[X86] Teach combineVectorShiftImm to constant fold undef elements to 0 not undef.

[craig.topper]

Summary

Shifts are supposed to always shift in zeros or sign bits regardless of their inputs. It's possible the input value may have been replaced with undef by SimplifyDemandedBits, but the shift in zeros are still demanded.

This issue was reported to me by ispc from 10.0. Unfortunately their failing test does not fail on trunk. Seems to be because the shl is optimized out earlier now and doesn't become VSHLI.

ispc bug https://github.com/ispc/ispc/issues/1771

[spatel]

LGTM - I double-checked that generic constant folding in IR and SDAG gets this right:

// shift undef, Y --> 0 (can always assume that the undef value is 0)

[RKSimon]

Possibly add a PSRAI test as well? Its more irregular than the logical shifts as it isn't necessarily zeros...

[craig.topper]

I think there may be a second bug. Noticed my fix only affected the 2i64 32-bit mode cases in vec_shift5.ll. Those are being handled by combineVectorShiftImm. There's still something happening in intrinsic lowering.

llvm/lib/Target/X86/X86ISelLowering.cpp

 
   // Constant Folding.
   APInt UndefElts;
   SmallVector<APInt, 32> EltBits;
   if (N->isOnlyUserOf(N0.getNode()) &&
       getTargetConstantBitsFromNode(N0, NumBitsPerElt, UndefElts, EltBits)) {
     assert(EltBits.size() == VT.getVectorNumElements() &&
            "Unexpected shift value type");
-    for (APInt &Elt : EltBits) {
-      if (X86ISD::VSHLI == Opcode)
+    // Undef elements need to fold to 0. It's possible SimplifyDemandedBits
+    // created an undef input due to no input bits being demanded, but user
+    // still expects 0 in other bits.
+    for (unsigned i = 0, e = EltBits.size(); i != e; ++i) {
+      APInt &Elt = EltBits[i];
+      if (UndefElts[i])
+        Elt = 0;
+      else if (X86ISD::VSHLI == Opcode)
         Elt <<= ShiftVal;
       else if (X86ISD::VSRAI == Opcode)
         Elt.ashrInPlace(ShiftVal);
       else
         Elt.lshrInPlace(ShiftVal);
     }
+    // Reset undef elements since they were zeroed above.
+    UndefElts = 0;
     return getConstVector(EltBits, UndefElts, VT.getSimpleVT(), DAG, SDLoc(N));
   }
 
   const TargetLowering &TLI = DAG.getTargetLoweringInfo();
   if (TLI.SimplifyDemandedBits(SDValue(N, 0),
                                APInt::getAllOnesValue(NumBitsPerElt), DCI))
     return SDValue(N, 0);
 

llvm/test/CodeGen/X86/vec_shift5.ll

 ; X64-NEXT:    retq
   %1 = tail call <4 x i32> @llvm.x86.sse2.psrai.d(<4 x i32> <i32 undef, i32 8, i32 undef, i32 32>, i32 3)
   ret <4 x i32> %1
 }
 
 define <2 x i64> @test11() {
 ; X32-LABEL: test11:
 ; X32:       # %bb.0:
-; X32-NEXT:    movaps {{.*#+}} xmm0 = <u,u,3,0>
+; X32-NEXT:    movaps {{.*#+}} xmm0 = [0,0,3,0]
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: test11:
 ; X64:       # %bb.0:
 ; X64-NEXT:    movaps {{.*#+}} xmm0 = <u,3>
 ; X64-NEXT:    retq
   %1 = tail call <2 x i64> @llvm.x86.sse2.psrli.q(<2 x i64> <i64 undef, i64 31>, i32 3)
   ret <2 x i64> %1

 ; X64-NEXT:    retq
   %1 = tail call <4 x i32> @llvm.x86.sse2.pslli.d(<4 x i32> <i32 undef, i32 8, i32 undef, i32 32>, i32 3)
   ret <4 x i32> %1
 }
 
 define <2 x i64> @test16() {
 ; X32-LABEL: test16:
 ; X32:       # %bb.0:
-; X32-NEXT:    movaps {{.*#+}} xmm0 = <u,u,248,0>
+; X32-NEXT:    movaps {{.*#+}} xmm0 = [0,0,248,0]
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: test16:
 ; X64:       # %bb.0:
 ; X64-NEXT:    movaps {{.*#+}} xmm0 = <u,248>
 ; X64-NEXT:    retq
   %1 = tail call <2 x i64> @llvm.x86.sse2.pslli.q(<2 x i64> <i64 undef, i64 31>, i32 3)
   ret <2 x i64> %1