corpus.blog
56,966 blogs · [ { "id": "01a08767-697e-716d-9a9d-e28f40ba18dc", "title": "MXFP8 GEMM: Up to 99% of cuBLAS performance using CUDA + PTX", "url": "https://danielvegamyhre.github.io/2026/03/29/mxfp8-gemm.html/2026/03/29/mxfp8-gemm.html", "published_at": "2026-03-29T00:00:00+00:00" }, { "id": "01a08767-697e-716d-9a9d-e28f4104d83d", "title": "Debugging deadlocks in warp-specialized GEMM kernels with CUDA-GDB", "url": "https://danielvegamyhre.github.io/2026/03/29/mxfp8-gemm.html/2026/02/02/cuda-gdb.html", "published_at": "2026-02-02T00:00:00+00:00" }, { "id": "01a08767-697e-716d-9a9d-e28f417f444b", "title": "An illustrated deep-dive into how the compute and comms in TP+SP are overlapped using Async TP", "url": "https://danielvegamyhre.github.io/2026/03/29/mxfp8-gemm.html/ml/performance/2025/05/26/async-tp.html", "published_at": "2025-05-26T19:45:51+00:00" }, { "id": "01a08767-697e-716d-9a9d-e28f41bda543", "title": "An illustrated deep-dive into Megatron-style tensor parallelism", "url": "https://danielvegamyhre.github.io/2026/03/29/mxfp8-gemm.html/ml/performance/2025/03/30/illustrated-megatron.html", "published_at": "2025-03-30T19:45:51+00:00" } ] posts Claim your blog
Back to danielvegamyhre.github.io
Blog · corpus.blog/blogs/danielvegamyhre.github.io/posts

danielvegamyhre.github.io

danielvegamyhre.github.io

2026

MXFP8 GEMM: Up to 99% of cuBLAS performance using CUDA + PTXoriginal ↗
29 Mar 2026
Debugging deadlocks in warp-specialized GEMM kernels with CUDA-GDBoriginal ↗
2 Feb 2026

2025

An illustrated deep-dive into how the compute and comms in TP+SP are overlapped using Async TPoriginal ↗
26 May 2025
An illustrated deep-dive into Megatron-style tensor parallelismoriginal ↗
30 Mar 2025