[SandboxIR] Remove tight-coupling with LLVM's SwitchInst::CaseHandle #167093
Conversation
|
Thank you for submitting a Pull Request (PR) to the LLVM Project! This PR will be automatically labeled and the relevant teams will be notified. If you wish to, you can add reviewers by using the "Reviewers" section on this page. If this is not working for you, it is probably because you do not have write permissions for the repository. In which case you can instead tag reviewers by name in a comment by using If you have received no comments on your PR for a week, you can request a review by "ping"ing the PR by adding a comment “Ping”. The common courtesy "ping" rate is once a week. Please remember that you are asking for valuable time from other developers. If you have further questions, they may be answered by the LLVM GitHub User Guide. You can also ask questions in a comment on this PR, on the LLVM Discord or on the forums. |
| public: | ||
| CaseHandle(Context &Ctx, llvm::SwitchInst::CaseIt LLVMCaseIt) | ||
| : Ctx(Ctx), LLVMCaseIt(LLVMCaseIt) {} | ||
| LLVM_ABI ConstantInt *getCaseValue() const; |
There was a problem hiding this comment.
why is this LLVM_ABI? The corresponding method in llvm::SwitchInst::CaseHandleImpl is not annotated like that.
There was a problem hiding this comment.
These are annotations for dll shared libraries if I am not mistaken (see https://llvm.org/docs/InterfaceExportAnnotations.html#llvm-shared-library). I think it's added to declarations that are defined in the .cpp file, so llvm::SwitchInst::CaseHandleImpl wouldn't need them (oh and I just noticed that it's missing from getCaseValue()). Or maybe we can just ignore them and let the folks who maintain these take care of them.
| CaseHandle &operator*() const { return const_cast<CaseIt *>(this)->CH; } | ||
| }; | ||
|
|
||
| // TODO: Implement a proper ConstCaseIt. |
There was a problem hiding this comment.
It should be possible to templatize CaseHandle and CaseIt similar to what the implementation in LLVM IR does to handle the const case, right?
| using CaseIt = llvm::SwitchInst::CaseIteratorImpl<CaseHandle>; | ||
| using ConstCaseIt = llvm::SwitchInst::CaseIteratorImpl<ConstCaseHandle>; | ||
| class CaseIt; | ||
| class CaseHandle { |
There was a problem hiding this comment.
It's kinda weird that the SandboxIR CaseHandle wraps an LLVM CaseIt and the SandboxIR CaseIt wraps an LLVM CaseHandle.
And then the CaseHandle methods have to dereference the iterator to get an LLVM CaseHandle and forward the method call, and most CaseIt methods access the iterator via CH.LLVMCaseIt.
There was a problem hiding this comment.
Yeah it is kind of weird. I did it this way because CaseHandle is basically the main interface of the iterator, so it might as well own it. But yeah, I could change it such that the iterator is owned by CaseIt, and pass it to CaseHandle in the CaseIt functions. Let me try it out.
There was a problem hiding this comment.
OK I tried wrapping an llvm::SwitchInst::CaseHandle within a SandboxIR CaseHandle, but I don't think it can be done. The reason is that llvm::SwitchInst::CaseHandle is not publicly default constructible because it would contain invalid state (as it's job is to point to a particular case inside the switch). But the problem is that SandboxIR CaseIt needs to own a default-constructible SandboxIR CaseHandle object because of SandboxIR CaseHandle & CaseIt::operator*() needs to return a reference to CaseHandle, not a copy (i.e., in CaseIt::operator*() we can't just create a local CaseHandle object and return it).
But design-wise having SandboxIR CaseHandle wrap an llvm::SwitchCase::iterator isn't actually too weird. The CaseHandle object needs to be able to point to a particular case in the switch, and to also provide the corresponding interface functions like getCaseIndex(), getSuccessor() etc. The "pointing to particular case" part is taken care of by the llvm::SwitchCase::iterator. Also modifying the CaseHandle's members from within CaseIt is pretty much what LLVM IR is doing too (e.g., operator++ increments CaseHandle::Index) because it is the LLVM CaseHandle that actually owns the state.
Wdyt?
There was a problem hiding this comment.
Sounds reasonable, thanks for the explanation. Maybe it would be worth adding a small comment explaining why it deviates from the usual pattern where a sandboxir::X wraps an llvm::X.
There was a problem hiding this comment.
Makes sense. I added a comment about this. I also added a couple minor comments (see fixup).
SandboxIR's SwitchInst CaseHandle was relying on LLVM IR's SwitchInst::CaseHandleImpl template, which may call private functions of SandboxIR's SwitchInst. This creates a dependency cycle which is against the design principles of Sandbox IR. The issue was exposed by: llvm#166842 Thanks to @aengelke for raising the issue.
|
@vporpo Congratulations on having your first Pull Request (PR) merged into the LLVM Project! Your changes will be combined with recent changes from other authors, then tested by our build bots. If there is a problem with a build, you may receive a report in an email or a comment on this PR. Please check whether problems have been caused by your change specifically, as the builds can include changes from many authors. It is not uncommon for your change to be included in a build that fails due to someone else's changes, or infrastructure issues. How to do this, and the rest of the post-merge process, is covered in detail here. If your change does cause a problem, it may be reverted, or you can revert it yourself. This is a normal part of LLVM development. You can fix your changes and open a new PR to merge them again. If you don't get any reports, no action is required from you. Your changes are working as expected, well done! |
3 participants
SandboxIR's SwitchInst CaseHandle was relying on LLVM IR's SwitchInst::CaseHandleImpl template, which may call private functions of SandboxIR's SwitchInst. This creates a dependency cycle which is against the design principles of Sandbox IR.
The issue was exposed by: #166842 Thanks to @aengelke for raising the issue.