[AVX512] Don't create SHRUNKBLEND SDNodes for 512-bit vectors.x

[craig.topper]

Summary

There are no 512-bit blend instructions so we shouldn't create SHRUNKBLEND for them.

On a side note, it looks like there may be a missed opportunity for constant folding TESTM when LHS and RHS are equal.

This fixes PR34139.

[delena]

test/CodeGen/X86/pr34139.ll

13 ↗

(On Diff #112097)

Could you, please, explain me how <16 x double> value is stored using one ZMM instruction?

[zvi]

lib/Target/X86/X86ISelLowering.cpp

30679 ↗

(On Diff #112097)

Any chance that due to the added bail-out we will be missing out on this combine?

[craig.topper]

lib/Target/X86/X86ISelLowering.cpp

30679 ↗

(On Diff #112097)

This combine runs on the very last DAG combine. The one above runs on earlier DAG combine. So I don't think there's an issue. If there was, I think the early out on BitWidth==1 above would be much worse.

test/CodeGen/X86/pr34139.ll

13 ↗

(On Diff #112097)

I think its because the IR is using a store to undef as its address. So I think we sort of merged the stores. If i put in a real address we get two stores. I'll try to unreduce the test case a little

[craig.topper]

Use a less reduced test case so that we still get multiple stores

llvm/trunk/lib/Target/X86/X86ISelLowering.cpp

if (N->getOpcode() == ISD::VSELECT && DCI.isBeforeLegalizeOps() &&
     if (VT.getVectorElementType() == MVT::i16)
       return SDValue();
     // Dynamic blending was only available from SSE4.1 onward.
     if (VT.is128BitVector() && !Subtarget.hasSSE41())
       return SDValue();
     // Byte blends are only available in AVX2
     if (VT == MVT::v32i8 && !Subtarget.hasAVX2())
       return SDValue();
+    // There are no 512-bit blend instructions that use sign bits.
+    if (VT.is512BitVector())
+      return SDValue();
 
     assert(BitWidth >= 8 && BitWidth <= 64 && "Invalid mask size");
     APInt DemandedMask(APInt::getSignMask(BitWidth));
     KnownBits Known;
     TargetLowering::TargetLoweringOpt TLO(DAG, !DCI.isBeforeLegalize(),
                                           !DCI.isBeforeLegalizeOps());
     if (TLI.ShrinkDemandedConstant(Cond, DemandedMask, TLO) ||
         TLI.SimplifyDemandedBits(Cond, DemandedMask, Known, TLO)) {

llvm/trunk/test/CodeGen/X86/pr34139.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc < %s -mtriple=x86_64-unknown-linux-gnu -mcpu=knl | FileCheck %s
+
+define void @f_f(<16 x double>* %ptr) {
+; CHECK-LABEL: f_f:
+; CHECK:       # BB#0:
+; CHECK-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; CHECK-NEXT:    vmovdqa %xmm0, (%rax)
+; CHECK-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0
+; CHECK-NEXT:    vmovapd (%rdi), %zmm1
+; CHECK-NEXT:    vmovapd 64(%rdi), %zmm2
+; CHECK-NEXT:    vptestmq %zmm0, %zmm0, %k1
+; CHECK-NEXT:    vmovapd %zmm0, %zmm1 {%k1}
+; CHECK-NEXT:    vmovapd %zmm0, %zmm2 {%k1}
+; CHECK-NEXT:    vmovapd %zmm2, 64(%rdi)
+; CHECK-NEXT:    vmovapd %zmm1, (%rdi)
+  store <16 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, <16 x i8>* undef
+  %load_mask8.i.i.i = load <16 x i8>, <16 x i8>* undef
+  %v.i.i.i.i = load <16 x double>, <16 x double>* %ptr
+  %mask_vec_i1.i.i.i51.i.i = icmp ne <16 x i8> %load_mask8.i.i.i, zeroinitializer
+  %v1.i.i.i.i = select <16 x i1> %mask_vec_i1.i.i.i51.i.i, <16 x double> undef, <16 x double> %v.i.i.i.i
+  store <16 x double> %v1.i.i.i.i, <16 x double>* %ptr
+  unreachable
+}