[ARM][CodeGen] Add integer support for complex deinterleaving

[dmgreen]

Sounds like a nice addition. LGTM

llvm/lib/CodeGen/ComplexDeinterleavingPass.cpp

638

These lines look pretty long

llvm/lib/Target/ARM/ARMISelLowering.cpp

21838

hasMVEFloatOps requires hasMVEIntegerOps, so this only need check for hasMVEIntegerOps

21858

Formatting

llvm/lib/CodeGen/ComplexDeinterleavingPass.cpp

 }
 
 ComplexDeinterleavingGraph::NodePtr
 ComplexDeinterleavingGraph::identifyAdd(Instruction *Real, Instruction *Imag) {
   LLVM_DEBUG(dbgs() << "identifyAdd " << *Real << " / " << *Imag << "\n");
 
   // Determine rotation
   ComplexDeinterleavingRotation Rotation;
-  if (Real->getOpcode() == Instruction::FSub &&
-      Imag->getOpcode() == Instruction::FAdd)
+    if ((Real->getOpcode() == Instruction::FSub && Imag->getOpcode() == Instruction::FAdd) ||
+        (Real->getOpcode() == Instruction::Sub && Imag->getOpcode() == Instruction::Add))
       Rotation = ComplexDeinterleavingRotation::Rotation_90;
-  else if (Real->getOpcode() == Instruction::FAdd &&
-           Imag->getOpcode() == Instruction::FSub)
+    else if ((Real->getOpcode() == Instruction::FAdd && Imag->getOpcode() == Instruction::FSub) ||
+             (Real->getOpcode() == Instruction::Add && Imag->getOpcode() == Instruction::Sub))
       Rotation = ComplexDeinterleavingRotation::Rotation_270;
     else {
       LLVM_DEBUG(dbgs() << " - Unhandled case, rotation is not assigned.\n");
       return nullptr;
     }
 
-  auto *AR = cast<Instruction>(Real->getOperand(0));
-  auto *BI = cast<Instruction>(Real->getOperand(1));
-  auto *AI = cast<Instruction>(Imag->getOperand(0));
-  auto *BR = cast<Instruction>(Imag->getOperand(1));
+  auto *AR = dyn_cast<Instruction>(Real->getOperand(0));
+  auto *BI = dyn_cast<Instruction>(Real->getOperand(1));
+  auto *AI = dyn_cast<Instruction>(Imag->getOperand(0));
+  auto *BR = dyn_cast<Instruction>(Imag->getOperand(1));
+
+  if(!AR || !AI || !BR || !BI) {
+    LLVM_DEBUG(dbgs() << " - Not all operands are instructions.\n");
+    return nullptr;
+  }
 
   NodePtr ResA = identifyNode(AR, AI);
   if (!ResA) {
     LLVM_DEBUG(dbgs() << " - AR/AI is not identified as a composite node.\n");
     return nullptr;
   }
   NodePtr ResB = identifyNode(BR, BI);
   if (!ResB) {
     LLVM_DEBUG(dbgs() << " - BR/BI is not identified as a composite node.\n");
     return nullptr;
   }
 
   NodePtr Node =
       prepareCompositeNode(ComplexDeinterleavingOperation::CAdd, Real, Imag);
   Node->Rotation = Rotation;
   Node->addOperand(ResA);
   Node->addOperand(ResB);
   return submitCompositeNode(Node);
 }
 
 static bool isInstructionPairAdd(Instruction *A, Instruction *B) {
   unsigned OpcA = A->getOpcode();
   unsigned OpcB = B->getOpcode();
+
   return (OpcA == Instruction::FSub && OpcB == Instruction::FAdd) ||
-         (OpcA == Instruction::FAdd && OpcB == Instruction::FSub);
+         (OpcA == Instruction::FAdd && OpcB == Instruction::FSub) ||
+         (OpcA == Instruction::Sub && OpcB == Instruction::Add) ||
+         (OpcA == Instruction::Add && OpcB == Instruction::Sub);
 }
 
 static bool isInstructionPairMul(Instruction *A, Instruction *B) {
   auto Pattern =
       m_BinOp(m_FMul(m_Value(), m_Value()), m_FMul(m_Value(), m_Value()));
 
   return match(A, Pattern) && match(B, Pattern);
 }
bool ComplexDeinterleavingCompositeNode::hasAllInternalUses(
   for (auto *I : InternalInstructions) {
     for (auto *User : I->users()) {
       if (!AllInstructions.contains(cast<Instruction>(User)))
         return false;
     }
   }
   return true;
 }
 No newline at end of file

