|
int r = width & MASK; \
|
|
int n = width & ~MASK; \
|
|
if (n > 0) { \
|
|
ANY_SIMD(r_buf, g_buf, b_buf, dst_ptr, depth, n); \
|
|
} \
|
|
ptrdiff_t np = n; \
|
|
memcpy(vin, r_buf + np, r * SBPP); \
|
|
memcpy(vin + 16, g_buf + np, r * SBPP); \
|
|
memcpy(vin + 32, b_buf + np, r * SBPP); \
|
|
ANY_SIMD(vin, vin + 16, vin + 32, vout, depth, MASK + 1); \
|
|
memcpy((uint8_t*)dst_ptr + np * BPP, vout, r * BPP); \
|
|
}
|
|
|
|
#ifdef HAS_MERGEXR30ROW_AVX2
|
|
ANY31PT(MergeXR30Row_Any_AVX2, MergeXR30Row_AVX2, uint16_t, 2, uint8_t, 4, 15)
|
|
#endif
|
|
|
|
#ifdef HAS_MERGEXR30ROW_NEON
|
|
ANY31PT(MergeXR30Row_Any_NEON, MergeXR30Row_NEON, uint16_t, 2, uint8_t, 4, 3)
|
|
ANY31PT(MergeXR30Row_10_Any_NEON,
|
|
MergeXR30Row_10_NEON,
|
|
uint16_t,
|
|
2,
|
|
uint8_t,
|
|
4,
|
|
7)
|
|
#endif
|
|
|
|
#ifdef HAS_MERGEXR64ROW_AVX2
|
|
ANY31PT(MergeXR64Row_Any_AVX2, MergeXR64Row_AVX2, uint16_t, 2, uint16_t, 8, 15)
|
|
#endif
|
|
|
|
#ifdef HAS_MERGEXR64ROW_NEON
|
|
ANY31PT(MergeXR64Row_Any_NEON, MergeXR64Row_NEON, uint16_t, 2, uint16_t, 8, 7)
|
|
#endif
|
|
|
|
#ifdef HAS_MERGEXRGB16TO8ROW_AVX2
|
|
ANY31PT(MergeXRGB16To8Row_Any_AVX2,
|
|
MergeXRGB16To8Row_AVX2,
|
|
uint16_t,
|
|
2,
|
|
uint8_t,
|
|
4,
|
|
15)
|
|
#endif
|
|
|
|
#ifdef HAS_MERGEXRGB16TO8ROW_NEON
|
|
ANY31PT(MergeXRGB16To8Row_Any_NEON,
|
|
MergeXRGB16To8Row_NEON,
|
|
uint16_t,
|
|
2,
|
|
uint8_t,
|
|
4,
|
|
7)
|
|
#endif
|
|
|
|
#undef ANY31PT
|
|
|
|
// Any 2 planes to 1.
|
|
#define ANY21(NAMEANY, ANY_SIMD, UVSHIFT, SBPP, SBPP2, BPP, MASK) \
|
|
void NAMEANY(const uint8_t* y_buf, const uint8_t* uv_buf, uint8_t* dst_ptr, \
|
|
int width) { \
|
|
SIMD_ALIGNED(uint8_t vin[128 * 2]); \
|
|
static_assert((MASK + 1) * SBPP <= sizeof(vin) / 2, \
|
|
"vin buffer too small"); \
|
|
static_assert(SS(MASK + 1, UVSHIFT) * SBPP2 <= sizeof(vin) / 2, \
|
|
"vin buffer too small"); \
|
|
SIMD_ALIGNED(uint8_t vout[128]); \
|
|
static_assert((MASK + 1) * BPP <= sizeof(vout), "vout buffer too small"); \
|
|
MSAN_MEMSET(vin, 0, sizeof(vin)); /* for msan */ \
|
|
MSAN_MEMSET(vout, 0, sizeof(vout)); /* for msan */ \
|
|
int r = width & MASK; \
|
|
int n = width & ~MASK; \
|
|
if (n > 0) { \
|
|
ANY_SIMD(y_buf, uv_buf, dst_ptr, n); \
|
|
} \
|
|
ptrdiff_t np = n; \
|
|
memcpy(vin, y_buf + np * SBPP, r * SBPP); \
|
|
memcpy(vin + 128, uv_buf + (np >> UVSHIFT) * SBPP2, \
|
|
SS(r, UVSHIFT) * SBPP2); \
|
|
ANY_SIMD(vin, vin + 128, vout, MASK + 1); \
|
|
memcpy(dst_ptr + np * BPP, vout, r * BPP); \
|
|
}
|
|
|
|
// 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, 15)
|
|
#endif
|
|
#ifdef HAS_MERGEUVROW_NEON
|
|
ANY21(MergeUVRow_Any_NEON, MergeUVRow_NEON, 0, 1, 1, 2, 15)
|
|
#endif
|
|
#ifdef HAS_MERGEUVROW_LSX
|
|
ANY21(MergeUVRow_Any_LSX, MergeUVRow_LSX, 0, 1, 1, 2, 15)
|
|
#endif
|
|
#ifdef HAS_MERGEUVROW_LASX
|
|
ANY21(MergeUVRow_Any_LASX, MergeUVRow_LASX, 0, 1, 1, 2, 31)
|
|
#endif
|
|
#ifdef HAS_NV21TOYUV24ROW_NEON
|
|
ANY21(NV21ToYUV24Row_Any_NEON, NV21ToYUV24Row_NEON, 1, 1, 2, 3, 15)
|
|
#endif
|
|
#ifdef HAS_NV21TOYUV24ROW_SSSE3
|
|
ANY21(NV21ToYUV24Row_Any_SSSE3, NV21ToYUV24Row_SSSE3, 1, 1, 2, 3, 15)
|
|
#endif
|
|
#ifdef HAS_NV21TOYUV24ROW_AVX2
|
|
ANY21(NV21ToYUV24Row_Any_AVX2, NV21ToYUV24Row_AVX2, 1, 1, 2, 3, 31)
|
|
#endif
|
|
// 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_ARGBBLENDROW_LSX
|
|
ANY21(ARGBBlendRow_Any_LSX, ARGBBlendRow_LSX, 0, 4, 4, 4, 3)
|
|
#endif
|
|
#ifdef HAS_ARGBBLENDROW_LASX
|
|
ANY21(ARGBBlendRow_Any_LASX, ARGBBlendRow_LASX, 0, 4, 4, 4, 7)
|
|
#endif
|
|
#ifdef HAS_ARGBMULTIPLYROW_LSX
|
|
ANY21(ARGBMultiplyRow_Any_LSX, ARGBMultiplyRow_LSX, 0, 4, 4, 4, 3)
|
|
#endif
|
|
#ifdef HAS_ARGBMULTIPLYROW_LASX
|
|
ANY21(ARGBMultiplyRow_Any_LASX, ARGBMultiplyRow_LASX, 0, 4, 4, 4, 7)
|
|
#endif
|
|
#ifdef HAS_ARGBADDROW_LSX
|
|
ANY21(ARGBAddRow_Any_LSX, ARGBAddRow_LSX, 0, 4, 4, 4, 3)
|
|
#endif
|
|
#ifdef HAS_ARGBADDROW_LASX
|
|
ANY21(ARGBAddRow_Any_LASX, ARGBAddRow_LASX, 0, 4, 4, 4, 7)
|
|
#endif
|
|
#ifdef HAS_ARGBSUBTRACTROW_LSX
|
|
ANY21(ARGBSubtractRow_Any_LSX, ARGBSubtractRow_LSX, 0, 4, 4, 4, 3)
|
|
#endif
|
|
#ifdef HAS_ARGBSUBTRACTROW_LASX
|
|
ANY21(ARGBSubtractRow_Any_LASX, ARGBSubtractRow_LASX, 0, 4, 4, 4, 7)
|
|
#endif
|
|
#ifdef HAS_ARGBALPHABLENDROW_SSSE3
|
|
ANY21(ARGBAlphaBlendRow_Any_SSSE3, ARGBAlphaBlendRow_SSSE3, 0, 4, 4, 4, 3)
|
|
#endif
|
|
#ifdef HAS_ARGBALPHABLENDROW_AVX2
|
|
ANY21(ARGBAlphaBlendRow_Any_AVX2, ARGBAlphaBlendRow_AVX2, 0, 4, 4, 4, 7)
|
|
#endif
|
|
#ifdef HAS_ARGBALPHABLENDROW_NEON
|
|
ANY21(ARGBAlphaBlendRow_Any_NEON, ARGBAlphaBlendRow_NEON, 0, 4, 4, 4, 7)
|
|
#endif
|
|
#ifdef HAS_ARGBALPHABLENDROW_LSX
|
|
ANY21(ARGBAlphaBlendRow_Any_LSX, ARGBAlphaBlendRow_LSX, 0, 4, 4, 4, 3)
|
|
#endif
|
|
#ifdef HAS_ARGBALPHABLENDROW_LASX
|
|
ANY21(ARGBAlphaBlendRow_Any_LASX, ARGBAlphaBlendRow_LASX, 0, 4, 4, 4, 7)
|
|
#endif
|
|
#ifdef HAS_ARGBALPHABLENDROW_WASMSIMD
|
|
ANY21(ARGBAlphaBlendRow_Any_WASMSIMD, ARGBAlphaBlendRow_WASMSIMD, 0, 4, 4, 4, 3)
|
|
#endif
|
|
#ifdef HAS_SOBELROW_AVX2
|
|
ANY21(SobelRow_Any_AVX2, SobelRow_AVX2, 0, 1, 1, 4, 15)
|
|
#endif
|
|
#ifdef HAS_SOBELROW_NEON
|
|
ANY21(SobelRow_Any_NEON, SobelRow_NEON, 0, 1, 1, 4, 7)
|
|
#endif
|
|
#ifdef HAS_SOBELROW_LSX
|
|
ANY21(SobelRow_Any_LSX, SobelRow_LSX, 0, 1, 1, 4, 7)
|
|
#endif
|
|
#ifdef HAS_SOBELROW_LASX
|
|
ANY21(SobelRow_Any_LASX, SobelRow_LASX, 0, 1, 1, 4, 15)
|
|
#endif
|
|
#ifdef HAS_SOBELTOPLANEROW_AVX2
|
|
ANY21(SobelToPlaneRow_Any_AVX2, SobelToPlaneRow_AVX2, 0, 1, 1, 1, 31)
|
|
#endif
|
|
#ifdef HAS_SOBELTOPLANEROW_NEON
|
|
ANY21(SobelToPlaneRow_Any_NEON, SobelToPlaneRow_NEON, 0, 1, 1, 1, 15)
|
|
#endif
|
|
#ifdef HAS_SOBELTOPLANEROW_LSX
|
|
ANY21(SobelToPlaneRow_Any_LSX, SobelToPlaneRow_LSX, 0, 1, 1, 1, 15)
|
|
#endif
|
|
#ifdef HAS_SOBELTOPLANEROW_LASX
|
|
ANY21(SobelToPlaneRow_Any_LASX, SobelToPlaneRow_LASX, 0, 1, 1, 1, 31)
|
|
#endif
|
|
#ifdef HAS_SOBELXYROW_AVX2
|
|
ANY21(SobelXYRow_Any_AVX2, SobelXYRow_AVX2, 0, 1, 1, 4, 15)
|
|
#endif
|
|
#ifdef HAS_SOBELXYROW_NEON
|
|
ANY21(SobelXYRow_Any_NEON, SobelXYRow_NEON, 0, 1, 1, 4, 7)
|
|
#endif
|
|
#ifdef HAS_SOBELXYROW_LSX
|
|
ANY21(SobelXYRow_Any_LSX, SobelXYRow_LSX, 0, 1, 1, 4, 7)
|
|
#endif
|
|
#ifdef HAS_SOBELXYROW_LASX
|
|
ANY21(SobelXYRow_Any_LASX, SobelXYRow_LASX, 0, 1, 1, 4, 15)
|
|
#endif
|
|
#ifdef HAS_MERGEUVROW_WASMSIMD
|
|
ANY21(MergeUVRow_Any_WASMSIMD, MergeUVRow_WASMSIMD, 0, 1, 1, 2, 15)
|
|
#endif
|
|
#undef ANY21
|
|
|
|
// Any 2 planes to 1 with stride
|
|
// width is measured in source pixels. 4 bytes contains 2 pixels
|
|
#define ANY21S(NAMEANY, ANY_SIMD, SBPP, BPP, MASK) \
|
|
void NAMEANY(const uint8_t* src_yuy2, int stride_yuy2, uint8_t* dst_uv, \
|
|
int width) { \
|
|
SIMD_ALIGNED(uint8_t vin[(MASK + 1) * SBPP * 2]); \
|
|
SIMD_ALIGNED(uint8_t vout[(MASK + 1) * BPP]); \
|
|
MSAN_MEMSET(vin, 0, sizeof(vin)); /* for msan */ \
|
|
MSAN_MEMSET(vout, 0, sizeof(vout)); /* for msan */ \
|
|
int awidth = (width + 1) / 2; \
|
|
int r = awidth & MASK; \
|
|
int n = awidth & ~MASK; \
|
|
if (n > 0) { \
|
|
ANY_SIMD(src_yuy2, stride_yuy2, dst_uv, n * 2); \
|
|
} \
|
|
ptrdiff_t np = n; \
|
|
memcpy(vin, src_yuy2 + np * SBPP, r * SBPP); \
|
|
memcpy(vin + (MASK + 1) * SBPP, src_yuy2 + stride_yuy2 + np * SBPP, \
|
|
r * SBPP); \
|
|
ANY_SIMD(vin, (MASK + 1) * SBPP, vout, (MASK + 1) * 2); \
|
|
memcpy(dst_uv + np * BPP, vout, r * BPP); \
|
|
}
|
|
|
|
#ifdef HAS_YUY2TONVUVROW_NEON
|
|
ANY21S(YUY2ToNVUVRow_Any_NEON, YUY2ToNVUVRow_NEON, 4, 2, 7)
|
|
#endif
|
|
#ifdef HAS_YUY2TONVUVROW_SSE2
|
|
ANY21S(YUY2ToNVUVRow_Any_SSE2, YUY2ToNVUVRow_SSE2, 4, 2, 7)
|
|
#endif
|
|
#ifdef HAS_YUY2TONVUVROW_AVX2
|
|
ANY21S(YUY2ToNVUVRow_Any_AVX2, YUY2ToNVUVRow_AVX2, 4, 2, 15)
|
|
#endif
|
|
#ifdef HAS_YUY2TONVUVROW_LSX
|
|
ANY21S(YUY2ToNVUVRow_Any_LSX, YUY2ToNVUVRow_LSX, 4, 2, 7)
|
|
#endif
|
|
#ifdef HAS_YUY2TONVUVROW_LASX
|
|
ANY21S(YUY2ToNVUVRow_Any_LASX, YUY2ToNVUVRow_LASX, 4, 2, 15)
|
|
#endif
|
|
#undef ANY21S
|
|
|
|
// Any 2 planes to 1 half-size merged UV row with stride (2 rows of U and V).
|
|
#define ANY21HMS(NAMEANY, ANY_SIMD, MASK) \
|
|
void NAMEANY(const uint8_t* src_u, int src_stride_u, const uint8_t* src_v, \
|
|
int src_stride_v, uint8_t* dst_uv, int width) { \
|
|
SIMD_ALIGNED(uint8_t vin[64 * 4]); \
|
|
static_assert(MASK + 1 <= sizeof(vin) / 4, "vin buffer too small"); \
|
|
SIMD_ALIGNED(uint8_t vout[64]); \
|
|
static_assert(SS(MASK + 1, 1) * 2 <= sizeof(vout), \
|
|
"vout buffer too small"); \
|
|
MSAN_MEMSET(vin, 0, sizeof(vin)); /* for msan */ \
|
|
MSAN_MEMSET(vout, 0, sizeof(vout)); /* for msan */ \
|
|
int r = width & MASK; \
|
|
int n = width & ~MASK; \
|
|
if (n > 0) { \
|
|
ANY_SIMD(src_u, src_stride_u, src_v, src_stride_v, dst_uv, n); \
|
|
} \
|
|
ptrdiff_t np = n; \
|
|
memcpy(vin, src_u + np, r); \
|
|
memcpy(vin + 64, src_u + src_stride_u + np, r); \
|
|
memcpy(vin + 128, src_v + np, r); \
|
|
memcpy(vin + 192, src_v + src_stride_v + np, r); \
|
|
if (r & 1) { \
|
|
vin[r] = vin[r - 1]; \
|
|
vin[64 + r] = vin[64 + r - 1]; \
|
|
vin[128 + r] = vin[128 + r - 1]; \
|
|
vin[192 + r] = vin[192 + r - 1]; \
|
|
} \
|
|
ANY_SIMD(vin, 64, vin + 128, 64, vout, MASK + 1); \
|
|
memcpy(dst_uv + (np >> 1) * 2, vout, SS(r, 1) * 2); \
|
|
}
|
|
|
|
#ifdef HAS_HALFMERGEUVROW_SSSE3
|
|
ANY21HMS(HalfMergeUVRow_Any_SSSE3, HalfMergeUVRow_SSSE3, 15)
|
|
#endif
|
|
#ifdef HAS_HALFMERGEUVROW_AVX2
|
|
ANY21HMS(HalfMergeUVRow_Any_AVX2, HalfMergeUVRow_AVX2, 31)
|
|
#endif
|
|
#ifdef HAS_HALFMERGEUVROW_NEON
|
|
ANY21HMS(HalfMergeUVRow_Any_NEON, HalfMergeUVRow_NEON, 15)
|
|
#endif
|
|
#ifdef HAS_HALFMERGEUVROW_LSX
|
|
ANY21HMS(HalfMergeUVRow_Any_LSX, HalfMergeUVRow_LSX, 15)
|
|
#endif
|
|
#ifdef HAS_HALFMERGEUVROW_LASX
|
|
ANY21HMS(HalfMergeUVRow_Any_LASX, HalfMergeUVRow_LASX, 31)
|
|
#endif
|
|
#undef ANY21HMS
|
|
|
|
// Any 2 planes to 1 with yuvconstants
|
|
#define ANY21C(NAMEANY, ANY_SIMD, UVSHIFT, SBPP, SBPP2, BPP, MASK) \
|
|
void NAMEANY(const uint8_t* y_buf, const uint8_t* uv_buf, uint8_t* dst_ptr, \
|
|
const struct YuvConstants* yuvconstants, int width) { \
|
|
SIMD_ALIGNED(uint8_t vin[128 * 2]); \
|
|
static_assert((MASK + 1) * SBPP <= sizeof(vin) / 2, \
|
|
"vin buffer too small"); \
|
|
static_assert(SS(MASK + 1, UVSHIFT) * SBPP2 <= sizeof(vin) / 2, \
|
|
"vin buffer too small"); \
|
|
SIMD_ALIGNED(uint8_t vout[128]); \
|
|
static_assert((MASK + 1) * BPP <= sizeof(vout), "vout buffer too small"); \
|
|
MSAN_MEMSET(vin, 0, sizeof(vin)); /* for msan */ \
|
|
MSAN_MEMSET(vout, 0, sizeof(vout)); /* for msan */ \
|
|
int r = width & MASK; \
|
|
int n = width & ~MASK; \
|
|
if (n > 0) { \
|
|
ANY_SIMD(y_buf, uv_buf, dst_ptr, yuvconstants, n); \
|
|
} \
|
|
ptrdiff_t np = n; \
|
|
memcpy(vin, y_buf + np * SBPP, r * SBPP); \
|
|
memcpy(vin + 128, uv_buf + (np >> UVSHIFT) * SBPP2, \
|
|
SS(r, UVSHIFT) * SBPP2); \
|
|
ANY_SIMD(vin, vin + 128, vout, yuvconstants, MASK + 1); \
|
|
memcpy(dst_ptr + np * BPP, vout, r * BPP); \
|
|
}
|
|
|
|
// Biplanar to RGB.
