ffi: shrink trampoline placement probe window - #64969
Merged
trivikr merged 1 commit intoAug 9, 2026
Merged
Conversation
Copy link
Copy Markdown
Member
Sorry, something went wrong.
Copy link
Copy Markdown
Collaborator
|
Review requested:
|
Sorry, something went wrong.
Copy link
Copy Markdown
Codecov Report✅ All modified and coverable lines are covered by tests. @@ Coverage Diff @@
## main #64969 +/- ##
=======================================
Coverage 90.33% 90.34%
=======================================
Files 760 760
Lines 248496 248496
Branches 46877 46901 +24
=======================================
+ Hits 224481 224492 +11
+ Misses 15454 15448 -6
+ Partials 8561 8556 -5 see 29 files with indirect coverage changes 🚀 New features to boost your workflow:
|
Sorry, something went wrong.
ShogunPanda
approved these changes
Aug 3, 2026
Copy link
Copy Markdown
Collaborator
Sorry, something went wrong.
AllocateCodeNear() searched 1024 pages in each direction for a free page next to the native target, one MAP_FIXED_NOREPLACE mmap per candidate. Every probe is expected to fail, so an exhausted window cost up to 2048 failing syscalls before taking the plain-mmap fallback that was already there, all to enable a jmp rel32 instead of movabs+jmp. Resolving a fast-eligible signature plateaued at ~790us, against ~28us for one that allocates no trampoline. Reduce the window to 16 pages per direction, bounding the exhausted case to 32 probes while still covering typical per-library symbol counts. Placement stays opportunistic: the caller already emits the absolute form when EmitJmpRel32() reports the target is out of range. Signed-off-by: Trivikram Kamat <16024985+trivikr@users.noreply.github.com> Assisted-by: claude:opus-5 PR-URL: nodejs#64969 Fixes: nodejs#64968 Reviewed-By: Paolo Insogna <paolo@cowtech.it>
trivikr
force-pushed
the
ffi-allocate-code-near-probe
branch
from
aa9a075 to
51a55d4
Compare
August 9, 2026 23:15
aduh95
pushed a commit
that referenced
this pull request
Aug 13, 2026
AllocateCodeNear() searched 1024 pages in each direction for a free page next to the native target, one MAP_FIXED_NOREPLACE mmap per candidate. Every probe is expected to fail, so an exhausted window cost up to 2048 failing syscalls before taking the plain-mmap fallback that was already there, all to enable a jmp rel32 instead of movabs+jmp. Resolving a fast-eligible signature plateaued at ~790us, against ~28us for one that allocates no trampoline. Reduce the window to 16 pages per direction, bounding the exhausted case to 32 probes while still covering typical per-library symbol counts. Placement stays opportunistic: the caller already emits the absolute form when EmitJmpRel32() reports the target is out of range. Signed-off-by: Trivikram Kamat <16024985+trivikr@users.noreply.github.com> Assisted-by: claude:opus-5 PR-URL: #64969 Fixes: #64968 Reviewed-By: Paolo Insogna <paolo@cowtech.it>
aduh95
pushed a commit
that referenced
this pull request
Aug 25, 2026
AllocateCodeNear() searched 1024 pages in each direction for a free page next to the native target, one MAP_FIXED_NOREPLACE mmap per candidate. Every probe is expected to fail, so an exhausted window cost up to 2048 failing syscalls before taking the plain-mmap fallback that was already there, all to enable a jmp rel32 instead of movabs+jmp. Resolving a fast-eligible signature plateaued at ~790us, against ~28us for one that allocates no trampoline. Reduce the window to 16 pages per direction, bounding the exhausted case to 32 probes while still covering typical per-library symbol counts. Placement stays opportunistic: the caller already emits the absolute form when EmitJmpRel32() reports the target is out of range. Signed-off-by: Trivikram Kamat <16024985+trivikr@users.noreply.github.com> Assisted-by: claude:opus-5 PR-URL: #64969 Fixes: #64968 Reviewed-By: Paolo Insogna <paolo@cowtech.it>
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
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.
Fixes: #64968
AllocateCodeNear() searched 1024 pages in each direction for a free page next to the native target, one MAP_FIXED_NOREPLACE mmap per candidate. Every probe is expected to fail, so an exhausted window cost up to 2048 failing syscalls before taking the plain-mmap fallback that was already there, all to enable a jmp rel32 instead of movabs+jmp. Resolving a fast-eligible signature plateaued at ~790us, against ~28us for one that allocates no trampoline.
Reduce the window to 16 pages per direction, bounding the exhausted case to 32 probes while still covering typical per-library symbol counts. Placement stays opportunistic: the caller already emits the absolute form when EmitJmpRel32() reports the target is out of range.
Add benchmark/ffi/get-function.js, which measures resolution rather than call throughput, so this path has coverage. Its fast arm reports 1,852 ops/sec at the old window and 37,836 at the new one, with the fast-ineligible control arm flat at ~80,000.
Assisted-by: claude:opus-5