vanilla-io_uring: pin V to latest + zero-alloc crud responses (default GC)#922
Draft
enghitalo wants to merge 2 commits into
Draft
vanilla-io_uring: pin V to latest + zero-alloc crud responses (default GC)#922enghitalo wants to merge 2 commits into
enghitalo wants to merge 2 commits into
Conversation
Contributor
Author
|
/benchmark -f vanilla-io_uring |
Contributor
|
👋 |
Contributor
Benchmark ResultsFramework:
Full log |
…ault GC) The crud cache-HIT path runs at ~195K RPS but ok_xcache() built a throwaway strings.Builder per hit and the caller then copied it into out (out << result) — two operations per request, pure GC pressure under the default Boehm GC. Convert the crud handlers (get/list/create/update) to write the response straight into the connection buffer `out` via the existing ws/wi helpers + a new write_ok_xcache / wb, so the cache-HIT path allocates nothing and the others drop the response-head builder and the body->out copy. Byte-identical output (validated end-to-end against Postgres: X-Cache MISS->HIT, 404, 201, invalidate-on-PUT, async-db, fortunes). Removes the now-dead ok()/ok_xcache(). No -gc none — fewer allocations make the default GC cheaper instead. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
enghitalo
changed the title
Contributor
Author
|
/benchmark -f vanilla-io_uring |
Contributor
|
👋 |
Contributor
Benchmark ResultsFramework:
Full log |
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
Pin the V toolchain used by frameworks/vanilla-io_uring from c0624b274 to the latest upstream V master commit 6c4d26f1f.
Why this can improve benchmark results
This V bump brings in upstream merges that should help allocation-heavy and high-concurrency benchmark paths:
a << x), fixes #27468 vlang/v#27470<<) 4-7x slower on master vs 0.5.1 (alloc, push_many, indexed-write unaffected) vlang/v#27468.Closed issue references
<<) 4-7x slower on master vs 0.5.1 (alloc, push_many, indexed-write unaffected) vlang/v#27468 (fixed by PR #27470)Scope
Update: zero-alloc crud response assembly (default GC, no
-gc none)The crud cache-HIT path runs at ~195K RPS, but
ok_xcache()built a throwawaystrings.Builderper hit and the caller then copied it intoout— pure GC pressure.The crud handlers (get/list/create/update) now write the response straight into the
connection buffer
out(via the existingws/wi+ a newwrite_ok_xcache/wb),so the cache-HIT path allocates nothing and the others drop the response-head builder
and the body→out copy. Byte-identical output, validated against Postgres (X-Cache
MISS→HIT, 404, 201, invalidate-on-PUT, async-db, fortunes). No
-gc none.Note: this reduces GC pressure and memory; the
async-db/fortunesthroughput ceilingis still architectural (blocking
db.pgon the io_uring loop) and is tracked separately— that part is unaffected by this change.