|
|
#ifdef HAS_NV12TOARGBROW_SSSE3
|
|
ANY21C(NV12ToARGBRow_Any_SSSE3, NV12ToARGBRow_SSSE3, 1, 1, 2, 4, 7)
|
|
#endif
|
|
#ifdef HAS_NV12TOARGBROW_AVX2
|
|
ANY21C(NV12ToARGBRow_Any_AVX2, NV12ToARGBRow_AVX2, 1, 1, 2, 4, 15)
|
|
#endif
|
|
#ifdef HAS_NV12TOARGBROW_NEON
|
|
ANY21C(NV12ToARGBRow_Any_NEON, NV12ToARGBRow_NEON, 1, 1, 2, 4, 7)
|
|
#endif
|
|
#ifdef HAS_NV12TOARGBROW_LSX
|
|
ANY21C(NV12ToARGBRow_Any_LSX, NV12ToARGBRow_LSX, 1, 1, 2, 4, 7)
|
|
#endif
|
|
#ifdef HAS_NV12TOARGBROW_LASX
|
|
ANY21C(NV12ToARGBRow_Any_LASX, NV12ToARGBRow_LASX, 1, 1, 2, 4, 15)
|
|
#endif
|
|
#ifdef HAS_NV21TOARGBROW_SSSE3
|
|
ANY21C(NV21ToARGBRow_Any_SSSE3, NV21ToARGBRow_SSSE3, 1, 1, 2, 4, 7)
|
|
#endif
|
|
#ifdef HAS_NV21TOARGBROW_AVX2
|
|
ANY21C(NV21ToARGBRow_Any_AVX2, NV21ToARGBRow_AVX2, 1, 1, 2, 4, 15)
|
|
#endif
|
|
#ifdef HAS_NV21TOARGBROW_NEON
|
|
ANY21C(NV21ToARGBRow_Any_NEON, NV21ToARGBRow_NEON, 1, 1, 2, 4, 7)
|
|
#endif
|
|
#ifdef HAS_NV21TOARGBROW_LSX
|
|
ANY21C(NV21ToARGBRow_Any_LSX, NV21ToARGBRow_LSX, 1, 1, 2, 4, 7)
|
|
#endif
|
|
#ifdef HAS_NV21TOARGBROW_LASX
|
|
ANY21C(NV21ToARGBRow_Any_LASX, NV21ToARGBRow_LASX, 1, 1, 2, 4, 15)
|
|
#endif
|
|
#ifdef HAS_NV12TORGB24ROW_NEON
|
|
ANY21C(NV12ToRGB24Row_Any_NEON, NV12ToRGB24Row_NEON, 1, 1, 2, 3, 7)
|
|
#endif
|
|
#ifdef HAS_NV21TORGB24ROW_NEON
|
|
ANY21C(NV21ToRGB24Row_Any_NEON, NV21ToRGB24Row_NEON, 1, 1, 2, 3, 7)
|
|
#endif
|
|
#ifdef HAS_NV12TORGB24ROW_SSSE3
|
|
ANY21C(NV12ToRGB24Row_Any_SSSE3, NV12ToRGB24Row_SSSE3, 1, 1, 2, 3, 15)
|
|
#endif
|
|
#ifdef HAS_NV21TORGB24ROW_SSSE3
|
|
ANY21C(NV21ToRGB24Row_Any_SSSE3, NV21ToRGB24Row_SSSE3, 1, 1, 2, 3, 15)
|
|
#endif
|
|
#ifdef HAS_NV12TORGB24ROW_AVX2
|
|
ANY21C(NV12ToRGB24Row_Any_AVX2, NV12ToRGB24Row_AVX2, 1, 1, 2, 3, 31)
|
|
#endif
|
|
#ifdef HAS_NV21TORGB24ROW_AVX2
|
|
ANY21C(NV21ToRGB24Row_Any_AVX2, NV21ToRGB24Row_AVX2, 1, 1, 2, 3, 31)
|
|
#endif
|
|
#ifdef HAS_NV12TORGB565ROW_SSSE3
|
|
ANY21C(NV12ToRGB565Row_Any_SSSE3, NV12ToRGB565Row_SSSE3, 1, 1, 2, 2, 7)
|
|
#endif
|
|
#ifdef HAS_NV12TORGB565ROW_AVX2
|
|
ANY21C(NV12ToRGB565Row_Any_AVX2, NV12ToRGB565Row_AVX2, 1, 1, 2, 2, 15)
|
|
#endif
|
|
#ifdef HAS_NV12TORGB565ROW_NEON
|
|
ANY21C(NV12ToRGB565Row_Any_NEON, NV12ToRGB565Row_NEON, 1, 1, 2, 2, 7)
|
|
#endif
|
|
#ifdef HAS_NV12TORGB565ROW_LSX
|
|
ANY21C(NV12ToRGB565Row_Any_LSX, NV12ToRGB565Row_LSX, 1, 1, 2, 2, 7)
|
|
#endif
|
|
#ifdef HAS_NV12TORGB565ROW_LASX
|
|
ANY21C(NV12ToRGB565Row_Any_LASX, NV12ToRGB565Row_LASX, 1, 1, 2, 2, 15)
|
|
#endif
|
|
#ifdef HAS_NV12TOARGBROW_WASMSIMD
|
|
ANY21C(NV12ToARGBRow_Any_WASMSIMD, NV12ToARGBRow_WASMSIMD, 1, 1, 2, 4, 7)
|
|
#endif
|
|
#ifdef HAS_NV21TOARGBROW_WASMSIMD
|
|
ANY21C(NV21ToARGBRow_Any_WASMSIMD, NV21ToARGBRow_WASMSIMD, 1, 1, 2, 4, 7)
|
|
#endif
|
|
#ifdef HAS_NV12TORGB565ROW_WASMSIMD
|
|
ANY21C(NV12ToRGB565Row_Any_WASMSIMD, NV12ToRGB565Row_WASMSIMD, 1, 1, 2, 2, 7)
|
|
#endif
|
|
#undef ANY21C
|
|
|
|
// Any 2 planes of 16 bit to 1 with yuvconstants
|
|
#define ANY21CT(NAMEANY, ANY_SIMD, UVSHIFT, DUVSHIFT, T, SBPP, BPP, MASK) \
|
|
void NAMEANY(const T* y_buf, const T* uv_buf, uint8_t* dst_ptr, \
|
|
const struct YuvConstants* yuvconstants, int width) { \
|
|
SIMD_ALIGNED(T vin[32 * 2]); \
|
|
static_assert(SS(MASK + 1, UVSHIFT) * SBPP * 2 <= sizeof(vin) / 2, \
|
|
"vin buffer too small"); \
|
|
SIMD_ALIGNED(uint8_t vout[64]); \
|
|
static_assert(SS(MASK + 1, DUVSHIFT) * BPP <= sizeof(vout), \
|
|
"vout buffer too small"); \
|
|
MSAN_MEMSET(vin, 0, sizeof(vin)); /* for msan */ \
|
|
MSAN_MEMSET(vout, 0, sizeof(vout)); /* for msan */ \
|
|
int r = width & MASK; \
|
|
int n = width & ~MASK; \
|
|
if (n > 0) { \
|
|
ANY_SIMD(y_buf, uv_buf, dst_ptr, yuvconstants, n); \
|
|
} \
|
|
ptrdiff_t np = n; \
|
|
memcpy(vin, y_buf + np, r * SBPP); \
|
|
memcpy(vin + 32, uv_buf + 2 * (np >> UVSHIFT), SS(r, UVSHIFT) * SBPP * 2); \
|
|
ANY_SIMD(vin, vin + 32, vout, yuvconstants, MASK + 1); \
|
|
memcpy(dst_ptr + (np >> DUVSHIFT) * BPP, vout, SS(r, DUVSHIFT) * BPP); \
|
|
}
|
|
|
|
#ifdef HAS_P210TOAR30ROW_SSSE3
|
|
ANY21CT(P210ToAR30Row_Any_SSSE3, P210ToAR30Row_SSSE3, 1, 0, uint16_t, 2, 4, 7)
|
|
#endif
|
|
#ifdef HAS_P210TOARGBROW_SSSE3
|
|
ANY21CT(P210ToARGBRow_Any_SSSE3, P210ToARGBRow_SSSE3, 1, 0, uint16_t, 2, 4, 7)
|
|
#endif
|
|
#ifdef HAS_P210TOARGBROW_AVX2
|
|
ANY21CT(P210ToARGBRow_Any_AVX2, P210ToARGBRow_AVX2, 1, 0, uint16_t, 2, 4, 15)
|
|
#endif
|
|
#ifdef HAS_P210TOAR30ROW_AVX2
|
|
ANY21CT(P210ToAR30Row_Any_AVX2, P210ToAR30Row_AVX2, 1, 0, uint16_t, 2, 4, 15)
|
|
#endif
|
|
#ifdef HAS_P210TOAR30ROW_NEON
|
|
ANY21CT(P210ToAR30Row_Any_NEON, P210ToAR30Row_NEON, 1, 0, uint16_t, 2, 4, 7)
|
|
#endif
|
|
#ifdef HAS_P210TOARGBROW_NEON
|
|
ANY21CT(P210ToARGBRow_Any_NEON, P210ToARGBRow_NEON, 1, 0, uint16_t, 2, 4, 7)
|
|
#endif
|
|
#ifdef HAS_P410TOAR30ROW_SSSE3
|
|
ANY21CT(P410ToAR30Row_Any_SSSE3, P410ToAR30Row_SSSE3, 0, 0, uint16_t, 2, 4, 7)
|
|
#endif
|
|
#ifdef HAS_P410TOARGBROW_SSSE3
|
|
ANY21CT(P410ToARGBRow_Any_SSSE3, P410ToARGBRow_SSSE3, 0, 0, uint16_t, 2, 4, 7)
|
|
#endif
|
|
#ifdef HAS_P410TOARGBROW_AVX2
|
|
ANY21CT(P410ToARGBRow_Any_AVX2, P410ToARGBRow_AVX2, 0, 0, uint16_t, 2, 4, 15)
|
|
#endif
|
|
#ifdef HAS_P410TOAR30ROW_AVX2
|
|
ANY21CT(P410ToAR30Row_Any_AVX2, P410ToAR30Row_AVX2, 0, 0, uint16_t, 2, 4, 15)
|
|
#endif
|
|
#ifdef HAS_P410TOAR30ROW_NEON
|
|
ANY21CT(P410ToAR30Row_Any_NEON, P410ToAR30Row_NEON, 0, 0, uint16_t, 2, 4, 7)
|
|
#endif
|
|
#ifdef HAS_P410TOARGBROW_NEON
|
|
ANY21CT(P410ToARGBRow_Any_NEON, P410ToARGBRow_NEON, 0, 0, uint16_t, 2, 4, 7)
|
|
#endif
|
|
|
|
#undef ANY21CT
|
|
|
|
// Any 2 16 bit planes with parameter to 1
|
|
#define ANY21PT(NAMEANY, ANY_SIMD, T, BPP, MASK) \
|
|
void NAMEANY(const T* src_u, const T* src_v, T* dst_uv, int depth, \
|
|
int width) { \
|
|
SIMD_ALIGNED(T vin[16 * 2]); \
|
|
static_assert((MASK + 1) * BPP <= sizeof(vin) / 2, \
|
|
"vin buffer too small"); \
|
|
SIMD_ALIGNED(T vout[16 * 2]); \
|
|
static_assert((MASK + 1) * BPP * 2 <= sizeof(vout), \
|
|
"vout buffer too small"); \
|
|
MSAN_MEMSET(vin, 0, sizeof(vin)); /* for msan */ \
|
|
MSAN_MEMSET(vout, 0, sizeof(vout)); /* for msan */ \
|
|
int r = width & MASK; \
|
|
int n = width & ~MASK; \
|
|
if (n > 0) { \
|
|
ANY_SIMD(src_u, src_v, dst_uv, depth, n); \
|
|
} \
|
|
ptrdiff_t np = n; \
|
|
memcpy(vin, src_u + np, r * BPP); \
|
|
memcpy(vin + 16, src_v + np, r * BPP); \
|
|
ANY_SIMD(vin, vin + 16, vout, depth, MASK + 1); \
|
|
memcpy(dst_uv + np * 2, vout, r * BPP * 2); \
|
|
}
|
|
|
|
#ifdef HAS_MERGEUVROW_16_AVX2
|
|
ANY21PT(MergeUVRow_16_Any_AVX2, MergeUVRow_16_AVX2, uint16_t, 2, 7)
|
|
#endif
|
|
#ifdef HAS_MERGEUVROW_16_NEON
|
|
ANY21PT(MergeUVRow_16_Any_NEON, MergeUVRow_16_NEON, uint16_t, 2, 7)
|
|
#endif
|
|
#ifdef HAS_MERGEUVROW_16_LSX
|
|
ANY21PT(MergeUVRow_16_Any_LSX, MergeUVRow_16_LSX, uint16_t, 2, 7)
|
|
#endif
|
|
#ifdef HAS_MERGEUVROW_16_LASX
|
|
ANY21PT(MergeUVRow_16_Any_LASX, MergeUVRow_16_LASX, uint16_t, 2, 15)
|
|
#endif
|
|
|
|
#undef ANY21CT
|
|
|
|
// Any 1 to 1.
|
|
#define ANY11(NAMEANY, ANY_SIMD, UVSHIFT, SBPP, BPP, MASK) \
|
|
void NAMEANY(const uint8_t* src_ptr, uint8_t* dst_ptr, int width) { \
|
|
SIMD_ALIGNED(uint8_t vin[256]); \
|
|
static_assert(SS(MASK + 1, UVSHIFT) * SBPP <= sizeof(vin), \
|
|
"vin buffer too small"); \
|
|
SIMD_ALIGNED(uint8_t vout[256]); \
|
|
static_assert((MASK + 1) * BPP <= sizeof(vout), "vout buffer too small"); \
|
|
MSAN_MEMSET(vin, 0, sizeof(vin)); /* for YUY2 and msan */ \
|
|
MSAN_MEMSET(vout, 0, sizeof(vout)); /* for msan */ \
|
|
int r = width & MASK; \
|
|
int n = width & ~MASK; \
|
|
if (n > 0) { \
|
|
ANY_SIMD(src_ptr, dst_ptr, n); \
|
|
} \
|
|
ptrdiff_t np = n; \
|
|
memcpy(vin, src_ptr + (np >> UVSHIFT) * SBPP, SS(r, UVSHIFT) * SBPP); \
|
|
ANY_SIMD(vin, vout, MASK + 1); \
|
|
memcpy(dst_ptr + np * BPP, vout, 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
|
|
#ifdef HAS_COPYROW_LSX
|
|
ANY11(CopyRow_Any_LSX, CopyRow_LSX, 0, 1, 1, 15)
|
|
#endif
|
|
#ifdef HAS_COPYROW_LASX
|
|
ANY11(CopyRow_Any_LASX, CopyRow_LASX, 0, 1, 1, 31)
|
|
#endif
|
|
#if defined(HAS_ARGBTORGB24ROW_SSSE3)
|
|
ANY11(ARGBToRGB24Row_Any_SSSE3, ARGBToRGB24Row_SSSE3, 0, 4, 3, 15)
|
|
ANY11(ARGBToRAWRow_Any_SSSE3, ARGBToRAWRow_SSSE3, 0, 4, 3, 15)
|
|
|
|
#endif
|
|
#if defined(HAS_ARGBTORGB24ROW_AVX2)
|
|
ANY11(ARGBToRGB24Row_Any_AVX2, ARGBToRGB24Row_AVX2, 0, 4, 3, 31)
|
|
#endif
|
|
#if defined(HAS_ARGBTORAWROW_AVX2)
|
|
ANY11(ARGBToRAWRow_Any_AVX2, ARGBToRAWRow_AVX2, 0, 4, 3, 31)
|
|
#endif
|
|
#if defined(HAS_ARGBTORGB565ROW_AVX2)
|
|
ANY11(ARGBToRGB565Row_Any_AVX2, ARGBToRGB565Row_AVX2, 0, 4, 2, 7)
|
|
#endif
|
|
#if defined(HAS_ARGBTOARGB1555ROW_AVX2)
|
|
ANY11(ARGBToARGB1555Row_Any_AVX2, ARGBToARGB1555Row_AVX2, 0, 4, 2, 7)
|
|
#endif
|
|
#if defined(HAS_ARGBTOARGB4444ROW_AVX2)
|
|
ANY11(ARGBToARGB4444Row_Any_AVX2, ARGBToARGB4444Row_AVX2, 0, 4, 2, 7)
|
|
#endif
|
|
#if defined(HAS_ABGRTOAR30ROW_SSSE3)
|
|
ANY11(ABGRToAR30Row_Any_SSSE3, ABGRToAR30Row_SSSE3, 0, 4, 4, 3)
|
|
#endif
|
|
#if defined(HAS_ABGRTOAR30ROW_NEON)
|
|
ANY11(ABGRToAR30Row_Any_NEON, ABGRToAR30Row_NEON, 0, 4, 4, 7)
|
|
#endif
|
|
#if defined(HAS_ARGBTOAR30ROW_NEON)
|
|
ANY11(ARGBToAR30Row_Any_NEON, ARGBToAR30Row_NEON, 0, 4, 4, 7)
|
|
#endif
|
|
#if defined(HAS_AR30TOARGBROW_NEON)
|
|
ANY11(AR30ToARGBRow_Any_NEON, AR30ToARGBRow_NEON, 0, 4, 4, 7)
|
|
#endif
|
|
#if defined(HAS_AR30TOABGRROW_NEON)
|
|
ANY11(AR30ToABGRRow_Any_NEON, AR30ToABGRRow_NEON, 0, 4, 4, 7)
|
|
#endif
|
|
#if defined(HAS_AR30TOAB30ROW_NEON)
|
|
ANY11(AR30ToAB30Row_Any_NEON, AR30ToAB30Row_NEON, 0, 4, 4, 7)
|
|
#endif
|
|
#if defined(HAS_AR30TOARGBROW_AVX2)
|
|
ANY11(AR30ToARGBRow_Any_AVX2, AR30ToARGBRow_AVX2, 0, 4, 4, 7)
|
|
#endif
|
|
#if defined(HAS_AR30TOABGRROW_AVX2)
|
|
ANY11(AR30ToABGRRow_Any_AVX2, AR30ToABGRRow_AVX2, 0, 4, 4, 7)
|
|
#endif
|
|
#if defined(HAS_AR30TOAB30ROW_AVX2)
|
|
ANY11(AR30ToAB30Row_Any_AVX2, AR30ToAB30Row_AVX2, 0, 4, 4, 7)
|
|
#endif
|
|
#if defined(HAS_ARGBTOAR30ROW_SSSE3)
|
|
ANY11(ARGBToAR30Row_Any_SSSE3, ARGBToAR30Row_SSSE3, 0, 4, 4, 3)
|
|
#endif
|
|
#if defined(HAS_ABGRTOAR30ROW_AVX2)
|
|
ANY11(ABGRToAR30Row_Any_AVX2, ABGRToAR30Row_AVX2, 0, 4, 4, 7)
|
|
#endif
|
|
#if defined(HAS_ARGBTOAR30ROW_AVX2)
|
|
ANY11(ARGBToAR30Row_Any_AVX2, ARGBToAR30Row_AVX2, 0, 4, 4, 7)
|
|
#endif
|
|
#if defined(HAS_ABGRTOAR30ROW_LSX)
|
|
ANY11(ABGRToAR30Row_Any_LSX, ABGRToAR30Row_LSX, 0, 4, 4, 3)
|
|
#endif
|
|
#if defined(HAS_ABGRTOAR30ROW_LASX)
|
|
ANY11(ABGRToAR30Row_Any_LASX, ABGRToAR30Row_LASX, 0, 4, 4, 7)
|
|
#endif
|
|
#if defined(HAS_ARGBTOAR30ROW_LSX)
|
|
ANY11(ARGBToAR30Row_Any_LSX, ARGBToAR30Row_LSX, 0, 4, 4, 3)
|
|
#endif
|
|
#if defined(HAS_ARGBTOAR30ROW_LASX)
|
|
ANY11(ARGBToAR30Row_Any_LASX, ARGBToAR30Row_LASX, 0, 4, 4, 7)
|
|
#endif
|
|
#if defined(HAS_AR30TOARGBROW_LSX)
|
|
ANY11(AR30ToARGBRow_Any_LSX, AR30ToARGBRow_LSX, 0, 4, 4, 3)
|
|
#endif
|
|
#if defined(HAS_AR30TOARGBROW_LASX)
|
|
ANY11(AR30ToARGBRow_Any_LASX, AR30ToARGBRow_LASX, 0, 4, 4, 7)
|
|
#endif
|
|
#if defined(HAS_AR30TOABGRROW_LSX)
|
|
ANY11(AR30ToABGRRow_Any_LSX, AR30ToABGRRow_LSX, 0, 4, 4, 3)
|
|
#endif
|
|
#if defined(HAS_AR30TOABGRROW_LASX)
|
|
ANY11(AR30ToABGRRow_Any_LASX, AR30ToABGRRow_LASX, 0, 4, 4, 7)
|
|
#endif
|
|
#if defined(HAS_AR30TOAB30ROW_LSX)
|
|
ANY11(AR30ToAB30Row_Any_LSX, AR30ToAB30Row_LSX, 0, 4, 4, 3)
|
|
#endif
|
|
#if defined(HAS_AR30TOAB30ROW_LASX)
|
|
ANY11(AR30ToAB30Row_Any_LASX, AR30ToAB30Row_LASX, 0, 4, 4, 7)
|
|
#endif
|
|
|
|
#if defined(HAS_J400TOARGBROW_AVX2)
|
|
ANY11(J400ToARGBRow_Any_AVX2, J400ToARGBRow_AVX2, 0, 1, 4, 15)
|
|
#endif
|
|
#if defined(HAS_RGB24TOARGBROW_SSSE3)
|
|
ANY11(RGB24ToARGBRow_Any_SSSE3, RGB24ToARGBRow_SSSE3, 0, 3, 4, 15)
|
|
ANY11(RAWToARGBRow_Any_SSSE3, RAWToARGBRow_SSSE3, 0, 3, 4, 15)
|
|
|
|
#endif
|
|
#if defined(HAS_RAWTOARGBROW_AVX2)
|
|
ANY11(RAWToARGBRow_Any_AVX2, RAWToARGBRow_AVX2, 0, 3, 4, 31)
|
|
#endif
|
|
#if defined(HAS_RGB24TOARGBROW_AVX2)
|
|
ANY11(RGB24ToARGBRow_Any_AVX2, RGB24ToARGBRow_AVX2, 0, 3, 4, 31)
|
|
#endif
|
|
#if defined(HAS_RAWTORGBAROW_AVX2)
|
|
ANY11(RAWToRGBARow_Any_AVX2, RAWToRGBARow_AVX2, 0, 3, 4, 15)
|
|
#endif
|
|
#if defined(HAS_RAWTORGB24ROW_SSSE3)
|
|
ANY11(RAWToRGB24Row_Any_SSSE3, RAWToRGB24Row_SSSE3, 0, 3, 3, 7)
|
|
#endif
|
|
#if defined(HAS_RAWTORGB24ROW_AVX2)
|
|
ANY11(RAWToRGB24Row_Any_AVX2, RAWToRGB24Row_AVX2, 0, 3, 3, 31)
|
|
#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_ARGBTORGB24ROW_NEON)
|
|
ANY11(ARGBToRGB24Row_Any_NEON, ARGBToRGB24Row_NEON, 0, 4, 3, 15)
|
|
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)
|
|
#endif
|
|
#if defined(HAS_ARGBTORGB24ROW_LSX)
|
|
ANY11(ARGBToRGB24Row_Any_LSX, ARGBToRGB24Row_LSX, 0, 4, 3, 15)
|
|
