[ownership] Add a DeadEndBlocksAnalysis that vends/saves DeadEndBlocks in between passes. #35477
Conversation
…s in between passes. Importantly this also lets us use the analysis framework to validate that we do properly invalidate DeadEndBlocks, preventing bugs. I did not thread this all over the compiler. Instead I just used it for now in SemanticARCOpts just to add some coverage without threading it into too many places.
|
The reason why I didn't thread it further is I am a little worried that it may expose more bugs than I want to fix in this PR X ). |
|
@swift-ci test |
|
Build failed |
|
@swift-ci smoke test macOS platform |
|
@swift-ci smoke test OS X platform |
|
thanks! lgtm |
|
@gottesmm post-merge review: Can you make the SetVector DeadEndBlocks::ReachableBlocks into a SmallPtrSet? A vector is only used as wordlist in the compute function. Especially when this is an analysis, which may be kept alive for a longer time, it doesn't make sense to store the vector. Also, are you planning to replace all occurrences of DeadEndBlocks with the analysis? If not (in the near future), can you add a comment for the DeadEndBlocks class, where you mention that one should better use the analysis? |
Importantly this also lets us use the analysis framework to validate that we do
properly invalidate DeadEndBlocks, preventing bugs.
I did not thread this all over the compiler. Instead I just used it for now in
SemanticARCOpts just to add some coverage without threading it into too many
places.