Fixes unit test failures by deriving nbit inference cache associativity from device warp size#161
Closed
aryaman-gupta wants to merge 1 commit into
Closed
Fixes unit test failures by deriving nbit inference cache associativity from device warp size#161aryaman-gupta wants to merge 1 commit into
aryaman-gupta wants to merge 1 commit into
Conversation
The nbit inference cache hardcoded associativity to 64 on ROCm, but the cache kernels scan one way per warp lane, so on wave32 (RDNA) half the ways were unreachable, corrupting UVM cache stats and codegen lookups. Derive cache_assoc from the device warp size instead, and make DEFAULT_ASSOC an unconditional 32 fallback. Co-Authored-By: Claude <noreply@anthropic.com>
Author
|
Merged into PR (pytorch#5804) branch in pytorch@bccfd22 |
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
The nbit inference cache hardcoded associativity to 64 on ROCm, but the cache kernels scan one way per warp lane, so on wave32 (RDNA) half the ways were unreachable, corrupting UVM cache stats and codegen lookups.
This led to failures in the following unit tests:
FBGEMM_TEST_WITH_ROCM=1 HIP_LAUNCH_BLOCKING=1 FBGEMM_TBE_ROCM_INFERENCE_PACKED_BAGS=1 pytest fbgemm_gpu/test/tbe/inference/nbit_cache_test.py::test_nbit_uvm_cache_statsFBGEMM_TEST_WITH_ROCM=1 HIP_LAUNCH_BLOCKING=1 FBGEMM_TBE_ROCM_INFERENCE_PACKED_BAGS=1 pytest fbgemm_gpu/test/tbe/inference/nbit_split_embeddings_test.py::test_int_nbit_split_embedding_uvm_caching_codegen_lookup_functionThis PR derives
cache_assocfrom the device warp size instead, and makeDEFAULT_ASSOCan unconditional 32 fallback.