ANY11(ARGBToRAWRow_Any_LSX, ARGBToRAWRow_LSX, 0, 4, 3, 15)
|
|
ANY11(ARGBToRGB565Row_Any_LSX, ARGBToRGB565Row_LSX, 0, 4, 2, 7)
|
|
ANY11(ARGBToARGB1555Row_Any_LSX, ARGBToARGB1555Row_LSX, 0, 4, 2, 7)
|
|
ANY11(ARGBToARGB4444Row_Any_LSX, ARGBToARGB4444Row_LSX, 0, 4, 2, 7)
|
|
#endif
|
|
#if defined(HAS_ARGBTORGB24ROW_LASX)
|
|
ANY11(ARGBToRGB24Row_Any_LASX, ARGBToRGB24Row_LASX, 0, 4, 3, 31)
|
|
ANY11(ARGBToRAWRow_Any_LASX, ARGBToRAWRow_LASX, 0, 4, 3, 31)
|
|
ANY11(ARGBToRGB565Row_Any_LASX, ARGBToRGB565Row_LASX, 0, 4, 2, 15)
|
|
ANY11(ARGBToARGB1555Row_Any_LASX, ARGBToARGB1555Row_LASX, 0, 4, 2, 15)
|
|
ANY11(ARGBToARGB4444Row_Any_LASX, ARGBToARGB4444Row_LASX, 0, 4, 2, 15)
|
|
#endif
|
|
#if defined(HAS_J400TOARGBROW_LSX)
|
|
ANY11(J400ToARGBRow_Any_LSX, J400ToARGBRow_LSX, 0, 1, 4, 7)
|
|
#endif
|
|
#if defined(HAS_J400TOARGBROW_LASX)
|
|
ANY11(J400ToARGBRow_Any_LASX, J400ToARGBRow_LASX, 0, 1, 4, 15)
|
|
#endif
|
|
#if defined(HAS_RAWTORGB24ROW_NEON)
|
|
ANY11(RAWToRGB24Row_Any_NEON, RAWToRGB24Row_NEON, 0, 3, 3, 15)
|
|
#endif
|
|
#if defined(HAS_RAWTORGB24ROW_LSX)
|
|
ANY11(RAWToRGB24Row_Any_LSX, RAWToRGB24Row_LSX, 0, 3, 3, 15)
|
|
#endif
|
|
#if defined(HAS_RAWTORGB24ROW_LASX)
|
|
ANY11(RAWToRGB24Row_Any_LASX, RAWToRGB24Row_LASX, 0, 3, 3, 31)
|
|
#endif
|
|
#if defined(HAS_RAWTORGB24ROW_WASMSIMD)
|
|
ANY11(RAWToRGB24Row_Any_WASMSIMD, RAWToRGB24Row_WASMSIMD, 0, 3, 3, 15)
|
|
#endif
|
|
|
|
#ifdef HAS_UYVYTOYROW_AVX2
|
|
ANY11(UYVYToYRow_Any_AVX2, UYVYToYRow_AVX2, 1, 4, 1, 31)
|
|
#endif
|
|
#ifdef HAS_YUY2TOYROW_AVX2
|
|
ANY11(YUY2ToYRow_Any_AVX2, YUY2ToYRow_AVX2, 1, 4, 1, 31)
|
|
#endif
|
|
#ifdef HAS_YUY2TOYROW_SSE2
|
|
ANY11(YUY2ToYRow_Any_SSE2, YUY2ToYRow_SSE2, 1, 4, 1, 15)
|
|
ANY11(UYVYToYRow_Any_SSE2, UYVYToYRow_SSE2, 1, 4, 1, 15)
|
|
#endif
|
|
|
|
#ifdef HAS_RGB565TOYROW_AVX2
|
|
ANY11(RGB565ToYRow_Any_AVX2, RGB565ToYRow_AVX2, 0, 2, 1, 31)
|
|
#endif
|
|
#ifdef HAS_ARGB1555TOYROW_AVX2
|
|
ANY11(ARGB1555ToYRow_Any_AVX2, ARGB1555ToYRow_AVX2, 0, 2, 1, 31)
|
|
#endif
|
|
#ifdef HAS_ARGB4444TOYROW_AVX2
|
|
ANY11(ARGB4444ToYRow_Any_AVX2, ARGB4444ToYRow_AVX2, 0, 2, 1, 31)
|
|
#endif
|
|
#ifdef HAS_RGB565TOYROW_NEON
|
|
ANY11(RGB565ToYRow_Any_NEON, RGB565ToYRow_NEON, 0, 2, 1, 15)
|
|
#endif
|
|
#ifdef HAS_RGB565TOYROW_LSX
|
|
ANY11(RGB565ToYRow_Any_LSX, RGB565ToYRow_LSX, 0, 2, 1, 7)
|
|
#endif
|
|
#ifdef HAS_RGB565TOYROW_LASX
|
|
ANY11(RGB565ToYRow_Any_LASX, RGB565ToYRow_LASX, 0, 2, 1, 15)
|
|
#endif
|
|
#ifdef HAS_ARGB1555TOYROW_NEON
|
|
#ifdef __aarch64__
|
|
ANY11(ARGB1555ToYRow_Any_NEON, ARGB1555ToYRow_NEON, 0, 2, 1, 15)
|
|
#else
|
|
ANY11(ARGB1555ToYRow_Any_NEON, ARGB1555ToYRow_NEON, 0, 2, 1, 7)
|
|
#endif
|
|
#endif
|
|
#ifdef HAS_ARGB1555TOYROW_LSX
|
|
ANY11(ARGB1555ToYRow_Any_LSX, ARGB1555ToYRow_LSX, 0, 2, 1, 7)
|
|
#endif
|
|
#ifdef HAS_ARGB1555TOYROW_LASX
|
|
ANY11(ARGB1555ToYRow_Any_LASX, ARGB1555ToYRow_LASX, 0, 2, 1, 15)
|
|
#endif
|
|
#ifdef HAS_ARGB4444TOYROW_NEON
|
|
#ifdef __aarch64__
|
|
ANY11(ARGB4444ToYRow_Any_NEON, ARGB4444ToYRow_NEON, 0, 2, 1, 15)
|
|
#else
|
|
ANY11(ARGB4444ToYRow_Any_NEON, ARGB4444ToYRow_NEON, 0, 2, 1, 7)
|
|
#endif
|
|
#endif
|
|
#ifdef HAS_YUY2TOYROW_NEON
|
|
ANY11(YUY2ToYRow_Any_NEON, YUY2ToYRow_NEON, 1, 4, 1, 15)
|
|
#endif
|
|
#ifdef HAS_UYVYTOYROW_NEON
|
|
ANY11(UYVYToYRow_Any_NEON, UYVYToYRow_NEON, 1, 4, 1, 15)
|
|
#endif
|
|
#ifdef HAS_YUY2TOYROW_LSX
|
|
ANY11(YUY2ToYRow_Any_LSX, YUY2ToYRow_LSX, 1, 4, 1, 15)
|
|
#endif
|
|
#ifdef HAS_YUY2TOYROW_LASX
|
|
ANY11(YUY2ToYRow_Any_LASX, YUY2ToYRow_LASX, 1, 4, 1, 31)
|
|
#endif
|
|
#ifdef HAS_UYVYTOYROW_LSX
|
|
ANY11(UYVYToYRow_Any_LSX, UYVYToYRow_LSX, 1, 4, 1, 15)
|
|
#endif
|
|
#ifdef HAS_UYVYTOYROW_LASX
|
|
ANY11(UYVYToYRow_Any_LASX, UYVYToYRow_LASX, 1, 4, 1, 31)
|
|
#endif
|
|
#ifdef HAS_AYUVTOYROW_AVX2
|
|
ANY11(AYUVToYRow_Any_AVX2, AYUVToYRow_AVX2, 0, 4, 1, 15)
|
|
#endif
|
|
#ifdef HAS_AYUVTOYROW_NEON
|
|
ANY11(AYUVToYRow_Any_NEON, AYUVToYRow_NEON, 0, 4, 1, 15)
|
|
#endif
|
|
#ifdef HAS_AYUVTOYROW_LSX
|
|
ANY11(AYUVToYRow_Any_LSX, AYUVToYRow_LSX, 0, 4, 1, 15)
|
|
#endif
|
|
#ifdef HAS_AYUVTOYROW_LASX
|
|
ANY11(AYUVToYRow_Any_LASX, AYUVToYRow_LASX, 0, 4, 1, 15)
|
|
#endif
|
|
#ifdef HAS_SWAPUVROW_SSSE3
|
|
ANY11(SwapUVRow_Any_SSSE3, SwapUVRow_SSSE3, 0, 2, 2, 15)
|
|
#endif
|
|
#ifdef HAS_SWAPUVROW_AVX2
|
|
ANY11(SwapUVRow_Any_AVX2, SwapUVRow_AVX2, 0, 2, 2, 31)
|
|
#endif
|
|
#ifdef HAS_SWAPUVROW_NEON
|
|
ANY11(SwapUVRow_Any_NEON, SwapUVRow_NEON, 0, 2, 2, 15)
|
|
#endif
|
|
#ifdef HAS_SWAPUVROW_LSX
|
|
ANY11(SwapUVRow_Any_LSX, SwapUVRow_LSX, 0, 2, 2, 15)
|
|
#endif
|
|
#ifdef HAS_SWAPUVROW_LASX
|
|
ANY11(SwapUVRow_Any_LASX, SwapUVRow_LASX, 0, 2, 2, 31)
|
|
#endif
|
|
#ifdef HAS_RGB24TOARGBROW_NEON
|
|
ANY11(RGB24ToARGBRow_Any_NEON, RGB24ToARGBRow_NEON, 0, 3, 4, 7)
|
|
#endif
|
|
#ifdef HAS_RGB24TOARGBROW_LSX
|
|
ANY11(RGB24ToARGBRow_Any_LSX, RGB24ToARGBRow_LSX, 0, 3, 4, 15)
|
|
#endif
|
|
#ifdef HAS_RGB24TOARGBROW_LASX
|
|
ANY11(RGB24ToARGBRow_Any_LASX, RGB24ToARGBRow_LASX, 0, 3, 4, 31)
|
|
#endif
|
|
#ifdef HAS_RAWTOARGBROW_NEON
|
|
ANY11(RAWToARGBRow_Any_NEON, RAWToARGBRow_NEON, 0, 3, 4, 7)
|
|
#endif
|
|
#ifdef HAS_RAWTORGBAROW_NEON
|
|
ANY11(RAWToRGBARow_Any_NEON, RAWToRGBARow_NEON, 0, 3, 4, 7)
|
|
#endif
|
|
#ifdef HAS_RAWTOARGBROW_LSX
|
|
ANY11(RAWToARGBRow_Any_LSX, RAWToARGBRow_LSX, 0, 3, 4, 15)
|
|
#endif
|
|
#ifdef HAS_RAWTOARGBROW_LASX
|
|
ANY11(RAWToARGBRow_Any_LASX, RAWToARGBRow_LASX, 0, 3, 4, 31)
|
|
#endif
|
|
#ifdef HAS_RAWTORGBAROW_LSX
|
|
ANY11(RAWToRGBARow_Any_LSX, RAWToRGBARow_LSX, 0, 3, 4, 15)
|
|
#endif
|
|
#ifdef HAS_RAWTORGBAROW_LASX
|
|
ANY11(RAWToRGBARow_Any_LASX, RAWToRGBARow_LASX, 0, 3, 4, 31)
|
|
#endif
|
|
#ifdef HAS_RGB565TOARGBROW_NEON
|
|
ANY11(RGB565ToARGBRow_Any_NEON, RGB565ToARGBRow_NEON, 0, 2, 4, 15)
|
|
#endif
|
|
#ifdef HAS_RGB565TOARGBROW_LSX
|
|
ANY11(RGB565ToARGBRow_Any_LSX, RGB565ToARGBRow_LSX, 0, 2, 4, 7)
|
|
#endif
|
|
#ifdef HAS_RGB565TOARGBROW_LASX
|
|
ANY11(RGB565ToARGBRow_Any_LASX, RGB565ToARGBRow_LASX, 0, 2, 4, 15)
|
|
#endif
|
|
#ifdef HAS_ARGB1555TOARGBROW_NEON
|
|
ANY11(ARGB1555ToARGBRow_Any_NEON, ARGB1555ToARGBRow_NEON, 0, 2, 4, 15)
|
|
#endif
|
|
#ifdef HAS_ARGB1555TOARGBROW_LSX
|
|
ANY11(ARGB1555ToARGBRow_Any_LSX, ARGB1555ToARGBRow_LSX, 0, 2, 4, 7)
|
|
#endif
|
|
#ifdef HAS_ARGB1555TOARGBROW_LASX
|
|
ANY11(ARGB1555ToARGBRow_Any_LASX, ARGB1555ToARGBRow_LASX, 0, 2, 4, 15)
|
|
#endif
|
|
#ifdef HAS_ARGB4444TOARGBROW_NEON
|
|
ANY11(ARGB4444ToARGBRow_Any_NEON, ARGB4444ToARGBRow_NEON, 0, 2, 4, 7)
|
|
#endif
|
|
#ifdef HAS_ARGB4444TOARGBROW_LSX
|
|
ANY11(ARGB4444ToARGBRow_Any_LSX, ARGB4444ToARGBRow_LSX, 0, 2, 4, 7)
|
|
#endif
|
|
#ifdef HAS_ARGB4444TOARGBROW_LASX
|
|
ANY11(ARGB4444ToARGBRow_Any_LASX, ARGB4444ToARGBRow_LASX, 0, 2, 4, 15)
|
|
#endif
|
|
#ifdef HAS_ARGBATTENUATEROW_SSSE3
|
|
ANY11(ARGBAttenuateRow_Any_SSSE3, ARGBAttenuateRow_SSSE3, 0, 4, 4, 3)
|
|
#endif
|
|
#ifdef HAS_ARGBUNATTENUATEROW_SSE2
|
|
ANY11(ARGBUnattenuateRow_Any_SSE2, ARGBUnattenuateRow_SSE2, 0, 4, 4, 3)
|
|
#endif
|
|
#ifdef HAS_ARGBATTENUATEROW_AVX2
|
|
ANY11(ARGBAttenuateRow_Any_AVX2, ARGBAttenuateRow_AVX2, 0, 4, 4, 7)
|
|
#endif
|
|
#ifdef HAS_ARGBUNATTENUATEROW_AVX2
|
|
ANY11(ARGBUnattenuateRow_Any_AVX2, ARGBUnattenuateRow_AVX2, 0, 4, 4, 7)
|
|
#endif
|
|
#ifdef HAS_ARGBATTENUATEROW_NEON
|
|
ANY11(ARGBAttenuateRow_Any_NEON, ARGBAttenuateRow_NEON, 0, 4, 4, 7)
|
|
#endif
|
|
#ifdef HAS_ARGBUNATTENUATEROW_NEON
|
|
ANY11(ARGBUnattenuateRow_Any_NEON, ARGBUnattenuateRow_NEON, 0, 4, 4, 3)
|
|
#endif
|
|
#ifdef HAS_ARGBATTENUATEROW_LSX
|
|
ANY11(ARGBAttenuateRow_Any_LSX, ARGBAttenuateRow_LSX, 0, 4, 4, 3)
|
|
#endif
|
|
#ifdef HAS_ARGBATTENUATEROW_LASX
|
|
ANY11(ARGBAttenuateRow_Any_LASX, ARGBAttenuateRow_LASX, 0, 4, 4, 7)
|
|
#endif
|
|
#ifdef HAS_ARGBEXTRACTALPHAROW_SSE2
|
|
ANY11(ARGBExtractAlphaRow_Any_SSE2, ARGBExtractAlphaRow_SSE2, 0, 4, 1, 7)
|
|
#endif
|
|
#ifdef HAS_ARGBEXTRACTALPHAROW_AVX2
|
|
ANY11(ARGBExtractAlphaRow_Any_AVX2, ARGBExtractAlphaRow_AVX2, 0, 4, 1, 31)
|
|
#endif
|
|
#ifdef HAS_ARGBEXTRACTALPHAROW_NEON
|
|
ANY11(ARGBExtractAlphaRow_Any_NEON, ARGBExtractAlphaRow_NEON, 0, 4, 1, 15)
|
|
#endif
|
|
#ifdef HAS_ARGBEXTRACTALPHAROW_LSX
|
|
ANY11(ARGBExtractAlphaRow_Any_LSX, ARGBExtractAlphaRow_LSX, 0, 4, 1, 7)
|
|
#endif
|
|
#ifdef HAS_ARGBEXTRACTALPHAROW_LASX
|
|
ANY11(ARGBExtractAlphaRow_Any_LASX, ARGBExtractAlphaRow_LASX, 0, 4, 1, 15)
|
|
#endif
|
|
#ifdef HAS_ARGBEXTRACTALPHAROW_WASMSIMD
|
|
ANY11(ARGBExtractAlphaRow_Any_WASMSIMD,
|
|
ARGBExtractAlphaRow_WASMSIMD,
|
|
0,
|
|
4,
|
|
1,
|
|
15)
|
|
#endif
|
|
#ifdef HAS_RAWTOARGBROW_WASMSIMD
|
|
ANY11(RAWToARGBRow_Any_WASMSIMD, RAWToARGBRow_WASMSIMD, 0, 3, 4, 15)
|
|
#endif
|
|
#ifdef HAS_RGB24TOARGBROW_WASMSIMD
|
|
ANY11(RGB24ToARGBRow_Any_WASMSIMD, RGB24ToARGBRow_WASMSIMD, 0, 3, 4, 15)
|
|
#endif
|
|
#ifdef HAS_SWAPUVROW_WASMSIMD
|
|
ANY11(SwapUVRow_Any_WASMSIMD, SwapUVRow_WASMSIMD, 0, 2, 2, 15)
|
|
#endif
|
|
#ifdef HAS_J400TOARGBROW_WASMSIMD
|
|
ANY11(J400ToARGBRow_Any_WASMSIMD, J400ToARGBRow_WASMSIMD, 0, 1, 4, 15)
|
|
#endif
|
|
#ifdef HAS_RGB565TOARGBROW_WASMSIMD
|
|
ANY11(RGB565ToARGBRow_Any_WASMSIMD, RGB565ToARGBRow_WASMSIMD, 0, 2, 4, 7)
|
|
#endif
|
|
#ifdef HAS_ARGB1555TOARGBROW_WASMSIMD
|
|
ANY11(ARGB1555ToARGBRow_Any_WASMSIMD, ARGB1555ToARGBRow_WASMSIMD, 0, 2, 4, 7)
|
|
#endif
|
|
#ifdef HAS_ARGB4444TOARGBROW_WASMSIMD
|
|
ANY11(ARGB4444ToARGBRow_Any_WASMSIMD, ARGB4444ToARGBRow_WASMSIMD, 0, 2, 4, 7)
|
|
#endif
|
|
#ifdef HAS_ARGBTORGB565ROW_WASMSIMD
|
|
ANY11(ARGBToRGB565Row_Any_WASMSIMD, ARGBToRGB565Row_WASMSIMD, 0, 4, 2, 7)
|
|
#endif
|
|
#ifdef HAS_ARGBTOARGB1555ROW_WASMSIMD
|
|
ANY11(ARGBToARGB1555Row_Any_WASMSIMD, ARGBToARGB1555Row_WASMSIMD, 0, 4, 2, 7)
|
|
#endif
|
|
#ifdef HAS_ARGBTOARGB4444ROW_WASMSIMD
|
|
ANY11(ARGBToARGB4444Row_Any_WASMSIMD, ARGBToARGB4444Row_WASMSIMD, 0, 4, 2, 7)
|
|
#endif
|
|
#ifdef HAS_ARGBGRAYROW_NEON
|
|
ANY11(ARGBGrayRow_Any_NEON, ARGBGrayRow_NEON, 0, 4, 4, 7)
|
|
#endif
|
|
#ifdef HAS_ARGBGRAYROW_NEON_DOTPROD
|
|
ANY11(ARGBGrayRow_Any_NEON_DotProd, ARGBGrayRow_NEON_DotProd, 0, 4, 4, 7)
|
|
#endif
|
|
#ifdef HAS_ARGBGRAYROW_LSX
|
|
ANY11(ARGBGrayRow_Any_LSX, ARGBGrayRow_LSX, 0, 4, 4, 7)
|
|
#endif
|
|
#ifdef HAS_ARGBGRAYROW_LASX
|
|
ANY11(ARGBGrayRow_Any_LASX, ARGBGrayRow_LASX, 0, 4, 4, 15)
|
|
#endif
|
|
#undef ANY11
|
|
|
|
// Any 1 to 1 blended. Destination is read, modify, write.
|
|
#define ANY11B(NAMEANY, ANY_SIMD, UVSHIFT, SBPP, BPP, MASK) \
|
|
void NAMEANY(const uint8_t* src_ptr, uint8_t* dst_ptr, int width) { \
|
|
SIMD_ALIGNED(uint8_t vin[64]); \
|
|
static_assert(SS(MASK + 1, UVSHIFT) * SBPP <= sizeof(vin), \
|
|
"vin buffer too small"); \
|
|
SIMD_ALIGNED(uint8_t vout[64]); \
|
|
static_assert((MASK + 1) * BPP <= sizeof(vout), "vout buffer too small"); \
|
|
MSAN_MEMSET(vin, 0, sizeof(vin)); /* for msan */ \
|
|
MSAN_MEMSET(vout, 0, sizeof(vout)); /* for msan */ \
|
|
int r = width & MASK; \
|
|
int n = width & ~MASK; \
|
|
if (n > 0) { \
|
|
ANY_SIMD(src_ptr, dst_ptr, n); \
|
|
} \
|
|
ptrdiff_t np = n; \
|
|
memcpy(vin, src_ptr + (np >> UVSHIFT) * SBPP, SS(r, UVSHIFT) * SBPP); \
|
|
memcpy(vout, dst_ptr + np * BPP, r * BPP); \
|
|
ANY_SIMD(vin, vout, MASK + 1); \
|
|
memcpy(dst_ptr + np * BPP, vout, r * BPP); \
|
|
}
|
|
|
|
#ifdef HAS_ARGBCOPYALPHAROW_AVX2
|
|
ANY11B(ARGBCopyAlphaRow_Any_AVX2, ARGBCopyAlphaRow_AVX2, 0, 4, 4, 15)
|
|
#endif
|
|
#ifdef HAS_ARGBCOPYALPHAROW_SSE2
|
|
ANY11B(ARGBCopyAlphaRow_Any_SSE2, ARGBCopyAlphaRow_SSE2, 0, 4, 4, 7)
|
|
#endif
|
|
#ifdef HAS_ARGBCOPYALPHAROW_NEON
|
|
ANY11B(ARGBCopyAlphaRow_Any_NEON, ARGBCopyAlphaRow_NEON, 0, 4, 4, 7)
|
|
#endif
|
|
#ifdef HAS_ARGBCOPYALPHAROW_LSX
|
|
ANY11B(ARGBCopyAlphaRow_Any_LSX, ARGBCopyAlphaRow_LSX, 0, 4, 4, 7)
|
|
#endif
|
|
#ifdef HAS_ARGBCOPYALPHAROW_LASX
|
|
ANY11B(ARGBCopyAlphaRow_Any_LASX, ARGBCopyAlphaRow_LASX, 0, 4, 4, 15)
|
|
#endif
|
|
#ifdef HAS_ARGBCOPYYTOALPHAROW_AVX2
|
|
ANY11B(ARGBCopyYToAlphaRow_Any_AVX2, ARGBCopyYToAlphaRow_AVX2, 0, 1, 4, 15)
|
|
#endif
|
|
#ifdef HAS_ARGBCOPYYTOALPHAROW_SSE2
|
|
ANY11B(ARGBCopyYToAlphaRow_Any_SSE2, ARGBCopyYToAlphaRow_SSE2, 0, 1, 4, 7)
|
|
#endif
|
|
#ifdef HAS_ARGBCOPYYTOALPHAROW_NEON
|
|
ANY11B(ARGBCopyYToAlphaRow_Any_NEON, ARGBCopyYToAlphaRow_NEON, 0, 1, 4, 15)
|
|
#endif
|
|
#ifdef HAS_ARGBCOPYYTOALPHAROW_LSX
|
|
ANY11B(ARGBCopyYToAlphaRow_Any_LSX, ARGBCopyYToAlphaRow_LSX, 0, 1, 4, 7)
|
|
#endif
|
|
#ifdef HAS_ARGBCOPYYTOALPHAROW_LASX
|
|
ANY11B(ARGBCopyYToAlphaRow_Any_LASX, ARGBCopyYToAlphaRow_LASX, 0, 1, 4, 15)
|
|
#endif
|
|
#ifdef HAS_ARGBCOPYYTOALPHAROW_WASMSIMD
|
|
ANY11B(ARGBCopyYToAlphaRow_Any_WASMSIMD,
|
|
ARGBCopyYToAlphaRow_WASMSIMD,
|
|
0,
|
|
1,
|
|
4,
|
|
15)
|
|
#endif
|
|
#undef ANY11B
|
|
|
|
// Any 1 to 1 with parameter.
