Kotlin / Kotlin/multik

Flaky: NativeLinAlg.inv ComplexFloat returns wrong result on Linux

Open
#286 0 comments 0 reactions 0 assignees View on GitHub
ai-generated bug
Dominant language
Kotlin
Stars
734
Forks
51
Avg merge
9h 52m
Merged PRs (30d)
6

Description

### Multik version

0.3.0

### Module

multik-openblas

### Platform

JVM

### Operating System

Linux

### Description

`NativeLinAlg.inv` for `ComplexFloat` matrices produces incorrect inverse intermittently on Linux (Ubuntu 22.04.5 LTS, gcc-12 12.3.0, JDK 21 Temurin). The test generates an 84x84 random ComplexFloat matrix (seeded with `Random(42)`), computes its inverse, and verifies `A * A_inv = I` with epsilon `1e-4f`. Failed in 2 out of 4 CI runs.

### Reproducer

```kotlin
val a = data.getComplexFloatM(84) // deterministic 84x84 ComplexFloat matrix from Random(42)
val ainv = NativeLinAlg.inv(a)

assertComplexFloatingNDArray(mk.identity(84), NativeLinAlg.dot(a, ainv), 1e-4f) // Fails intermittently
```

### Additional information

```text
Flaky — failed in 2/4 CI runs.
AssertionFailedError at NativeLinAlgTest.kt:344.
CI runs: 22511673326, 22511861936.
Environment: Ubuntu 22.04.5 LTS, gcc-12 (12.3.0), g++-12, JDK 21 Temurin, x86_64.
```

Contributor guide

Open the contributing guide

Research direction

Start with the failing case at NativeLinAlgTest.kt:344 and run the 84x84 seeded ComplexFloat inverse check in the reported Linux environment. Trace NativeLinAlg.inv and its OpenBLAS-backed path to identify why the result varies between runs. Done means the test consistently verifies A * A_inv = I within 1e-4f.

Written by the indexing model from the issue text.

Assessment

Tech stack
kotlin
Domain
data
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
45/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.