Skip to content

CSparse: Adjust letter-case of CMake target to other libraries.#1042

Open
mmuetzel wants to merge 7 commits into
DrTimothyAldenDavis:dev2from
mmuetzel:csparse
Open

CSparse: Adjust letter-case of CMake target to other libraries.#1042
mmuetzel wants to merge 7 commits into
DrTimothyAldenDavis:dev2from
mmuetzel:csparse

Conversation

@mmuetzel

Copy link
Copy Markdown
Contributor

The letter-case of the targets as they are used in the top-level CMakeLists.txt file didn't match the letter case in the CMakeLists.txt file for the CSparse library.

Use the letter-case of the targets in the top-level CMakeLists.txt file because it is more consistent with the letter-case of most other library targets (most importantly CXSparse).

Additionally, build the CSparse library as part of the CI using the root CMakeLists.txt file on Ubuntu.

Fixes #1041.

mmuetzel added 4 commits June 13, 2026 17:55
The letter-case of the targets as they are used in the top-level CMakeLists.txt file didn't match the letter case in the CMakeLists.txt file for the CSparse library.

Use the letter-case of the targets in the top-level CMakeLists.txt file because it is more consistent with the letter-case of most other library targets (most importantly CXSparse).

Fixes DrTimothyAldenDavis#1041.
Rename demo executables and targets so their names are different from the ones
of the CXSparse library.
Add flags `--config Release` to `cmake` command in "install" step.
@mmuetzel mmuetzel changed the title CSparse: Adjust case of CMake target to other libraries. CSparse: Adjust letter-case of CMake target to other libraries. Jun 13, 2026
mmuetzel added 2 commits June 13, 2026 20:39
GitHub recently updated to Visual Studio 2026 on their `windows-latest` image.
See: actions/runner-images#14017

CUDA 12.5 (as installed by Jimver/cuda-toolkit) does not support Visual Studio
2026.
Update the installed CUDA version to 13.2.0.
Support for the architectures 52 (Maxwell) and 60 (Pascal) have been removed
in CUDA 13.2.0. Build for newer architectures instead.
According to Google AI these are:
75: Turing (e.g., RTX 20-series, GTX 16-series, T4)
80: Ampere Datacenter (e.g., A100)
86: Ampere Consumer (e.g., RTX 30-series)
89: Ada Lovelace (e.g., RTX 40-series)
@mmuetzel

Copy link
Copy Markdown
Contributor Author

This needed a couple of follow-up changes to get the CI configurations to build again.
I hope it is ok that I made those (somewhat unrelated) changes in this PR.

Some tests seem to be hanging (indefinitely?).
Reduce the timeout to 5 minutes (from the default 25 minutes) per test to
unblock the CI runners a bit faster.
@mmuetzel

Copy link
Copy Markdown
Contributor Author

It looks like some tests are failing on the MSVC runners with CUDA:
E.g., with static libraries: https://github.com/DrTimothyAldenDavis/SuiteSparse/actions/runs/27478164546/job/81238881128?pr=1042#step:20:335

The following tests FAILED:
	 13 - CHOLMOD_int64_double_bcsstk01 (Timeout)
	 15 - CHOLMOD_int64_single_bcsstk01 (Timeout)
	 17 - CHOLMOD_int64_double_lp_afiro (Timeout)
	 19 - CHOLMOD_int64_single_lp_afiro (Timeout)
	 21 - CHOLMOD_int64_double_can24 (Timeout)
	 23 - CHOLMOD_int64_single_can24 (Timeout)
	 25 - CHOLMOD_int64_double_complex (Timeout)
	 27 - CHOLMOD_int64_single_complex (Timeout)
	 29 - CHOLMOD_int64_double_supernodal (Timeout)
	 31 - CHOLMOD_int64_single_supernodal (Timeout)

Even more tests (maybe all tests for CHOLMOD?) are timing out with shared libraries. The entire CI run seems to have failed for a different(?) reason. (It's still showing as incomplete.)
https://github.com/DrTimothyAldenDavis/SuiteSparse/actions/runs/27478164546/job/81238881094?pr=1042#step:20:90

Maybe an incompatibility with newer CUDA versions? That would be out of my expertise.

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.

1 participant