Skip to content

Conversation

@oleksandr-pavlyk
Copy link
Contributor

This PR backports gh-1567 to 'maintenance/0.16.x' branch.

  • 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?
  • If this PR is a work in progress, are you opening the PR as a draft?

…is kernel

This resolves

```
RuntimeError: Exceeded the number of registers available on the hardware.
        The number registers per work-group cannot exceed 65536 for this kernel on this device.
        The kernel uses 108 registers per work-item for a total of 1024 work-items per work-group.
 -54 (PI_ERROR_INVALID_WORK_GROUP_SIZE)
```

when running example:

```python
import dpctl.tensor as dpt
m1 = dpt.ones((1000, 1000), dtype="i4", device="cuda")
m2 = dpt.ones((1000, 1003), dtype="i4", device="cuda")
r = dpt.matmul(m1[:, :900], m2[:900, :])
```
@oleksandr-pavlyk oleksandr-pavlyk changed the title Use kernel device-specific descriptor to determine max-wg-size for th… Backport gh-1567 to 0.16.x maintenance branch Mar 26, 2024
@github-actions
Copy link

github-actions bot commented Mar 26, 2024

Deleted rendered PR docs from intelpython.github.com/dpctl, latest should be updated shortly. 🤞

@github-actions
Copy link

Array API standard conformance tests for dpctl=0.16.0=py310h15de555_2 ran successfully.
Passed: 837
Failed: 2
Skipped: 92

@oleksandr-pavlyk oleksandr-pavlyk merged commit 2ce17a2 into maintenance/0.16.x Mar 27, 2024
@oleksandr-pavlyk oleksandr-pavlyk deleted the backport-gh-1567 branch March 27, 2024 01:07
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.

2 participants