onnxruntime
[MLAS] Simplify & optimize Arm64 NCHWc Convolution kernels
#26691

Merged

Commits

Simplify the Fp32 Pointwise Kernel to use GEMM

Rohanjames1997 committed 221 days ago
Simplify the Fp32 Depthwise kernel

Rohanjames1997 committed 221 days ago
Simplify the Fp32 Depthwise Conv kernel

Rohanjames1997 committed 221 days ago
Make fp32 Depthwise Conv branchless

Rohanjames1997 committed 221 days ago
Make fp32 Depthwise Conv branchless

Rohanjames1997 committed 221 days ago
Make MlasConvFloatKernelNeonImpl branchless

Rohanjames1997 committed 221 days ago
Remove redundant code

Rohanjames1997 committed 221 days ago
Merge two loops

Rohanjames1997 committed 221 days ago
Make ReLU branchless in Fp32 Pointwise Conv

Rohanjames1997 committed 220 days ago
Merge branch 'microsoft:main' into simplify-fp32-conv

Rohanjames1997 committed 213 days ago
Fix the GEMM implementation and expand the test coverage.

Rohanjames1997 committed 211 days ago
Fix segfault in ConvNoBiasAddFusion

Rohanjames1997 committed 211 days ago
Eliminate potential segfault

Rohanjames1997 committed 211 days ago
Copilot's suggestion for boundary checks

Rohanjames1997 committed 211 days ago