|
|
#define ANY11P(NAMEANY, ANY_SIMD, T, SBPP, BPP, MASK) \
|
|
void NAMEANY(const uint8_t* src_ptr, uint8_t* dst_ptr, T param, int width) { \
|
|
SIMD_ALIGNED(uint8_t vin[(MASK + 1) * SBPP]); \
|
|
SIMD_ALIGNED(uint8_t vout[(MASK + 1) * BPP]); \
|
|
MSAN_MEMSET(vin, 0, sizeof(vin)); /* for msan */ \
|
|
MSAN_MEMSET(vout, 0, sizeof(vout)); /* for msan */ \
|
|
int r = width & MASK; \
|
|
int n = width & ~MASK; \
|
|
if (n > 0) { \
|
|
ANY_SIMD(src_ptr, dst_ptr, param, n); \
|
|
} \
|
|
ptrdiff_t np = n; \
|
|
memcpy(vin, src_ptr + np * SBPP, r * SBPP); \
|
|
ANY_SIMD(vin, vout, param, MASK + 1); \
|
|
memcpy(dst_ptr + np * BPP, vout, r * BPP); \
|
|
}
|
|
|
|
#if defined(HAS_I400TOARGBROW_SSE2)
|
|
ANY11P(I400ToARGBRow_Any_SSE2,
|
|
I400ToARGBRow_SSE2,
|
|
const struct YuvConstants*,
|
|
1,
|
|
4,
|
|
7)
|
|
#endif
|
|
#if defined(HAS_I400TOARGBROW_AVX2)
|
|
ANY11P(I400ToARGBRow_Any_AVX2,
|
|
I400ToARGBRow_AVX2,
|
|
const struct YuvConstants*,
|
|
1,
|
|
4,
|
|
15)
|
|
#endif
|
|
#if defined(HAS_I400TOARGBROW_NEON)
|
|
ANY11P(I400ToARGBRow_Any_NEON,
|
|
I400ToARGBRow_NEON,
|
|
const struct YuvConstants*,
|
|
1,
|
|
4,
|
|
7)
|
|
#endif
|
|
#if defined(HAS_I400TOARGBROW_LSX)
|
|
ANY11P(I400ToARGBRow_Any_LSX,
|
|
I400ToARGBRow_LSX,
|
|
const struct YuvConstants*,
|
|
1,
|
|
4,
|
|
7)
|
|
#endif
|
|
#if defined(HAS_I400TOARGBROW_LASX)
|
|
ANY11P(I400ToARGBRow_Any_LASX,
|
|
I400ToARGBRow_LASX,
|
|
const struct YuvConstants*,
|
|
1,
|
|
4,
|
|
15)
|
|
#endif
|
|
|
|
#if defined(HAS_ARGBTORGB565DITHERROW_AVX2)
|
|
ANY11P(ARGBToRGB565DitherRow_Any_AVX2,
|
|
ARGBToRGB565DitherRow_AVX2,
|
|
const uint32_t,
|
|
4,
|
|
2,
|
|
7)
|
|
#endif
|
|
#if defined(HAS_ARGBTORGB565DITHERROW_NEON)
|
|
ANY11P(ARGBToRGB565DitherRow_Any_NEON,
|
|
ARGBToRGB565DitherRow_NEON,
|
|
const uint32_t,
|
|
4,
|
|
2,
|
|
7)
|
|
#endif
|
|
#if defined(HAS_ARGBTORGB565DITHERROW_LSX)
|
|
ANY11P(ARGBToRGB565DitherRow_Any_LSX,
|
|
ARGBToRGB565DitherRow_LSX,
|
|
const uint32_t,
|
|
4,
|
|
2,
|
|
3)
|
|
#endif
|
|
#if defined(HAS_ARGBTORGB565DITHERROW_LASX)
|
|
ANY11P(ARGBToRGB565DitherRow_Any_LASX,
|
|
ARGBToRGB565DitherRow_LASX,
|
|
const uint32_t,
|
|
4,
|
|
2,
|
|
7)
|
|
#endif
|
|
#ifdef HAS_ARGBSHUFFLEROW_SSSE3
|
|
ANY11P(ARGBShuffleRow_Any_SSSE3, ARGBShuffleRow_SSSE3, const uint8_t*, 4, 4, 7)
|
|
#endif
|
|
#ifdef HAS_ARGBSHUFFLEROW_AVX2
|
|
ANY11P(ARGBShuffleRow_Any_AVX2, ARGBShuffleRow_AVX2, const uint8_t*, 4, 4, 15)
|
|
#endif
|
|
#ifdef HAS_ARGBSHUFFLEROW_NEON
|
|
ANY11P(ARGBShuffleRow_Any_NEON, ARGBShuffleRow_NEON, const uint8_t*, 4, 4, 3)
|
|
#endif
|
|
#ifdef HAS_ARGBSHUFFLEROW_LSX
|
|
ANY11P(ARGBShuffleRow_Any_LSX, ARGBShuffleRow_LSX, const uint8_t*, 4, 4, 3)
|
|
#endif
|
|
#ifdef HAS_ARGBSHUFFLEROW_LASX
|
|
ANY11P(ARGBShuffleRow_Any_LASX, ARGBShuffleRow_LASX, const uint8_t*, 4, 4, 7)
|
|
#endif
|
|
#if defined(HAS_ARGBTORGB565DITHERROW_WASMSIMD)
|
|
ANY11P(ARGBToRGB565DitherRow_Any_WASMSIMD,
|
|
ARGBToRGB565DitherRow_WASMSIMD,
|
|
const uint32_t,
|
|
4,
|
|
2,
|
|
7)
|
|
#endif
|
|
#ifdef HAS_ARGBSHUFFLEROW_WASMSIMD
|
|
ANY11P(ARGBShuffleRow_Any_WASMSIMD,
|
|
ARGBShuffleRow_WASMSIMD,
|
|
const uint8_t*,
|
|
4,
|
|
4,
|
|
3)
|
|
#endif
|
|
#ifdef HAS_ARGBORROW_SSE2
|
|
ANY11P(ARGBOrRow_Any_SSE2, ARGBOrRow_SSE2, const uint32_t, 4, 4, 3)
|
|
#endif
|
|
#ifdef HAS_ARGBORROW_AVX2
|
|
ANY11P(ARGBOrRow_Any_AVX2, ARGBOrRow_AVX2, const uint32_t, 4, 4, 7)
|
|
#endif
|
|
#ifdef HAS_ARGBORROW_NEON
|
|
ANY11P(ARGBOrRow_Any_NEON, ARGBOrRow_NEON, const uint32_t, 4, 4, 7)
|
|
#endif
|
|
#ifdef HAS_ARGBORROW_LSX
|
|
ANY11P(ARGBOrRow_Any_LSX, ARGBOrRow_LSX, const uint32_t, 4, 4, 3)
|
|
#endif
|
|
#ifdef HAS_ARGBORROW_LASX
|
|
ANY11P(ARGBOrRow_Any_LASX, ARGBOrRow_LASX, const uint32_t, 4, 4, 7)
|
|
#endif
|
|
#ifdef HAS_ARGBORROW_WASMSIMD
|
|
ANY11P(ARGBOrRow_Any_WASMSIMD, ARGBOrRow_WASMSIMD, const uint32_t, 4, 4, 3)
|
|
#endif
|
|
#ifdef HAS_ARGBPOLYNOMIALROW_SSE2
|
|
ANY11P(ARGBPolynomialRow_Any_SSE2,
|
|
ARGBPolynomialRow_SSE2,
|
|
const float*,
|
|
4,
|
|
4,
|
|
1)
|
|
#endif
|
|
#ifdef HAS_ARGBPOLYNOMIALROW_AVX2
|
|
ANY11P(ARGBPolynomialRow_Any_AVX2,
|
|
ARGBPolynomialRow_AVX2,
|
|
const float*,
|
|
4,
|
|
4,
|
|
1)
|
|
#endif
|
|
#ifdef HAS_ARGBPOLYNOMIALROW_NEON
|
|
ANY11P(ARGBPolynomialRow_Any_NEON,
|
|
ARGBPolynomialRow_NEON,
|
|
const float*,
|
|
4,
|
|
4,
|
|
1)
|
|
#endif
|
|
#ifdef HAS_ARGBCOLORMATRIXROW_NEON
|
|
ANY11P(ARGBColorMatrixRow_Any_NEON,
|
|
ARGBColorMatrixRow_NEON,
|
|
const int8_t*,
|
|
4,
|
|
4,
|
|
7)
|
|
#endif
|
|
#ifdef HAS_ARGBCOLORMATRIXROW_NEON_I8MM
|
|
ANY11P(ARGBColorMatrixRow_Any_NEON_I8MM,
|
|
ARGBColorMatrixRow_NEON_I8MM,
|
|
const int8_t*,
|
|
4,
|
|
4,
|
|
7)
|
|
#endif
|
|
#ifdef HAS_ARGBCOLORMATRIXROW_LSX
|
|
ANY11P(ARGBColorMatrixRow_Any_LSX,
|
|
ARGBColorMatrixRow_LSX,
|
|
const int8_t*,
|
|
4,
|
|
4,
|
|
3)
|
|
#endif
|
|
#ifdef HAS_ARGBCOLORMATRIXROW_LASX
|
|
ANY11P(ARGBColorMatrixRow_Any_LASX,
|
|
ARGBColorMatrixRow_LASX,
|
|
const int8_t*,
|
|
4,
|
|
4,
|
|
7)
|
|
#endif
|
|
#undef ANY11P
|
|
|
|
// Any 1 in-place.
|
|
#define ANY1I(NAMEANY, ANY_SIMD, BPP, MASK) \
|
|
void NAMEANY(uint8_t* dst_ptr, int width) { \
|
|
SIMD_ALIGNED(uint8_t vout[(MASK + 1) * BPP]); \
|
|
MSAN_MEMSET(vout, 0, sizeof(vout)); /* for msan */ \
|
|
int r = width & MASK; \
|
|
int n = width & ~MASK; \
|
|
if (n > 0) { \
|
|
ANY_SIMD(dst_ptr, n); \
|
|
} \
|
|
ptrdiff_t np = n; \
|
|
memcpy(vout, dst_ptr + np * BPP, r * BPP); \
|
|
ANY_SIMD(vout, MASK + 1); \
|
|
memcpy(dst_ptr + np * BPP, vout, r * BPP); \
|
|
}
|
|
|
|
#ifdef HAS_ARGBSEPIAROW_NEON
|
|
ANY1I(ARGBSepiaRow_Any_NEON, ARGBSepiaRow_NEON, 4, 7)
|
|
#endif
|
|
#ifdef HAS_ARGBSEPIAROW_NEON_DOTPROD
|
|
ANY1I(ARGBSepiaRow_Any_NEON_DotProd, ARGBSepiaRow_NEON_DotProd, 4, 7)
|
|
#endif
|
|
#ifdef HAS_ARGBSEPIAROW_LSX
|
|
ANY1I(ARGBSepiaRow_Any_LSX, ARGBSepiaRow_LSX, 4, 7)
|
|
#endif
|
|
#ifdef HAS_ARGBSEPIAROW_LASX
|
|
ANY1I(ARGBSepiaRow_Any_LASX, ARGBSepiaRow_LASX, 4, 15)
|
|
#endif
|
|
#undef ANY1I
|
|
|
|
// Any 1 in-place with 3 int parameters.
|
|
#define ANY1I3(NAMEANY, ANY_SIMD, BPP, MASK) \
|
|
void NAMEANY(uint8_t* dst_ptr, int p0, int p1, int p2, int width) { \
|
|
SIMD_ALIGNED(uint8_t vout[(MASK + 1) * BPP]); \
|
|
MSAN_MEMSET(vout, 0, sizeof(vout)); /* for msan */ \
|
|
int r = width & MASK; \
|
|
int n = width & ~MASK; \
|
|
if (n > 0) { \
|
|
ANY_SIMD(dst_ptr, p0, p1, p2, n); \
|
|
} \
|
|
ptrdiff_t np = n; \
|
|
memcpy(vout, dst_ptr + np * BPP, r * BPP); \
|
|
ANY_SIMD(vout, p0, p1, p2, MASK + 1); \
|
|
memcpy(dst_ptr + np * BPP, vout, r * BPP); \
|
|
}
|
|
|
|
#ifdef HAS_ARGBQUANTIZEROW_NEON
|
|
ANY1I3(ARGBQuantizeRow_Any_NEON, ARGBQuantizeRow_NEON, 4, 7)
|
|
#endif
|
|
#ifdef HAS_ARGBQUANTIZEROW_LSX
|
|
ANY1I3(ARGBQuantizeRow_Any_LSX, ARGBQuantizeRow_LSX, 4, 3)
|
|
#endif
|
|
#ifdef HAS_ARGBQUANTIZEROW_LASX
|
|
ANY1I3(ARGBQuantizeRow_Any_LASX, ARGBQuantizeRow_LASX, 4, 7)
|
|
#endif
|
|
#undef ANY1I3
|
|
|
|
// Any 1 to 1 with trailing parameter.
|
|
#define ANY11PT(NAMEANY, ANY_SIMD, T, SBPP, BPP, MASK) \
|
|
void NAMEANY(const uint8_t* src_ptr, uint8_t* dst_ptr, int width, T param) { \
|
|
SIMD_ALIGNED(uint8_t vin[(MASK + 1) * SBPP]); \
|
|
SIMD_ALIGNED(uint8_t vout[(MASK + 1) * BPP]); \
|
|
MSAN_MEMSET(vin, 0, sizeof(vin)); /* for msan */ \
|
|
MSAN_MEMSET(vout, 0, sizeof(vout)); /* for msan */ \
|
|
int r = width & MASK; \
|
|
int n = width & ~MASK; \
|
|
if (n > 0) { \
|
|
ANY_SIMD(src_ptr, dst_ptr, n, param); \
|
|
} \
|
|
ptrdiff_t np = n; \
|
|
memcpy(vin, src_ptr + np * SBPP, r * SBPP); \
|
|
ANY_SIMD(vin, vout, MASK + 1, param); \
|
|
memcpy(dst_ptr + np * BPP, vout, r * BPP); \
|
|
}
|
|
|
|
#ifdef HAS_ARGBSHADEROW_NEON
|
|
ANY11PT(ARGBShadeRow_Any_NEON, ARGBShadeRow_NEON, uint32_t, 4, 4, 7)
|
|
#endif
|
|
#ifdef HAS_ARGBSHADEROW_LSX
|
|
ANY11PT(ARGBShadeRow_Any_LSX, ARGBShadeRow_LSX, uint32_t, 4, 4, 3)
|
|
#endif
|
|
#ifdef HAS_ARGBSHADEROW_LASX
|
|
ANY11PT(ARGBShadeRow_Any_LASX, ARGBShadeRow_LASX, uint32_t, 4, 4, 7)
|
|
#endif
|
|
#undef ANY11PT
|
|
|
|
#ifdef HAS_ARGBLUMACOLORTABLEROW_NEON
|
|
void ARGBLumaColorTableRow_Any_NEON(const uint8_t* src_argb,
|
|
uint8_t* dst_argb,
|
|
int width,
|
|
const uint8_t* luma,
|
|
const uint32_t lumacoeff) {
|
|
SIMD_ALIGNED(uint8_t vin[4 * 4]);
|
|
SIMD_ALIGNED(uint8_t vout[4 * 4]);
|
|
MSAN_MEMSET(vin, 0, sizeof(vin)); /* for msan */
|
|
MSAN_MEMSET(vout, 0, sizeof(vout)); /* for msan */
|
|
int r = width & 3;
|
|
int n = width & ~3;
|
|
if (n > 0) {
|
|
ARGBLumaColorTableRow_NEON(src_argb, dst_argb, n, luma, lumacoeff);
|
|
}
|
|
ptrdiff_t np = n;
|
|
memcpy(vin, src_argb + np * 4, r * 4);
|
|
ARGBLumaColorTableRow_NEON(vin, vout, 4, luma, lumacoeff);
|
|
memcpy(dst_argb + np * 4, vout, r * 4);
|
|
}
|
|
#endif
|
|
|
|
// Any 1 to 1 with type
|
|
#define ANY11T(NAMEANY, ANY_SIMD, SBPP, BPP, STYPE, DTYPE, MASK) \
|
|
void NAMEANY(const STYPE* src_ptr, DTYPE* dst_ptr, int width) { \
|
|
SIMD_ALIGNED(uint8_t vin[(MASK + 1) * SBPP]); \
|
|
SIMD_ALIGNED(uint8_t vout[(MASK + 1) * BPP]); \
|
|
MSAN_MEMSET(vin, 0, sizeof(vin)); /* for msan */ \
|
|
MSAN_MEMSET(vout, 0, sizeof(vout)); /* for msan */ \
|
|
int r = width & MASK; \
|
|
int n = width & ~MASK; \
|
|
if (n > 0) { \
|
|
ANY_SIMD(src_ptr, dst_ptr, n); \
|
|
} \
|
|
ptrdiff_t np = n; \
|
|
memcpy(vin, (uint8_t*)(src_ptr) + np * SBPP, r * SBPP); \
|
|
ANY_SIMD((STYPE*)vin, (DTYPE*)vout, MASK + 1); \
|
|
memcpy((uint8_t*)(dst_ptr) + np * BPP, vout, r * BPP); \
|
|
}
|
|
|
|
#ifdef HAS_ARGBTOAR64ROW_SSSE3
|
|
ANY11T(ARGBToAR64Row_Any_SSSE3, ARGBToAR64Row_SSSE3, 4, 8, uint8_t, uint16_t, 3)
|
|
#endif
|
|
|
|
#ifdef HAS_ARGBTOAB64ROW_SSSE3
|
|
ANY11T(ARGBToAB64Row_Any_SSSE3, ARGBToAB64Row_SSSE3, 4, 8, uint8_t, uint16_t, 3)
|
|
#endif
|
|
|
|
#ifdef HAS_AR64TOARGBROW_SSSE3
|
|
ANY11T(AR64ToARGBRow_Any_SSSE3, AR64ToARGBRow_SSSE3, 8, 4, uint16_t, uint8_t, 3)
|
|
#endif
|
|
|
|
#ifdef HAS_AB64TOARGBROW_SSSE3
|
|
ANY11T(AB64ToARGBRow_Any_SSSE3, AB64ToARGBRow_SSSE3, 8, 4, uint16_t, uint8_t, 3)
|
|
#endif
|
|
|
|
#ifdef HAS_ARGBTOAR64ROW_AVX2
|
|
ANY11T(ARGBToAR64Row_Any_AVX2, ARGBToAR64Row_AVX2, 4, 8, uint8_t, uint16_t, 7)
|
|
#endif
|
|
|
|
#ifdef HAS_ARGBTOAB64ROW_AVX2
|
|
ANY11T(ARGBToAB64Row_Any_AVX2, ARGBToAB64Row_AVX2, 4, 8, uint8_t, uint16_t, 7)
|
|
#endif
|
|
|
|
#ifdef HAS_AR64TOARGBROW_AVX2
|
|
ANY11T(AR64ToARGBRow_Any_AVX2, AR64ToARGBRow_AVX2, 8, 4, uint16_t, uint8_t, 7)
|
|
#endif
|
|
|
|
#ifdef HAS_AB64TOARGBROW_AVX2
|
|
ANY11T(AB64ToARGBRow_Any_AVX2, AB64ToARGBRow_AVX2, 8, 4, uint16_t, uint8_t, 7)
|
|
#endif
|
|
|
|
#ifdef HAS_ARGBTOAR64ROW_NEON
|
|
ANY11T(ARGBToAR64Row_Any_NEON, ARGBToAR64Row_NEON, 4, 8, uint8_t, uint16_t, 7)
|
|
#endif
|
|
|
|
#ifdef HAS_ARGBTOAB64ROW_NEON
|
|
ANY11T(ARGBToAB64Row_Any_NEON, ARGBToAB64Row_NEON, 4, 8, uint8_t, uint16_t, 7)
|
|
#endif
|
|
|
|
#ifdef HAS_AR64TOARGBROW_NEON
|
|
ANY11T(AR64ToARGBRow_Any_NEON, AR64ToARGBRow_NEON, 8, 4, uint16_t, uint8_t, 7)
|
|
#endif
|
|
|
|
#ifdef HAS_AB64TOARGBROW_NEON
|
|
ANY11T(AB64ToARGBRow_Any_NEON, AB64ToARGBRow_NEON, 8, 4, uint16_t, uint8_t, 7)
|
|
#endif
|
|
#ifdef HAS_ARGBTOAR64ROW_LSX
|
|
ANY11T(ARGBToAR64Row_Any_LSX, ARGBToAR64Row_LSX, 4, 8, uint8_t, uint16_t, 3)
|
|
#endif
|
|
#ifdef HAS_ARGBTOAR64ROW_LASX
|
|
ANY11T(ARGBToAR64Row_Any_LASX, ARGBToAR64Row_LASX, 4, 8, uint8_t, uint16_t, 7)
|
|
#endif
|
|
#ifdef HAS_ARGBTOAB64ROW_LSX
|
|
ANY11T(ARGBToAB64Row_Any_LSX, ARGBToAB64Row_LSX, 4, 8, uint8_t, uint16_t, 3)
|
|
#endif
|
|
#ifdef HAS_ARGBTOAB64ROW_LASX
|
|
ANY11T(ARGBToAB64Row_Any_LASX, ARGBToAB64Row_LASX, 4, 8, uint8_t, uint16_t, 7)
|
|
#endif
|
|
#ifdef HAS_AR64TOARGBROW_LSX
|
|
ANY11T(AR64ToARGBRow_Any_LSX, AR64ToARGBRow_LSX, 8, 4, uint16_t, uint8_t, 3)
|
|
#endif
|
|
#ifdef HAS_AR64TOARGBROW_LASX
|
|
ANY11T(AR64ToARGBRow_Any_LASX, AR64ToARGBRow_LASX, 8, 4, uint16_t, uint8_t, 7)
|
|
#endif
|
|
#ifdef HAS_AB64TOARGBROW_LSX
|
|
ANY11T(AB64ToARGBRow_Any_LSX, AB64ToARGBRow_LSX, 8, 4, uint16_t, uint8_t, 3)
|
|
#endif
|
|
#ifdef HAS_AB64TOARGBROW_LASX
|
|
ANY11T(AB64ToARGBRow_Any_LASX, AB64ToARGBRow_LASX, 8, 4, uint16_t, uint8_t, 7)
|
|
#endif
|
|
|
|
#ifdef HAS_COPYROW_16_AVX2
|
|
ANY11T(CopyRow_16_Any_AVX2, CopyRow_16_AVX2, 2, 2, uint16_t, uint16_t, 31)
|
|
#endif
|
|
#ifdef HAS_COPYROW_16_NEON
|
|
ANY11T(CopyRow_16_Any_NEON, CopyRow_16_NEON, 2, 2, uint16_t, uint16_t, 15)
|
|
#endif
|
|
#ifdef HAS_COPYROW_16_LSX
|
|
ANY11T(CopyRow_16_Any_LSX, CopyRow_16_LSX, 2, 2, uint16_t, uint16_t, 15)
|
|
#endif
|
|
#ifdef HAS_COPYROW_16_LASX
|
|
ANY11T(CopyRow_16_Any_LASX, CopyRow_16_LASX, 2, 2, uint16_t, uint16_t, 31)
|
|
#endif
|
|
|
|
#undef ANY11T
|
|
|
|
// Any 1 to 1 with parameter and shorts. BPP measures in shorts.