llvm/lib/Target/ARM/ARMISelLowering.cpp

 }
 
 void ARMTargetLowering::finalizeLowering(MachineFunction &MF) const {
   MF.getFrameInfo().computeMaxCallFrameSize(MF);
   TargetLoweringBase::finalizeLowering(MF);
 }
 
 bool ARMTargetLowering::isComplexDeinterleavingSupported() const {
-  return Subtarget->hasMVEFloatOps();
+  return Subtarget->hasMVEFloatOps() || Subtarget->hasMVEIntegerOps();
 }
 
 bool ARMTargetLowering::isComplexDeinterleavingOperationSupported(
     ComplexDeinterleavingOperation Operation, Type *Ty) const {
   auto *VTy = dyn_cast<FixedVectorType>(Ty);
   if (!VTy)
     return false;
 
   auto *ScalarTy = VTy->getScalarType();
   unsigned NumElements = VTy->getNumElements();
 
   unsigned VTyWidth = VTy->getScalarSizeInBits() * NumElements;
   if (VTyWidth < 128 || !llvm::isPowerOf2_32(VTyWidth))
     return false;
 
   // Both VCADD and VCMUL/VCMLA support the same types, F16 and F32
-  return ScalarTy->isHalfTy() || ScalarTy->isFloatTy();
+  if (ScalarTy->isHalfTy() || ScalarTy->isFloatTy())
+    return Subtarget->hasMVEFloatOps();
+
+  if(Operation != ComplexDeinterleavingOperation::CAdd)
+    return false;
+
+  return Subtarget->hasMVEIntegerOps() &&
+        (ScalarTy->isIntegerTy(8)  ||
+         ScalarTy->isIntegerTy(16) ||
+         ScalarTy->isIntegerTy(32));
+
 }
 
 Value *ARMTargetLowering::createComplexDeinterleavingIR(
     Instruction *I, ComplexDeinterleavingOperation OperationType,
     ComplexDeinterleavingRotation Rotation, Value *InputA, Value *InputB,
     Value *Accumulator) const {
 
   FixedVectorType *Ty = cast<FixedVectorType>(InputA->getType());

llvm/test/CodeGen/Thumb2/mve-complex-deinterleaving-i16-add.ll

entry:
   %interleaved.vec = shufflevector <2 x i16> %0, <2 x i16> %1, <4 x i32> <i32 0, i32 2, i32 1, i32 3>
   ret <4 x i16> %interleaved.vec
 }
 
 ; Expected to transform
 define arm_aapcs_vfpcc <8 x i16> @complex_add_v8i16(<8 x i16> %a, <8 x i16> %b) {
 ; CHECK-LABEL: complex_add_v8i16:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vrev32.16 q2, q1
-; CHECK-NEXT:    vadd.i32 q2, q2, q0
-; CHECK-NEXT:    vrev32.16 q0, q0
-; CHECK-NEXT:    vsub.i32 q0, q1, q0
-; CHECK-NEXT:    vmovnt.i32 q0, q2
+; CHECK-NEXT:    vcadd.i16 q0, q1, q0, #90
 ; CHECK-NEXT:    bx lr
 entry:
   %a.real = shufflevector <8 x i16> %a, <8 x i16> zeroinitializer, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
   %a.imag = shufflevector <8 x i16> %a, <8 x i16> zeroinitializer, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
   %b.real = shufflevector <8 x i16> %b, <8 x i16> zeroinitializer, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
   %b.imag = shufflevector <8 x i16> %b, <8 x i16> zeroinitializer, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
   %0 = sub <4 x i16> %b.real, %a.imag
   %1 = add <4 x i16> %b.imag, %a.real
   %interleaved.vec = shufflevector <4 x i16> %0, <4 x i16> %1, <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
   ret <8 x i16> %interleaved.vec
 }
 
