Search results

support varlen api for triton / cuda backend

Status: Open.

#225 In xlite-dev/ffpa-attn;

· DefTruth opened on May 24, 2026

remove BUILD_FFPA_ATTN_MMA_L20 like build flags

Status: Open.

#223 In xlite-dev/ffpa-attn;

· DefTruth opened on May 24, 2026

varlen API support for sm89, sm80

Status: Open.

#221 In xlite-dev/ffpa-attn;

· DefTruth opened on May 22, 2026

support headdim up to 1024 for cutedsl backend

Status: Open.

#219 In xlite-dev/ffpa-attn;

· DefTruth opened on May 22, 2026

Persistent tune max is too slow!

Status: Open.

#218 In xlite-dev/ffpa-attn;

· DefTruth opened on May 22, 2026

CuteDSL backend not support blackwell, arch error!

Status: Open.

#194 In xlite-dev/ffpa-attn;

· DefTruth opened on May 18, 2026

E2E train/inference examples for FFPA

Status: Open.

#189 In xlite-dev/ffpa-attn;

· DefTruth opened on May 17, 2026

[RFC] ffpa-attn 0.2.0 roadmap

enhancementNew feature or requestNew feature or request

Status: Open.

#85 In xlite-dev/ffpa-attn;

· DefTruth opened on Aug 5, 2025