[SelectionDAG] Teach getNode to simplify a couple easy cases of EXTRACT_SUBVECTOR

[craig.topper]

Summary

This teaches getNode to simplify extracting from Undef. This is similar to what is done for EXTRACT_VECTOR_ELT. It also adds support for extracting from CONCAT_VECTOR when we can reused one of the inputs to the concat. These seem like simple non-target specific optimizations.

For X86 we currently handle undef in extractSubvector, but not all EXTRACT_SUBVECTOR creations go through there.

Ultimately, my motivation here is to simplify extractSubvector and remove custom lowering for EXTRACT_SUBVECTOR since we don't do anything but handle undef and BUILD_VECTOR optimizations, but those should be DAG combines.

[RKSimon]

LGTM

lib/CodeGen/SelectionDAG/SelectionDAG.cpp

4094 ↗

(On Diff #85310)

Not sure if this will ever be needed? A concat of 0 vectors should have asserted at creation surely?

llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAG.cpp

if (VT.isSimple() && N1.getValueType().isSimple()) {
                 <= N1.getValueType().getVectorNumElements())
                && "Extract subvector overflow!");
       }
 
       // Trivial extraction.
       if (VT.getSimpleVT() == N1.getSimpleValueType())
         return N1;
 
+      // EXTRACT_SUBVECTOR of an UNDEF is an UNDEF.
+      if (N1.isUndef())
+        return getUNDEF(VT);
+
+      // EXTRACT_SUBVECTOR of CONCAT_VECTOR can be simplified if the pieces of
+      // the concat have the same type as the extract.
+      if (N2C && N1.getOpcode() == ISD::CONCAT_VECTORS &&
+          N1.getNumOperands() > 0 &&
+          VT == N1.getOperand(0).getValueType()) {
+        unsigned Factor = VT.getVectorNumElements();
+        return N1.getOperand(N2C->getZExtValue() / Factor);
+      }
+
       // EXTRACT_SUBVECTOR of INSERT_SUBVECTOR is often created
       // during shuffle legalization.
       if (N1.getOpcode() == ISD::INSERT_SUBVECTOR && N2 == N1.getOperand(2) &&
           VT == N1.getOperand(1).getValueType())
         return N1.getOperand(1);
     }
     break;
   }

llvm/trunk/test/CodeGen/X86/avx512-ext.ll

; ALL-NEXT:    retq
   %c = sext <8 x i8> %sign_load to <8 x i64>
   store <8 x i64> %c, <8 x i64>* %res
   ret void
 }
 
