Morph MORPH
SIGN IN SIGN UP
dhruvhead / executorch UNCLAIMED
0 0 1 Python

COMMITS

/ kernels/optimized/blas/BlasKernel.cpp
test
May 16, 2025
P
Mostly sync BlasKernel.cpp with ATen ReducedPrecisionGemvFastPathKernel (#10941)
pytorchbot committed
7719d31
September 21, 2024
S
Don't require -march compiler flags to use bfdot (#5444)
Scott Wolchok committed
0eee42a
September 16, 2024
S
Reapply D62466496: Build optimized kernels with bf16 support and gate usage at runtime (#5376)
Scott Wolchok committed
2b3cc27
September 14, 2024
D
Revert D62466496: Multisect successfully blamed "D62466496: [ExecuTorch] Build optimized kernels with bf16 support and gate usage at runtime" for one build failure
Dark Knight committed
e0c9312
S
Build optimized kernels with bf16 support and gate usage at runtime (#5251)
Scott Wolchok committed
248ba52
S
Use bfdot if compiled with ARM_FEATURE_BF16 (#5249)
Scott Wolchok committed
69ae1e1
September 12, 2024
S
Make ForcedUnroll usage in bf16 BlasKernel actually work for -Oz builds (#5247)
Scott Wolchok committed
88508c5
S
port bf16 dot product kernel from ATen CPUBlas (#5245)
Scott Wolchok committed
5e7efe6