Add vtr_reg_weekly_no_he that drops the high effort Titan tests and gaussian blur

[litghost]

Description

This PR is attempting to create a weekly that can provide the vast majority of the signal, without the excessive runtime.

Related Issue

Motivation and Context

This PR is attempting to do two things:

• Return QoR / wallclock / memory usage in ~12 - 24 for purposes of quality comparisons

How Has This Been Tested?

Types of changes

• Bug fix (change which fixes an issue)
• New feature (change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to change)

Checklist:

• My change requires a change to the documentation
• I have updated the documentation accordingly
• I have added tests to cover my changes
• All new and existing tests passed

[litghost]

@kmurray Dropping the high effort titan tests (and gaussian blur) and running with 3 cores means that the "weekly" test set returns in ~40 hrs, which isn't bad!

I've added a CPU and memory load log every 5 minutes in this PR, which will hopefully identify why the CI takes "forever" if a large memory regression is introduced.

[litghost]

@kmurray : Is there a better way to compose the vtr_reg_weekly and vtr_reg_weekly_no_he with less duplication?

[kmurray]

Is there a better way to compose the vtr_reg_weekly and vtr_reg_weekly_no_he with less duplication?

I've considered that, and it's potentially possible but would need some scripting work.

How I'd approach it would be to extend the task_list.txt format to support #include like behaviour, similar to the parse/pass results files. We could the define vtr_reg_weekly_he task_list.txt something like:

%include vtr_reg_weekly/task_lists.basic.txt
%include vtr_reg_weekly/task_lists.he.txt

This would support selectively not running the HE titan test by using the appropriate task list. Selectively dropping a particular benchmark (e.g. gaussianblur) would be more difficult.

[litghost]

This would support selectively not running the HE titan test by using the appropriate task list. Selectively dropping a particular benchmark (e.g. gaussianblur) would be more difficult.

Once we have the include feature, it is straightforward to make 3 task lists.

Do you need the deduplication effort solved right now, or is this PR acceptable?

[vaughnbetz]

Discussed in meeting. Don't need this PR, and instead just stop doing weekly as a pre-submit (do as post-merge only).
Action to Keith