41554 blogs · 4,417,066 posts Claim your blog

Record · corpus.blog/posts/01a08767-697e-716d-9a9d-e28f40ba18dc

MXFP8 GEMM: Up to 99% of cuBLAS performance using CUDA + PTX

danielvegamyhre.github.io · published 29 March 2026

A record is what was published and who pointed at it. The text of the post is not held here: read it at the source, or at its Wayback capture.

Read the post at danielvegamyhre.github.io Wayback capture

0blogs citing

First cited: —
Most recent: —
Rank this month: not ranked
Links from this post: 18
Words captured: 18,827

Cited by

No blog we hold has cited this post yet.

Links from this post

Link

Host

https://github.com/danielvegamyhre/gemm/blob/main/blackwell/mxfp8/3_tcgen05_mxfp8_2cta_256n_overlap128_tma_store/tcgen05_mxfp8_2cta_256n_overlap128_tma_store.cu

github.com

How to Optimize a CUDA Matmul Kernel for cuBLAS-like Performance: a Worklog Read at the source ↗

Outperforming cuBLAS on H100: a Worklog Read at the source ↗

cudaforfun.substack.com

tcgen05 for dummies Read at the source ↗

gau-nernst.github.io

https://github.com/danielvegamyhre/gemm/blob/main/blackwell/mxfp8/

github.com

https://www.opencompute.org/documents/ocp-microscaling-formats-mx-v1-0-spec-final-pdf

opencompute.org

https://docs.nvidia.com/cuda/parallel-thread-execution/

docs.nvidia.com

https://github.com/danielvegamyhre/gemm/tree/main/blackwell/bf16

github.com

https://x.com/gaunernst/status/2008183108999487627?s=20

x.com

Modular: Matrix Multiplication on Blackwell: Part 2 - Using Hardware Features to Optimize Matmul Read at the source ↗

CUTLASS Tutorial: Fast Matrix-Multiplication with WGMMA on NVIDIA® Hopper™ GPUs Read at the source ↗

research.colfax-intl.com

Modular: Matrix Multiplication on Blackwell: Part 2 - Using Hardware Features to Optimize Matmul Read at the source ↗

https://docs.nvidia.com/cuda/cuda-driver-api/group__CUDA__TENSOR__MEMORY.html

docs.nvidia.com

https://github.com/danielvegamyhre/gemm/blob/b62be1f47aade2a16a19b54e77e2be1c491b5df4/hopper/1_wgmma_no_swizzle_tma_pipeline/no_swizzle_pipeline_tma_wgmma.cu

github.com

https://docs.nvidia.com/cuda/parallel-thread-execution/index.html

docs.nvidia.com

https://github.com/danielvegamyhre/gemm/blob/main/blackwell/mxfp8/1_tcgen05_mxfp8_2cta_256n_overlap128/tcgen05_mxfp8_2cta_256n_overlap128.cu

github.com

https://en.wikipedia.org/wiki/Hilbert_curve

en.wikipedia.org

https://github.com/deepseek-ai/DeepGEMM/blob/d30fc36c8f229f4f873b90a492f6e19e6e610923/deep_gemm/include/deep_gemm/impls/sm100_fp8_gemm_1d1d.cuh

github.com