too narrow constraints in a unit test?

[alexanderguzhva]

I've noticed a failure on a master branch (20caafd) in Knowhere unit tests. UT were compiled with clang-17 on an AVX-512 machine. gcc-9 shows no problems.

It seems like a 'noise' in floating-point operations.

I0703 14:11:50.064527  3057 hnsw.cc:139] [KNOWHERE][operator()][knowhere_build6] HNSW build progress: 80%
I0703 14:11:50.069895  3056 hnsw.cc:139] [KNOWHERE][operator()][knowhere_build5] HNSW build progress: 90%
I0703 14:11:50.074909  3035 time_recorder.cc:49] [KNOWHERE][PrintTimeRecord][knowhere_tests] Building HNSW cost: graph build (42.129970 ms)
I0703 14:11:50.075062  3035 hnsw.cc:150] [KNOWHERE][Add][knowhere_tests] there are 0 points can not be reached
I0703 14:11:50.075078  3035 time_recorder.cc:49] [KNOWHERE][PrintTimeRecord][knowhere_tests] Building HNSW cost: graph repair (0.176110 ms)
I0703 14:11:50.075088  3035 hnsw.cc:160] [KNOWHERE][Add][knowhere_tests] HNSW built with #points num:1000 #M:128 #max level:1 #ef_construction:200 #dim:128
I0703 14:11:50.075103  3035 time_recorder.cc:49] [KNOWHERE][PrintTimeRecord][knowhere_tests] Build index: done (42.780289 ms)

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
knowhere_tests is a Catch2 v3.3.1 host application.
Run with -? for options

-------------------------------------------------------------------------------
Test Iterator Mem Index With Float Vector
  Test Search with Bitset using iterator insufficient results
-------------------------------------------------------------------------------
/home/ubuntu/zilliz/knowhere_ut/knowhere/tests/ut/test_iterator.cc:205
...............................................................................

/home/ubuntu/zilliz/knowhere_ut/knowhere/tests/ut/test_iterator.cc:238: FAILED:
  REQUIRE( recall > kKnnRecallThreshold )
with expansion:
  0.8f > 0.8f
with messages:
  name := "HNSW_SQ8"
  cfg_json := "{"M":128,"dim":128,"ef":64,"efConstruction":200,"k":5,
  "metric_type":"COSINE"}"

[github-actions]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions. Rotten issues close after 30d of inactivity. Reopen the issue with /reopen.

[adarshs1310]

@alexanderguzhva How did you fix this?

[liliu-z]

/reopen

[sre-ci-robot]

@liliu-z: Reopened this issue.

In response to this:

/reopen

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository.

[github-actions]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions. Rotten issues close after 30d of inactivity. Reopen the issue with /reopen.