[PATCH] D48330: [GVN] Avoid casting a vector of size less than 8 bits to i8

[Eli Friedman via Phabricator via llvm-commits]

efriedma added inline comments.


================
Comment at: lib/Transforms/Utils/VNCoercion.cpp:25
+
+  if (llvm::alignTo(StoreSize, 8) != StoreSize)
+    return false;
----------------
Please add a brief comment here explaining why we bail out.


================
Comment at: test/Transforms/GVN/load-load-odd-vector.ll:3
+
+define <7 x i1> @f(<7 x i1>* %a) {
+entry:
----------------
Please put this test into the same file as the other test.


Repository:
  rL LLVM

https://reviews.llvm.org/D48330