[UR][CUDA][HIP] Add UR_{CUDA,HIP}_STACK_SIZE env var for per-thread s… - #22763
Open
zjin-lcf wants to merge 1 commit into
Open
[UR][CUDA][HIP] Add UR_{CUDA,HIP}_STACK_SIZE env var for per-thread s…#22763zjin-lcf wants to merge 1 commit into
zjin-lcf wants to merge 1 commit into
Conversation
…tack size Add support for configuring the per-thread device stack size limit at device initialization via the UR_CUDA_STACK_SIZE and UR_HIP_STACK_SIZE environment variables. These map to cuCtxSetLimit(CU_LIMIT_STACK_SIZE) and hipDeviceSetLimit(hipLimitStackSize) respectively (the driver-level equivalent of cudaDeviceSetLimit(cudaLimitStackSize)), which is useful for kernels with deep recursion or large per-thread private data that would otherwise overflow the small default stack. An invalid value (non-positive or non-numeric) is rejected with a diagnostic at device initialization. Documentation and an E2E test covering the invalid/valid cases are included. Addresses intel#12271 Co-authored-by: Cursor <cursoragent@cursor.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
…tack size
Add support for configuring the per-thread device stack size limit at device initialization via the UR_CUDA_STACK_SIZE and UR_HIP_STACK_SIZE environment variables. These map to cuCtxSetLimit(CU_LIMIT_STACK_SIZE) and hipDeviceSetLimit(hipLimitStackSize) respectively (the driver-level equivalent of cudaDeviceSetLimit(cudaLimitStackSize)), which is useful for kernels with deep recursion or large per-thread private data that would otherwise overflow the small default stack.
An invalid value (non-positive or non-numeric) is rejected with a diagnostic at device initialization. Documentation and an E2E test covering the invalid/valid cases are included.
Addresses #12271