Skip to content

branch-4.1: [codex] fix ANN OpenMP build budget and add concurrency test #61313#61652

Merged
yiguolei merged 1 commit intobranch-4.1from
auto-pick-61313-branch-4.1
Mar 24, 2026
Merged

branch-4.1: [codex] fix ANN OpenMP build budget and add concurrency test #61313#61652
yiguolei merged 1 commit intobranch-4.1from
auto-pick-61313-branch-4.1

Conversation

@github-actions
Copy link
Copy Markdown
Contributor

Cherry-picked from #61313

## Summary
This PR fixes ANN index build OpenMP thread budgeting and adds a BE unit
test for the concurrency cap.

## Problem
`ScopedOmpThreadBudget` computed available budget from
`config::omp_threads_limit` directly. With the default
`omp_threads_limit = -1`, each builder effectively reserved 1 thread and
did not follow the documented auto behavior (80% of CPU cores). In
addition, concurrent builders could overrun the intended global limit
because there was no wait/coordination when the budget was exhausted.

## Root Cause
In `faiss_ann_index.cpp`:
- The constructor used `config::omp_threads_limit` directly instead of
the auto-resolved limit from `get_omp_threads_limit()`.
- No blocking mechanism existed when all OpenMP budget was already in
use.

## Fix
- Use `get_omp_threads_limit()` as the actual global limit for
budgeting.
- Add a condition variable to block builders until at least one OpenMP
slot is available.
- Keep the existing policy of reserving up to half of remaining budget
(minimum 1).
- Add concise comments to explain wait/wakeup behavior.

## Test
- Added `VectorSearchTest.OmpThreadBudgetNeverExceedsLimit` in
`be/test/storage/index/ann/faiss_vector_index_test.cpp`.
- The test sets `config::omp_threads_limit = 1`, runs multiple
concurrent ANN `add()` builds, samples `ann_index_build_index_threads`,
and asserts peak usage never exceeds 1 and finally returns to 0.

## Validation
- Local compilation/test run was intentionally skipped per request.
@github-actions github-actions bot requested a review from yiguolei as a code owner March 24, 2026 05:44
@Thearas
Copy link
Copy Markdown
Contributor

Thearas commented Mar 24, 2026

Thank you for your contribution to Apache Doris.
Don't know what should be done next? See How to process your PR.

Please clearly describe your PR:

  1. What problem was fixed (it's best to include specific error reporting information). How it was fixed.
  2. Which behaviors were modified. What was the previous behavior, what is it now, why was it modified, and what possible impacts might there be.
  3. What features were added. Why was this function added?
  4. Which code was refactored and why was this part of the code refactored?
  5. Which functions were optimized and what is the difference before and after the optimization?

@dataroaring dataroaring reopened this Mar 24, 2026
@Thearas
Copy link
Copy Markdown
Contributor

Thearas commented Mar 24, 2026

run buildall

@hello-stephen
Copy link
Copy Markdown
Contributor

BE UT Coverage Report

Increment line coverage 100.00% (4/4) 🎉

Increment coverage report
Complete coverage report

Category Coverage
Function Coverage 52.66% (19505/37042)
Line Coverage 36.09% (181951/504095)
Region Coverage 32.36% (140077/432895)
Branch Coverage 33.53% (61346/182948)

@hello-stephen
Copy link
Copy Markdown
Contributor

BE Regression && UT Coverage Report

Increment line coverage 100.00% (4/4) 🎉

Increment coverage report
Complete coverage report

Category Coverage
Function Coverage 70.99% (25735/36254)
Line Coverage 53.76% (270071/502387)
Region Coverage 51.03% (222952/436867)
Branch Coverage 52.59% (96479/183468)

@yiguolei yiguolei merged commit 8a166f8 into branch-4.1 Mar 24, 2026
27 of 29 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

5 participants