[PATCH] D50051: AMDGPU: Push fcanonicalize through partially constant build_vector

[Stanislav Mekhanoshin via Phabricator via llvm-commits]

rampitec added inline comments.


================
Comment at: lib/Target/AMDGPU/SIISelLowering.cpp:6940
+    if (vectorEltWillFoldAway(Lo) || vectorEltWillFoldAway(Hi)) {
+      for (unsigned I = 0; I != NumOps; ++I) {
+        SDValue Op = N0.getOperand(I);
----------------
Mixing variable NumOps and static array size 2 if potentially messy. Nothing except 2 can come here right now given a check for v2f16, but I would suggest to do it uniformly, either use 2 here or use SmallVector.


https://reviews.llvm.org/D50051