 ; Expected to transform
 define arm_aapcs_vfpcc <16 x i16> @complex_add_v16i16(<16 x i16> %a, <16 x i16> %b) {
 ; CHECK-LABEL: complex_add_v16i16:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13}
-; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13}
-; CHECK-NEXT:    vmov q4, q0
-; CHECK-NEXT:    vmov.u16 r0, q0[1]
-; CHECK-NEXT:    vmov.16 q0[0], r0
-; CHECK-NEXT:    vmov.u16 r0, q4[3]
-; CHECK-NEXT:    vmov.16 q0[1], r0
-; CHECK-NEXT:    vmov.u16 r0, q4[5]
-; CHECK-NEXT:    vmov.16 q0[2], r0
-; CHECK-NEXT:    vmov.u16 r0, q4[7]
-; CHECK-NEXT:    vmov.16 q0[3], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[1]
-; CHECK-NEXT:    vmov.16 q0[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[3]
-; CHECK-NEXT:    vmov.16 q0[5], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[5]
-; CHECK-NEXT:    vmov.16 q0[6], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[7]
-; CHECK-NEXT:    vmov.16 q0[7], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[0]
-; CHECK-NEXT:    vmov.16 q5[0], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[2]
-; CHECK-NEXT:    vmov.16 q5[1], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[4]
-; CHECK-NEXT:    vmov.16 q5[2], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[6]
-; CHECK-NEXT:    vmov.16 q5[3], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[0]
-; CHECK-NEXT:    vmov.16 q5[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[2]
-; CHECK-NEXT:    vmov.16 q5[5], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[4]
-; CHECK-NEXT:    vmov.16 q5[6], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[6]
-; CHECK-NEXT:    vmov.16 q5[7], r0
-; CHECK-NEXT:    vsub.i16 q5, q5, q0
-; CHECK-NEXT:    vmov.u16 r0, q5[0]
-; CHECK-NEXT:    vmov.16 q0[0], r0
-; CHECK-NEXT:    vmov.u16 r0, q4[0]
-; CHECK-NEXT:    vmov.16 q6[0], r0
-; CHECK-NEXT:    vmov.u16 r0, q4[2]
-; CHECK-NEXT:    vmov.16 q6[1], r0
-; CHECK-NEXT:    vmov.u16 r0, q4[4]
-; CHECK-NEXT:    vmov.16 q6[2], r0
-; CHECK-NEXT:    vmov.u16 r0, q4[6]
-; CHECK-NEXT:    vmov.16 q6[3], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[0]
-; CHECK-NEXT:    vmov.16 q6[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[2]
-; CHECK-NEXT:    vmov.16 q6[5], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[4]
-; CHECK-NEXT:    vmov.16 q6[6], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[6]
-; CHECK-NEXT:    vmov.16 q6[7], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[1]
-; CHECK-NEXT:    vmov.16 q1[0], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[3]
-; CHECK-NEXT:    vmov.16 q1[1], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[5]
-; CHECK-NEXT:    vmov.16 q1[2], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[7]
-; CHECK-NEXT:    vmov.16 q1[3], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[1]
-; CHECK-NEXT:    vmov.16 q1[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[3]
-; CHECK-NEXT:    vmov.16 q1[5], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[5]
-; CHECK-NEXT:    vmov.16 q1[6], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[7]
-; CHECK-NEXT:    vmov.16 q1[7], r0
-; CHECK-NEXT:    vadd.i16 q2, q1, q6
-; CHECK-NEXT:    vmov.u16 r0, q2[0]
-; CHECK-NEXT:    vmov.16 q0[1], r0
-; CHECK-NEXT:    vmov.u16 r0, q5[1]
-; CHECK-NEXT:    vmov.16 q0[2], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[1]
-; CHECK-NEXT:    vmov.16 q0[3], r0
-; CHECK-NEXT:    vmov.u16 r0, q5[2]
-; CHECK-NEXT:    vmov.16 q0[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[2]
-; CHECK-NEXT:    vmov.16 q0[5], r0
-; CHECK-NEXT:    vmov.u16 r0, q5[3]
-; CHECK-NEXT:    vmov.16 q0[6], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[3]
-; CHECK-NEXT:    vmov.16 q0[7], r0
-; CHECK-NEXT:    vmov.u16 r0, q5[4]
-; CHECK-NEXT:    vmov.16 q1[0], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[4]
-; CHECK-NEXT:    vmov.16 q1[1], r0
-; CHECK-NEXT:    vmov.u16 r0, q5[5]
-; CHECK-NEXT:    vmov.16 q1[2], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[5]
-; CHECK-NEXT:    vmov.16 q1[3], r0
-; CHECK-NEXT:    vmov.u16 r0, q5[6]
-; CHECK-NEXT:    vmov.16 q1[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[6]
-; CHECK-NEXT:    vmov.16 q1[5], r0
-; CHECK-NEXT:    vmov.u16 r0, q5[7]
-; CHECK-NEXT:    vmov.16 q1[6], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[7]
-; CHECK-NEXT:    vmov.16 q1[7], r0
-; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13}
+; CHECK-NEXT:    vcadd.i16 q0, q2, q0, #90
+; CHECK-NEXT:    vcadd.i16 q1, q3, q1, #90
 ; CHECK-NEXT:    bx lr
 entry:
   %a.real = shufflevector <16 x i16> %a, <16 x i16> zeroinitializer, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
   %a.imag = shufflevector <16 x i16> %a, <16 x i16> zeroinitializer, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
   %b.real = shufflevector <16 x i16> %b, <16 x i16> zeroinitializer, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
   %b.imag = shufflevector <16 x i16> %b, <16 x i16> zeroinitializer, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
   %0 = sub <8 x i16> %b.real, %a.imag
   %1 = add <8 x i16> %b.imag, %a.real
   %interleaved.vec = shufflevector <8 x i16> %0, <8 x i16> %1, <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
   ret <16 x i16> %interleaved.vec
 }
 