 define <64 x i16> @test21(<64 x i16> %x , <64 x i1> %mask) nounwind readnone {
 ; KNL-LABEL: test21:
 ; KNL:       ## BB#0:
-; KNL-NEXT:    pushq %rbp
-; KNL-NEXT:    pushq %r15
-; KNL-NEXT:    pushq %r14
-; KNL-NEXT:    pushq %r13
-; KNL-NEXT:    pushq %r12
-; KNL-NEXT:    pushq %rbx
-; KNL-NEXT:    vpmovsxbd %xmm7, %zmm7
-; KNL-NEXT:    vpslld $31, %zmm7, %zmm7
-; KNL-NEXT:    vpmovsxbd %xmm6, %zmm6
-; KNL-NEXT:    vpslld $31, %zmm6, %zmm6
-; KNL-NEXT:    vpmovsxbd %xmm5, %zmm5
-; KNL-NEXT:    vpslld $31, %zmm5, %zmm5
-; KNL-NEXT:    vpmovsxbd %xmm4, %zmm4
-; KNL-NEXT:    vpslld $31, %zmm4, %zmm4
-; KNL-NEXT:    vptestmd %zmm4, %zmm4, %k0
-; KNL-NEXT:    kshiftlw $14, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    kshiftlw $15, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %r13d
-; KNL-NEXT:    kshiftlw $13, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %ecx
-; KNL-NEXT:    kshiftlw $12, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %r11d
-; KNL-NEXT:    kshiftlw $11, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %r8d
-; KNL-NEXT:    kshiftlw $10, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %edi
-; KNL-NEXT:    kshiftlw $9, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %edx
-; KNL-NEXT:    kshiftlw $8, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %esi
-; KNL-NEXT:    kshiftlw $7, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %ebx
-; KNL-NEXT:    kshiftlw $6, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %ebp
-; KNL-NEXT:    kshiftlw $5, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %r14d
-; KNL-NEXT:    kshiftlw $4, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %r15d
-; KNL-NEXT:    kshiftlw $3, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %r12d
-; KNL-NEXT:    kshiftlw $2, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %r10d
-; KNL-NEXT:    kshiftlw $1, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    kmovw %k1, %r9d
-; KNL-NEXT:    vptestmd %zmm5, %zmm5, %k1
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    vmovd %r13d, %xmm4
-; KNL-NEXT:    kmovw %k0, %r13d
-; KNL-NEXT:    kshiftlw $14, %k1, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    vpinsrb $1, %eax, %xmm4, %xmm4
-; KNL-NEXT:    kmovw %k0, %eax
-; KNL-NEXT:    kshiftlw $15, %k1, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    vpinsrb $2, %ecx, %xmm4, %xmm4
-; KNL-NEXT:    kmovw %k0, %ecx
-; KNL-NEXT:    kshiftlw $13, %k1, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    vpinsrb $3, %r11d, %xmm4, %xmm4
-; KNL-NEXT:    kmovw %k0, %r11d
-; KNL-NEXT:    kshiftlw $12, %k1, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    vpinsrb $4, %r8d, %xmm4, %xmm4
-; KNL-NEXT:    kmovw %k0, %r8d
-; KNL-NEXT:    kshiftlw $11, %k1, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    vpinsrb $5, %edi, %xmm4, %xmm4
-; KNL-NEXT:    kmovw %k0, %edi
-; KNL-NEXT:    movl %edi, -{{[0-9]+}}(%rsp) ## 4-byte Spill
-; KNL-NEXT:    kshiftlw $10, %k1, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    vpinsrb $6, %edx, %xmm4, %xmm4
-; KNL-NEXT:    kmovw %k0, %edx
-; KNL-NEXT:    kshiftlw $9, %k1, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    vpinsrb $7, %esi, %xmm4, %xmm4
-; KNL-NEXT:    kmovw %k0, %esi
-; KNL-NEXT:    kshiftlw $8, %k1, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    vpinsrb $8, %ebx, %xmm4, %xmm4
