[PATCH] D19675: [InstCombine][AVX2] Add support for simplifying AVX2 per-element shifts to native shifts
Quentin Colombet via llvm-commits
llvm-commits at lists.llvm.org
Mon May 16 10:48:09 PDT 2016
qcolombet added inline comments.
================
Comment at: lib/Transforms/InstCombine/InstCombineCalls.cpp:379
@@ +378,3 @@
+ if (!COp || (!isa<UndefValue>(COp) && !isa<ConstantInt>(COp)))
+ break;
+
----------------
return nullptr?
Also, shouldn’t the type of the argument of the intrinsic + the fact that CShift is constant make sure this could be an assert?
================
Comment at: lib/Transforms/InstCombine/InstCombineCalls.cpp:414
@@ +413,3 @@
+
+ // We can't handle only some out of range values with generic logical shifts.
+ if (AnyOutOfRange)
----------------
Could you rephrase? I have trouble parsing the sentence :P.
Repository:
rL LLVM
http://reviews.llvm.org/D19675
More information about the llvm-commits
mailing list