|
|
#define ANY11C(NAMEANY, ANY_SIMD, SBPP, BPP, STYPE, DTYPE, MASK) \
|
|
void NAMEANY(const STYPE* src_ptr, DTYPE* dst_ptr, int value, int width) { \
|
|
SIMD_ALIGNED(STYPE vin[64]); \
|
|
static_assert((MASK + 1) * SBPP <= sizeof(vin), "vin buffer too small"); \
|
|
SIMD_ALIGNED(DTYPE vout[64]); \
|
|
static_assert((MASK + 1) * BPP <= sizeof(vout), "vout buffer too small"); \
|
|
MSAN_MEMSET(vin, 0, sizeof(vin)); /* for msan */ \
|
|
MSAN_MEMSET(vout, 0, sizeof(vout)); /* for msan */ \
|
|
int r = width & MASK; \
|
|
int n = width & ~MASK; \
|
|
if (n > 0) { \
|
|
ANY_SIMD(src_ptr, dst_ptr, value, n); \
|
|
} \
|
|
ptrdiff_t np = n; \
|
|
memcpy(vin, src_ptr + np, r * SBPP); \
|
|
ANY_SIMD(vin, vout, value, MASK + 1); \
|
|
memcpy(dst_ptr + np, vout, r * BPP); \
|
|
}
|
|
|
|
#ifdef HAS_CONVERT16TO8ROW_SSSE3
|
|
ANY11C(Convert16To8Row_Any_SSSE3,
|
|
Convert16To8Row_SSSE3,
|
|
2,
|
|
1,
|
|
uint16_t,
|
|
uint8_t,
|
|
15)
|
|
#endif
|
|
#ifdef HAS_CONVERT16TO8ROW_AVX2
|
|
ANY11C(Convert16To8Row_Any_AVX2,
|
|
Convert16To8Row_AVX2,
|
|
2,
|
|
1,
|
|
uint16_t,
|
|
uint8_t,
|
|
31)
|
|
#endif
|
|
#ifdef HAS_CONVERT16TO8ROW_NEON
|
|
ANY11C(Convert16To8Row_Any_NEON,
|
|
Convert16To8Row_NEON,
|
|
2,
|
|
1,
|
|
uint16_t,
|
|
uint8_t,
|
|
15)
|
|
#endif
|
|
#ifdef HAS_CONVERT16TO8ROW_LSX
|
|
ANY11C(Convert16To8Row_Any_LSX,
|
|
Convert16To8Row_LSX,
|
|
2,
|
|
1,
|
|
uint16_t,
|
|
uint8_t,
|
|
15)
|
|
#endif
|
|
#ifdef HAS_CONVERT16TO8ROW_LASX
|
|
ANY11C(Convert16To8Row_Any_LASX,
|
|
Convert16To8Row_LASX,
|
|
2,
|
|
1,
|
|
uint16_t,
|
|
uint8_t,
|
|
31)
|
|
#endif
|
|
#ifdef HAS_CONVERT8TO16ROW_SSE2
|
|
ANY11C(Convert8To16Row_Any_SSE2,
|
|
Convert8To16Row_SSE2,
|
|
1,
|
|
2,
|
|
uint8_t,
|
|
uint16_t,
|
|
15)
|
|
#endif
|
|
#ifdef HAS_CONVERT8TO16ROW_AVX2
|
|
ANY11C(Convert8To16Row_Any_AVX2,
|
|
Convert8To16Row_AVX2,
|
|
1,
|
|
2,
|
|
uint8_t,
|
|
uint16_t,
|
|
31)
|
|
#endif
|
|
#ifdef HAS_CONVERT8TO16ROW_NEON
|
|
ANY11C(Convert8To16Row_Any_NEON,
|
|
Convert8To16Row_NEON,
|
|
1,
|
|
2,
|
|
uint8_t,
|
|
uint16_t,
|
|
15)
|
|
#endif
|
|
#ifdef HAS_CONVERT8TO16ROW_LSX
|
|
ANY11C(Convert8To16Row_Any_LSX,
|
|
Convert8To16Row_LSX,
|
|
1,
|
|
2,
|
|
uint8_t,
|
|
uint16_t,
|
|
15)
|
|
#endif
|
|
#ifdef HAS_CONVERT8TO16ROW_LASX
|
|
ANY11C(Convert8To16Row_Any_LASX,
|
|
Convert8To16Row_LASX,
|
|
1,
|
|
2,
|
|
uint8_t,
|
|
uint16_t,
|
|
31)
|
|
#endif
|
|
#ifdef HAS_MULTIPLYROW_16_AVX2
|
|
ANY11C(MultiplyRow_16_Any_AVX2,
|
|
MultiplyRow_16_AVX2,
|
|
2,
|
|
2,
|
|
uint16_t,
|
|
uint16_t,
|
|
31)
|
|
#endif
|
|
#ifdef HAS_MULTIPLYROW_16_NEON
|
|
ANY11C(MultiplyRow_16_Any_NEON,
|
|
MultiplyRow_16_NEON,
|
|
2,
|
|
2,
|
|
uint16_t,
|
|
uint16_t,
|
|
15)
|
|
#endif
|
|
#ifdef HAS_MULTIPLYROW_16_LSX
|
|
ANY11C(MultiplyRow_16_Any_LSX,
|
|
MultiplyRow_16_LSX,
|
|
2,
|
|
2,
|
|
uint16_t,
|
|
uint16_t,
|
|
15)
|
|
#endif
|
|
#ifdef HAS_MULTIPLYROW_16_LASX
|
|
ANY11C(MultiplyRow_16_Any_LASX,
|
|
MultiplyRow_16_LASX,
|
|
2,
|
|
2,
|
|
uint16_t,
|
|
uint16_t,
|
|
31)
|
|
#endif
|
|
#ifdef HAS_DIVIDEROW_16_AVX2
|
|
ANY11C(DivideRow_16_Any_AVX2, DivideRow_16_AVX2, 2, 2, uint16_t, uint16_t, 31)
|
|
#endif
|
|
#ifdef HAS_DIVIDEROW_16_NEON
|
|
ANY11C(DivideRow_16_Any_NEON, DivideRow_16_NEON, 2, 2, uint16_t, uint16_t, 15)
|
|
#endif
|
|
#ifdef HAS_DIVIDEROW_16_LSX
|
|
ANY11C(DivideRow_16_Any_LSX, DivideRow_16_LSX, 2, 2, uint16_t, uint16_t, 15)
|
|
#endif
|
|
#ifdef HAS_DIVIDEROW_16_LASX
|
|
ANY11C(DivideRow_16_Any_LASX, DivideRow_16_LASX, 2, 2, uint16_t, uint16_t, 31)
|
|
#endif
|
|
#ifdef HAS_CONVERT8TO16ROW_WASMSIMD
|
|
ANY11C(Convert8To16Row_Any_WASMSIMD,
|
|
Convert8To16Row_WASMSIMD,
|
|
1,
|
|
2,
|
|
uint8_t,
|
|
uint16_t,
|
|
15)
|
|
#endif
|
|
#ifdef HAS_MULTIPLYROW_16_WASMSIMD
|
|
ANY11C(MultiplyRow_16_Any_WASMSIMD,
|
|
MultiplyRow_16_WASMSIMD,
|
|
2,
|
|
2,
|
|
uint16_t,
|
|
uint16_t,
|
|
15)
|
|
#endif
|
|
#undef ANY11C
|
|
|
|
// Any 1 to 1 with parameter and shorts. BPP measures in shorts.
|
|
#define ANY11SB(NAMEANY, ANY_SIMD, SBPP, BPP, STYPE, DTYPE, MASK) \
|
|
void NAMEANY(const STYPE* src_ptr, DTYPE* dst_ptr, int scale, int bias, \
|
|
int width) { \
|
|
SIMD_ALIGNED(STYPE vin[64]); \
|
|
static_assert((MASK + 1) * SBPP <= sizeof(vin), "vin buffer too small"); \
|
|
SIMD_ALIGNED(DTYPE vout[64]); \
|
|
static_assert((MASK + 1) * BPP <= sizeof(vout), "vout buffer too small"); \
|
|
MSAN_MEMSET(vin, 0, sizeof(vin)); /* for msan */ \
|
|
MSAN_MEMSET(vout, 0, sizeof(vout)); /* for msan */ \
|
|
int r = width & MASK; \
|
|
int n = width & ~MASK; \
|
|
if (n > 0) { \
|
|
ANY_SIMD(src_ptr, dst_ptr, scale, bias, n); \
|
|
} \
|
|
ptrdiff_t np = n; \
|
|
memcpy(vin, src_ptr + np, r * SBPP); \
|
|
ANY_SIMD(vin, vout, scale, bias, MASK + 1); \
|
|
memcpy(dst_ptr + np, vout, r * BPP); \
|
|
}
|
|
|
|
#ifdef HAS_CONVERT8TO8ROW_NEON
|
|
ANY11SB(Convert8To8Row_Any_NEON,
|
|
Convert8To8Row_NEON,
|
|
1,
|
|
1,
|
|
uint8_t,
|
|
uint8_t,
|
|
31)
|
|
#endif
|
|
#ifdef HAS_CONVERT8TO8ROW_AVX2
|
|
ANY11SB(Convert8To8Row_Any_AVX2,
|
|
Convert8To8Row_AVX2,
|
|
1,
|
|
1,
|
|
uint8_t,
|
|
uint8_t,
|
|
31)
|
|
#endif
|
|
#ifdef HAS_CONVERT8TO8ROW_LSX
|
|
ANY11SB(Convert8To8Row_Any_LSX,
|
|
Convert8To8Row_LSX,
|
|
1,
|
|
1,
|
|
uint8_t,
|
|
uint8_t,
|
|
15)
|
|
#endif
|
|
#ifdef HAS_CONVERT8TO8ROW_LASX
|
|
ANY11SB(Convert8To8Row_Any_LASX,
|
|
Convert8To8Row_LASX,
|
|
1,
|
|
1,
|
|
uint8_t,
|
|
uint8_t,
|
|
31)
|
|
#endif
|
|
#undef ANY11B
|
|
|
|
// Any 1 to 1 with parameter and shorts to byte. BPP measures in shorts.
|
|
#define ANY11P16(NAMEANY, ANY_SIMD, ST, T, SBPP, BPP, MASK) \
|
|
void NAMEANY(const ST* src_ptr, T* dst_ptr, float param, int width) { \
|
|
SIMD_ALIGNED(ST vin[32]); \
|
|
static_assert((MASK + 1) * SBPP <= sizeof(vin), "vin buffer too small"); \
|
|
SIMD_ALIGNED(T vout[32]); \
|
|
static_assert((MASK + 1) * BPP <= sizeof(vout), "vout buffer too small"); \
|
|
MSAN_MEMSET(vin, 0, sizeof(vin)); /* for msan */ \
|
|
MSAN_MEMSET(vout, 0, sizeof(vout)); /* for msan */ \
|
|
int r = width & MASK; \
|
|
int n = width & ~MASK; \
|
|
if (n > 0) { \
|
|
ANY_SIMD(src_ptr, dst_ptr, param, n); \
|
|
} \
|
|
ptrdiff_t np = n; \
|
|
memcpy(vin, src_ptr + np, r * SBPP); \
|
|
ANY_SIMD(vin, vout, param, MASK + 1); \
|
|
memcpy(dst_ptr + np, vout, r * BPP); \
|
|
}
|
|
|
|
#ifdef HAS_HALFFLOATROW_AVX2
|
|
ANY11P16(HalfFloatRow_Any_AVX2, HalfFloatRow_AVX2, uint16_t, uint16_t, 2, 2, 15)
|
|
#endif
|
|
#ifdef HAS_HALFFLOATROW_F16C
|
|
ANY11P16(HalfFloatRow_Any_F16C, HalfFloatRow_F16C, uint16_t, uint16_t, 2, 2, 15)
|
|
ANY11P16(HalfFloat1Row_Any_F16C,
|
|
HalfFloat1Row_F16C,
|
|
uint16_t,
|
|
uint16_t,
|
|
2,
|
|
2,
|
|
15)
|
|
#endif
|
|
#ifdef HAS_HALFFLOATROW_NEON
|
|
ANY11P16(HalfFloatRow_Any_NEON, HalfFloatRow_NEON, uint16_t, uint16_t, 2, 2, 15)
|
|
#endif
|
|
#ifdef HAS_BYTETOFLOATROW_AVX2
|
|
ANY11P16(ByteToFloatRow_Any_AVX2, ByteToFloatRow_AVX2, uint8_t, float, 1, 4, 15)
|
|
#endif
|
|
#ifdef HAS_BYTETOFLOATROW_NEON
|
|
ANY11P16(ByteToFloatRow_Any_NEON, ByteToFloatRow_NEON, uint8_t, float, 1, 4, 7)
|
|
#endif
|
|
#ifdef HAS_HALFFLOATROW_LSX
|
|
ANY11P16(HalfFloatRow_Any_LSX, HalfFloatRow_LSX, uint16_t, uint16_t, 2, 2, 7)
|
|
#endif
|
|
#ifdef HAS_HALFFLOATROW_LASX
|
|
ANY11P16(HalfFloatRow_Any_LASX, HalfFloatRow_LASX, uint16_t, uint16_t, 2, 2, 15)
|
|
#endif
|
|
#undef ANY11P16
|
|
|
|
// Any 1 to 1 with yuvconstants
|
|
#define ANY11C(NAMEANY, ANY_SIMD, UVSHIFT, SBPP, BPP, MASK) \
|
|
void NAMEANY(const uint8_t* src_ptr, uint8_t* dst_ptr, \
|
|
const struct YuvConstants* yuvconstants, int width) { \
|
|
SIMD_ALIGNED(uint8_t vin[256]); \
|
|
static_assert(SS(MASK + 1, UVSHIFT) * SBPP <= sizeof(vin), \
|
|
"vin buffer too small"); \
|
|
SIMD_ALIGNED(uint8_t vout[256]); \
|
|
static_assert((MASK + 1) * BPP <= sizeof(vout), "vout buffer too small"); \
|
|
MSAN_MEMSET(vin, 0, sizeof(vin)); /* for YUY2 and msan */ \
|
|
MSAN_MEMSET(vout, 0, sizeof(vout)); /* for msan */ \
|
|
int r = width & MASK; \
|
|
int n = width & ~MASK; \
|
|
if (n > 0) { \
|
|
ANY_SIMD(src_ptr, dst_ptr, yuvconstants, n); \
|
|
} \
|
|
ptrdiff_t np = n; \
|
|
memcpy(vin, src_ptr + (np >> UVSHIFT) * SBPP, SS(r, UVSHIFT) * SBPP); \
|
|
ANY_SIMD(vin, vout, yuvconstants, MASK + 1); \
|
|
memcpy(dst_ptr + np * BPP, vout, r * BPP); \
|
|
}
|
|
|
|
#if defined(HAS_YUY2TOARGBROW_SSSE3)
|
|
ANY11C(YUY2ToARGBRow_Any_SSSE3, YUY2ToARGBRow_SSSE3, 1, 4, 4, 7)
|
|
ANY11C(UYVYToARGBRow_Any_SSSE3, UYVYToARGBRow_SSSE3, 1, 4, 4, 7)
|
|
#endif
|
|
#if defined(HAS_YUY2TOARGBROW_AVX2)
|
|
ANY11C(YUY2ToARGBRow_Any_AVX2, YUY2ToARGBRow_AVX2, 1, 4, 4, 31)
|
|
ANY11C(UYVYToARGBRow_Any_AVX2, UYVYToARGBRow_AVX2, 1, 4, 4, 31)
|
|
#endif
|
|
#if defined(HAS_YUY2TOARGBROW_NEON)
|
|
ANY11C(YUY2ToARGBRow_Any_NEON, YUY2ToARGBRow_NEON, 1, 4, 4, 7)
|
|
ANY11C(UYVYToARGBRow_Any_NEON, UYVYToARGBRow_NEON, 1, 4, 4, 7)
|
|
#endif
|
|
#if defined(HAS_YUY2TOARGBROW_LSX)
|
|
ANY11C(YUY2ToARGBRow_Any_LSX, YUY2ToARGBRow_LSX, 1, 4, 4, 7)
|
|
ANY11C(UYVYToARGBRow_Any_LSX, UYVYToARGBRow_LSX, 1, 4, 4, 7)
|
|
#endif
|
|
#if defined(HAS_YUY2TOARGBROW_LASX)
|
|
ANY11C(YUY2ToARGBRow_Any_LASX, YUY2ToARGBRow_LASX, 1, 4, 4, 15)
|
|
ANY11C(UYVYToARGBRow_Any_LASX, UYVYToARGBRow_LASX, 1, 4, 4, 15)
|
|
#endif
|
|
#undef ANY11C
|
|
|
|
// Any 1 to 1 interpolate. Takes 2 rows of source via stride.
