-
Notifications
You must be signed in to change notification settings - Fork 40
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
Add Navi3x/Navi4x hardware to the weekly run #1768
base: develop
Are you sure you want to change the base?
Conversation
mlir/utils/jenkins/Jenkinsfile
Outdated
@@ -685,7 +688,7 @@ pipeline { | |||
axes { | |||
axis { | |||
name 'ARCH' | |||
values 'gfx908', 'gfx90a', 'gfx1030' | |||
values 'gfx908', 'gfx90a', 'gfx1030', 'gfx1100', 'gfx1101', 'gfx1201' |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
gfx1100 and gfx1101 both are Navi3x based cards. I don't think we need to run on both of them.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
You would need to use same label later in this stage for Navi3x:
rocMLIR/mlir/utils/jenkins/Jenkinsfile
Line 815 in a8ac1fb
values 'gfx908', 'gfx90a', 'gfx942', 'gfx1030', 'gfx1101' |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@umangyadav regarding gfx1100 and gfx1101, I talked with Stefan so we thought that it would be good to do it on both and see which is more stable.
But if you think it's better, we can run it only on one of them.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
it would be good to do it on both and see which is more stable.
To me it seems we are not correctly making use of the Jenkins Tuning and Nightly Runs as of today. i.e. we are not really analyzing perf results on daily basis.
Therefore it seems to me we are better off keeping just one of the Navi3x on the CI right now. otherwise we increase likelihood of failing weekly CI due to random errors and it will cause more problems.
Can you also add Navi3x to parameterSweeps stage ? rocMLIR/mlir/utils/jenkins/Jenkinsfile Line 626 in a8ac1fb
|
@@ -812,7 +814,7 @@ pipeline { | |||
axes { | |||
axis { | |||
name 'CHIP' | |||
values 'gfx908', 'gfx90a', 'gfx1030', 'gfx1101' | |||
values 'gfx908', 'gfx90a', 'gfx1030', 'gfx1100', 'gfx1201' |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
can you trigger nightly and weekly tuning runs manually on this branch and see if it is tuning and generating perf reports ?
Do weekly tuning on Navi3x/Navi4x and use those perfDBs for nightly performance runs
Resolves #1727