 ; Expected to transform
 define arm_aapcs_vfpcc <32 x i16> @complex_add_v32i16(<32 x i16> %a, <32 x i16> %b) {
 ; CHECK-LABEL: complex_add_v32i16:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13, d14, d15}
-; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13, d14, d15}
-; CHECK-NEXT:    .pad #32
-; CHECK-NEXT:    sub sp, #32
-; CHECK-NEXT:    vmov q4, q0
-; CHECK-NEXT:    vmov.u16 r0, q0[1]
-; CHECK-NEXT:    vmov.16 q0[0], r0
-; CHECK-NEXT:    vmov.u16 r0, q4[3]
-; CHECK-NEXT:    vmov.16 q0[1], r0
-; CHECK-NEXT:    vmov.u16 r0, q4[5]
-; CHECK-NEXT:    vmov.16 q0[2], r0
-; CHECK-NEXT:    vmov.u16 r0, q4[7]
-; CHECK-NEXT:    vmov.16 q0[3], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[1]
-; CHECK-NEXT:    vmov.16 q0[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[3]
-; CHECK-NEXT:    vmov.16 q0[5], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[5]
-; CHECK-NEXT:    vmov.16 q0[6], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[7]
-; CHECK-NEXT:    vmov.16 q0[7], r0
-; CHECK-NEXT:    add r0, sp, #96
-; CHECK-NEXT:    vldrw.u32 q6, [r0]
-; CHECK-NEXT:    vstrw.32 q3, [sp] @ 16-byte Spill
-; CHECK-NEXT:    vmov q3, q2
-; CHECK-NEXT:    vmov.u16 r0, q6[0]
-; CHECK-NEXT:    vmov.16 q2[0], r0
-; CHECK-NEXT:    vmov.u16 r0, q6[2]
-; CHECK-NEXT:    vmov.16 q2[1], r0
-; CHECK-NEXT:    vmov.u16 r0, q6[4]
-; CHECK-NEXT:    vmov.16 q2[2], r0
-; CHECK-NEXT:    vmov.u16 r0, q6[6]
-; CHECK-NEXT:    vmov.16 q2[3], r0
-; CHECK-NEXT:    add r0, sp, #112
-; CHECK-NEXT:    vldrw.u32 q7, [r0]
-; CHECK-NEXT:    vmov.u16 r0, q7[0]
-; CHECK-NEXT:    vmov.16 q2[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q7[2]
-; CHECK-NEXT:    vmov.16 q2[5], r0
-; CHECK-NEXT:    vmov.u16 r0, q7[4]
-; CHECK-NEXT:    vmov.16 q2[6], r0
-; CHECK-NEXT:    vmov.u16 r0, q7[6]
-; CHECK-NEXT:    vmov.16 q2[7], r0
-; CHECK-NEXT:    vsub.i16 q5, q2, q0
-; CHECK-NEXT:    vmov.u16 r0, q5[0]
-; CHECK-NEXT:    vmov.16 q2[0], r0
-; CHECK-NEXT:    vmov.u16 r0, q4[0]
-; CHECK-NEXT:    vmov.16 q0[0], r0
-; CHECK-NEXT:    vmov.u16 r0, q4[2]
-; CHECK-NEXT:    vmov.16 q0[1], r0
-; CHECK-NEXT:    vmov.u16 r0, q4[4]
-; CHECK-NEXT:    vmov.16 q0[2], r0
-; CHECK-NEXT:    vmov.u16 r0, q4[6]
-; CHECK-NEXT:    vmov.16 q0[3], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[0]
-; CHECK-NEXT:    vmov.16 q0[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[2]
-; CHECK-NEXT:    vmov.16 q0[5], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[4]
-; CHECK-NEXT:    vmov.16 q0[6], r0
-; CHECK-NEXT:    vmov.u16 r0, q1[6]
-; CHECK-NEXT:    vmov.16 q0[7], r0
-; CHECK-NEXT:    vmov.u16 r0, q6[1]
-; CHECK-NEXT:    vmov.16 q1[0], r0
-; CHECK-NEXT:    vmov.u16 r0, q6[3]
-; CHECK-NEXT:    vmov.16 q1[1], r0
-; CHECK-NEXT:    vmov.u16 r0, q6[5]
-; CHECK-NEXT:    vmov.16 q1[2], r0
-; CHECK-NEXT:    vmov.u16 r0, q6[7]
-; CHECK-NEXT:    vmov.16 q1[3], r0
-; CHECK-NEXT:    vmov.u16 r0, q7[1]
-; CHECK-NEXT:    vmov.16 q1[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q7[3]
-; CHECK-NEXT:    vmov.16 q1[5], r0
-; CHECK-NEXT:    vmov.u16 r0, q7[5]
-; CHECK-NEXT:    vmov.16 q1[6], r0
-; CHECK-NEXT:    vmov.u16 r0, q7[7]
-; CHECK-NEXT:    vmov.16 q1[7], r0
-; CHECK-NEXT:    vadd.i16 q0, q1, q0
-; CHECK-NEXT:    vmov.u16 r0, q0[0]
-; CHECK-NEXT:    vmov.16 q2[1], r0
-; CHECK-NEXT:    vmov.u16 r0, q5[1]
-; CHECK-NEXT:    vmov.16 q2[2], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[1]
