[PATCH] D65692: [ValueTracking] When calculating known bits for integer abs, make sure we're looking at a negate and not just any instruction with the nsw flag set.

Fri Aug 2 23:52:57 PDT 2019

nikic added a comment.

As we only have a handful of SPF_ABS uses, it might make sense to introduce a separate abs flavor for the nsw case (which would also allow to handle the x-y case more accurately). SPF_PABS? SPF_ABS_NSW?

(Not in this revision of course.)

================
Comment at: llvm/lib/Analysis/ValueTracking.cpp:1098
       // will be 0 because that makes abs(INT_MIN) undefined.
-      if (Q.IIQ.hasNoSignedWrap(cast<Instruction>(RHS)))
+      if (match(RHS, m_NSWSub(m_ZeroInt(), m_Specific(LHS))))
         MaxHighZeros = 1;
----------------
I think this should still be going through IIQ, otherwise NewGVN might be unhappy.

Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D65692/new/

https://reviews.llvm.org/D65692