-; KNL-NEXT:    kmovw %k0, %ebx
-; KNL-NEXT:    kshiftlw $7, %k1, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    vpinsrb $9, %ebp, %xmm4, %xmm4
-; KNL-NEXT:    kmovw %k0, %ebp
-; KNL-NEXT:    kshiftlw $6, %k1, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    vpinsrb $10, %r14d, %xmm4, %xmm4
-; KNL-NEXT:    kmovw %k0, %r14d
-; KNL-NEXT:    kshiftlw $5, %k1, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    vpinsrb $11, %r15d, %xmm4, %xmm4
-; KNL-NEXT:    kmovw %k0, %r15d
-; KNL-NEXT:    kshiftlw $4, %k1, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    vpinsrb $12, %r12d, %xmm4, %xmm4
-; KNL-NEXT:    kmovw %k0, %edi
-; KNL-NEXT:    kshiftlw $3, %k1, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    vpinsrb $13, %r10d, %xmm4, %xmm4
-; KNL-NEXT:    kmovw %k0, %r10d
-; KNL-NEXT:    kshiftlw $2, %k1, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    vpinsrb $14, %r9d, %xmm4, %xmm4
-; KNL-NEXT:    kmovw %k0, %r9d
-; KNL-NEXT:    kshiftlw $1, %k1, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    vpinsrb $15, %r13d, %xmm4, %xmm4
-; KNL-NEXT:    kmovw %k0, %r12d
-; KNL-NEXT:    vptestmd %zmm6, %zmm6, %k0
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    vmovd %ecx, %xmm5
-; KNL-NEXT:    kmovw %k1, %r13d
-; KNL-NEXT:    kshiftlw $14, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    vpinsrb $1, %eax, %xmm5, %xmm5
-; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    movl %eax, -{{[0-9]+}}(%rsp) ## 4-byte Spill
-; KNL-NEXT:    kshiftlw $15, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    vpinsrb $2, %r11d, %xmm5, %xmm5
-; KNL-NEXT:    kmovw %k1, %eax
-; KNL-NEXT:    kshiftlw $13, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    vpinsrb $3, %r8d, %xmm5, %xmm5
-; KNL-NEXT:    kmovw %k1, %ecx
-; KNL-NEXT:    movl %ecx, -{{[0-9]+}}(%rsp) ## 4-byte Spill
-; KNL-NEXT:    kshiftlw $12, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    vpinsrb $4, -{{[0-9]+}}(%rsp), %xmm5, %xmm5 ## 4-byte Folded Reload
-; KNL-NEXT:    kmovw %k1, %ecx
-; KNL-NEXT:    kshiftlw $11, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    vpinsrb $5, %edx, %xmm5, %xmm5
-; KNL-NEXT:    kmovw %k1, %r8d
-; KNL-NEXT:    kshiftlw $10, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    vpinsrb $6, %esi, %xmm5, %xmm5
-; KNL-NEXT:    kmovw %k1, %edx
-; KNL-NEXT:    kshiftlw $9, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    vpinsrb $7, %ebx, %xmm5, %xmm5
-; KNL-NEXT:    kmovw %k1, %esi
-; KNL-NEXT:    kshiftlw $8, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    vpinsrb $8, %ebp, %xmm5, %xmm5
-; KNL-NEXT:    kmovw %k1, %ebp
-; KNL-NEXT:    kshiftlw $7, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    vpinsrb $9, %r14d, %xmm5, %xmm5
-; KNL-NEXT:    kmovw %k1, %ebx
-; KNL-NEXT:    kshiftlw $6, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    vpinsrb $10, %r15d, %xmm5, %xmm5
-; KNL-NEXT:    kmovw %k1, %r11d
-; KNL-NEXT:    kshiftlw $5, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    vpinsrb $11, %edi, %xmm5, %xmm5
-; KNL-NEXT:    kmovw %k1, %edi
-; KNL-NEXT:    kshiftlw $4, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    vpinsrb $12, %r10d, %xmm5, %xmm5
-; KNL-NEXT:    kmovw %k1, %r10d
