Files
foxhunt/crates
jgrusewski 1f36d2fd24 perf: q_denoise_backward cuBLAS pipeline + attn/IQL 2-phase bias grad
q_denoise_backward (63.5% GPU time, 169ms/call): decomposed into
cuBLAS forward replay + backward GEMMs. 8 cuBLAS GEMMs + elementwise
kernels replace 1800-thread serial loop. Target: <1ms/call.

attn_bias_grad_reduce + iql_bias_grad_reduce: converted from serial
batch loops to 2-phase shared-memory reduction (same pattern as IQN).

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-04-19 13:50:32 +02:00
..