Skip to content

Fix: avoid hooking cuMemGetInfo_v2 to prevent memory conflict. #80

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
wants to merge 1 commit into from
Closed

Fix: avoid hooking cuMemGetInfo_v2 to prevent memory conflict. #80

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Fix: avoid hooking cuMemGetInfo_v2 to prevent memory conflict with … 
…total and limit memory.

Signed-off-by: yangshiqi <[email protected]>
  • Loading branch information
@yangshiqi
yangshiqi committed Jun 11, 2025
commit fcfc20ddd7563012e5f7186a8abc8fe3df87df1f
14 changes: 8 additions & 6 deletions src/cuda/memory.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -499,19 +499,21 @@ CUresult cuMemGetInfo_v2(size_t* free, size_t* total) {
size_t limit = get_current_device_memory_limit(dev);
if (limit == 0) {
CUDA_OVERRIDE_CALL(cuda_library_entry,cuMemGetInfo_v2, free, total);
//LOG_INFO("orig free=%ld total=%ld\n",*free,*total);
LOG_INFO("orig free=%ld total=%ld",*free,*total);
*free= *total - usage;
LOG_INFO("after free=%ld total=%ld",*free,*total);
LOG_MSG("orig free=%ld total=%ld",*free,*total);
*free = *total - usage;
LOG_MSG("after free=%ld total=%ld",*free,*total);
return CUDA_SUCCESS;
} else if (limit < usage) {
LOG_WARN("limit < usage; usage=%ld, limit=%ld",usage,limit);
return CUDA_ERROR_INVALID_VALUE;
} else {
CUDA_OVERRIDE_CALL(cuda_library_entry,cuMemGetInfo_v2, free, total);
LOG_MSG("orig free=%ld total=%ld limit=%ld usage=%ld",*free,*total,limit,usage);
*free = limit - usage;
*total = limit;
// Ensure total memory does not exceed the physical or imposed limit.
size_t actual_limit = (limit > *total) ? *total : limit;
*free = (actual_limit > usage) ? (actual_limit - usage) : 0;
*total = actual_limit;
LOG_MSG("after free=%ld total=%ld limit=%ld usage=%ld",*free,*total,limit,usage);
return CUDA_SUCCESS;
}
}
Expand Down