cuda: enable the BF16 WMMA matmul path only for qwen4exp - #75

Merged

pwilkin merged 6 commits into

perf/mmb-qwen4exp-only

Sep 26, 2026

Commits on Sep 17, 2026

cuda: enable the BF16 WMMA matmul path only for qwen4exp

dzannotti
and
claude
committed
42f7e89

Commits on Sep 25, 2026

Merge branch 'master' into perf/mmb-qwen4exp-only
pwilkin
committed
95ae7a2
cuda : gate mmb per backend context instead of per process
pwilkin
committed
bc866d5

Commits on Sep 26, 2026

tests : add --mmb on|off to test-backend-ops
pwilkin
committed
4208491
Merge halo-box master into perf/mmb-qwen4exp-only
pwilkin
committed
c3ee03a
cuda : zero quantized padding without a synchronous memset
pwilkin
committed
cb7dfec