Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Fix windows test failures currently observed in syevx, sygvx and getrf_large #731

Draft
wants to merge 16 commits into
base: develop
Choose a base branch
from

Conversation

jmachado-amd
Copy link
Collaborator

This PR is meant to investigate and fix the current test failures observed in syevx, sygvx and getrf_large, which appear to be specific to rocSOLVER's windows builds.

An attempt will be made to keep the tests as close as possible to their current state.

This commit investigates how the randomicity inherent with sampling the
tests' input data might be interfering with their outcome.
@jmachado-amd jmachado-amd added the noOptimizations Disable optimized kernels for small sizes for some routines label May 31, 2024
@jmachado-amd jmachado-amd marked this pull request as draft May 31, 2024 17:27
Small bug fix to force the use of current thread id to generate seed.
@jmachado-amd jmachado-amd changed the base branch from develop to release-staging/rocm-rel-6.2 May 31, 2024 20:05
@jmachado-amd jmachado-amd changed the base branch from release-staging/rocm-rel-6.2 to develop May 31, 2024 20:15
jmachado-amd and others added 14 commits May 31, 2024 16:56
Inspect seed choice during tests.
This commit is meant to facilitate debugging the tests for syevx/sygvx
by printing, as part of those tests' outputs, hashes of the matrices
used to store input, eigenvalues and eigenvectors (for both LAPACK and
rocSOLVER).
… fix-windows-test-failures-syevx-sygvx-getrf_large-rocsolver3.26
Change brittle comparison between number of computed eigenvalues for
cases in which a value range is provided as input.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
noOptimizations Disable optimized kernels for small sizes for some routines
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants