[inductor] remove RBLOCK from persistent reduction kernel's parameter list #98653
Conversation
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
… list [ghstack-poisoned]
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/98653
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit 2a03229: This comment was automatically generated by Dr. CI and updates every 15 minutes. |
jansel
approved these changes
Apr 8, 2023
…s parameter list" This PR resolves comments #97203 (comment) . Send a separate PR since it's easier to test and make sure there is no perf impact. Tests: 1. python test/inductor/test_torchinductor.py 2. run `python benchmarks/dynamo/torchbench.py --backend inductor --amp --performance --dashboard --only hf_Bert --disable-cudagraphs --training` before and after the change to make sure the perf change is neutral. Now a persistent reduction kernel in hf_Bert looks like: ``` persistent_reduction( size_hints=[4096, 1024], reduction_hint=ReductionHint.INNER, filename=__file__, meta={'signature': {0: '*fp32', 1: '*i64', 2: '*fp16', 3: '*i64', 4: '*fp16', 5: '*i64', 6: '*fp16', 7: '*fp16', 8: '*fp16', 9: '*fp16', 10: 'i32', 11: 'i32'}, 'device': 0, 'constants': {}, 'mutated_arg_names': ['in_out_ptr0'], 'configs': [instance_descriptor(divisible_by_16=(0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11), equal_to_1=())]} ) triton.jit def triton_(in_out_ptr0, in_ptr0, in_ptr1, in_ptr2, in_ptr3, in_ptr4, in_ptr5, in_ptr6, in_ptr7, out_ptr2, xnumel, rnumel, XBLOCK : tl.constexpr): xnumel = 4096 rnumel = 768 RBLOCK: tl.constexpr = 1024 xoffset = tl.program_id(0) * XBLOCK xindex = xoffset + tl.arange(0, XBLOCK)[:, None] ``` [ghstack-poisoned]
|
@pytorchbot merge |
Merge startedYour change will be merged once all checks pass (ETA 0-4 Hours). Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
skotapati
pushed a commit
to kulinseth/pytorch
that referenced
this pull request
Apr 10, 2023
… list (pytorch#98653) This PR resolves comments pytorch#97203 (comment) . Send a separate PR since it's easier to test and make sure there is no perf impact. Tests: 1. python test/inductor/test_torchinductor.py 2. run `python benchmarks/dynamo/torchbench.py --backend inductor --amp --performance --dashboard --only hf_Bert --disable-cudagraphs --training` before and after the change to make sure the perf change is neutral. Now a persistent reduction kernel in hf_Bert looks like: ``` @persistent_reduction( size_hints=[4096, 1024], reduction_hint=ReductionHint.INNER, filename=__file__, meta={'signature': {0: '*fp32', 1: '*i64', 2: '*fp16', 3: '*i64', 4: '*fp16', 5: '*i64', 6: '*fp16', 7: '*fp16', 8: '*fp16', 9: '*fp16', 10: 'i32', 11: 'i32'}, 'device': 0, 'constants': {}, 'mutated_arg_names': ['in_out_ptr0'], 'configs': [instance_descriptor(divisible_by_16=(0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11), equal_to_1=())]} ) @triton.jit def triton_(in_out_ptr0, in_ptr0, in_ptr1, in_ptr2, in_ptr3, in_ptr4, in_ptr5, in_ptr6, in_ptr7, out_ptr2, xnumel, rnumel, XBLOCK : tl.constexpr): xnumel = 4096 rnumel = 768 RBLOCK: tl.constexpr = 1024 xoffset = tl.program_id(0) * XBLOCK xindex = xoffset + tl.arange(0, XBLOCK)[:, None] ``` Pull Request resolved: pytorch#98653 Approved by: https://github.com/jansel
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
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.
Stack from ghstack (oldest at bottom):
This PR resolves comments #97203 (comment) . Send a separate PR since it's easier to test and make sure there is no perf impact.
Tests:
python benchmarks/dynamo/torchbench.py --backend inductor --amp --performance --dashboard --only hf_Bert --disable-cudagraphs --trainingbefore and after the change to make sure the perf change is neutral.Now a persistent reduction kernel in hf_Bert looks like:
cc @soumith @voznesenskym @penguinwu @anijain2305 @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @Xia-Weiwen @wenzhe-nrv @jiayisunx @peterbell10 @desertfire