[DO NOT MERGE] minimum change reproducing the CUDA memory bug in direct to APC #3458
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
4 tasks
qwang98
commented
Nov 27, 2025
Comment on lines
+260
to
+268
| // let mut output = DeviceMatrix::<BabyBear>::with_capacity(height, width); | ||
| use openvm_stark_backend::p3_field::FieldAlgebra; | ||
| let zeros = vec![BabyBear::ZERO; height * width]; | ||
| let device_buffer = zeros | ||
| .to_device() | ||
| .expect("copy zero trace to device failed"); | ||
| println!("output len: {}", device_buffer.len()); | ||
| let mut output = | ||
| DeviceMatrix::<BabyBear>::new(Arc::new(device_buffer), height, width); |
Collaborator
Author
There was a problem hiding this comment.
Zero'ing out the output buffer so that we are sure any illegal memory access isn't due to the cells being empty.
qwang98
commented
Nov 27, 2025
Comment on lines
230
to
232
| if air_name == "VmAirWrapper<Rv32BaseAluAdapterAir, BaseAluCoreAir<4, 8>" { | ||
| return None; | ||
| } |
Collaborator
Author
There was a problem hiding this comment.
Skip dummy trace generation.
qwang98
commented
Nov 27, 2025
Comment on lines
288
to
290
| if *air_name == "VmAirWrapper<Rv32BaseAluAdapterAir, BaseAluCoreAir<4, 8>" { | ||
| return (airs, substitutions) | ||
| } |
Collaborator
Author
There was a problem hiding this comment.
Skip creating Subst, so dummy trace aren't written to APC trace and we have an incorrect APC trace, which should normally panic in prover but in some runs it panics with CUDA memory access at RangeTupleCheckerChipGPU, which is exactly the bug this PR's tries to reproduce.
Collaborator
Author
|
Close as fixed. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
Reproduces the bug as described in this comment: powdr-labs/openvm#50 (comment)
Note that the CUDA memory error "very unrelatedly" happens on
RangeTupleChecker.It also non-deterministically happens in some runs only (might need to run
guest_prove_simplea few times before it's encountered).