-; KNL-NEXT:    kshiftlw $3, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    vpinsrb $13, %r9d, %xmm5, %xmm5
-; KNL-NEXT:    kmovw %k1, %r9d
-; KNL-NEXT:    kshiftlw $2, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    vpinsrb $14, %r12d, %xmm5, %xmm5
-; KNL-NEXT:    kmovw %k1, %r14d
-; KNL-NEXT:    kshiftlw $1, %k0, %k1
-; KNL-NEXT:    kshiftrw $15, %k1, %k1
-; KNL-NEXT:    vpinsrb $15, %r13d, %xmm5, %xmm5
-; KNL-NEXT:    kmovw %k1, %r15d
-; KNL-NEXT:    vptestmd %zmm7, %zmm7, %k1
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    vmovd %eax, %xmm6
-; KNL-NEXT:    kmovw %k0, %r12d
-; KNL-NEXT:    kshiftlw $14, %k1, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    vpinsrb $1, -{{[0-9]+}}(%rsp), %xmm6, %xmm6 ## 4-byte Folded Reload
-; KNL-NEXT:    kmovw %k0, %r13d
-; KNL-NEXT:    kshiftlw $15, %k1, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    vpinsrb $2, -{{[0-9]+}}(%rsp), %xmm6, %xmm6 ## 4-byte Folded Reload
-; KNL-NEXT:    kmovw %k0, %eax
-; KNL-NEXT:    kshiftlw $13, %k1, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    vpinsrb $3, %ecx, %xmm6, %xmm6
-; KNL-NEXT:    kmovw %k0, %ecx
-; KNL-NEXT:    kshiftlw $12, %k1, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    vpinsrb $4, %r8d, %xmm6, %xmm6
-; KNL-NEXT:    kmovw %k0, %r8d
-; KNL-NEXT:    kshiftlw $11, %k1, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    vpinsrb $5, %edx, %xmm6, %xmm6
-; KNL-NEXT:    kmovw %k0, %edx
-; KNL-NEXT:    kshiftlw $10, %k1, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    vpinsrb $6, %esi, %xmm6, %xmm6
-; KNL-NEXT:    kmovw %k0, %esi
-; KNL-NEXT:    kshiftlw $9, %k1, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    vpinsrb $7, %ebp, %xmm6, %xmm6
-; KNL-NEXT:    kmovw %k0, %ebp
-; KNL-NEXT:    kshiftlw $8, %k1, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    vpinsrb $8, %ebx, %xmm6, %xmm6
-; KNL-NEXT:    kmovw %k0, %ebx
-; KNL-NEXT:    kshiftlw $7, %k1, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    vpinsrb $9, %r11d, %xmm6, %xmm6
-; KNL-NEXT:    kmovw %k0, %r11d
-; KNL-NEXT:    kshiftlw $6, %k1, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    vpinsrb $10, %edi, %xmm6, %xmm6
-; KNL-NEXT:    kmovw %k0, %edi
-; KNL-NEXT:    kshiftlw $5, %k1, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    vpinsrb $11, %r10d, %xmm6, %xmm6
-; KNL-NEXT:    kmovw %k0, %r10d
-; KNL-NEXT:    kshiftlw $4, %k1, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    vpinsrb $12, %r9d, %xmm6, %xmm6
-; KNL-NEXT:    kmovw %k0, %r9d
-; KNL-NEXT:    kshiftlw $3, %k1, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    vpinsrb $13, %r14d, %xmm6, %xmm6
-; KNL-NEXT:    kmovw %k0, %r14d
-; KNL-NEXT:    kshiftlw $2, %k1, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    vpinsrb $14, %r15d, %xmm6, %xmm6
-; KNL-NEXT:    kmovw %k0, %r15d
-; KNL-NEXT:    kshiftlw $1, %k1, %k0
-; KNL-NEXT:    kshiftrw $15, %k0, %k0
-; KNL-NEXT:    vpinsrb $15, %r12d, %xmm6, %xmm6
-; KNL-NEXT:    kmovw %k0, %r12d
-; KNL-NEXT:    kshiftrw $15, %k1, %k0
-; KNL-NEXT:    vmovd %eax, %xmm7
-; KNL-NEXT:    kmovw %k0, %eax
-; KNL-NEXT:    vpinsrb $1, %r13d, %xmm7, %xmm7
-; KNL-NEXT:    vpinsrb $2, %ecx, %xmm7, %xmm7
-; KNL-NEXT:    vpinsrb $3, %r8d, %xmm7, %xmm7
-; KNL-NEXT:    vpinsrb $4, %edx, %xmm7, %xmm7
-; KNL-NEXT:    vpinsrb $5, %esi, %xmm7, %xmm7
