Skip to content
This repository has been archived by the owner on Mar 21, 2024. It is now read-only.

Fix 32-bit MSVC builds. #1582

Merged
merged 1 commit into from
Dec 17, 2021

Conversation

alliepiper
Copy link
Collaborator

CUDA runtime functions are annotated with CUDARTAPI, which expands to
__stdcall on MSVC.

thrust/system/cuda/memory_resource.h defines some
function pointer types intended for use with CUDART memory
allocation/deallocation functions, but they lack the CUDARTAPI markup.

Added this markup to fix 32-bit builds. Repro'd issue and validated
fix on MSVC 2019.

Fixes #1458.

CUDA runtime functions are annotated with CUDARTAPI, which expands to
__stdcall on MSVC.

`thrust/system/cuda/memory_resource.h` defines some
function pointer types intended for use with CUDART memory
allocation/deallocation functions, but they lack the CUDARTAPI markup.

Added this markup to fix 32-bit builds. Repro'd issue and validated
fix on MSVC 2019.

Fixes NVIDIA#1458.
@alliepiper
Copy link
Collaborator Author

DVS CL: 30787474

run tests

@alliepiper alliepiper added testing: gpuCI in progress Started gpuCI testing. testing: internal ci in progress Currently testing on internal NVIDIA CI (DVS). labels Dec 16, 2021
@alliepiper alliepiper merged commit c72c701 into NVIDIA:main Dec 17, 2021
@alliepiper alliepiper deleted the gh.1458/msvc_x86_CUDARTAPI_fixes branch December 17, 2021 17:36
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
P1: should have Necessary, but not critical. testing: gpuCI in progress Started gpuCI testing. testing: internal ci in progress Currently testing on internal NVIDIA CI (DVS).
Projects
None yet
Development

Successfully merging this pull request may close these issues.

thrust::system::cuda::detail::deallocation_fn type mismatch with cudaFree
2 participants