| /dports/misc/mnn/MNN-1.2.0/source/backend/cpu/compute/ |
| H A D | WinogradInt8Helper.cpp | 53 VecType::load(srcStart + 0 * srcZStep), in _sourceTransUnit4x4Pack4x4()
54 VecType::load(srcStart + 1 * srcZStep), in _sourceTransUnit4x4Pack4x4()
55 VecType::load(srcStart + 2 * srcZStep), in _sourceTransUnit4x4Pack4x4()
56 VecType::load(srcStart + 3 * srcZStep) in _sourceTransUnit4x4Pack4x4()
74 srcStart += srcZStep * 4; in _sourceTransUnit4x4Pack4x4()
91 srcStart += srcZStep; in _sourceTransUnit4x4Pack4x4()
112 srcStart += srcZStep * 4; in _sourceTransUnit4x4Pack4x16()
161 srcStart += srcZStep * 4; in _sourceTransUnit4x4Pack16x4()
194 srcStart += srcZStep; in _sourceTransUnit4x4Pack16x16()
227 srcStart += srcZStep; in _destTransformUnit4x2()
[all …]
|
| H A D | Int8FunctionsOpt.cpp | 152 const int srcZStep = im2colParameter->iw * im2colParameter->ih * GEMM_INT8_UNIT; in _fastIm2Col() local
158 auto inputZ0 = inputK + srcZStep * (2 * sz + 0); in _fastIm2Col()
159 auto inputZ1 = inputK + srcZStep * (2 * sz + 1); in _fastIm2Col()
230 auto srcZStep = im2colParameter->srcZStep; in _im2colCommon() local
262 inputK += srcZStep; in _im2colCommon()
274 …fastIm2Col &= (im2colParam->srcYStep == iw * GEMM_INT8_UNIT && im2colParam->srcZStep == ih * iw * … in chooseIm2Col()
318 auto srcChannleStride = im2colParameter->srcZStep; in _im2colCommonSdot()
359 const int srcZStep = im2colParameter->iw * im2colParameter->ih * GEMM_INT8_UNIT; in _fastIm2ColSdot() local
365 auto inputZ0 = inputK + srcZStep * sz; in _fastIm2ColSdot()
377 …fastIm2Col &= (im2colParam->srcYStep == iw * GEMM_INT8_UNIT && im2colParam->srcZStep == ih * iw * … in chooseIm2ColSdot()
|
| H A D | WinogradInt8Helper.hpp | 26 …typedef void(*SrcTransFunc)(const int8_t* srcStart, int8_t* dstStart, size_t srcZStep, size_t dstX…
27 …nsFunc)(const float* srcStart, float* dstStart, size_t srcXStep, size_t srcZStep, size_t dstXStep,…
|
| H A D | ConvolutionWinograd.cpp | 284 int srcZStep = xC * pack; in onExecute() local
306 auto srcZ = srcXi + z * srcZStep * bytes; in onExecute()
322 auto srcZ = srcXi + z * srcZStep * bytes; in onExecute()
|
| H A D | ConvInt8Winograd.cpp | 470 int srcZStep = xC * UNIT; in onExecute() local
500 … alphaX * unitStep, srcZStep, alphaX * UNIT, alphaX * mUnitY * UNIT, dc_4); in onExecute()
508 … mDestTransformY(srcXi, _dstStart, unitStep, srcZStep, dstYStep, dstZStep, dc_4); in onExecute()
512 unitStep, srcZStep, UNIT, dstZStep, dc_4); in onExecute()
|
| H A D | ConvInt8TiledExecutor.cpp | 185 mIm2ColParamter.srcZStep = input->stride(1) * UNIT; in onResize()
|
| H A D | ConvolutionInt8Executor.cpp | 145 mIm2ColParamter.srcZStep = inputs[0]->stride(1) * UNIT; in onResize()
|
| H A D | CommonOptFunction.h | 60 size_t depthQuad, size_t dstZStep, size_t srcZStep);
|
| H A D | CommonOptFunction.cpp | 1431 size_t depthQuad, size_t dstZStep, size_t srcZStep) { in MNNUInt8ToInt16WithOffsetC4Fast() argument
1433 srcZStep /= sizeof(uint8_t); in MNNUInt8ToInt16WithOffsetC4Fast()
1436 auto srcZ = srcStart + sz * srcZStep; in MNNUInt8ToInt16WithOffsetC4Fast()
|
| /dports/misc/mnn/MNN-1.2.0/source/backend/cpu/ |
| H A D | CPUTFQuantizedConv2D.cpp | 44 …Sum(int32_t* inputSum, int8_t* colAddr, const uint8_t* inputOrigin, size_t srcZStep, size_t icDiv8,
80 …Sum(int32_t* inputSum, int8_t* colAddr, const uint8_t* inputOrigin, size_t srcZStep, size_t icDiv8, in MNNLoadU8AndSum() argument
87 auto inputZ0 = inputK + srcZStep * (SRC_C4_UNIT * sz + 0); in MNNLoadU8AndSum()
88 auto inputZ1 = inputK + srcZStep * (SRC_C4_UNIT * sz + 1); in MNNLoadU8AndSum()
89 auto inputZ2 = inputK + srcZStep * (SRC_C4_UNIT * sz + 2); in MNNLoadU8AndSum()
90 auto inputZ3 = inputK + srcZStep * (SRC_C4_UNIT * sz + 3); in MNNLoadU8AndSum()
340 auto srcZStep = iw * ih * UNIT; in _im2ColCommon() local
365 auto inputZ = inputK + srcZStep * sz; in _im2ColCommon()
|
| /dports/misc/mnn/MNN-1.2.0/source/core/ |
| H A D | ConvolutionCommon.hpp | 51 int32_t srcZStep; member
|
| /dports/misc/mnn/MNN-1.2.0/backupcode/cpubackend/ |
| H A D | ConvolutionWinograd3D.cpp | 197 int srcZStep = xC * 4; in onExecute() local
213 auto srcZ = srcXi + z * srcZStep; in onExecute()
230 auto srcZ = srcXi + z * srcZStep; in onExecute()
|