-; KNL-NEXT:    vpinsrb $6, %ebp, %xmm7, %xmm7
-; KNL-NEXT:    vpinsrb $7, %ebx, %xmm7, %xmm7
-; KNL-NEXT:    vpinsrb $8, %r11d, %xmm7, %xmm7
-; KNL-NEXT:    vpinsrb $9, %edi, %xmm7, %xmm7
-; KNL-NEXT:    vpinsrb $10, %r10d, %xmm7, %xmm7
-; KNL-NEXT:    vpinsrb $11, %r9d, %xmm7, %xmm7
-; KNL-NEXT:    vpinsrb $12, %r14d, %xmm7, %xmm7
-; KNL-NEXT:    vpinsrb $13, %r15d, %xmm7, %xmm7
+; KNL-NEXT:    vpmovzxbw {{.*#+}} ymm7 = xmm7[0],zero,xmm7[1],zero,xmm7[2],zero,xmm7[3],zero,xmm7[4],zero,xmm7[5],zero,xmm7[6],zero,xmm7[7],zero,xmm7[8],zero,xmm7[9],zero,xmm7[10],zero,xmm7[11],zero,xmm7[12],zero,xmm7[13],zero,xmm7[14],zero,xmm7[15],zero
+; KNL-NEXT:    vpmovzxbw {{.*#+}} ymm6 = xmm6[0],zero,xmm6[1],zero,xmm6[2],zero,xmm6[3],zero,xmm6[4],zero,xmm6[5],zero,xmm6[6],zero,xmm6[7],zero,xmm6[8],zero,xmm6[9],zero,xmm6[10],zero,xmm6[11],zero,xmm6[12],zero,xmm6[13],zero,xmm6[14],zero,xmm6[15],zero
+; KNL-NEXT:    vpmovzxbw {{.*#+}} ymm5 = xmm5[0],zero,xmm5[1],zero,xmm5[2],zero,xmm5[3],zero,xmm5[4],zero,xmm5[5],zero,xmm5[6],zero,xmm5[7],zero,xmm5[8],zero,xmm5[9],zero,xmm5[10],zero,xmm5[11],zero,xmm5[12],zero,xmm5[13],zero,xmm5[14],zero,xmm5[15],zero
 ; KNL-NEXT:    vpmovzxbw {{.*#+}} ymm4 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero,xmm4[4],zero,xmm4[5],zero,xmm4[6],zero,xmm4[7],zero,xmm4[8],zero,xmm4[9],zero,xmm4[10],zero,xmm4[11],zero,xmm4[12],zero,xmm4[13],zero,xmm4[14],zero,xmm4[15],zero
 ; KNL-NEXT:    vpsllw $15, %ymm4, %ymm4
 ; KNL-NEXT:    vpsraw $15, %ymm4, %ymm4
 ; KNL-NEXT:    vpand %ymm0, %ymm4, %ymm0
-; KNL-NEXT:    vpmovzxbw {{.*#+}} ymm4 = xmm5[0],zero,xmm5[1],zero,xmm5[2],zero,xmm5[3],zero,xmm5[4],zero,xmm5[5],zero,xmm5[6],zero,xmm5[7],zero,xmm5[8],zero,xmm5[9],zero,xmm5[10],zero,xmm5[11],zero,xmm5[12],zero,xmm5[13],zero,xmm5[14],zero,xmm5[15],zero
-; KNL-NEXT:    vpsllw $15, %ymm4, %ymm4
+; KNL-NEXT:    vpsllw $15, %ymm5, %ymm4
 ; KNL-NEXT:    vpsraw $15, %ymm4, %ymm4
 ; KNL-NEXT:    vpand %ymm1, %ymm4, %ymm1
-; KNL-NEXT:    vpmovzxbw {{.*#+}} ymm4 = xmm6[0],zero,xmm6[1],zero,xmm6[2],zero,xmm6[3],zero,xmm6[4],zero,xmm6[5],zero,xmm6[6],zero,xmm6[7],zero,xmm6[8],zero,xmm6[9],zero,xmm6[10],zero,xmm6[11],zero,xmm6[12],zero,xmm6[13],zero,xmm6[14],zero,xmm6[15],zero
-; KNL-NEXT:    vpsllw $15, %ymm4, %ymm4
+; KNL-NEXT:    vpsllw $15, %ymm6, %ymm4
 ; KNL-NEXT:    vpsraw $15, %ymm4, %ymm4
 ; KNL-NEXT:    vpand %ymm2, %ymm4, %ymm2
-; KNL-NEXT:    vpinsrb $14, %r12d, %xmm7, %xmm4
-; KNL-NEXT:    vpinsrb $15, %eax, %xmm4, %xmm4
-; KNL-NEXT:    vpmovzxbw {{.*#+}} ymm4 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero,xmm4[4],zero,xmm4[5],zero,xmm4[6],zero,xmm4[7],zero,xmm4[8],zero,xmm4[9],zero,xmm4[10],zero,xmm4[11],zero,xmm4[12],zero,xmm4[13],zero,xmm4[14],zero,xmm4[15],zero
-; KNL-NEXT:    vpsllw $15, %ymm4, %ymm4
+; KNL-NEXT:    vpsllw $15, %ymm7, %ymm4
 ; KNL-NEXT:    vpsraw $15, %ymm4, %ymm4
 ; KNL-NEXT:    vpand %ymm3, %ymm4, %ymm3
-; KNL-NEXT:    popq %rbx
-; KNL-NEXT:    popq %r12
-; KNL-NEXT:    popq %r13
-; KNL-NEXT:    popq %r14
-; KNL-NEXT:    popq %r15
-; KNL-NEXT:    popq %rbp
 ; KNL-NEXT:    retq
 ;
 ; SKX-LABEL: test21:
 ; SKX:       ## BB#0:
 ; SKX-NEXT:    vpsllw $7, %zmm2, %zmm2
 ; SKX-NEXT:    vpmovb2m %zmm2, %k1
 ; SKX-NEXT:    vmovdqu16 %zmm0, %zmm0 {%k1} {z}
 ; SKX-NEXT:    kshiftrq $32, %k1, %k1

