]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commit
Avoid PDL race conditions by disabling __restrict__ when PDL is used (#24030)
authorAndreas Kieslinger <redacted>
Wed, 3 Jun 2026 11:56:42 +0000 (13:56 +0200)
committerGitHub <redacted>
Wed, 3 Jun 2026 11:56:42 +0000 (13:56 +0200)
commit9e58d4d692ed3d350591cc86d06c73c61c122509
tree0cf5e574727793b7453ab0ebb783401697b161ec
parent3571fa5435ac9ff243662b1caabc407e8d433c9d
Avoid PDL race conditions by disabling __restrict__ when PDL is used (#24030)

* Removes __restrict__ from PDL kernel headers due to incompatibility with
PDL. Adds preprocessor directives based on arch in kernel body to add
__restrict__ to retain performance on older architectures.

* Simplifies new __restrict__ usage via macro

* Add hopper to PDL __restrict__ fix.

Co-authored-by: Oliver Simons <redacted>
---------

Co-authored-by: Oliver Simons <redacted>
14 files changed:
ggml/src/ggml-cuda/common.cuh
ggml/src/ggml-cuda/fattn-common.cuh
ggml/src/ggml-cuda/fattn-mma-f16.cuh
ggml/src/ggml-cuda/fattn-tile.cuh
ggml/src/ggml-cuda/fattn-vec.cuh
ggml/src/ggml-cuda/fattn-wmma-f16.cu
ggml/src/ggml-cuda/getrows.cu
ggml/src/ggml-cuda/mmvf.cu
ggml/src/ggml-cuda/mmvq.cu
ggml/src/ggml-cuda/quantize.cu
ggml/src/ggml-cuda/reduce_rows.cuh
ggml/src/ggml-cuda/set-rows.cu
ggml/src/ggml-cuda/ssm-conv.cu
ggml/src/ggml-cuda/ssm-scan.cu