The performance of the dav_subspace algorithm has declined compared to the previous version.

[dyzheng]

Details

Test case:
task.19258.zip

The time of same case cost 3339 s with version 3.8.0 on 4*DCU device:

log-previous.zip

and 5550 s with latest version:

log-new.zip

Please find out the reason of performance declining with GPU/DCU device.

Task list for Issue attackers (only for developers)

• Reproduce the performance issue on a similar system or environment.
• Identify the specific section of the code causing the performance issue.
• Investigate the issue and determine the root cause.
• Research best practices and potential solutions for the identified performance issue.
• Implement the chosen solution to address the performance issue.
• Test the implemented solution to ensure it improves performance without introducing new issues.
• Optimize the solution if necessary, considering trade-offs between performance and other factors (e.g., code complexity, readability, maintainability).
• Review and incorporate any relevant feedback from users or developers.
• Merge the improved solution into the main codebase and notify the issue reporter.

[dyzheng]

It caused by #5732 , default setting of diago_smooth_ethr is false. The default should be true with ks_solver dav_subspace before.

[haozhihan]

It caused by #5732 , default setting of diago_smooth_ethr is false. The default should be true with ks_solver dav_subspace before.

Yes, this PR added the smooth ethr feature to both Davidson and CG methods. At the same time, the default setting is to turn off smooth ethr, while the previous version dav-subspace method was turned on by default.

The configuration of different iterative methods is the same now.