perf(codegen): write indent in chunks of 32 bytes #12745
Merged
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
github-actions
bot
added
the
C-performance
Member
Author
How to use the Graphite Merge QueueAdd either label to this PR to merge it via the merge queue:
You must have a Graphite account in order to use the merge queue. Sign up using this link. An organization admin has enabled the Graphite Merge Queue in this repository. Please do not merge from GitHub as this will restart CI on PRs being processed by the merge queue. This stack of pull requests is managed by Graphite. Learn more about stacking. |
CodSpeed Instrumentation Performance ReportMerging #12745 will not alter performanceComparing Summary
|
Member
Merge activity
|
#12691 added the option to write indentation as either tabs or spaces. This changed 2 things about writing indentation: 1. On x86_64, it now takes 3 SIMD instructions to "splat" the indent character across a 16-byte XMM register. Previously it was a 16-byte XMM read from a static. 2. If using spaces as indent, it'll usually be in width of 2 or 4. The indent printer has a fast path for writing less than 16 bytes of indentation. When using a single tab as indent, it's rare you'd write more than 16 bytes, because code is rarely that deeply nested. But if using 4 spaces as indent, depth only needs to get to 5 to exceed 16 bytes, which is not uncommon. This makes hitting the slow path far more likely. Therefore, increase the size of chunks of indentation that's written on the fast path to 32 bytes. This adds only one more XMM write, because it can reuse the "splatted" XMM register, and makes it less likely to hit the slow path. It's not possible to optimize for all cases because tabs and spaces indentation are quite different in this respect, but I think increasing the chunk size to 32 is probably a decent "middle of the road" option. Seems to have no ill effects on our benchmarks, which use tabs for indentation. So this seems not to hurt the tabs case, but should help the 4 x spaces case significantly.
graphite-app
bot
force-pushed
the
08-02-perf_codegen_write_indent_in_chunks_of_32_bytes
branch
from
244676d to
e8ac1a5
Compare
graphite-app
bot
deleted the
08-02-perf_codegen_write_indent_in_chunks_of_32_bytes
branch
This was referenced Aug 6, 2025
taearls
pushed a commit
to taearls/oxc
that referenced
this pull request
Aug 12, 2025
oxc-project#12691 added the option to write indentation as either tabs or spaces. This changed 2 things about writing indentation: 1. On x86_64, it now takes 3 SIMD instructions to "splat" the indent character across a 16-byte XMM register. Previously it was a 16-byte XMM read from a static. 2. If using spaces as indent, it'll usually be in width of 2 or 4. The indent printer has a fast path for writing less than 16 bytes of indentation. When using a single tab as indent, it's rare you'd write more than 16 bytes, because code is rarely that deeply nested. But if using 4 spaces as indent, depth only needs to get to 5 to exceed 16 bytes, which is not uncommon. This makes hitting the slow path far more likely. Therefore, increase the size of chunks of indentation that's written on the fast path to 32 bytes. This adds only one more XMM write, because it can reuse the "splatted" XMM register, and makes it less likely to hit the slow path. It's not possible to optimize for all cases because tabs and spaces indentation are quite different in this respect, but I think increasing the chunk size to 32 is probably a decent "middle of the road" option. Seems to have no ill effects on our benchmarks, which use tabs for indentation. So this seems not to hurt the tabs case, but should help the 4 x spaces case significantly.
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.
#12691 added the option to write indentation as either tabs or spaces.
This changed 2 things about writing indentation:
On x86_64, it now takes 3 SIMD instructions to "splat" the indent character across a 16-byte XMM register. Previously it was a 16-byte XMM read from a static.
If using spaces as indent, it'll usually be in width of 2 or 4. The indent printer has a fast path for writing less than 16 bytes of indentation. When using a single tab as indent, it's rare you'd write more than 16 bytes, because code is rarely that deeply nested. But if using 4 spaces as indent, depth only needs to get to 5 to exceed 16 bytes, which is not uncommon. This makes hitting the slow path far more likely.
Therefore, increase the size of chunks of indentation that's written on the fast path to 32 bytes. This adds only one more XMM write, because it can reuse the "splatted" XMM register, and makes it less likely to hit the slow path.
It's not possible to optimize for all cases because tabs and spaces indentation are quite different in this respect, but I think increasing the chunk size to 32 is probably a decent "middle of the road" option.
Seems to have no ill effects on our benchmarks, which use tabs for indentation. So this seems not to hurt the tabs case, but should help the 4 x spaces case significantly.