|
|
#define ANY11I(NAMEANY, ANY_SIMD, TD, TS, SBPP, BPP, MASK) \
|
|
void NAMEANY(TD* dst_ptr, const TS* src_ptr, ptrdiff_t src_stride, \
|
|
int width, int source_y_fraction) { \
|
|
SIMD_ALIGNED(TS vin[64 * 2]); \
|
|
static_assert((MASK + 1) * SBPP * sizeof(TS) <= sizeof(vin) / 2, \
|
|
"vin buffer too small"); \
|
|
SIMD_ALIGNED(TD vout[64]); \
|
|
static_assert((MASK + 1) * BPP * sizeof(TD) <= sizeof(vout), \
|
|
"vout buffer too small"); \
|
|
MSAN_MEMSET(vin, 0, sizeof(vin)); /* for msan */ \
|
|
MSAN_MEMSET(vout, 0, sizeof(vout)); /* for msan */ \
|
|
int r = width & MASK; \
|
|
int n = width & ~MASK; \
|
|
if (n > 0) { \
|
|
ANY_SIMD(dst_ptr, src_ptr, src_stride, n, source_y_fraction); \
|
|
} \
|
|
ptrdiff_t np = n; \
|
|
memcpy(vin, src_ptr + np * SBPP, r * SBPP * sizeof(TS)); \
|
|
if (source_y_fraction) { \
|
|
memcpy(vin + 64, src_ptr + src_stride + np * SBPP, \
|
|
r * SBPP * sizeof(TS)); \
|
|
} \
|
|
ANY_SIMD(vout, vin, 64, MASK + 1, source_y_fraction); \
|
|
memcpy(dst_ptr + np * BPP, vout, r * BPP * sizeof(TD)); \
|
|
}
|
|
|
|
#if defined(HAS_INTERPOLATEROW_AVX2)
|
|
ANY11I(InterpolateRow_Any_AVX2, InterpolateRow_AVX2, uint8_t, uint8_t, 1, 1, 31)
|
|
#endif
|
|
#ifdef HAS_INTERPOLATEROW_NEON
|
|
ANY11I(InterpolateRow_Any_NEON, InterpolateRow_NEON, uint8_t, uint8_t, 1, 1, 15)
|
|
#endif
|
|
#ifdef HAS_INTERPOLATEROW_LSX
|
|
ANY11I(InterpolateRow_Any_LSX, InterpolateRow_LSX, uint8_t, uint8_t, 1, 1, 15)
|
|
#endif
|
|
#ifdef HAS_INTERPOLATEROW_LASX
|
|
ANY11I(InterpolateRow_Any_LASX, InterpolateRow_LASX, uint8_t, uint8_t, 1, 1, 31)
|
|
#endif
|
|
|
|
#ifdef HAS_INTERPOLATEROW_16_NEON
|
|
ANY11I(InterpolateRow_16_Any_NEON,
|
|
InterpolateRow_16_NEON,
|
|
uint16_t,
|
|
uint16_t,
|
|
1,
|
|
1,
|
|
7)
|
|
#endif
|
|
#ifdef HAS_INTERPOLATEROW_16_AVX2
|
|
ANY11I(InterpolateRow_16_Any_AVX2,
|
|
InterpolateRow_16_AVX2,
|
|
uint16_t,
|
|
uint16_t,
|
|
1,
|
|
1,
|
|
15)
|
|
#endif
|
|
#ifdef HAS_INTERPOLATEROW_16_LSX
|
|
ANY11I(InterpolateRow_16_Any_LSX,
|
|
InterpolateRow_16_LSX,
|
|
uint16_t,
|
|
uint16_t,
|
|
1,
|
|
1,
|
|
7)
|
|
#endif
|
|
#ifdef HAS_INTERPOLATEROW_16_LASX
|
|
ANY11I(InterpolateRow_16_Any_LASX,
|
|
InterpolateRow_16_LASX,
|
|
uint16_t,
|
|
uint16_t,
|
|
1,
|
|
1,
|
|
15)
|
|
#endif
|
|
#ifdef HAS_INTERPOLATEROW_WASMSIMD
|
|
ANY11I(InterpolateRow_Any_WASMSIMD,
|
|
InterpolateRow_WASMSIMD,
|
|
uint8_t,
|
|
uint8_t,
|
|
1,
|
|
1,
|
|
15)
|
|
#endif
|
|
#ifdef HAS_INTERPOLATEROW_16_WASMSIMD
|
|
ANY11I(InterpolateRow_16_Any_WASMSIMD,
|
|
InterpolateRow_16_WASMSIMD,
|
|
uint16_t,
|
|
uint16_t,
|
|
1,
|
|
1,
|
|
7)
|
|
#endif
|
|
#undef ANY11I
|
|
|
|
// Any 1 to 1 half row
|
|
#define ANY11H(NAMEANY, ANY_SIMD, T, MASK) \
|
|
void NAMEANY(const T* src_ptr, ptrdiff_t src_stride, T* dst_ptr, \
|
|
int width) { \
|
|
SIMD_ALIGNED(T vin[64 * 2]); \
|
|
static_assert((MASK + 1) * sizeof(T) <= sizeof(vin) / 2, \
|
|
"vin buffer too small"); \
|
|
SIMD_ALIGNED(T vout[64]); \
|
|
static_assert((MASK + 1) * sizeof(T) <= sizeof(vout), \
|
|
"vout buffer too small"); \
|
|
MSAN_MEMSET(vin, 0, sizeof(vin)); /* for msan */ \
|
|
MSAN_MEMSET(vout, 0, sizeof(vout)); /* for msan */ \
|
|
int r = width & MASK; \
|
|
int n = width & ~MASK; \
|
|
if (n > 0) { \
|
|
ANY_SIMD(src_ptr, src_stride, dst_ptr, n); \
|
|
} \
|
|
ptrdiff_t np = n; \
|
|
memcpy(vin, src_ptr + np, r * sizeof(T)); \
|
|
memcpy(vin + 64, src_ptr + src_stride + np, r * sizeof(T)); \
|
|
ANY_SIMD(vin, 64, vout, MASK + 1); \
|
|
memcpy(dst_ptr + np, vout, r * sizeof(T)); \
|
|
}
|
|
|
|
#ifdef HAS_HALFROW_AVX2
|
|
ANY11H(HalfRow_Any_AVX2, HalfRow_AVX2, uint8_t, 31)
|
|
#endif
|
|
#ifdef HAS_HALFROW_16_AVX2
|
|
ANY11H(HalfRow_16_Any_AVX2, HalfRow_16_AVX2, uint16_t, 15)
|
|
#endif
|
|
#ifdef HAS_HALFROW_LSX
|
|
ANY11H(HalfRow_Any_LSX, HalfRow_LSX, uint8_t, 15)
|
|
#endif
|
|
#ifdef HAS_HALFROW_LASX
|
|
ANY11H(HalfRow_Any_LASX, HalfRow_LASX, uint8_t, 31)
|
|
#endif
|
|
#ifdef HAS_HALFROW_16_LSX
|
|
ANY11H(HalfRow_16_Any_LSX, HalfRow_16_LSX, uint16_t, 7)
|
|
#endif
|
|
#ifdef HAS_HALFROW_16_LASX
|
|
ANY11H(HalfRow_16_Any_LASX, HalfRow_16_LASX, uint16_t, 15)
|
|
#endif
|
|
|
|
#undef ANY11H
|
|
|
|
// Any 1 to 1 half row with scale param
|
|
#define ANY11HS(NAMEANY, ANY_SIMD, TS, TD, MASK) \
|
|
void NAMEANY(const TS* src_ptr, ptrdiff_t src_stride, TD* dst_ptr, \
|
|
int scale, int width) { \
|
|
SIMD_ALIGNED(TS vin[64 * 2]); \
|
|
static_assert((MASK + 1) * sizeof(TS) <= sizeof(vin) / 2, \
|
|
"vin buffer too small"); \
|
|
SIMD_ALIGNED(TD vout[64]); \
|
|
static_assert((MASK + 1) * sizeof(TD) <= sizeof(vout), \
|
|
"vout buffer too small"); \
|
|
MSAN_MEMSET(vin, 0, sizeof(vin)); /* for msan */ \
|
|
MSAN_MEMSET(vout, 0, sizeof(vout)); /* for msan */ \
|
|
int r = width & MASK; \
|
|
int n = width & ~MASK; \
|
|
if (n > 0) { \
|
|
ANY_SIMD(src_ptr, src_stride, dst_ptr, scale, n); \
|
|
} \
|
|
ptrdiff_t np = n; \
|
|
memcpy(vin, src_ptr + np, r * sizeof(TS)); \
|
|
memcpy(vin + 64, src_ptr + src_stride + np, r * sizeof(TS)); \
|
|
ANY_SIMD(vin, 64, vout, scale, MASK + 1); \
|
|
memcpy(dst_ptr + np, vout, r * sizeof(TD)); \
|
|
}
|
|
|
|
#ifdef HAS_HALFROW_16TO8_SSSE3
|
|
ANY11HS(HalfRow_16To8_Any_SSSE3, HalfRow_16To8_SSSE3, uint16_t, uint8_t, 15)
|
|
#endif
|
|
#ifdef HAS_HALFROW_16TO8_AVX2
|
|
ANY11HS(HalfRow_16To8_Any_AVX2, HalfRow_16To8_AVX2, uint16_t, uint8_t, 31)
|
|
#endif
|
|
#ifdef HAS_HALFROW_16TO8_NEON
|
|
ANY11HS(HalfRow_16To8_Any_NEON, HalfRow_16To8_NEON, uint16_t, uint8_t, 15)
|
|
#endif
|
|
#ifdef HAS_HALFROW_16TO8_LSX
|
|
ANY11HS(HalfRow_16To8_Any_LSX, HalfRow_16To8_LSX, uint16_t, uint8_t, 15)
|
|
#endif
|
|
#ifdef HAS_HALFROW_16TO8_LASX
|
|
ANY11HS(HalfRow_16To8_Any_LASX, HalfRow_16To8_LASX, uint16_t, uint8_t, 31)
|
|
#endif
|
|
|
|
#undef ANY11HS
|
|
|
|
// Any 2 to 1 half-width row with scale param
|
|
#define ANY11HWS(NAMEANY, ANY_SIMD, TS, TD, MASK) \
|
|
void NAMEANY(const TS* src_ptr, ptrdiff_t src_stride, TD* dst_ptr, \
|
|
int scale, int width) { \
|
|
SIMD_ALIGNED(TS vin[128 * 2]); \
|
|
static_assert((MASK + 1) * 2 * sizeof(TS) <= sizeof(vin), \
|
|
"vin buffer too small"); \
|
|
SIMD_ALIGNED(TD vout[64]); \
|
|
static_assert((MASK + 1) * sizeof(TD) <= sizeof(vout), \
|
|
"vout buffer too small"); \
|
|
MSAN_MEMSET(vin, 0, sizeof(vin)); /* for msan */ \
|
|
MSAN_MEMSET(vout, 0, sizeof(vout)); /* for msan */ \
|
|
int r = width & MASK; \
|
|
int n = width & ~MASK; \
|
|
if (n > 0) { \
|
|
ANY_SIMD(src_ptr, src_stride, dst_ptr, scale, n); \
|
|
} \
|
|
ptrdiff_t np = n; \
|
|
memcpy(vin, src_ptr + np * 2, r * 2 * sizeof(TS)); \
|
|
memcpy(vin + 128, src_ptr + src_stride + np * 2, r * 2 * sizeof(TS)); \
|
|
ANY_SIMD(vin, 128, vout, scale, MASK + 1); \
|
|
memcpy(dst_ptr + np, vout, r * sizeof(TD)); \
|
|
}
|
|
|
|
#ifdef HAS_HALFWIDTHROW_16TO8_SSSE3
|
|
ANY11HWS(HalfWidthRow_16To8_Any_SSSE3,
|
|
HalfWidthRow_16To8_SSSE3,
|
|
uint16_t,
|
|
uint8_t,
|
|
7)
|
|
#endif
|
|
#ifdef HAS_HALFWIDTHROW_16TO8_AVX2
|
|
ANY11HWS(HalfWidthRow_16To8_Any_AVX2,
|
|
HalfWidthRow_16To8_AVX2,
|
|
uint16_t,
|
|
uint8_t,
|
|
15)
|
|
#endif
|
|
#ifdef HAS_HALFWIDTHROW_16TO8_NEON
|
|
ANY11HWS(HalfWidthRow_16To8_Any_NEON,
|
|
HalfWidthRow_16To8_NEON,
|
|
uint16_t,
|
|
uint8_t,
|
|
7)
|
|
void HalfWidthRow_16To8_Odd_NEON(const uint16_t* src_uv,
|
|
ptrdiff_t src_uv_stride,
|
|
uint8_t* dst_uv,
|
|
int scale,
|
|
int width) {
|
|
int n = width - 1;
|
|
if (n > 0) {
|
|
if (IS_ALIGNED(n, 8)) {
|
|
HalfWidthRow_16To8_NEON(src_uv, src_uv_stride, dst_uv, scale, n);
|
|
} else {
|
|
HalfWidthRow_16To8_Any_NEON(src_uv, src_uv_stride, dst_uv, scale, n);
|
|
}
|
|
src_uv += (ptrdiff_t)n * 2;
|
|
dst_uv += n;
|
|
}
|
|
const uint16_t* src_uv1 = src_uv + src_uv_stride;
|
|
uint32_t avg = (uint32_t)src_uv[0] + (uint32_t)src_uv1[0] + 1u;
|
|
dst_uv[0] = (uint8_t)((avg * (uint32_t)scale) >> 17);
|
|
}
|
|
#endif
|
|
|
|
#undef ANY11HWS
|
|
|
|
// Any 1 to 1 mirror.
|
|
#define ANY11M(NAMEANY, ANY_SIMD, BPP, MASK) \
|
|
void NAMEANY(const uint8_t* src_ptr, uint8_t* dst_ptr, int width) { \
|
|
SIMD_ALIGNED(uint8_t vin[128]); \
|
|
static_assert((MASK + 1) * BPP <= sizeof(vin), "vin buffer too small"); \
|
|
SIMD_ALIGNED(uint8_t vout[128]); \
|
|
static_assert((MASK + 1) * BPP <= sizeof(vout), "vout buffer too small"); \
|
|
MSAN_MEMSET(vin, 0, sizeof(vin)); /* for msan */ \
|
|
MSAN_MEMSET(vout, 0, sizeof(vout)); /* for msan */ \
|
|
int r = width & MASK; \
|
|
int n = width & ~MASK; \
|
|
if (n > 0) { \
|
|
ANY_SIMD(src_ptr + r * BPP, dst_ptr, n); \
|
|
} \
|
|
ptrdiff_t np = n; \
|
|
memcpy(vin, src_ptr, r * BPP); \
|
|
ANY_SIMD(vin, vout, MASK + 1); \
|
|
memcpy(dst_ptr + np * BPP, vout + (MASK + 1 - r) * BPP, r * BPP); \
|
|
}
|
|
|
|
#ifdef HAS_MIRRORROW_AVX2
|
|
ANY11M(MirrorRow_Any_AVX2, MirrorRow_AVX2, 1, 31)
|
|
#endif
|
|
#ifdef HAS_MIRRORROW_SSSE3
|
|
ANY11M(MirrorRow_Any_SSSE3, MirrorRow_SSSE3, 1, 15)
|
|
#endif
|
|
#ifdef HAS_MIRRORROW_NEON
|
|
ANY11M(MirrorRow_Any_NEON, MirrorRow_NEON, 1, 31)
|
|
#endif
|
|
#ifdef HAS_MIRRORROW_LSX
|
|
ANY11M(MirrorRow_Any_LSX, MirrorRow_LSX, 1, 15)
|
|
#endif
|
|
#ifdef HAS_MIRRORROW_LASX
|
|
ANY11M(MirrorRow_Any_LASX, MirrorRow_LASX, 1, 31)
|
|
#endif
|
|
#ifdef HAS_MIRRORUVROW_AVX2
|
|
ANY11M(MirrorUVRow_Any_AVX2, MirrorUVRow_AVX2, 2, 15)
|
|
#endif
|
|
#ifdef HAS_MIRRORUVROW_SSSE3
|
|
ANY11M(MirrorUVRow_Any_SSSE3, MirrorUVRow_SSSE3, 2, 7)
|
|
#endif
|
|
#ifdef HAS_MIRRORUVROW_NEON
|
|
ANY11M(MirrorUVRow_Any_NEON, MirrorUVRow_NEON, 2, 31)
|
|
#endif
|
|
#ifdef HAS_MIRRORUVROW_LSX
|
|
ANY11M(MirrorUVRow_Any_LSX, MirrorUVRow_LSX, 2, 7)
|
|
#endif
|
|
#ifdef HAS_MIRRORUVROW_LASX
|
|
ANY11M(MirrorUVRow_Any_LASX, MirrorUVRow_LASX, 2, 15)
|
|
#endif
|
|
#ifdef HAS_ARGBMIRRORROW_AVX2
|
|
ANY11M(ARGBMirrorRow_Any_AVX2, ARGBMirrorRow_AVX2, 4, 7)
|
|
#endif
|
|
#ifdef HAS_ARGBMIRRORROW_SSE2
|
|
ANY11M(ARGBMirrorRow_Any_SSE2, ARGBMirrorRow_SSE2, 4, 3)
|
|
#endif
|
|
#ifdef HAS_ARGBMIRRORROW_NEON
|
|
ANY11M(ARGBMirrorRow_Any_NEON, ARGBMirrorRow_NEON, 4, 7)
|
|
#endif
|
|
#ifdef HAS_ARGBMIRRORROW_LSX
|
|
ANY11M(ARGBMirrorRow_Any_LSX, ARGBMirrorRow_LSX, 4, 3)
|
|
#endif
|
|
#ifdef HAS_ARGBMIRRORROW_LASX
|
|
ANY11M(ARGBMirrorRow_Any_LASX, ARGBMirrorRow_LASX, 4, 7)
|
|
#endif
|
|
#ifdef HAS_RGB24MIRRORROW_AVX2
|
|
ANY11M(RGB24MirrorRow_Any_AVX2, RGB24MirrorRow_AVX2, 3, 31)
|
|
#endif
|
|
#ifdef HAS_RGB24MIRRORROW_SSSE3
|
|
ANY11M(RGB24MirrorRow_Any_SSSE3, RGB24MirrorRow_SSSE3, 3, 15)
|
|
#endif
|
|
#ifdef HAS_RGB24MIRRORROW_NEON
|
|
ANY11M(RGB24MirrorRow_Any_NEON, RGB24MirrorRow_NEON, 3, 15)
|
|
#endif
|
|
#ifdef HAS_RGB24MIRRORROW_LSX
|
|
ANY11M(RGB24MirrorRow_Any_LSX, RGB24MirrorRow_LSX, 3, 15)
|
|
#endif
|
|
#ifdef HAS_RGB24MIRRORROW_LASX
|
|
ANY11M(RGB24MirrorRow_Any_LASX, RGB24MirrorRow_LASX, 3, 31)
|
|
#endif
|
|
#ifdef HAS_MIRRORROW_WASMSIMD
|
|
ANY11M(MirrorRow_Any_WASMSIMD, MirrorRow_WASMSIMD, 1, 15)
|
|
#endif
|
|
#ifdef HAS_MIRRORUVROW_WASMSIMD
|
|
ANY11M(MirrorUVRow_Any_WASMSIMD, MirrorUVRow_WASMSIMD, 2, 7)
|
|
#endif
|
|
#ifdef HAS_ARGBMIRRORROW_WASMSIMD
|
|
ANY11M(ARGBMirrorRow_Any_WASMSIMD, ARGBMirrorRow_WASMSIMD, 4, 3)
|
|
#endif
|
|
#ifdef HAS_RGB24MIRRORROW_WASMSIMD
|
|
ANY11M(RGB24MirrorRow_Any_WASMSIMD, RGB24MirrorRow_WASMSIMD, 3, 15)
|
|
#endif
|
|
#undef ANY11M
|
|
|
|
// Any 1 to 1 mirror with type.
|
|
#define ANY11MT(NAMEANY, ANY_SIMD, T, BPP, MASK) \
|
|
void NAMEANY(const T* src_ptr, T* dst_ptr, int width) { \
|
|
SIMD_ALIGNED(T vin[64]); \
|
|
static_assert((MASK + 1) * BPP <= sizeof(vin), "vin buffer too small"); \
|
|
SIMD_ALIGNED(T vout[64]); \
|
|
static_assert((MASK + 1) * BPP <= sizeof(vout), "vout buffer too small"); \
|
|
MSAN_MEMSET(vin, 0, sizeof(vin)); /* for msan */ \
|
|
MSAN_MEMSET(vout, 0, sizeof(vout)); /* for msan */ \
|
|
int r = width & MASK; \
|
|
int n = width & ~MASK; \
|
|
if (n > 0) { \
|
|
ANY_SIMD(src_ptr + r, dst_ptr, n); \
|
|
} \
|
|
ptrdiff_t np = n; \
|
|
memcpy(vin, src_ptr, r * BPP); \
|
|
ANY_SIMD(vin, vout, MASK + 1); \
|
|
memcpy(dst_ptr + np, vout + (MASK + 1 - r), r * BPP); \
|
|
}
|
|
|
|
#ifdef HAS_MIRRORROW_16_NEON
|
|
ANY11MT(MirrorRow_16_Any_NEON, MirrorRow_16_NEON, uint16_t, 2, 15)
|
|
#endif
|
|
#undef ANY11MT
|
|
|
|
// Any 1 to 2 mirror split UV.
|
|
#define ANY12MR(NAMEANY, ANY_SIMD, MASK) \
|
|
void NAMEANY(const uint8_t* src_uv, uint8_t* dst_u, uint8_t* dst_v, \
|
|
int width) { \
|
|
SIMD_ALIGNED(uint8_t vin[128]); \
|
|
static_assert((MASK + 1) * 2 <= sizeof(vin), "vin buffer too small"); \
|
|
SIMD_ALIGNED(uint8_t vout[64 * 2]); \
|
|
static_assert(MASK + 1 <= sizeof(vout) / 2, "vout buffer too small"); \
|
|
MSAN_MEMSET(vin, 0, sizeof(vin)); /* for msan */ \
|
|
MSAN_MEMSET(vout, 0, sizeof(vout)); /* for msan */ \
|
|
int r = width & MASK; \
|
|
int n = width & ~MASK; \
|
|
if (n > 0) { \
|
|
ANY_SIMD(src_uv + r * 2, dst_u, dst_v, n); \
|
|
} \
|
|
ptrdiff_t np = n; \
|
|
memcpy(vin, src_uv, r * 2); \
|
|
ANY_SIMD(vin, vout, vout + 64, MASK + 1); \
|
|
memcpy(dst_u + np, vout + (MASK + 1 - r), r); \
|
|
memcpy(dst_v + np, vout + 64 + (MASK + 1 - r), r); \
|
|
}
|
|
|
|
#ifdef HAS_MIRRORSPLITUVROW_SSSE3
|
|
ANY12MR(MirrorSplitUVRow_Any_SSSE3, MirrorSplitUVRow_SSSE3, 7)
|
|
#endif
|
|
#ifdef HAS_MIRRORSPLITUVROW_AVX2
|
|
ANY12MR(MirrorSplitUVRow_Any_AVX2, MirrorSplitUVRow_AVX2, 15)
|
|
#endif
|
|
#ifdef HAS_MIRRORSPLITUVROW_NEON
|
|
ANY12MR(MirrorSplitUVRow_Any_NEON, MirrorSplitUVRow_NEON, 15)
|
|
#endif
|
|
#ifdef HAS_MIRRORSPLITUVROW_LSX
|
|
ANY12MR(MirrorSplitUVRow_Any_LSX, MirrorSplitUVRow_LSX, 15)
|
|
#endif
|
|
#ifdef HAS_MIRRORSPLITUVROW_LASX
|
|
ANY12MR(MirrorSplitUVRow_Any_LASX, MirrorSplitUVRow_LASX, 31)
|
|
#endif
|
|
#ifdef HAS_MIRRORSPLITUVROW_WASMSIMD
|
|
ANY12MR(MirrorSplitUVRow_Any_WASMSIMD, MirrorSplitUVRow_WASMSIMD, 15)
|
|
#endif
|
|
#undef ANY12MR
|
|
|
|
// Any 1 plane. (memset)
|
|
#define ANY1(NAMEANY, ANY_SIMD, T, BPP, MASK) \
|
|
void NAMEANY(uint8_t* dst_ptr, T v32, int width) { \
|
|
SIMD_ALIGNED(uint8_t vout[64]); \
|
|
static_assert((MASK + 1) * BPP <= sizeof(vout), "vout buffer too small"); \
|
|
MSAN_MEMSET(vout, 0, sizeof(vout)); /* for msan */ \
|
|
int r = width & MASK; \
|
|
int n = width & ~MASK; \
|
|
if (n > 0) { \
|
|
ANY_SIMD(dst_ptr, v32, n); \
|
|
} \
|
|
ptrdiff_t np = n; \
|
|
ANY_SIMD(vout, v32, MASK + 1); \
|
|
memcpy(dst_ptr + np * BPP, vout, r * BPP); \
|
|
}
|
|
|
|
#ifdef HAS_SETROW_AVX2
|
|
ANY1(SetRow_Any_AVX2, SetRow_AVX2, uint8_t, 1, 31)
|
|
#endif
|
|
#ifdef HAS_SETROW_NEON
|
|
ANY1(SetRow_Any_NEON, SetRow_NEON, uint8_t, 1, 15)
|
|
#endif
|
|
#ifdef HAS_SETROW_LSX
|
|
ANY1(SetRow_Any_LSX, SetRow_LSX, uint8_t, 1, 15)
|
|
#endif
|
|
#ifdef HAS_SETROW_LASX
|
|
ANY1(SetRow_Any_LASX, SetRow_LASX, uint8_t, 1, 31)
|
|
#endif
|
|
#ifdef HAS_ARGBSETROW_AVX2
|
|
ANY1(ARGBSetRow_Any_AVX2, ARGBSetRow_AVX2, uint32_t, 4, 7)
|
|
#endif
|
|
#ifdef HAS_ARGBSETROW_NEON
|
|
ANY1(ARGBSetRow_Any_NEON, ARGBSetRow_NEON, uint32_t, 4, 3)
|
|
#endif
|
|
#ifdef HAS_ARGBSETROW_LSX
|
|
ANY1(ARGBSetRow_Any_LSX, ARGBSetRow_LSX, uint32_t, 4, 3)
|
|
#endif
|
|
#ifdef HAS_ARGBSETROW_LASX
|
|
ANY1(ARGBSetRow_Any_LASX, ARGBSetRow_LASX, uint32_t, 4, 7)
|
|
#endif
|
|
#undef ANY1
|
|
|
|
// Any 1 to 2. Outputs UV planes.