-; CHECK-NEXT:    vmov.16 q2[3], r0
-; CHECK-NEXT:    vmov.u16 r0, q5[2]
-; CHECK-NEXT:    vmov.16 q2[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[2]
-; CHECK-NEXT:    vmov.16 q2[5], r0
-; CHECK-NEXT:    vmov.u16 r0, q5[3]
-; CHECK-NEXT:    vmov.16 q2[6], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[3]
-; CHECK-NEXT:    vmov.16 q2[7], r0
-; CHECK-NEXT:    vmov.u16 r0, q5[4]
-; CHECK-NEXT:    vmov.16 q1[0], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[4]
-; CHECK-NEXT:    vmov.16 q1[1], r0
-; CHECK-NEXT:    vmov.u16 r0, q5[5]
-; CHECK-NEXT:    vmov.16 q1[2], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[5]
-; CHECK-NEXT:    vmov.16 q1[3], r0
-; CHECK-NEXT:    vmov.u16 r0, q5[6]
-; CHECK-NEXT:    vmov.16 q1[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[6]
-; CHECK-NEXT:    vmov.16 q1[5], r0
-; CHECK-NEXT:    vmov.u16 r0, q5[7]
-; CHECK-NEXT:    vmov.16 q1[6], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[7]
-; CHECK-NEXT:    vmov.16 q1[7], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[1]
-; CHECK-NEXT:    vmov.16 q4[0], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[3]
-; CHECK-NEXT:    vstrw.32 q2, [sp, #16] @ 16-byte Spill
-; CHECK-NEXT:    vmov.16 q4[1], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[5]
-; CHECK-NEXT:    vldrw.u32 q2, [sp] @ 16-byte Reload
-; CHECK-NEXT:    vmov.16 q4[2], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[7]
-; CHECK-NEXT:    vmov.16 q4[3], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[1]
-; CHECK-NEXT:    vmov.16 q4[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[3]
-; CHECK-NEXT:    vmov.16 q4[5], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[5]
-; CHECK-NEXT:    vmov.16 q4[6], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[7]
-; CHECK-NEXT:    vmov.16 q4[7], r0
-; CHECK-NEXT:    add r0, sp, #128
-; CHECK-NEXT:    vldrw.u32 q6, [r0]
-; CHECK-NEXT:    vmov.u16 r0, q6[0]
-; CHECK-NEXT:    vmov.16 q0[0], r0
-; CHECK-NEXT:    vmov.u16 r0, q6[2]
-; CHECK-NEXT:    vmov.16 q0[1], r0
-; CHECK-NEXT:    vmov.u16 r0, q6[4]
-; CHECK-NEXT:    vmov.16 q0[2], r0
-; CHECK-NEXT:    vmov.u16 r0, q6[6]
-; CHECK-NEXT:    vmov.16 q0[3], r0
-; CHECK-NEXT:    add r0, sp, #144
-; CHECK-NEXT:    vldrw.u32 q7, [r0]
-; CHECK-NEXT:    vmov.u16 r0, q7[0]
-; CHECK-NEXT:    vmov.16 q0[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q7[2]
-; CHECK-NEXT:    vmov.16 q0[5], r0
-; CHECK-NEXT:    vmov.u16 r0, q7[4]
-; CHECK-NEXT:    vmov.16 q0[6], r0
-; CHECK-NEXT:    vmov.u16 r0, q7[6]
-; CHECK-NEXT:    vmov.16 q0[7], r0
-; CHECK-NEXT:    vsub.i16 q5, q0, q4
-; CHECK-NEXT:    vmov.u16 r0, q5[0]
-; CHECK-NEXT:    vmov.16 q4[0], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[0]
-; CHECK-NEXT:    vmov.16 q0[0], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[2]
-; CHECK-NEXT:    vmov.16 q0[1], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[4]
-; CHECK-NEXT:    vmov.16 q0[2], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[6]
-; CHECK-NEXT:    vmov.16 q0[3], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[0]
-; CHECK-NEXT:    vmov.16 q0[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[2]
-; CHECK-NEXT:    vmov.16 q0[5], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[4]
-; CHECK-NEXT:    vmov.16 q0[6], r0
-; CHECK-NEXT:    vmov.u16 r0, q2[6]
-; CHECK-NEXT:    vmov.16 q0[7], r0
