-
Notifications
You must be signed in to change notification settings - Fork 53
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
Static Systolic Array #1740
Static Systolic Array #1740
Conversation
@@ -0,0 +1,245 @@ | |||
{ |
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.
Just wondering if we should have the same .expect
file for the dynamic and static array to ensure they are producing the same value? The tricky thing is the cycles
field but we can remove it using jq
.
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.
I think it would be nice to do this: one runt question I'm having trouble with: is there a way for two different input files to have the same ".expect" and ".data" file?
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.
Yeah, the trick is using the expect_dir
location for the test suite so that the .expect
files are generated in the same place. For example, here is one for a Filament test suite: https://github.com/cucapra/filament/blob/main/runt.toml#L82
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.
Thanks
Awesome stuff! One comment about testing but otherwise merge whenever! |
* added static option for systolic array * flake8 * tests don't duplicate data * reformat test
Systolic array generator now has option to generate static systolic array (i.e., fixed contraction dimension). Default is still dynamic contraction dimension, but you can specify the "-s" or "-static" argument to make it non-dynamic. Having this option might be helpful to quantify the cost of having a dynamic contraction dimension.
(In our evaluation we will also, separately from this, have another comparison between the same post-op implemented statically and dynamically, in order to quantify the cost of dynamism in post-ops).