ggml-cuda: add chunked SSD matmul for Mamba-2 prefill acceleration - #22675

Merged

gaugarg-nv merged 10 commits into

ggml-org:masterfrom

BLSharda:bhaviksharda/mamba2_SSD

Jul 28, 2026

Commits on Jul 21, 2026

ggml-cuda: add chunked SSD matmul for Mamba-2 prefill acceleration
BLSharda
committed
8180cdb
cuda: added SSD CICD fixes for CUDA / HIP / MUSA / MSVC.
BLSharda
committed
e9633ae
ggml-cuda: review comments fixed.
BLSharda
committed
3f93219
ggml-cuda: Fuse M matrix materialization into pre_matmul kernel and enabled test.
BLSharda
committed
21f3745
ggml-cuda: test updates and fixes
BLSharda
committed
b0af8f4
ggml-cuda: test updates to remove hardcoding of tensor initialise data limits.
BLSharda
committed
f6ed777
ggml-cuda: ssd minor review comment fixed.
BLSharda
committed
f35681f
ggml-cuda: ssd minor CICD fixed.
BLSharda
committed
a942d1b
CUDA SSD: Fixes correctness by promoting s0_stride_seq to int64_t, improves memory coalescing in ssm_ssd_prepare_dt_kernel, and boosts efficiency by merging B_weighted and C_scaled; also addresses …
BLSharda
committed
3c6938c

Commits on Jul 28, 2026

cuda: fix sdata read-write race in prepare_dt fallback scan loop
BLSharda
committed
f96b0e5