Dense shared memory workspaces #302
Open
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
Wanted to add Changwan as a reviewer, but I don't know his GitHub username. |
Oh, I saw it just now. My Github username is "hochawa". Could you add me? |
For whatever reason, it's not letting me add you as a reviewer, weird. I think you can review it anyway, though. |
rohany
added a commit
to rohany/taco
that referenced
this pull request
Jan 22, 2021
This commit adds a function `debugCompileSource` to a `Tensor` that allows for the `Tensor` to use a kernel from a provided source file instead of generating a new one. This allows developers to add prints/assertions to TACO generated code to debug faster. Inspired by Amalee's PR (tensor-compiler#302), I would have found a command like this very useful for debugging generated code.
rohany
added a commit
to rohany/taco
that referenced
this pull request
Jan 22, 2021
This commit adds a function `debugCompileSource` to a `Tensor` that allows for the `Tensor` to use a kernel from a provided source file instead of generating a new one. This allows developers to add prints/assertions to TACO generated code to debug faster. Inspired by Amalee's PR (tensor-compiler#302), I would have found a command like this very useful for debugging generated code.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This PR adds support for dense shared memory workspaces on GPUs. There's still a little bit of work to be done for the case where multiple precomputed temporaries are used in one kernel, though none of the tests explicitly check this case. Would be nice to get feedback on the choice to add the GPUWorkspace enums, especially since I'm attaching them to Var (similar to is_ptr and is_ tensor, but the Var is assigned a GPUWorkspace enum). Another important question I have is whether the precomputed temporary (at least, in the GPU case) needs the loop that initializes it to zero.
Also, this PR includes some debugging functions that I set up for myself but thought might be useful to others. After generating a kernel / taco temporary file, I wanted to be able to edit the just-generated file and then compile and run everything it again with my handwritten changes.