|
|
#define ANY12(NAMEANY, ANY_SIMD, UVSHIFT, BPP, DUVSHIFT, MASK) \
|
|
void NAMEANY(const uint8_t* src_ptr, uint8_t* dst_u, uint8_t* dst_v, \
|
|
int width) { \
|
|
SIMD_ALIGNED(uint8_t vin[256]); \
|
|
static_assert(SS(MASK + 1, UVSHIFT) * BPP <= sizeof(vin), \
|
|
"vin buffer too small"); \
|
|
SIMD_ALIGNED(uint8_t vout[256 * 2]); \
|
|
static_assert(SS(MASK + 1, DUVSHIFT) <= sizeof(vout) / 2, \
|
|
"vout buffer too small"); \
|
|
MSAN_MEMSET(vin, 0, sizeof(vin)); /* for msan */ \
|
|
MSAN_MEMSET(vout, 0, sizeof(vout)); /* for msan */ \
|
|
int r = width & MASK; \
|
|
int n = width & ~MASK; \
|
|
if (n > 0) { \
|
|
ANY_SIMD(src_ptr, dst_u, dst_v, n); \
|
|
} \
|
|
ptrdiff_t np = n; \
|
|
memcpy(vin, src_ptr + (np >> UVSHIFT) * BPP, SS(r, UVSHIFT) * BPP); \
|
|
ANY_SIMD(vin, vout, vout + 256, MASK + 1); \
|
|
memcpy(dst_u + (np >> DUVSHIFT), vout, SS(r, DUVSHIFT)); \
|
|
memcpy(dst_v + (np >> DUVSHIFT), vout + 256, SS(r, DUVSHIFT)); \
|
|
}
|
|
|
|
#ifdef HAS_SPLITUVROW_SSE2
|
|
ANY12(SplitUVRow_Any_SSE2, SplitUVRow_SSE2, 0, 2, 0, 15)
|
|
#endif
|
|
#ifdef HAS_SPLITUVROW_AVX2
|
|
ANY12(SplitUVRow_Any_AVX2, SplitUVRow_AVX2, 0, 2, 0, 31)
|
|
#endif
|
|
#ifdef HAS_SPLITUVROW_NEON
|
|
ANY12(SplitUVRow_Any_NEON, SplitUVRow_NEON, 0, 2, 0, 15)
|
|
#endif
|
|
#ifdef HAS_SPLITUVROW_LSX
|
|
ANY12(SplitUVRow_Any_LSX, SplitUVRow_LSX, 0, 2, 0, 15)
|
|
#endif
|
|
#ifdef HAS_SPLITUVROW_LASX
|
|
ANY12(SplitUVRow_Any_LASX, SplitUVRow_LASX, 0, 2, 0, 31)
|
|
#endif
|
|
#ifdef HAS_YUY2TOUV422ROW_AVX2
|
|
ANY12(YUY2ToUV422Row_Any_AVX2, YUY2ToUV422Row_AVX2, 1, 4, 1, 31)
|
|
#endif
|
|
#ifdef HAS_UYVYTOUV422ROW_AVX2
|
|
ANY12(UYVYToUV422Row_Any_AVX2, UYVYToUV422Row_AVX2, 1, 4, 1, 31)
|
|
#endif
|
|
#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
|
|
#ifdef HAS_YUY2TOUV422ROW_NEON
|
|
ANY12(YUY2ToUV422Row_Any_NEON, YUY2ToUV422Row_NEON, 1, 4, 1, 15)
|
|
ANY12(UYVYToUV422Row_Any_NEON, UYVYToUV422Row_NEON, 1, 4, 1, 15)
|
|
#endif
|
|
#ifdef HAS_YUY2TOUV422ROW_LSX
|
|
ANY12(YUY2ToUV422Row_Any_LSX, YUY2ToUV422Row_LSX, 1, 4, 1, 15)
|
|
ANY12(UYVYToUV422Row_Any_LSX, UYVYToUV422Row_LSX, 1, 4, 1, 15)
|
|
#endif
|
|
#ifdef HAS_YUY2TOUV422ROW_LASX
|
|
ANY12(YUY2ToUV422Row_Any_LASX, YUY2ToUV422Row_LASX, 1, 4, 1, 31)
|
|
ANY12(UYVYToUV422Row_Any_LASX, UYVYToUV422Row_LASX, 1, 4, 1, 31)
|
|
#endif
|
|
#undef ANY12
|
|
|
|
// Any 2 16 bit planes with parameter to 1
|
|
#define ANY12PT(NAMEANY, ANY_SIMD, T, BPP, MASK) \
|
|
void NAMEANY(const T* src_uv, T* dst_u, T* dst_v, int depth, int width) { \
|
|
SIMD_ALIGNED(T vin[16 * 2]); \
|
|
static_assert((MASK + 1) * BPP * 2 <= sizeof(vin), \
|
|
"vin buffer too small"); \
|
|
SIMD_ALIGNED(T vout[16 * 2]); \
|
|
static_assert((MASK + 1) * BPP <= sizeof(vout) / 2, \
|
|
"vout buffer too small"); \
|
|
MSAN_MEMSET(vin, 0, sizeof(vin)); /* for msan */ \
|
|
MSAN_MEMSET(vout, 0, sizeof(vout)); /* for msan */ \
|
|
int r = width & MASK; \
|
|
int n = width & ~MASK; \
|
|
if (n > 0) { \
|
|
ANY_SIMD(src_uv, dst_u, dst_v, depth, n); \
|
|
} \
|
|
ptrdiff_t np = n; \
|
|
memcpy(vin, src_uv + np * 2, r * BPP * 2); \
|
|
ANY_SIMD(vin, vout, vout + 16, depth, MASK + 1); \
|
|
memcpy(dst_u + np, vout, r * BPP); \
|
|
memcpy(dst_v + np, vout + 16, r * BPP); \
|
|
}
|
|
|
|
#ifdef HAS_SPLITUVROW_16_AVX2
|
|
ANY12PT(SplitUVRow_16_Any_AVX2, SplitUVRow_16_AVX2, uint16_t, 2, 15)
|
|
#endif
|
|
|
|
#ifdef HAS_SPLITUVROW_16_NEON
|
|
ANY12PT(SplitUVRow_16_Any_NEON, SplitUVRow_16_NEON, uint16_t, 2, 7)
|
|
#endif
|
|
#ifdef HAS_SPLITUVROW_16_LSX
|
|
ANY12PT(SplitUVRow_16_Any_LSX, SplitUVRow_16_LSX, uint16_t, 2, 7)
|
|
#endif
|
|
#ifdef HAS_SPLITUVROW_16_LASX
|
|
ANY12PT(SplitUVRow_16_Any_LASX, SplitUVRow_16_LASX, uint16_t, 2, 15)
|
|
#endif
|
|
|
|
#undef ANY21CT
|
|
|
|
// Any 1 to 3. Outputs RGB planes.
|
|
#define ANY13(NAMEANY, ANY_SIMD, BPP, MASK) \
|
|
void NAMEANY(const uint8_t* src_ptr, uint8_t* dst_r, uint8_t* dst_g, \
|
|
uint8_t* dst_b, int width) { \
|
|
SIMD_ALIGNED(uint8_t vin[(MASK + 1) * BPP]); \
|
|
SIMD_ALIGNED(uint8_t vout[(MASK + 1) * 3]); \
|
|
MSAN_MEMSET(vin, 0, sizeof(vin)); /* for msan */ \
|
|
MSAN_MEMSET(vout, 0, sizeof(vout)); /* for msan */ \
|
|
int r = width & MASK; \
|
|
int n = width & ~MASK; \
|
|
if (n > 0) { \
|
|
ANY_SIMD(src_ptr, dst_r, dst_g, dst_b, n); \
|
|
} \
|
|
ptrdiff_t np = n; \
|
|
memcpy(vin, src_ptr + np * BPP, r * BPP); \
|
|
ANY_SIMD(vin, vout, vout + MASK + 1, vout + (MASK + 1) * 2, MASK + 1); \
|
|
memcpy(dst_r + np, vout, r); \
|
|
memcpy(dst_g + np, vout + MASK + 1, r); \
|
|
memcpy(dst_b + np, vout + (MASK + 1) * 2, r); \
|
|
}
|
|
|
|
#ifdef HAS_SPLITRGBROW_SSSE3
|
|
ANY13(SplitRGBRow_Any_SSSE3, SplitRGBRow_SSSE3, 3, 15)
|
|
#endif
|
|
#ifdef HAS_SPLITRGBROW_SSE41
|
|
ANY13(SplitRGBRow_Any_SSE41, SplitRGBRow_SSE41, 3, 15)
|
|
#endif
|
|
#ifdef HAS_SPLITRGBROW_AVX2
|
|
ANY13(SplitRGBRow_Any_AVX2, SplitRGBRow_AVX2, 3, 31)
|
|
#endif
|
|
#ifdef HAS_SPLITRGBROW_NEON
|
|
ANY13(SplitRGBRow_Any_NEON, SplitRGBRow_NEON, 3, 15)
|
|
#endif
|
|
#ifdef HAS_SPLITXRGBROW_SSE2
|
|
ANY13(SplitXRGBRow_Any_SSE2, SplitXRGBRow_SSE2, 4, 7)
|
|
#endif
|
|
#ifdef HAS_SPLITXRGBROW_SSSE3
|
|
ANY13(SplitXRGBRow_Any_SSSE3, SplitXRGBRow_SSSE3, 4, 7)
|
|
#endif
|
|
#ifdef HAS_SPLITXRGBROW_AVX2
|
|
ANY13(SplitXRGBRow_Any_AVX2, SplitXRGBRow_AVX2, 4, 15)
|
|
#endif
|
|
#ifdef HAS_SPLITXRGBROW_NEON
|
|
ANY13(SplitXRGBRow_Any_NEON, SplitXRGBRow_NEON, 4, 15)
|
|
#endif
|
|
#ifdef HAS_SPLITXRGBROW_LSX
|
|
ANY13(SplitXRGBRow_Any_LSX, SplitXRGBRow_LSX, 4, 15)
|
|
#endif
|
|
#ifdef HAS_SPLITXRGBROW_LASX
|
|
ANY13(SplitXRGBRow_Any_LASX, SplitXRGBRow_LASX, 4, 15)
|
|
#endif
|
|
|
|
// Any 1 to 4. Outputs ARGB planes.
|
|
#define ANY14(NAMEANY, ANY_SIMD, BPP, MASK) \
|
|
void NAMEANY(const uint8_t* src_ptr, uint8_t* dst_r, uint8_t* dst_g, \
|
|
uint8_t* dst_b, uint8_t* dst_a, int width) { \
|
|
SIMD_ALIGNED(uint8_t vin[16 * 4]); \
|
|
static_assert((MASK + 1) * BPP <= sizeof(vin), "vin buffer too small"); \
|
|
SIMD_ALIGNED(uint8_t vout[16 * 4]); \
|
|
static_assert(MASK + 1 <= sizeof(vout) / 4, "vout buffer too small"); \
|
|
MSAN_MEMSET(vin, 0, sizeof(vin)); /* for msan */ \
|
|
MSAN_MEMSET(vout, 0, sizeof(vout)); /* for msan */ \
|
|
int r = width & MASK; \
|
|
int n = width & ~MASK; \
|
|
if (n > 0) { \
|
|
ANY_SIMD(src_ptr, dst_r, dst_g, dst_b, dst_a, n); \
|
|
} \
|
|
ptrdiff_t np = n; \
|
|
memcpy(vin, src_ptr + np * BPP, r * BPP); \
|
|
ANY_SIMD(vin, vout, vout + 16, vout + 32, vout + 48, MASK + 1); \
|
|
memcpy(dst_r + np, vout, r); \
|
|
memcpy(dst_g + np, vout + 16, r); \
|
|
memcpy(dst_b + np, vout + 32, r); \
|
|
memcpy(dst_a + np, vout + 48, r); \
|
|
}
|
|
|
|
#ifdef HAS_SPLITARGBROW_SSE2
|
|
ANY14(SplitARGBRow_Any_SSE2, SplitARGBRow_SSE2, 4, 7)
|
|
#endif
|
|
#ifdef HAS_SPLITARGBROW_SSSE3
|
|
ANY14(SplitARGBRow_Any_SSSE3, SplitARGBRow_SSSE3, 4, 7)
|
|
#endif
|
|
#ifdef HAS_SPLITARGBROW_AVX2
|
|
ANY14(SplitARGBRow_Any_AVX2, SplitARGBRow_AVX2, 4, 15)
|
|
#endif
|
|
#ifdef HAS_SPLITARGBROW_NEON
|
|
ANY14(SplitARGBRow_Any_NEON, SplitARGBRow_NEON, 4, 15)
|
|
#endif
|
|
#ifdef HAS_SPLITARGBROW_LSX
|
|
ANY14(SplitARGBRow_Any_LSX, SplitARGBRow_LSX, 4, 15)
|
|
#endif
|
|
#ifdef HAS_SPLITARGBROW_LASX
|
|
ANY14(SplitARGBRow_Any_LASX, SplitARGBRow_LASX, 4, 15)
|
|
#endif
|
|
|
|
// 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_t* src_ptr, int src_stride, uint8_t* dst_u, \
|
|
uint8_t* dst_v, int width) { \
|
|
SIMD_ALIGNED(uint8_t vin[256 * 2]); \
|
|
static_assert(SS(MASK + 1, UVSHIFT) * BPP <= sizeof(vin) / 2, \
|
|
"vin buffer too small"); \
|
|
SIMD_ALIGNED(uint8_t vout[256 * 2]); \
|
|
static_assert(SS(MASK + 1, 1) <= sizeof(vout) / 2, \
|
|
"vout buffer too small"); \
|
|
MSAN_MEMSET(vin, 0, sizeof(vin)); /* for msan */ \
|
|
MSAN_MEMSET(vout, 0, sizeof(vout)); /* for msan */ \
|
|
int r = width & MASK; \
|
|
int n = width & ~MASK; \
|
|
if (n > 0) { \
|
|
ANY_SIMD(src_ptr, src_stride, dst_u, dst_v, n); \
|
|
} \
|
|
ptrdiff_t np = n; \
|
|
memcpy(vin, src_ptr + (np >> UVSHIFT) * BPP, SS(r, UVSHIFT) * BPP); \
|
|
memcpy(vin + 256, src_ptr + src_stride + (np >> UVSHIFT) * BPP, \
|
|
SS(r, UVSHIFT) * BPP); \
|
|
if ((width & 1) && UVSHIFT == 0) { /* repeat last pixel for subsample */ \
|
|
memcpy(vin + SS(r, UVSHIFT) * BPP, vin + SS(r, UVSHIFT) * BPP - BPP, \
|
|
BPP); \
|
|
memcpy(vin + 256 + SS(r, UVSHIFT) * BPP, \
|
|
vin + 256 + SS(r, UVSHIFT) * BPP - BPP, BPP); \
|
|
} \
|
|
ANY_SIMD(vin, 256, vout, vout + 256, MASK + 1); \
|
|
memcpy(dst_u + (np >> 1), vout, SS(r, 1)); \
|
|
memcpy(dst_v + (np >> 1), vout + 256, SS(r, 1)); \
|
|
}
|
|
|
|
#define ANY12M(NAMEANY, ANY_SIMD, BPP, MASK) \
|
|
void NAMEANY(const uint8_t* src_ptr, uint8_t* dst_u, uint8_t* dst_v, \
|
|
int width, const struct ArgbConstants* c) { \
|
|
SIMD_ALIGNED(uint8_t vin[256]); \
|
|
static_assert((MASK + 1) * BPP <= sizeof(vin), "vin buffer too small"); \
|
|
SIMD_ALIGNED(uint8_t vout[256 * 2]); \
|
|
static_assert(MASK + 1 <= sizeof(vout) / 2, "vout buffer too small"); \
|
|
MSAN_MEMSET(vin, 0, sizeof(vin)); /* for msan */ \
|
|
MSAN_MEMSET(vout, 0, sizeof(vout)); /* for msan */ \
|
|
int r = width & MASK; \
|
|
int n = width & ~MASK; \
|
|
if (n > 0) { \
|
|
ANY_SIMD(src_ptr, dst_u, dst_v, n, c); \
|
|
} \
|
|
ptrdiff_t np = n; \
|
|
memcpy(vin, src_ptr + np * BPP, r * BPP); \
|
|
ANY_SIMD(vin, vout, vout + 256, MASK + 1, c); \
|
|
memcpy(dst_u + np, vout, r); \
|
|
memcpy(dst_v + np, vout + 256, r); \
|
|
}
|
|
|
|
#define ANY12MS(NAMEANY, ANY_SIMD, UVSHIFT, BPP, MASK) \
|
|
void NAMEANY(const uint8_t* src_ptr, int src_stride, uint8_t* dst_u, \
|
|
uint8_t* dst_v, int width, const struct ArgbConstants* c) { \
|
|
SIMD_ALIGNED(uint8_t vin[256 * 2]); \
|
|
static_assert(SS(MASK + 1, UVSHIFT) * BPP <= sizeof(vin) / 2, \
|
|
"vin buffer too small"); \
|
|
SIMD_ALIGNED(uint8_t vout[256 * 2]); \
|
|
static_assert(SS(MASK + 1, 1) <= sizeof(vout) / 2, \
|
|
"vout buffer too small"); \
|
|
MSAN_MEMSET(vin, 0, sizeof(vin)); /* for msan */ \
|
|
MSAN_MEMSET(vout, 0, sizeof(vout)); /* for msan */ \
|
|
int r = width & MASK; \
|
|
int n = width & ~MASK; \
|
|
if (n > 0) { \
|
|
ANY_SIMD(src_ptr, src_stride, dst_u, dst_v, n, c); \
|
|
} \
|
|
ptrdiff_t np = n; \
|
|
memcpy(vin, src_ptr + (np >> UVSHIFT) * BPP, SS(r, UVSHIFT) * BPP); \
|
|
memcpy(vin + 256, src_ptr + src_stride + (np >> UVSHIFT) * BPP, \
|
|
SS(r, UVSHIFT) * BPP); \
|
|
if ((width & 1) && UVSHIFT == 0) { /* repeat last pixel for subsample */ \
|
|
memcpy(vin + SS(r, UVSHIFT) * BPP, vin + SS(r, UVSHIFT) * BPP - BPP, \
|
|
BPP); \
|
|
memcpy(vin + 256 + SS(r, UVSHIFT) * BPP, \
|
|
vin + 256 + SS(r, UVSHIFT) * BPP - BPP, BPP); \
|
|
} \
|
|
ANY_SIMD(vin, 256, vout, vout + 256, MASK + 1, c); \
|
|
memcpy(dst_u + (np >> 1), vout, SS(r, 1)); \
|
|
memcpy(dst_v + (np >> 1), vout + 256, SS(r, 1)); \
|
|
}
|
|
|
|
#ifdef HAS_ARGBTOUVMATRIXROW_NEON
|
|
ANY12MS(ARGBToUVMatrixRow_Any_NEON, ARGBToUVMatrixRow_NEON, 0, 4, 15)
|
|
#endif
|
|
#ifdef HAS_ARGBTOUVMATRIXROW_NEON_I8MM
|
|
ANY12MS(ARGBToUVMatrixRow_Any_NEON_I8MM, ARGBToUVMatrixRow_NEON_I8MM, 0, 4, 15)
|
|
#endif
|
|
#ifdef HAS_RGBTOUVMATRIXROW_NEON
|
|
ANY12MS(RGBToUVMatrixRow_Any_NEON, RGBToUVMatrixRow_NEON, 0, 3, 15)
|
|
#endif
|
|
#ifdef HAS_RGB565TOUVMATRIXROW_NEON
|
|
ANY12MS(RGB565ToUVMatrixRow_Any_NEON, RGB565ToUVMatrixRow_NEON, 0, 2, 15)
|
|
#endif
|
|
#ifdef HAS_ARGB1555TOUVMATRIXROW_NEON
|
|
ANY12MS(ARGB1555ToUVMatrixRow_Any_NEON, ARGB1555ToUVMatrixRow_NEON, 0, 2, 15)
|
|
#endif
|
|
#ifdef HAS_ARGB4444TOUVMATRIXROW_NEON
|
|
ANY12MS(ARGB4444ToUVMatrixRow_Any_NEON, ARGB4444ToUVMatrixRow_NEON, 0, 2, 15)
|
|
#endif
|
|
#ifdef HAS_ARGBTOUVMATRIXROW_AVX2
|
|
ANY12MS(ARGBToUVMatrixRow_Any_AVX2, ARGBToUVMatrixRow_AVX2, 0, 4, 15)
|
|
ANY12MS(RGBToUVMatrixRow_Any_AVX2, RGBToUVMatrixRow_AVX2, 0, 3, 31)
|
|
ANY12MS(RGB565ToUVMatrixRow_Any_AVX2, RGB565ToUVMatrixRow_AVX2, 0, 2, 15)
|
|
#ifdef HAS_ARGB1555TOARGBROW_AVX2
|
|
ANY12MS(ARGB1555ToUVMatrixRow_Any_AVX2, ARGB1555ToUVMatrixRow_AVX2, 0, 2, 15)
|
|
#endif
|
|
#ifdef HAS_ARGB4444TOARGBROW_AVX2
|
|
ANY12MS(ARGB4444ToUVMatrixRow_Any_AVX2, ARGB4444ToUVMatrixRow_AVX2, 0, 2, 15)
|
|
#endif
|
|
#endif
|
|
#ifdef HAS_ARGBTOUVMATRIXROW_SSSE3
|
|
ANY12MS(ARGBToUVMatrixRow_Any_SSSE3, ARGBToUVMatrixRow_SSSE3, 0, 4, 7)
|
|
#endif
|
|
#ifdef HAS_RGBTOUVMATRIXROW_SSSE3
|
|
ANY12MS(RGBToUVMatrixRow_Any_SSSE3, RGBToUVMatrixRow_SSSE3, 0, 3, 15)
|
|
#endif
|
|
#ifdef HAS_ARGBTOUV444MATRIXROW_AVX2
|
|
ANY12M(ARGBToUV444MatrixRow_Any_AVX2, ARGBToUV444MatrixRow_AVX2, 4, 31)
|
|
#endif
|
|
#ifdef HAS_RGBTOUV444MATRIXROW_AVX2
|
|
ANY12M(RGBToUV444MatrixRow_Any_AVX2, RGBToUV444MatrixRow_AVX2, 3, 31)
|
|
#endif
|
|
#ifdef HAS_ARGBTOUV444MATRIXROW_SSSE3
|
|
ANY12M(ARGBToUV444MatrixRow_Any_SSSE3, ARGBToUV444MatrixRow_SSSE3, 4, 15)
|
|
#endif
|
|
#ifdef HAS_RGBTOUV444MATRIXROW_SSSE3
|
|
ANY12M(RGBToUV444MatrixRow_Any_SSSE3, RGBToUV444MatrixRow_SSSE3, 3, 15)
|
|
#endif
|
|
#ifdef HAS_ARGBTOUV444MATRIXROW_NEON
|
|
ANY12M(ARGBToUV444MatrixRow_Any_NEON, ARGBToUV444MatrixRow_NEON, 4, 7)
|
|
#endif
|
|
#ifdef HAS_RGBTOUV444MATRIXROW_NEON
|
|
ANY12M(RGBToUV444MatrixRow_Any_NEON, RGBToUV444MatrixRow_NEON, 3, 7)
|
|
#endif
|
|
#ifdef HAS_ARGBTOUV444MATRIXROW_NEON_I8MM
|
|
ANY12M(ARGBToUV444MatrixRow_Any_NEON_I8MM, ARGBToUV444MatrixRow_NEON_I8MM, 4, 7)
|
|
#endif
|
|
#ifdef HAS_ARGBTOUVMATRIXROW_WASMSIMD
|
|
ANY12MS(ARGBToUVMatrixRow_Any_WASMSIMD, ARGBToUVMatrixRow_WASMSIMD, 0, 4, 15)
|
|
#endif
|
|
#ifdef HAS_RGBTOUVMATRIXROW_WASMSIMD
|
|
ANY12MS(RGBToUVMatrixRow_Any_WASMSIMD, RGBToUVMatrixRow_WASMSIMD, 0, 3, 15)
|
|
#endif
|
|
