kokkos / kokkos/stdBLAS

`trmm` test failing on Kokkos impl

Open
#226 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

bug
Dominant language
C++
Stars
167
Forks
31
PR merge metrics
No merged PRs in 30d

Description

@mhoemmen @fnrizzi
With Kokkos implementations connected by #209, trmm native test fails when executed on Kokkos implementation (with LINALG_ENABLE_KOKKOS_DEFAULT=ON), e.g. see this CI job. There seem to be two issues:

  1. Kokkos View crashing:
unknown file: Failure
C++ exception with description "Constructing View and initializing data with uninitialized execution space" thrown in the test body.
  1. incorrect results:
[ RUN      ] BLAS3_trmm.left_lower_tri_implicit_diag
/home/runner/work/stdBLAS/stdBLAS/src/stdblas/tests/native/trmm.cpp:140: Failure
The difference between gs(i,j) and C(i,j) is nan, which exceeds tol, where
gs(i,j) evaluates to -4.4000000000000004,
C(i,j) evaluates to nan, and
tol evaluates to 1.0000000000000001e-09.
Matrices differ at index (0,0)

/home/runner/work/stdBLAS/stdBLAS/src/stdblas/tests/native/trmm.cpp:140: Failure
The difference between gs(i,j) and C(i,j) is nan, which exceeds tol, where
gs(i,j) evaluates to 7.4000000000000004,
C(i,j) evaluates to nan, and
tol evaluates to 1.0000000000000001e-09.
Matrices differ at index (1,0)

(...)

See full output: full-out.txt

Contributor guide

No contributing guide indexed for this repository

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start with src/stdblas/tests/native/trmm.cpp and reproduce the native test with LINALG_ENABLE_KOKKOS_DEFAULT=ON, using the linked CI output and full-out.txt. Investigate the Kokkos View initialization exception and the NaN results in trmm. Done means the Kokkos-enabled trmm test runs without the exception and its matrix comparisons pass.

Written by the indexing model from the issue text.

Assessment

Tech stack
cpp
Domain
backend, testing-qa
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
35/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.