llvm/trunk/test/CodeGen/X86/avx512-pmovxrm.ll

; X64-NEXT:    retq
   %2 = zext <16 x i8> %1 to <16 x i32>
   ret <16 x i32> %2
 }
 
 define <8 x i64> @test_llvm_x86_avx512_pmovzxbq(<16 x i8>* %a) {
 ; X32-LABEL: test_llvm_x86_avx512_pmovzxbq:
 ; X32:       ## BB#0:
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; X32-NEXT:    vmovdqu (%eax), %xmm0
-; X32-NEXT:    vpmovzxbq {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero,xmm0[2],zero,zero,zero,zero,zero,zero,zero,xmm0[3],zero,zero,zero,zero,zero,zero,zero,xmm0[4],zero,zero,zero,zero,zero,zero,zero,xmm0[5],zero,zero,zero,zero,zero,zero,zero,xmm0[6],zero,zero,zero,zero,zero,zero,zero,xmm0[7],zero,zero,zero,zero,zero,zero,zero
+; X32-NEXT:    vpmovzxbq {{.*#+}} zmm0 = mem[0],zero,zero,zero,zero,zero,zero,zero,mem[1],zero,zero,zero,zero,zero,zero,zero,mem[2],zero,zero,zero,zero,zero,zero,zero,mem[3],zero,zero,zero,zero,zero,zero,zero,mem[4],zero,zero,zero,zero,zero,zero,zero,mem[5],zero,zero,zero,zero,zero,zero,zero,mem[6],zero,zero,zero,zero,zero,zero,zero,mem[7],zero,zero,zero,zero,zero,zero,zero
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: test_llvm_x86_avx512_pmovzxbq:
 ; X64:       ## BB#0:
-; X64-NEXT:    vmovdqu (%rdi), %xmm0
-; X64-NEXT:    vpmovzxbq {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero,xmm0[2],zero,zero,zero,zero,zero,zero,zero,xmm0[3],zero,zero,zero,zero,zero,zero,zero,xmm0[4],zero,zero,zero,zero,zero,zero,zero,xmm0[5],zero,zero,zero,zero,zero,zero,zero,xmm0[6],zero,zero,zero,zero,zero,zero,zero,xmm0[7],zero,zero,zero,zero,zero,zero,zero
+; X64-NEXT:    vpmovzxbq {{.*#+}} zmm0 = mem[0],zero,zero,zero,zero,zero,zero,zero,mem[1],zero,zero,zero,zero,zero,zero,zero,mem[2],zero,zero,zero,zero,zero,zero,zero,mem[3],zero,zero,zero,zero,zero,zero,zero,mem[4],zero,zero,zero,zero,zero,zero,zero,mem[5],zero,zero,zero,zero,zero,zero,zero,mem[6],zero,zero,zero,zero,zero,zero,zero,mem[7],zero,zero,zero,zero,zero,zero,zero
 ; X64-NEXT:    retq
   %1 = load <16 x i8>, <16 x i8>* %a, align 1
   %2 = shufflevector <16 x i8> %1, <16 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
   %3 = zext <8 x i8> %2 to <8 x i64>
   ret <8 x i64> %3
 }
 