-; CHECK-NEXT:    vmov.u16 r0, q6[1]
-; CHECK-NEXT:    vmov.16 q3[0], r0
-; CHECK-NEXT:    vmov.u16 r0, q6[3]
-; CHECK-NEXT:    vmov.16 q3[1], r0
-; CHECK-NEXT:    vmov.u16 r0, q6[5]
-; CHECK-NEXT:    vmov.16 q3[2], r0
-; CHECK-NEXT:    vmov.u16 r0, q6[7]
-; CHECK-NEXT:    vmov.16 q3[3], r0
-; CHECK-NEXT:    vmov.u16 r0, q7[1]
-; CHECK-NEXT:    vmov.16 q3[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q7[3]
-; CHECK-NEXT:    vmov.16 q3[5], r0
-; CHECK-NEXT:    vmov.u16 r0, q7[5]
-; CHECK-NEXT:    vmov.16 q3[6], r0
-; CHECK-NEXT:    vmov.u16 r0, q7[7]
-; CHECK-NEXT:    vmov.16 q3[7], r0
-; CHECK-NEXT:    vadd.i16 q0, q3, q0
-; CHECK-NEXT:    vmov.u16 r0, q0[0]
-; CHECK-NEXT:    vmov.16 q4[1], r0
-; CHECK-NEXT:    vmov.u16 r0, q5[1]
-; CHECK-NEXT:    vmov.16 q4[2], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[1]
-; CHECK-NEXT:    vmov.16 q4[3], r0
-; CHECK-NEXT:    vmov.u16 r0, q5[2]
-; CHECK-NEXT:    vmov.16 q4[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[2]
-; CHECK-NEXT:    vmov.16 q4[5], r0
-; CHECK-NEXT:    vmov.u16 r0, q5[3]
-; CHECK-NEXT:    vmov.16 q4[6], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[3]
-; CHECK-NEXT:    vmov.16 q4[7], r0
-; CHECK-NEXT:    vmov.u16 r0, q5[4]
-; CHECK-NEXT:    vmov.16 q3[0], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[4]
-; CHECK-NEXT:    vmov.16 q3[1], r0
-; CHECK-NEXT:    vmov.u16 r0, q5[5]
-; CHECK-NEXT:    vmov.16 q3[2], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[5]
-; CHECK-NEXT:    vmov.16 q3[3], r0
-; CHECK-NEXT:    vmov.u16 r0, q5[6]
-; CHECK-NEXT:    vmov.16 q3[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[6]
-; CHECK-NEXT:    vmov.16 q3[5], r0
-; CHECK-NEXT:    vmov.u16 r0, q5[7]
-; CHECK-NEXT:    vmov.16 q3[6], r0
-; CHECK-NEXT:    vmov.u16 r0, q0[7]
-; CHECK-NEXT:    vmov.16 q3[7], r0
-; CHECK-NEXT:    vldrw.u32 q0, [sp, #16] @ 16-byte Reload
-; CHECK-NEXT:    vmov q2, q4
-; CHECK-NEXT:    add sp, #32
-; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13, d14, d15}
+; CHECK-NEXT:    .vsave {d8, d9}
+; CHECK-NEXT:    vpush {d8, d9}
+; CHECK-NEXT:    add r0, sp, #16
+; CHECK-NEXT:    vldrw.u32 q4, [r0]
+; CHECK-NEXT:    add r0, sp, #32
+; CHECK-NEXT:    vcadd.i16 q0, q4, q0, #90
+; CHECK-NEXT:    vldrw.u32 q4, [r0]
+; CHECK-NEXT:    add r0, sp, #48
+; CHECK-NEXT:    vcadd.i16 q1, q4, q1, #90
+; CHECK-NEXT:    vldrw.u32 q4, [r0]
+; CHECK-NEXT:    add r0, sp, #64
+; CHECK-NEXT:    vcadd.i16 q2, q4, q2, #90
+; CHECK-NEXT:    vldrw.u32 q4, [r0]
+; CHECK-NEXT:    vcadd.i16 q3, q4, q3, #90
+; CHECK-NEXT:    vpop {d8, d9}
 ; CHECK-NEXT:    bx lr
 entry:
   %a.real = shufflevector <32 x i16> %a, <32 x i16> zeroinitializer, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30>
   %a.imag = shufflevector <32 x i16> %a, <32 x i16> zeroinitializer, <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
   %b.real = shufflevector <32 x i16> %b, <32 x i16> zeroinitializer, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30>
   %b.imag = shufflevector <32 x i16> %b, <32 x i16> zeroinitializer, <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
   %0 = sub <16 x i16> %b.real, %a.imag
   %1 = add <16 x i16> %b.imag, %a.real
   %interleaved.vec = shufflevector <16 x i16> %0, <16 x i16> %1, <32 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23, i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
   ret <32 x i16> %interleaved.vec
 }

