HLS: Initial conversion of llvm intrinsics generated by Clang #483
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
Currently only llvm.smin.i32 is handled
phate
requested changes
May 14, 2024
There was a problem hiding this comment.
I stopped reviewing after I saw that this is extremely limited and would only cover cases for the intrinsic being in the lambda subregion (and not even in any gamma/theta node of the lambda).
The absolutely right way to go about this would be to:
- Introduce an smin operation similar like we did for the memcpy intrinsic
- Detect the smin intrinsic in the frontend when converting LLVM to RVSDG similarly like we do for memcpy.
- Add support for converting the smin operation/node back to an LLVM intrinsic in the backend.
- Add a pass in the HLS part that checks for the smin/node and converts it to a gamma.
If you want to restrict it to the HLS part of the compiler (and not touch the other parts of the compiler), then you should:
- Add an HLS pass that iterates through all the (sub-)regions of an RVSDG/lambda (you only have one lambda I guess). There is no need for a toplogical sorting, so taking the nodes directly should be enough. For every call node that you find, you :
- Compute the CallTypeClassifier
- If the call type classifier is an external call and the argument is a smin intrisinc, then replace this node with a gamma node.
| { | ||
| auto arg = root->argument(i); | ||
| auto imp = dynamic_cast<const llvm::impport *>(&arg->port()); | ||
| std::cout << "impport " << imp->name() << ": " << imp->type().debug_string() << "\n"; |
There was a problem hiding this comment.
I do not think that that it is intended to print stuff here, no?
| std::cout << "impport " << imp->name() << ": " << imp->type().debug_string() << "\n"; | ||
| if (imp->name().rfind("llvm.", 0) == 0) | ||
| { | ||
| std::cout << "replacing intrinsic " << imp->name() << std::endl; |
| @@ -413,9 +414,81 @@ rvsdg2ref(llvm::RvsdgModule & rhls, std::string path) | |||
| dump_ref(rhls, path); | |||
| } | |||
|
|
|||
| void | |||
| } | ||
| } | ||
|
|
||
| void |
Comment on lines
+420
to
+425
| for (auto user : *out) | ||
| { | ||
| if (auto ni = dynamic_cast<rvsdg::node_input *>(user)) | ||
| { | ||
| if (dynamic_cast<const llvm::CallOperation *>(&ni->node()->operation())) | ||
| { |
There was a problem hiding this comment.
This will fail to detect the smin if the call node is not in the lambda subregion (i.e., in a gamma or theta of any further nested region)
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.
Currently only llvm.smin.i32 is handled