#ifdef HAS_RGB565TOUVMATRIXROW_WASMSIMD
|
|
ANY12MS(RGB565ToUVMatrixRow_Any_WASMSIMD,
|
|
RGB565ToUVMatrixRow_WASMSIMD,
|
|
0,
|
|
2,
|
|
15)
|
|
#endif
|
|
#ifdef HAS_ARGB1555TOUVMATRIXROW_WASMSIMD
|
|
ANY12MS(ARGB1555ToUVMatrixRow_Any_WASMSIMD,
|
|
ARGB1555ToUVMatrixRow_WASMSIMD,
|
|
0,
|
|
2,
|
|
15)
|
|
#endif
|
|
#ifdef HAS_ARGB4444TOUVMATRIXROW_WASMSIMD
|
|
ANY12MS(ARGB4444ToUVMatrixRow_Any_WASMSIMD,
|
|
ARGB4444ToUVMatrixRow_WASMSIMD,
|
|
0,
|
|
2,
|
|
15)
|
|
#endif
|
|
#ifdef HAS_ARGBTOUV444MATRIXROW_WASMSIMD
|
|
ANY12M(ARGBToUV444MatrixRow_Any_WASMSIMD, ARGBToUV444MatrixRow_WASMSIMD, 4, 15)
|
|
#endif
|
|
#ifdef HAS_RGBTOUV444MATRIXROW_WASMSIMD
|
|
ANY12M(RGBToUV444MatrixRow_Any_WASMSIMD, RGBToUV444MatrixRow_WASMSIMD, 3, 15)
|
|
#endif
|
|
|
|
#define ANY11MC(NAMEANY, ANY_SIMD, BPP, MASK) \
|
|
void NAMEANY(const uint8_t* src_ptr, uint8_t* dst_ptr, int width, \
|
|
const struct ArgbConstants* c) { \
|
|
SIMD_ALIGNED(uint8_t vin[256]); \
|
|
static_assert((MASK + 1) * BPP <= sizeof(vin), "vin buffer too small"); \
|
|
SIMD_ALIGNED(uint8_t vout[256]); \
|
|
static_assert(MASK + 1 <= sizeof(vout), "vout buffer too small"); \
|
|
MSAN_MEMSET(vin, 0, sizeof(vin)); /* for msan */ \
|
|
MSAN_MEMSET(vout, 0, sizeof(vout)); /* for msan */ \
|
|
int r = width & MASK; \
|
|
int n = width & ~MASK; \
|
|
if (n > 0) { \
|
|
ANY_SIMD(src_ptr, dst_ptr, n, c); \
|
|
} \
|
|
ptrdiff_t np = n; \
|
|
memcpy(vin, src_ptr + np * BPP, r * BPP); \
|
|
ANY_SIMD(vin, vout, MASK + 1, c); \
|
|
memcpy(dst_ptr + np, vout, r); \
|
|
}
|
|
|
|
#ifdef HAS_ARGBTOYMATRIXROW_SSSE3
|
|
ANY11MC(ARGBToYMatrixRow_Any_SSSE3, ARGBToYMatrixRow_SSSE3, 4, 15)
|
|
ANY11MC(RGBToYMatrixRow_Any_SSSE3, RGBToYMatrixRow_SSSE3, 3, 15)
|
|
#endif
|
|
#ifdef HAS_ARGBTOYMATRIXROW_AVX2
|
|
ANY11MC(ARGBToYMatrixRow_Any_AVX2, ARGBToYMatrixRow_AVX2, 4, 31)
|
|
ANY11MC(RGBToYMatrixRow_Any_AVX2, RGBToYMatrixRow_AVX2, 3, 31)
|
|
ANY11MC(RGB565ToYMatrixRow_Any_AVX2, RGB565ToYMatrixRow_AVX2, 2, 31)
|
|
#ifdef HAS_ARGB1555TOYMATRIXROW_AVX2
|
|
ANY11MC(ARGB1555ToYMatrixRow_Any_AVX2, ARGB1555ToYMatrixRow_AVX2, 2, 31)
|
|
#endif
|
|
#ifdef HAS_ARGB4444TOYMATRIXROW_AVX2
|
|
ANY11MC(ARGB4444ToYMatrixRow_Any_AVX2, ARGB4444ToYMatrixRow_AVX2, 2, 31)
|
|
#endif
|
|
#endif
|
|
#ifdef HAS_ARGBTOYMATRIXROW_NEON
|
|
ANY11MC(ARGBToYMatrixRow_Any_NEON, ARGBToYMatrixRow_NEON, 4, 15)
|
|
#endif
|
|
#ifdef HAS_ARGBTOYMATRIXROW_NEON_DOTPROD
|
|
ANY11MC(ARGBToYMatrixRow_Any_NEON_DotProd, ARGBToYMatrixRow_NEON_DotProd, 4, 15)
|
|
ANY11MC(RGBToYMatrixRow_Any_NEON_DotProd, RGBToYMatrixRow_NEON_DotProd, 3, 15)
|
|
#endif
|
|
#ifdef HAS_RGBTOYMATRIXROW_NEON
|
|
ANY11MC(RGBToYMatrixRow_Any_NEON, RGBToYMatrixRow_NEON, 3, 15)
|
|
#endif
|
|
#ifdef HAS_RGB565TOYMATRIXROW_NEON
|
|
ANY11MC(RGB565ToYMatrixRow_Any_NEON, RGB565ToYMatrixRow_NEON, 2, 15)
|
|
#endif
|
|
#ifdef HAS_ARGB1555TOYMATRIXROW_NEON
|
|
ANY11MC(ARGB1555ToYMatrixRow_Any_NEON, ARGB1555ToYMatrixRow_NEON, 2, 15)
|
|
#endif
|
|
#ifdef HAS_ARGB4444TOYMATRIXROW_NEON
|
|
ANY11MC(ARGB4444ToYMatrixRow_Any_NEON, ARGB4444ToYMatrixRow_NEON, 2, 15)
|
|
#endif
|
|
#ifdef HAS_ARGBTOYMATRIXROW_LSX
|
|
ANY11MC(ARGBToYMatrixRow_Any_LSX, ARGBToYMatrixRow_LSX, 4, 7)
|
|
#endif
|
|
#ifdef HAS_ARGBTOYMATRIXROW_LASX
|
|
ANY11MC(ARGBToYMatrixRow_Any_LASX, ARGBToYMatrixRow_LASX, 4, 15)
|
|
#endif
|
|
#ifdef HAS_RGBTOYMATRIXROW_LSX
|
|
ANY11MC(RGBToYMatrixRow_Any_LSX, RGBToYMatrixRow_LSX, 3, 15)
|
|
#endif
|
|
#ifdef HAS_RGBTOYMATRIXROW_LASX
|
|
ANY11MC(RGBToYMatrixRow_Any_LASX, RGBToYMatrixRow_LASX, 3, 31)
|
|
#endif
|
|
#ifdef HAS_ARGBTOYMATRIXROW_WASMSIMD
|
|
ANY11MC(ARGBToYMatrixRow_Any_WASMSIMD, ARGBToYMatrixRow_WASMSIMD, 4, 15)
|
|
#endif
|
|
#ifdef HAS_RGBTOYMATRIXROW_WASMSIMD
|
|
ANY11MC(RGBToYMatrixRow_Any_WASMSIMD, RGBToYMatrixRow_WASMSIMD, 3, 15)
|
|
#endif
|
|
#ifdef HAS_RGB565TOYMATRIXROW_WASMSIMD
|
|
ANY11MC(RGB565ToYMatrixRow_Any_WASMSIMD, RGB565ToYMatrixRow_WASMSIMD, 2, 15)
|
|
#endif
|
|
#ifdef HAS_ARGB1555TOYMATRIXROW_WASMSIMD
|
|
ANY11MC(ARGB1555ToYMatrixRow_Any_WASMSIMD, ARGB1555ToYMatrixRow_WASMSIMD, 2, 15)
|
|
#endif
|
|
#ifdef HAS_ARGB4444TOYMATRIXROW_WASMSIMD
|
|
ANY11MC(ARGB4444ToYMatrixRow_Any_WASMSIMD, ARGB4444ToYMatrixRow_WASMSIMD, 2, 15)
|
|
#endif
|
|
#undef ANY11MC
|
|
|
|
#ifdef HAS_YUY2TOUVROW_AVX2
|
|
ANY12S(YUY2ToUVRow_Any_AVX2, YUY2ToUVRow_AVX2, 1, 4, 31)
|
|
#endif
|
|
#ifdef HAS_UYVYTOUVROW_AVX2
|
|
ANY12S(UYVYToUVRow_Any_AVX2, UYVYToUVRow_AVX2, 1, 4, 31)
|
|
#endif
|
|
#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_RGB24TOUVROW_NEON
|
|
ANY12S(RGB24ToUVRow_Any_NEON, RGB24ToUVRow_NEON, 0, 3, 15)
|
|
#endif
|
|
#ifdef HAS_RGB24TOUVROW_LSX
|
|
ANY12S(RGB24ToUVRow_Any_LSX, RGB24ToUVRow_LSX, 0, 3, 15)
|
|
#endif
|
|
#ifdef HAS_RGB24TOUVROW_LASX
|
|
ANY12S(RGB24ToUVRow_Any_LASX, RGB24ToUVRow_LASX, 0, 3, 31)
|
|
#endif
|
|
#ifdef HAS_RAWTOUVROW_NEON
|
|
ANY12S(RAWToUVRow_Any_NEON, RAWToUVRow_NEON, 0, 3, 15)
|
|
#endif
|
|
#ifdef HAS_RAWTOUVROW_LSX
|
|
ANY12S(RAWToUVRow_Any_LSX, RAWToUVRow_LSX, 0, 3, 15)
|
|
#endif
|
|
#ifdef HAS_RAWTOUVROW_LASX
|
|
ANY12S(RAWToUVRow_Any_LASX, RAWToUVRow_LASX, 0, 3, 31)
|
|
#endif
|
|
#ifdef HAS_RGB565TOUVROW_AVX2
|
|
ANY12S(RGB565ToUVRow_Any_AVX2, RGB565ToUVRow_AVX2, 0, 2, 31)
|
|
#endif
|
|
#ifdef HAS_ARGB1555TOUVROW_AVX2
|
|
ANY12S(ARGB1555ToUVRow_Any_AVX2, ARGB1555ToUVRow_AVX2, 0, 2, 31)
|
|
#endif
|
|
#ifdef HAS_ARGB4444TOUVROW_AVX2
|
|
ANY12S(ARGB4444ToUVRow_Any_AVX2, ARGB4444ToUVRow_AVX2, 0, 2, 31)
|
|
#endif
|
|
#ifdef HAS_RGB565TOUVROW_NEON
|
|
ANY12S(RGB565ToUVRow_Any_NEON, RGB565ToUVRow_NEON, 0, 2, 15)
|
|
#endif
|
|
#ifdef HAS_RGB565TOUVROW_LSX
|
|
ANY12S(RGB565ToUVRow_Any_LSX, RGB565ToUVRow_LSX, 0, 2, 15)
|
|
#endif
|
|
#ifdef HAS_RGB565TOUVROW_LASX
|
|
ANY12S(RGB565ToUVRow_Any_LASX, RGB565ToUVRow_LASX, 0, 2, 31)
|
|
#endif
|
|
#ifdef HAS_ARGB1555TOUVROW_NEON
|
|
ANY12S(ARGB1555ToUVRow_Any_NEON, ARGB1555ToUVRow_NEON, 0, 2, 15)
|
|
#endif
|
|
#ifdef HAS_ARGB1555TOUVROW_LSX
|
|
ANY12S(ARGB1555ToUVRow_Any_LSX, ARGB1555ToUVRow_LSX, 0, 2, 15)
|
|
#endif
|
|
#ifdef HAS_ARGB1555TOUVROW_LASX
|
|
ANY12S(ARGB1555ToUVRow_Any_LASX, ARGB1555ToUVRow_LASX, 0, 2, 31)
|
|
#endif
|
|
#ifdef HAS_ARGB4444TOUVROW_NEON
|
|
ANY12S(ARGB4444ToUVRow_Any_NEON, ARGB4444ToUVRow_NEON, 0, 2, 15)
|
|
#endif
|
|
#ifdef HAS_YUY2TOUVROW_NEON
|
|
ANY12S(YUY2ToUVRow_Any_NEON, YUY2ToUVRow_NEON, 1, 4, 15)
|
|
#endif
|
|
#ifdef HAS_UYVYTOUVROW_NEON
|
|
ANY12S(UYVYToUVRow_Any_NEON, UYVYToUVRow_NEON, 1, 4, 15)
|
|
#endif
|
|
#ifdef HAS_YUY2TOUVROW_LSX
|
|
ANY12S(YUY2ToUVRow_Any_LSX, YUY2ToUVRow_LSX, 1, 4, 15)
|
|
#endif
|
|
#ifdef HAS_YUY2TOUVROW_LASX
|
|
ANY12S(YUY2ToUVRow_Any_LASX, YUY2ToUVRow_LASX, 1, 4, 31)
|
|
#endif
|
|
#ifdef HAS_UYVYTOUVROW_LSX
|
|
ANY12S(UYVYToUVRow_Any_LSX, UYVYToUVRow_LSX, 1, 4, 15)
|
|
#endif
|
|
#ifdef HAS_UYVYTOUVROW_LASX
|
|
ANY12S(UYVYToUVRow_Any_LASX, UYVYToUVRow_LASX, 1, 4, 31)
|
|
#endif
|
|
#undef ANY12S
|
|
|
|
// Any 1 to 1 with source stride (2 rows of source). Outputs UV plane.
|
|
// 128 byte row allows for 32 avx ARGB pixels.
|
|
#define ANY11S(NAMEANY, ANY_SIMD, UVSHIFT, BPP, MASK) \
|
|
void NAMEANY(const uint8_t* src_ptr, int src_stride, uint8_t* dst_vu, \
|
|
int width) { \
|
|
SIMD_ALIGNED(uint8_t vin[128 * 2]); \
|
|
static_assert(SS(MASK + 1, UVSHIFT) * BPP <= sizeof(vin) / 2, \
|
|
"vin buffer too small"); \
|
|
SIMD_ALIGNED(uint8_t vout[128]); \
|
|
static_assert(SS(MASK + 1, 1) * 2 <= sizeof(vout), \
|
|
"vout buffer too small"); \
|
|
MSAN_MEMSET(vin, 0, sizeof(vin)); /* for msan */ \
|
|
MSAN_MEMSET(vout, 0, sizeof(vout)); /* for msan */ \
|
|
int r = width & MASK; \
|
|
int n = width & ~MASK; \
|
|
if (n > 0) { \
|
|
ANY_SIMD(src_ptr, src_stride, dst_vu, n); \
|
|
} \
|
|
ptrdiff_t np = n; \
|
|
memcpy(vin, src_ptr + (np >> UVSHIFT) * BPP, SS(r, UVSHIFT) * BPP); \
|
|
memcpy(vin + 128, src_ptr + src_stride + (np >> UVSHIFT) * BPP, \
|
|
SS(r, UVSHIFT) * BPP); \
|
|
if ((width & 1) && UVSHIFT == 0) { /* repeat last pixel for subsample */ \
|
|
memcpy(vin + SS(r, UVSHIFT) * BPP, vin + SS(r, UVSHIFT) * BPP - BPP, \
|
|
BPP); \
|
|
memcpy(vin + 128 + SS(r, UVSHIFT) * BPP, \
|
|
vin + 128 + SS(r, UVSHIFT) * BPP - BPP, BPP); \
|
|
} \
|
|
ANY_SIMD(vin, 128, vout, MASK + 1); \
|
|
memcpy(dst_vu + (np >> 1) * 2, vout, SS(r, 1) * 2); \
|
|
}
|
|
|
|
#ifdef HAS_AYUVTOUVROW_AVX2
|
|
ANY11S(AYUVToUVRow_Any_AVX2, AYUVToUVRow_AVX2, 0, 4, 15)
|
|
#endif
|
|
#ifdef HAS_AYUVTOVUROW_AVX2
|
|
ANY11S(AYUVToVURow_Any_AVX2, AYUVToVURow_AVX2, 0, 4, 15)
|
|
#endif
|
|
#ifdef HAS_AYUVTOVUROW_NEON
|
|
ANY11S(AYUVToUVRow_Any_NEON, AYUVToUVRow_NEON, 0, 4, 15)
|
|
ANY11S(AYUVToVURow_Any_NEON, AYUVToVURow_NEON, 0, 4, 15)
|
|
#endif
|
|
#ifdef HAS_AYUVTOUVROW_LSX
|
|
ANY11S(AYUVToUVRow_Any_LSX, AYUVToUVRow_LSX, 0, 4, 15)
|
|
#endif
|
|
#ifdef HAS_AYUVTOUVROW_LASX
|
|
ANY11S(AYUVToUVRow_Any_LASX, AYUVToUVRow_LASX, 0, 4, 15)
|
|
#endif
|
|
#ifdef HAS_AYUVTOVUROW_LSX
|
|
ANY11S(AYUVToVURow_Any_LSX, AYUVToVURow_LSX, 0, 4, 15)
|
|
#endif
|
|
#ifdef HAS_AYUVTOVUROW_LASX
|
|
ANY11S(AYUVToVURow_Any_LASX, AYUVToVURow_LASX, 0, 4, 15)
|
|
#endif
|
|
#undef ANY11S
|
|
|
|
#define ANYDETILE(NAMEANY, ANY_SIMD, T, BPP, MASK) \
|
|
void NAMEANY(const T* src, ptrdiff_t src_tile_stride, T* dst, int width) { \
|
|
SIMD_ALIGNED(T vin[16]); \
|
|
static_assert((MASK + 1) * BPP <= sizeof(vin), "vin buffer too small"); \
|
|
SIMD_ALIGNED(T vout[16]); \
|
|
static_assert((MASK + 1) * BPP <= sizeof(vout), "vout buffer too small"); \
|
|
MSAN_MEMSET(vin, 0, sizeof(vin)); /* for msan */ \
|
|
MSAN_MEMSET(vout, 0, sizeof(vout)); /* for msan */ \
|
|
int r = width & MASK; \
|
|
int n = width & ~MASK; \
|
|
if (n > 0) { \
|
|
ANY_SIMD(src, src_tile_stride, dst, n); \
|
|
} \
|
|
ptrdiff_t np = n; \
|
|
memcpy(vin, src + (np / 16) * src_tile_stride, r * BPP); \
|
|
ANY_SIMD(vin, src_tile_stride, vout, MASK + 1); \
|
|
memcpy(dst + np, vout, r * BPP); \
|
|
}
|
|
|
|
#ifdef HAS_DETILEROW_NEON
|
|
ANYDETILE(DetileRow_Any_NEON, DetileRow_NEON, uint8_t, 1, 15)
|
|
#endif
|
|
#ifdef HAS_DETILEROW_AVX2
|
|
ANYDETILE(DetileRow_Any_AVX2, DetileRow_AVX2, uint8_t, 1, 15)
|
|
#endif
|
|
#ifdef HAS_DETILEROW_16_NEON
|
|
ANYDETILE(DetileRow_16_Any_NEON, DetileRow_16_NEON, uint16_t, 2, 15)
|
|
#endif
|
|
#ifdef HAS_DETILEROW_16_AVX2
|
|
ANYDETILE(DetileRow_16_Any_AVX2, DetileRow_16_AVX2, uint16_t, 2, 15)
|
|
#endif
|
|
#ifdef HAS_DETILEROW_LSX
|
|
ANYDETILE(DetileRow_Any_LSX, DetileRow_LSX, uint8_t, 1, 15)
|
|
#endif
|
|
#ifdef HAS_DETILEROW_LASX
|
|
ANYDETILE(DetileRow_Any_LASX, DetileRow_LASX, uint8_t, 1, 15)
|
|
#endif
|
|
#ifdef HAS_DETILEROW_16_LSX
|
|
ANYDETILE(DetileRow_16_Any_LSX, DetileRow_16_LSX, uint16_t, 2, 15)
|
|
#endif
|
|
#ifdef HAS_DETILEROW_16_LASX
|
|
ANYDETILE(DetileRow_16_Any_LASX, DetileRow_16_LASX, uint16_t, 2, 15)
|
|
#endif
|
|
|
|
// DetileSplitUVRow width is in bytes
|
|
#define ANYDETILESPLITUV(NAMEANY, ANY_SIMD, MASK) \
|
|
void NAMEANY(const uint8_t* src_uv, ptrdiff_t src_tile_stride, \
|
|
uint8_t* dst_u, uint8_t* dst_v, int width) { \
|
|
SIMD_ALIGNED(uint8_t vin[16]); \
|
|
static_assert(MASK + 1 <= sizeof(vin), "vin buffer too small"); \
|
|
SIMD_ALIGNED(uint8_t vout[8 * 2]); \
|
|
static_assert((MASK + 1) / 2 <= sizeof(vout) / 2, \
|
|
"vout buffer too small"); \
|
|
MSAN_MEMSET(vin, 0, sizeof(vin)); /* for msan */ \
|
|
MSAN_MEMSET(vout, 0, sizeof(vout)); /* for msan */ \
|
|
int r = width & MASK; \
|
|
int n = width & ~MASK; \
|
|
if (n > 0) { \
|
|
ANY_SIMD(src_uv, src_tile_stride, dst_u, dst_v, n); \
|
|
} \
|
|
ptrdiff_t np = n; \
|
|
memcpy(vin, src_uv + (np / 16) * src_tile_stride, r); \
|
|
ANY_SIMD(vin, src_tile_stride, vout, vout + 8, r); \
|
|
memcpy(dst_u + np / 2, vout, (r + 1) / 2); \
|
|
memcpy(dst_v + np / 2, vout + 8, (r + 1) / 2); \
|
|
}
|
|
|
|
#ifdef HAS_DETILESPLITUVROW_NEON
|
|
ANYDETILESPLITUV(DetileSplitUVRow_Any_NEON, DetileSplitUVRow_NEON, 15)
|
|
#endif
|
|
#ifdef HAS_DETILESPLITUVROW_AVX2
|
|
ANYDETILESPLITUV(DetileSplitUVRow_Any_AVX2, DetileSplitUVRow_AVX2, 15)
|
|
#endif
|
|
|
|
#define ANYDETILEMERGE(NAMEANY, ANY_SIMD, MASK) \
|
|
void NAMEANY(const uint8_t* src_y, ptrdiff_t src_y_tile_stride, \
|
|
const uint8_t* src_uv, ptrdiff_t src_uv_tile_stride, \
|
|
uint8_t* dst_yuy2, int width) { \
|
|
SIMD_ALIGNED(uint8_t vin[32 * 2]); \
|
|
static_assert(MASK + 1 <= sizeof(vin) / 2, "vin buffer too small"); \
|
|
SIMD_ALIGNED(uint8_t vout[32 * 2]); \
|
|
static_assert(2 * (MASK + 1) <= sizeof(vout), "vout buffer too small"); \
|
|
MSAN_MEMSET(vin, 0, sizeof(vin)); /* for msan */ \
|
|
MSAN_MEMSET(vout, 0, sizeof(vout)); /* for msan */ \
|
|
int r = width & MASK; \
|
|
int n = width & ~MASK; \
|
|
if (n > 0) { \
|
|
ANY_SIMD(src_y, src_y_tile_stride, src_uv, src_uv_tile_stride, dst_yuy2, \
|
|
n); \
|
|
} \
|
|
ptrdiff_t np = n; \
|
|
int r0 = (r < 16) ? r : 16; \
|
|
memcpy(vin, src_y + (np / 16) * src_y_tile_stride, r0); \
|
|
memcpy(vin + 32, src_uv + (np / 16) * src_uv_tile_stride, r0); \
|
|
if (r > 16) { \
|
|
memcpy(vin + 16, src_y + (np / 16 + 1) * src_y_tile_stride, r - 16); \
|
|
memcpy(vin + 48, src_uv + (np / 16 + 1) * src_uv_tile_stride, r - 16); \
|
|
} \
|
|
ANY_SIMD(vin, 16, vin + 32, 16, vout, MASK + 1); \
|
|
memcpy(dst_yuy2 + 2 * np, vout, 2 * r); \
|
|
}
|
|
|
|
#ifdef HAS_DETILETOYUY2_NEON
|
|
ANYDETILEMERGE(DetileToYUY2_Any_NEON, DetileToYUY2_NEON, 15)
|
|
#endif
|
|
|
|
#ifdef HAS_DETILETOYUY2_AVX2
|
|
ANYDETILEMERGE(DetileToYUY2_Any_AVX2, DetileToYUY2_AVX2, 31)
|
|
#endif
|
|
|
|
#undef ANYDETILEMERGE
|
|
|
|
#undef MSAN_MEMSET
|
|
|
|
#ifdef __cplusplus
|
|
} // extern "C"
|
|
} // namespace libyuv
|
|
#endif
|