 define <16 x i32> @test_llvm_x86_avx512_pmovzxwd(<16 x i16>* %a) {

llvm/trunk/test/CodeGen/X86/avx512dq-intrinsics-upgrade.ll

; CHECK-NEXT:    retq
   ret <16 x float> %res4
 }
 
 declare <8 x double> @llvm.x86.avx512.mask.insertf64x2.512(<8 x double>, <2 x double>, i32, <8 x double>, i8)
 
 define <8 x double>@test_int_x86_avx512_mask_insertf64x2_512(<8 x double> %x0, <2 x double> %x1,<8 x double> %x3, i8 %x4) {
 ; CHECK-LABEL: test_int_x86_avx512_mask_insertf64x2_512:
 ; CHECK:       ## BB#0:
-; CHECK-NEXT:    ## kill: %XMM1<def> %XMM1<kill> %ZMM1<def>
 ; CHECK-NEXT:    vinsertf64x2 $1, %xmm1, %zmm0, %zmm3
 ; CHECK-NEXT:    kmovb %edi, %k1
 ; CHECK-NEXT:    vinsertf64x2 $1, %xmm1, %zmm0, %zmm2 {%k1}
 ; CHECK-NEXT:    vinsertf64x2 $1, %xmm1, %zmm0, %zmm0 {%k1} {z}
 ; CHECK-NEXT:    vaddpd %zmm0, %zmm2, %zmm0
 ; CHECK-NEXT:    vaddpd %zmm3, %zmm0, %zmm0
 ; CHECK-NEXT:    retq
   %res = call <8 x double> @llvm.x86.avx512.mask.insertf64x2.512(<8 x double> %x0, <2 x double> %x1, i32 1, <8 x double> %x3, i8 %x4)
; CHECK-NEXT:    retq
   ret <16 x i32> %res4
 }
 
 declare <8 x i64> @llvm.x86.avx512.mask.inserti64x2.512(<8 x i64>, <2 x i64>, i32, <8 x i64>, i8)
 
 define <8 x i64>@test_int_x86_avx512_mask_inserti64x2_512(<8 x i64> %x0, <2 x i64> %x1, <8 x i64> %x3, i8 %x4) {
 ; CHECK-LABEL: test_int_x86_avx512_mask_inserti64x2_512:
 ; CHECK:       ## BB#0:
-; CHECK-NEXT:    ## kill: %XMM1<def> %XMM1<kill> %ZMM1<def>
 ; CHECK-NEXT:    vinserti64x2 $1, %xmm1, %zmm0, %zmm3
 ; CHECK-NEXT:    kmovb %edi, %k1
 ; CHECK-NEXT:    vinserti64x2 $1, %xmm1, %zmm0, %zmm2 {%k1}
 ; CHECK-NEXT:    vinserti64x2 $1, %xmm1, %zmm0, %zmm0 {%k1} {z}
 ; CHECK-NEXT:    vpaddq %zmm0, %zmm2, %zmm0
 ; CHECK-NEXT:    vpaddq %zmm0, %zmm3, %zmm0
 ; CHECK-NEXT:    retq
   %res = call <8 x i64> @llvm.x86.avx512.mask.inserti64x2.512(<8 x i64> %x0, <2 x i64> %x1, i32 1, <8 x i64> %x3, i8 %x4)
   %res1 = call <8 x i64> @llvm.x86.avx512.mask.inserti64x2.512(<8 x i64> %x0, <2 x i64> %x1, i32 1, <8 x i64> zeroinitializer, i8 %x4)
   %res2 = call <8 x i64> @llvm.x86.avx512.mask.inserti64x2.512(<8 x i64> %x0, <2 x i64> %x1, i32 1, <8 x i64> %x3, i8 -1)
   %res3 = add <8 x i64> %res, %res1
   %res4 = add <8 x i64> %res2, %res3
   ret <8 x i64> %res4
 }

