Originally published at: GitHub - NVIDIA/cutlass: CUDA Templates for Linear Algebra Subroutines
Provides support for the NVIDIA Blackwell SM100 architecture. CUTLASS is a collection of CUDA C++ templates and abstractions for implementing high-performance GEMM computations.
jwitsoe
1
Related topics
| Topic | Replies | Views | Activity | |
|---|---|---|---|---|
| CUTLASS: Fast Linear Algebra in CUDA C++ | 13 | 2244 | September 9, 2024 | |
| CUTLASS: Principled Abstractions for Handling Multidimensional Data Through Tensors and Spatial Microkernels | 1 | 62 | June 6, 2026 | |
| cuBLAS convolution does not use Tensor Cores | 6 | 2423 | June 8, 2021 | |
| Where is cute's gemm code? | 20 | 2873 | October 13, 2024 | |
| Cutlass cute-dsl error | 1 | 106 | July 31, 2025 | |
| Implementing High Performance Matrix Multiplication Using CUTLASS v2.8 | 0 | 563 | November 23, 2021 | |
| Just Released: CUTLASS 3.8 | 0 | 147 | January 31, 2025 | |
| How to benchmark on Thor to get the real FP4/FP8 performance TFOPS | 10 | 535 | March 16, 2026 | |
| Is it correct that my Pascal card is calling Maxwell_gemm kernels through cublas? And if so, why is cublas unusably slow for me? | 6 | 1070 | August 23, 2018 | |
| my speedy SGEMM | 91 | 277223 | May 29, 2013 |