llvm/test/CodeGen/Thumb2/mve-complex-deinterleaving-i32-add.ll

entry:
   %interleaved.vec = shufflevector <1 x i32> %0, <1 x i32> %1, <2 x i32> <i32 0, i32 1>
   ret <2 x i32> %interleaved.vec
 }
 
 ; Expected to transform
 define arm_aapcs_vfpcc <4 x i32> @complex_add_v4i32(<4 x i32> %a, <4 x i32> %b) {
 ; CHECK-LABEL: complex_add_v4i32:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vrev64.32 q2, q0
-; CHECK-NEXT:    vmov r1, s6
-; CHECK-NEXT:    vmov r0, s10
-; CHECK-NEXT:    vrev64.32 q3, q1
-; CHECK-NEXT:    vmov r2, s4
-; CHECK-NEXT:    subs r0, r1, r0
-; CHECK-NEXT:    vmov r1, s8
-; CHECK-NEXT:    subs r1, r2, r1
-; CHECK-NEXT:    vmov r2, s0
-; CHECK-NEXT:    vmov q2[2], q2[0], r1, r0
-; CHECK-NEXT:    vmov r0, s14
-; CHECK-NEXT:    vmov r1, s2
-; CHECK-NEXT:    add r0, r1
-; CHECK-NEXT:    vmov r1, s12
-; CHECK-NEXT:    add r1, r2
-; CHECK-NEXT:    vmov q2[3], q2[1], r1, r0
+; CHECK-NEXT:    vcadd.i32 q2, q1, q0, #90
 ; CHECK-NEXT:    vmov q0, q2
 ; CHECK-NEXT:    bx lr
 entry:
   %a.real = shufflevector <4 x i32> %a, <4 x i32> zeroinitializer, <2 x i32> <i32 0, i32 2>
   %a.imag = shufflevector <4 x i32> %a, <4 x i32> zeroinitializer, <2 x i32> <i32 1, i32 3>
   %b.real = shufflevector <4 x i32> %b, <4 x i32> zeroinitializer, <2 x i32> <i32 0, i32 2>
   %b.imag = shufflevector <4 x i32> %b, <4 x i32> zeroinitializer, <2 x i32> <i32 1, i32 3>
   %0 = sub <2 x i32> %b.real, %a.imag
   %1 = add <2 x i32> %b.imag, %a.real
   %interleaved.vec = shufflevector <2 x i32> %0, <2 x i32> %1, <4 x i32> <i32 0, i32 2, i32 1, i32 3>
   ret <4 x i32> %interleaved.vec
 }
 
