skidrow

8 ms·

91.
Beating cuBLAS in Single-Precision General Matrix Multiplication(salykova.github.io)
3 points by skidrow 2y ago|0 comments
92.
Show HN: Beating cuBLAS in Single-Precision General Matrix Multiplication(salykova.github.io)
2 points by skidrow 2y ago|0 comments
93.
Beating OpenBLAS in FP32 Matrix Multiplication(salykova.github.io)
2 points by skidrow 2y ago|0 comments
94.
Beating cuBLAS in Single-Precision General Matrix Multiplication(salykova.github.io)
98 points by skidrow 2y ago|8 comments
95.
CUDA Matrix Multiplication Optimization(leimao.github.io)
2 points by skidrow 2y ago|0 comments
96.
Tutorials on Tinygrad(mesozoic-egg.github.io)
2 points by skidrow 2y ago|0 comments
97.
Matrix Multiplication on GPU: Faster Than Nvidia, Sometimes(indii.org)
4 points by skidrow 2y ago|0 comments
98.
CUDA Matrix Multiplication Optimization(leimao.github.io)
4 points by skidrow 2y ago|0 comments
99.
What Shapes Do Matrix Multiplications Like?(thonking.ai)
164 points by skidrow 2y ago|11 comments
100.
What shapes do matrix multiplications like?(thonking.ai)
2 points by skidrow 2y ago|0 comments
101.
by skidrow 2y ago
SIMD intrinsics and manually unrolled loops are surely needed. That's the reason why all BLAS libraries vectorize and unroll loops manually. Even modern compilers can't properly auto-vectorize and unroll with 100% success rate.
102.
by skidrow 2y ago
are OpenBLAS and MKL not well optimized lol? They literally compared against OpenBLAS/MKL and posted the results in the article. As someone already mentioned, this implementation is faster than MKL even on a Intel Xeon with 96 cores. M
103.
by skidrow 2y ago
Look, it's indeed a resonable comparison. They use matrix sizes up to M=N=K=5000, so the ffi overhead is neglectable. What's the point of compairing NumPy with BLAS if NumPy does use BLAS under the hood?
104.
by skidrow 2y ago
Their implementation outperforms not only the recent version of OpenBLAS but also MKL on their machine (these are DEFAULT BLAS libraries shipped with numpy). What's the point of compairing against BLIS if numpy doesn't use it by d