Use alloca to address memory layout effects #744
Conversation
This file contains 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
the iter collect pollutes the call stack (10 levels off calls)
samueltardieu
suggested changes
Mar 29, 2024
| @@ -277,6 +293,8 @@ where | |||
| } | |||
|
|
|||
| b.iters = b.iters.wrapping_mul(2); | |||
| b.iters = b.iters.min(64); // To make sure we offset the test at least with 0-64 bytes | |||
| // wit alloca | |||
There was a problem hiding this comment.
Nit: typo ("wit" -> "with"). Also, the comment is typically placed above the line, not on the side.
Comment on lines
+252
to
+253
| stack_alloc, /* how much bytes we want to allocate */ | ||
| |_memory: &mut [core::mem::MaybeUninit<u8>] /* dynamically stack allocated slice itself */| { |
There was a problem hiding this comment.
Maybe stack_alloc should be inlined here, to prevent getting a warning on platforms not selected by the cfg attribute. Using a name such as _shifting_stack_space or similar instead of _memory would be self-documenting.
|
Seems quite straightforward and if this helps in reducing random changes between runs I'm all for this. |
|
@PSeitz Thanks for your contribution! Are you interested in updating this PR for the review comments or should we take it on? |
Open
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Outdated suggestions cannot be applied.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.
the iter collect pollutes the call stack (10 levels of calls)
As mentioned in Eliminate memory layout bias when measuring (with LLVM stabilizer) #334, randomness in memory layout has a huge impact on the bench results. When I benchmark with criterion it's not uncommon to see effects of +-30%. Tests seem much more stable with alloca, there's still random 6% jumps in one test I've been running.
Adresses #334