Only call triton in worker process; run async_compile.triton ahead of time in Scheduler.codegen #146342
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
… time in Scheduler.codegen ### Big idea This PR extends #144288 by combining calling triton in worker processes with the future cache: we kick off triton compilation in the worker processes earlier, during inductor codegen. Basically instead of calling async_compile.triton for the first time only after the entire code has been generated, we start compiling as soon as we know we'll need to compile the kernel. Then, when loading the generated inductor code, we can simply read from our in memory future cache, considerably increasing the parallelism. ### Implementation Overview In total, the diff does the following: - Converts TritonFuture to LambdaFuture, only calling triton.compile on worker processes - Now that triton.compile() isn't called on the main process, we call TritonBundler on all compiled kernels when we get them back from workers - Extend @eellison's future cache to a class, mostly as a refactor - Finally, call async_compile.triton ahead of time in Scheduler.codegen if workers are warmed up. This causes the subsequent async_compile.triton call that occurs after codegen to cache hit on cold start. In the diffs after this, I will add more to CompiledTritonKernels so that TritonBundler, on a warm start, automatically populates the in memory cache on warm start with the existing triton kernels, avoiding calling triton altogether on warm starts. Because LambdaFutures are much faster to kick off than TritonFutures, due to not needing to load from TritonCodeCache at all, the time spent kicking off these worker jobs is pretty minimal for inductor codegen. ### Can we split the diff for easier review? It's best if this diff lands atomically with all of these changes, as doing the ahead of time codegen compile is only performant if we replace TritonFuture with LambdaFuture(as we don't need to load the triton kernel on the main process). However, I've made a diff stack for easier reviewing here: D69070048 - Run async_compile.triton ahead of time in Scheduler.codegen D68633454 - Only call triton in worker process Differential Revision: [D69013710](https://our.internmc.facebook.com/intern/diff/D69013710/) [ghstack-poisoned]
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/146342
Note: Links to docs will display an error until the docs builds have been completed. This comment was automatically generated by Dr. CI and updates every 15 minutes. |
This PR needs a
|
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.
Stack from ghstack (oldest at bottom):
Big idea
This PR extends #144288 by combining calling triton in worker processes with the future cache: we kick off triton compilation in the worker processes earlier, during inductor codegen. Basically instead of calling async_compile.triton for the first time only after the entire code has been generated, we start compiling as soon as we know we'll need to compile the kernel. Then, when loading the generated inductor code, we can simply read from our in memory future cache, considerably increasing the parallelism.
Implementation Overview
In total, the diff does the following:
In the diffs after this, I will add more to CompiledTritonKernels so that TritonBundler, on a warm start, automatically populates the in memory cache on warm start with the existing triton kernels, avoiding calling triton altogether on warm starts.
Because LambdaFutures are much faster to kick off than TritonFutures, due to not needing to load from TritonCodeCache at all, the time spent kicking off these worker jobs is pretty minimal for inductor codegen.
Can we split the diff for easier review?
It's best if this diff lands atomically with all of these changes, as doing the ahead of time codegen compile is only performant if we replace TritonFuture with LambdaFuture(as we don't need to load the triton kernel on the main process). However, I've made a diff stack for easier reviewing here:
D69070048 - Run async_compile.triton ahead of time in Scheduler.codegen
D68633454 - Only call triton in worker process
Differential Revision: D69013710
cc @voznesenskym @penguinwu @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @ipiszy @yf225 @chenyang78 @kadeng @muchulee8 @amjames @desertfire @chauhang @aakhundov