whisper.cpp/ggml
Hongqiang Wang cd340ea832 opencl: port fused ssm_scan kernel (Mamba-2, d_state in {128, 256}) to GPU (llama/26439)
* opencl: port fused ssm_scan kernel (Mamba-2, d_state in {128, 256})

Fold the fused per-token SSM_SCAN recurrent step from opencl/gdn-qwen36-35b
onto the unified base. Previously SSM_SCAN fell back to CPU here; now scalar-A
Mamba-2 with d_state in {128,256}, all-f32, runs on GPU. Other shapes (incl.
Mamba-1 element-wise A) still fall back. test-backend-ops -o SSM_SCAN passes on
Adreno X2-90. opt-out via GGML_OPENCL_DISABLE_SSM_SCAN=1.

* opencl: cleanup

* opencl: require K == 1

---------

Co-authored-by: Li He <lih@qti.qualcomm.com>
2026-08-21 19:23:26 +03:00
..
cmake cmake : add config version support (ggml/1582) 2026-08-14 22:16:06 +03:00
include ggml: add ggml_rope_set_offset (+ metal support) (llama/27120) 2026-08-21 19:23:26 +03:00
src opencl: port fused ssm_scan kernel (Mamba-2, d_state in {128, 256}) to GPU (llama/26439) 2026-08-21 19:23:26 +03:00
.gitignore whisper : reorganize source code + improve CMake (#2256) 2024-06-26 19:34:09 +03:00
CMakeLists.txt ggml : bump version to 0.20.2 (ggml/1589) 2026-08-18 15:33:21 +03:00