[PATCH] D66987: [InlineCost] Perform caller store analysis

[Matt Denton via Phabricator via llvm-commits]

mpdenton marked an inline comment as done.
mpdenton added a comment.

I mostly would like feedback on the idea--there are one or two TODOs in the code and I spotted a bug. But, the change has some useful consequences for Chrome nonetheless.



================
Comment at: llvm/lib/Analysis/InlineCost.cpp:1268
+      // Store either the final constant or a nullptr.
+      CallerStackConstants[PtrAndOffset] = StoreC;
+    }
----------------
Ah, spotted a bug as soon as I posted this--a store to a stack location that overlaps (but is not equal to) an existing constant won't invalidate the original constant. I'll have to figure out how to fix that.


Repository:
  rL LLVM

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D66987/new/

https://reviews.llvm.org/D66987