Skip to content

Fix multi-pass tree reductions and matmul, vecdot strides in reductions - #3082

Merged
antonwolfy merged 5 commits into
masterfrom
fix-tree-reductions
Oct 2, 2026
Merged

antonwolfy merged 5 commits into
masterfrom
fix-tree-reductions

Conversation

@ndgrigorian

@ndgrigorian ndgrigorian commented Oct 1, 2026 •

Copy link
Copy Markdown
Collaborator

This PR fixes bugs in tree reductions, as well as a in vecdot for strided outputs

  • tree reductions which would take multiple passes over a row could, with very large rows, produce incorrect results
  • vecdot passed the strides of the second input as the strides of the output
  • A bug with GEMM kernels/matmul where intermediate reductions could use the wrong strides
  • Have you provided a meaningful PR description?
  • Have you added a test, reproducer or referred to an issue with a reproducer?
  • Have you tested your changes locally for CPU and GPU devices?
  • Have you made sure that new changes do not introduce compiler warnings?
  • Have you checked performance impact of proposed changes?
  • Have you added documentation for your changes, if necessary?
  • Have you added your changes to the changelog?

@github-actions

github-actions Bot commented Oct 1, 2026 •

Copy link
Copy Markdown
Contributor

View rendered docs @ https://intelpython.github.io/dpnp/index.html

@github-actions

github-actions Bot commented Oct 1, 2026 •

Copy link
Copy Markdown
Contributor

Array API standard conformance tests for dpnp=0.21.0dev11=np2py314h8d9cdd5_18 ran successfully.
Passed: 1376
Failed: 0
Skipped: 6

@coveralls

coveralls commented Oct 1, 2026 •

Copy link
Copy Markdown
Collaborator

Coverage Status

coverage: 78.622% (+0.002%) from 78.62% — fix-tree-reductions into master

@antonwolfy antonwolfy added this to the 0.21.0 release milestone Oct 2, 2026
Comment thread dpnp/tensor/libtensor/include/kernels/linalg_functions/gemm.hpp
Comment thread dpnp/tensor/libtensor/include/kernels/linalg_functions/gemm.hpp
@ndgrigorian
ndgrigorian requested a review from antonwolfy October 2, 2026 17:08
@ndgrigorian ndgrigorian changed the title Fix multi-pass tree reductions and vecdot result strides Fix multi-pass tree reductions and matmul, vecdot strides in reductions Oct 2, 2026

@antonwolfy antonwolfy left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thank you @ndgrigorian

@antonwolfy
antonwolfy merged commit 6c3fca4 into master Oct 2, 2026
111 of 113 checks passed
@antonwolfy
antonwolfy deleted the fix-tree-reductions branch October 2, 2026 21:26
github-actions Bot added a commit that referenced this pull request Oct 2, 2026
…ctions (#3082)

This PR fixes bugs in tree reductions, as well as a in `vecdot` for
strided outputs
* tree reductions which would take multiple passes over a row could,
with very large rows, produce incorrect results
* `vecdot` passed the strides of the second input as the strides of the
output
* A bug with GEMM kernels/`matmul` where intermediate reductions could
use the wrong strides 6c3fca4
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.

3 participants