[PATCH] D101778: [CPG][ARM] Optimize towards branch on zero in codegenprepare

[Nikita Popov via Phabricator via llvm-commits]

nikic added inline comments.


================
Comment at: llvm/lib/CodeGen/CodeGenPrepare.cpp:7710
+  Value *X = Cmp->getOperand(0);
+  unsigned CmpC = cast<ConstantInt>(Cmp->getOperand(1))->getZExtValue();
+
----------------
This is going to assert for large (> 64-bit) icmps.


================
Comment at: llvm/lib/CodeGen/CodeGenPrepare.cpp:7731
+    if ((Cmp->getPredicate() == ICmpInst::ICMP_EQ ||
+         Cmp->getPredicate() == ICmpInst::ICMP_NE) &&
+        (match(UI, m_Add(m_Specific(X), m_SpecificInt(-CmpC))) ||
----------------
Style nit: `Cmp->isEquality()`


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D101778/new/

https://reviews.llvm.org/D101778