Search results

Speed regression in rmsnorm_fwd when used with torch compile

Status: Open.

#77 In Dao-AILab/quack;

· Maharshi-Pandya opened on Mar 3, 2026

B200 Cross Entropy Backward Performance

Status: Open.

#68 In Dao-AILab/quack;

· Jokeren opened on Jan 26, 2026

incorrect mbarrier.arrive used for sm100 gather_A gemm

Status: Open.

#63 In Dao-AILab/quack;

· Yang-YiFan opened on Jan 6, 2026

Skip SMEM for small $N$ in softmax kernel

Status: Open.

#62 In Dao-AILab/quack;

· GiftedNovaHD opened on Jan 4, 2026

could you offer an example of fp8 gemm with scale for SM90

Status: Open.

#58 In Dao-AILab/quack;

· kousakawang opened on Dec 4, 2025

Fused Add + RMSNorm pattern

Status: Open.

#56 In Dao-AILab/quack;

· AndreSlavescu opened on Nov 9, 2025

Question: _set_cluster_n() in RMSNorm fwd

Status: Open.

#51 In Dao-AILab/quack;

· karthik-man opened on Oct 16, 2025

Is FP8 tensor supported in GEMM?

Status: Open.

#45 In Dao-AILab/quack;

· akaitsuki-ii opened on Sep 28, 2025

Is it possible for A100/A800?

Status: Open.

#41 In Dao-AILab/quack;

· airlsyn opened on Sep 12, 2025

Support CPU mCuSeqlensM in Grouped Gemm?

Status: Open.

#34 In Dao-AILab/quack;

· zfan2356 opened on Sep 4, 2025

RMSNorm how to use residual param?

Status: Open.

#26 In Dao-AILab/quack;

· lengrongfu opened on Jul 24, 2025

Does quack support Ada lovelace?

Status: Open.

#21 In Dao-AILab/quack;

· cailun01 opened on Jul 17, 2025