llvm/trunk/test/CodeGen/X86/vector-zext.ll

 ; AVX2-LABEL: zext_16i8_to_8i64:
 ; AVX2:       # BB#0: # %entry
 ; AVX2-NEXT:    vpmovzxbq {{.*#+}} ymm2 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero,xmm0[2],zero,zero,zero,zero,zero,zero,zero,xmm0[3],zero,zero,zero,zero,zero,zero,zero
 ; AVX2-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[1,1,2,3]
 ; AVX2-NEXT:    vpmovzxbq {{.*#+}} ymm1 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero,xmm0[2],zero,zero,zero,zero,zero,zero,zero,xmm0[3],zero,zero,zero,zero,zero,zero,zero
 ; AVX2-NEXT:    vmovdqa %ymm2, %ymm0
 ; AVX2-NEXT:    retq
 ;
-; AVX512F-LABEL: zext_16i8_to_8i64:
-; AVX512F:       # BB#0: # %entry
-; AVX512F-NEXT:    vpmovzxbq {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero,xmm0[2],zero,zero,zero,zero,zero,zero,zero,xmm0[3],zero,zero,zero,zero,zero,zero,zero,xmm0[4],zero,zero,zero,zero,zero,zero,zero,xmm0[5],zero,zero,zero,zero,zero,zero,zero,xmm0[6],zero,zero,zero,zero,zero,zero,zero,xmm0[7],zero,zero,zero,zero,zero,zero,zero
-; AVX512F-NEXT:    retq
-;
-; AVX512BW-LABEL: zext_16i8_to_8i64:
-; AVX512BW:       # BB#0: # %entry
-; AVX512BW-NEXT:    # kill: %XMM0<def> %XMM0<kill> %ZMM0<def>
-; AVX512BW-NEXT:    vpmovzxbq {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero,xmm0[2],zero,zero,zero,zero,zero,zero,zero,xmm0[3],zero,zero,zero,zero,zero,zero,zero,xmm0[4],zero,zero,zero,zero,zero,zero,zero,xmm0[5],zero,zero,zero,zero,zero,zero,zero,xmm0[6],zero,zero,zero,zero,zero,zero,zero,xmm0[7],zero,zero,zero,zero,zero,zero,zero
-; AVX512BW-NEXT:    retq
+; AVX512-LABEL: zext_16i8_to_8i64:
+; AVX512:       # BB#0: # %entry
+; AVX512-NEXT:    vpmovzxbq {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero,xmm0[2],zero,zero,zero,zero,zero,zero,zero,xmm0[3],zero,zero,zero,zero,zero,zero,zero,xmm0[4],zero,zero,zero,zero,zero,zero,zero,xmm0[5],zero,zero,zero,zero,zero,zero,zero,xmm0[6],zero,zero,zero,zero,zero,zero,zero,xmm0[7],zero,zero,zero,zero,zero,zero,zero
+; AVX512-NEXT:    retq
 entry:
   %B = shufflevector <16 x i8> %A, <16 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
   %C = zext <8 x i8> %B to <8 x i64>
   ret <8 x i64> %C
 }
 
 define <4 x i32> @zext_8i16_to_4i32(<8 x i16> %A) nounwind uwtable readnone ssp {
 ; SSE2-LABEL: zext_8i16_to_4i32: