]> git.djapps.eu Git - pkg/ggml/sources/ggml/commit
CUDA: use shared mem for ssm_conv (llama/20128)
authorAman Gupta <redacted>
Fri, 6 Mar 2026 15:09:59 +0000 (23:09 +0800)
committerGeorgi Gerganov <redacted>
Sun, 15 Mar 2026 19:50:13 +0000 (21:50 +0200)
commit405b24b00ebe670c35b1d255ffcb9befe06e726c
treec25f3efbf125adc1c0c1cc4ec30979a7d366c26e
parent084b8c6fc9e1649248ab339ef719d66514aafc00
CUDA: use shared mem for ssm_conv (llama/20128)

* CUDA: use shared mem for ssm_conv

* fuse silu + ssm_conv

* fuse unary + mul

* enable for fp16

* formatting

Co-authored-by: Johannes Gäßler <redacted>
---------

Co-authored-by: Johannes Gäßler <redacted>
src/ggml-cuda/ggml-cuda.cu
src/ggml-cuda/ssm-conv.cu
src/ggml-cuda/ssm-conv.cuh
src/ggml-cuda/unary.cu
src/ggml-cuda/unary.cuh
tests/test-backend-ops.cpp