56,966 blogs · 4,417,066 posts Claim your blog

Record · corpus.blog/posts/01a08727-0bde-705d-b50f-27b8f4545f14

Reimplementing FlashAttention for performance and giggles

aminediro.com · published 4 December 2025

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 aminediro.com Wayback capture

1blogs citing

First cited: 4 Jun 2026
Most recent: 4 Jun 2026
Rank this month: not ranked
Links from this post: 13
Words captured: 12,604

Cited by

Blog

In the post

Date

Making Local LLM Go Brrr Read at the source ↗

4 Jun 2026

Links from this post

Link

Host

https://fleetwood.dev/

fleetwood.dev

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

https://en.wikipedia.org/wiki/Single_instruction,_multiple_threads

en.wikipedia.org

https://modal.com/gpu-glossary/device-software/memory-hierarchy

modal.com

https://github.com/AmineDiro/nano-llama-flash/blob/0d2b59e74574b32d03c25dca43c1c34c2735433e/kernels/triton_flash_att.py

github.com

https://www.youtube.com/watch?v=F_BazucyCMw&t=6910s

youtube.com

https://modal.com/gpu-glossary/perf

modal.com

https://github.com/AmineDiro/nano-llama-flash/blob/main/kernels/triton_flash_att_v2.py

github.com

https://github.com/AmineDiro/nano-llama-flash/blob/main/kernels/triton_flash_att_v2_padded.py

github.com

https://github.com/AmineDiro/nano-llama-flash/blob/main/kernels/triton_flash_att_v2_transpose.py

github.com

https://modal.com/gpu-glossary/perf/roofline-model

modal.com

https://github.com/AmineDiro/nano-llama-flash/blob/main/kernels/triton_flash_att_v2_transpose_fp16.py

github.com

https://github.com/AmineDiro/nano-llama-flash/tree/main/kernels

github.com