[PATCH] [X86][SSE] pslldq/psrldq byte shifts/rotation for SSE2

Chandler Carruth chandlerc at gmail.com
Sun Oct 12 18:33:15 PDT 2014

Comment at: lib/Target/X86/X86ISelLowering.cpp:7545
@@ +7544,3 @@
+  bool LoIsZero = ISD::isBuildVectorAllZeros(Lo.getNode());
+  bool HiIsZero = ISD::isBuildVectorAllZeros(Hi.getNode());
Please use the zeroable computation routine that the rest of the vector shuffle lowering uses. There are lots of other patterns that show up which this doesn't necessarily catch.


More information about the llvm-commits mailing list