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

CUDA regression in tensor.recompile, expr.reduction0 and others #456

Closed
Infinoid opened this issue May 7, 2021 · 3 comments
Closed

CUDA regression in tensor.recompile, expr.reduction0 and others #456

Infinoid opened this issue May 7, 2021 · 3 comments
Assignees
Labels
bug Indicates an unexpected problem or unintended behavior

Comments

@Infinoid
Copy link
Contributor

Infinoid commented May 7, 2021

e94e68e fixed #454, but introduced some new failures.

Here is the list of new failures:

	 15 - expr.reduction0 (Failed)
	 26 - */expr.storage/* (Failed)
	 59 - */lower.compile/* (Failed)
	 67 - */parafac.eval/* (Failed)
	 68 - qcd.mul0 (Failed)
	 69 - qcd.mul1 (Failed)
	 70 - qcd.mul2 (Failed)
	139 - tensor.skip_recompile (Failed)
	140 - tensor.recompile (Failed)
	191 - pytaco (Child aborted)

There are some existing failures in the windowing tests, see #426 for some of those.

The full test log is here: https://gist.github.com/Infinoid/0a56eb158356ae499c88bd9be0366937

I am running with gcc 10.2.0 and cuda 11.2. TACO is configured with -DCMAKE_BUILD_TYPE=Debug -DCUDA=ON -DOPENMP=ON -DPYTHON=ON.

@stephenchouca stephenchouca self-assigned this May 7, 2021
@stephenchouca stephenchouca added the bug Indicates an unexpected problem or unintended behavior label May 7, 2021
@stephenchouca
Copy link
Contributor

The latest commit should fix all the CUDA tests other than the windowing tests (in particular, */stride.windowing/*, */indexSetVectors.windowing/*, and */indexSetMatrices.windowing/*), which I believe are all related to #426. @rohany let me know if that's not the case.

@rohany
Copy link
Contributor

rohany commented May 16, 2021

stride.windowing is definitely related to #426, but I haven't looked into the other two. It's definitely possible that the break/continue problem in 426 is the cause of those failures though.

@Infinoid
Copy link
Contributor Author

I re-ran tests with -DCUDA=ON and confirmed that all of the new regressions have been fixed. Only the windowing failures remain.

98% tests passed, 3 tests failed out of 181

Total Test time (real) = 523.25 sec

The following tests did not run:
	 17 - expr.accumulate (Disabled)
	 27 - format.mm_permute_formats (Disabled)
	 71 - qcd.mul3 (Disabled)
	 94 - generate_evaluation_files.cpu (Disabled)
	 95 - generate_evaluation_files.gpu (Disabled)
	 96 - generate_figures.cpu (Disabled)
	151 - tensor_types.complex_mul_scalar (Disabled)
	156 - tensor_types.coordinate_types (Disabled)
	162 - lower.transpose (Disabled)
	163 - lower.transpose2 (Disabled)
	164 - lower.transpose3 (Disabled)

The following tests FAILED:
	183 - */stride.windowing/* (Failed)
	184 - */indexSetVectors.windowing/* (Failed)
	185 - */indexSetMatrices.windowing/* (Failed)

Thanks!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
bug Indicates an unexpected problem or unintended behavior
Projects
None yet
Development

No branches or pull requests

3 participants