[llvm] r355208 - Revert "[WebAssembly] Lower SIMD shifts since they are fixed in V8"

Thomas Lively via llvm-commits llvm-commits at lists.llvm.org
Fri Mar 1 09:43:56 PST 2019


Author: tlively
Date: Fri Mar  1 09:43:55 2019
New Revision: 355208

URL: http://llvm.org/viewvc/llvm-project?rev=355208&view=rev
Log:
Revert "[WebAssembly] Lower SIMD shifts since they are fixed in V8"

They weren't fixed in V8. Oops.

Modified:
    llvm/trunk/lib/Target/WebAssembly/WebAssemblyISelLowering.cpp

Modified: llvm/trunk/lib/Target/WebAssembly/WebAssemblyISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/WebAssembly/WebAssemblyISelLowering.cpp?rev=355208&r1=355207&r2=355208&view=diff
==============================================================================
--- llvm/trunk/lib/Target/WebAssembly/WebAssemblyISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/WebAssembly/WebAssemblyISelLowering.cpp Fri Mar  1 09:43:55 2019
@@ -1291,6 +1291,11 @@ SDValue WebAssemblyTargetLowering::Lower
   // Only manually lower vector shifts
   assert(Op.getSimpleValueType().isVector());
 
+  // Expand all vector shifts until V8 fixes its implementation
+  // TODO: remove this once V8 is fixed
+  if (!Subtarget->hasUnimplementedSIMD128())
+    return unrollVectorShift(Op, DAG);
+
   // Unroll non-splat vector shifts
   BuildVectorSDNode *ShiftVec;
   SDValue SplatVal;




More information about the llvm-commits mailing list