 ; Expected to transform
 define arm_aapcs_vfpcc <8 x i32> @complex_add_v8i32(<8 x i32> %a, <8 x i32> %b) {
 ; CHECK-LABEL: complex_add_v8i32:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .vsave {d8, d9, d10, d11}
-; CHECK-NEXT:    vpush {d8, d9, d10, d11}
-; CHECK-NEXT:    vmov q4, q1
-; CHECK-NEXT:    vmov.f32 s4, s0
-; CHECK-NEXT:    vmov.f32 s5, s2
-; CHECK-NEXT:    vmov.f32 s20, s9
-; CHECK-NEXT:    vmov.f32 s21, s11
-; CHECK-NEXT:    vmov.f32 s0, s1
-; CHECK-NEXT:    vmov.f32 s1, s3
-; CHECK-NEXT:    vmov.f32 s9, s10
-; CHECK-NEXT:    vmov.f32 s6, s16
-; CHECK-NEXT:    vmov.f32 s7, s18
-; CHECK-NEXT:    vmov.f32 s22, s13
-; CHECK-NEXT:    vmov.f32 s23, s15
-; CHECK-NEXT:    vmov.f32 s2, s17
-; CHECK-NEXT:    vadd.i32 q1, q5, q1
-; CHECK-NEXT:    vmov.f32 s3, s19
-; CHECK-NEXT:    vmov.f32 s10, s12
-; CHECK-NEXT:    vmov.f32 s11, s14
-; CHECK-NEXT:    vsub.i32 q2, q2, q0
-; CHECK-NEXT:    vmov.f32 s1, s4
-; CHECK-NEXT:    vmov.f32 s3, s5
-; CHECK-NEXT:    vmov.f32 s5, s6
-; CHECK-NEXT:    vmov.f32 s0, s8
-; CHECK-NEXT:    vmov.f32 s2, s9
-; CHECK-NEXT:    vmov.f32 s4, s10
-; CHECK-NEXT:    vmov.f32 s6, s11
-; CHECK-NEXT:    vpop {d8, d9, d10, d11}
+; CHECK-NEXT:    .vsave {d8, d9}
+; CHECK-NEXT:    vpush {d8, d9}
+; CHECK-NEXT:    vcadd.i32 q4, q2, q0, #90
+; CHECK-NEXT:    vcadd.i32 q2, q3, q1, #90
+; CHECK-NEXT:    vmov q0, q4
+; CHECK-NEXT:    vmov q1, q2
+; CHECK-NEXT:    vpop {d8, d9}
 ; CHECK-NEXT:    bx lr
 entry:
   %a.real = shufflevector <8 x i32> %a, <8 x i32> zeroinitializer, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
   %a.imag = shufflevector <8 x i32> %a, <8 x i32> zeroinitializer, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
   %b.real = shufflevector <8 x i32> %b, <8 x i32> zeroinitializer, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
   %b.imag = shufflevector <8 x i32> %b, <8 x i32> zeroinitializer, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
   %0 = sub <4 x i32> %b.real, %a.imag
   %1 = add <4 x i32> %b.imag, %a.real
   %interleaved.vec = shufflevector <4 x i32> %0, <4 x i32> %1, <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
   ret <8 x i32> %interleaved.vec
 }
 
 ; Expected to transform
 define arm_aapcs_vfpcc <16 x i32> @complex_add_v16i32(<16 x i32> %a, <16 x i32> %b) {
 ; CHECK-LABEL: complex_add_v16i32:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13, d14, d15}
 ; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13, d14, d15}
-; CHECK-NEXT:    add r0, sp, #80
-; CHECK-NEXT:    vmov q4, q1
-; CHECK-NEXT:    vldrw.u32 q5, [r0]
 ; CHECK-NEXT:    add r0, sp, #64
-; CHECK-NEXT:    vldrw.u32 q6, [r0]
-; CHECK-NEXT:    add r0, sp, #112
-; CHECK-NEXT:    vmov.f32 s4, s0
-; CHECK-NEXT:    vmov.f32 s5, s2
-; CHECK-NEXT:    vmov.f32 s28, s25
-; CHECK-NEXT:    vmov.f32 s29, s27
-; CHECK-NEXT:    vmov.f32 s0, s1
-; CHECK-NEXT:    vmov.f32 s1, s3
-; CHECK-NEXT:    vmov.f32 s25, s26
-; CHECK-NEXT:    vmov.f32 s30, s21
-; CHECK-NEXT:    vmov.f32 s31, s23
-; CHECK-NEXT:    vmov.f32 s2, s17
-; CHECK-NEXT:    vmov.f32 s3, s19
-; CHECK-NEXT:    vmov.f32 s26, s20
-; CHECK-NEXT:    vmov.f32 s27, s22
 ; CHECK-NEXT:    vldrw.u32 q5, [r0]
+; CHECK-NEXT:    add r0, sp, #80
+; CHECK-NEXT:    vcadd.i32 q4, q5, q0, #90
+; CHECK-NEXT:    vldrw.u32 q0, [r0]
 ; CHECK-NEXT:    add r0, sp, #96
-; CHECK-NEXT:    vmov.f32 s6, s16
-; CHECK-NEXT:    vmov.f32 s7, s18
-; CHECK-NEXT:    vsub.i32 q4, q6, q0
-; CHECK-NEXT:    vldrw.u32 q6, [r0]
-; CHECK-NEXT:    vadd.i32 q1, q7, q1
-; CHECK-NEXT:    vmov.f32 s0, s16
-; CHECK-NEXT:    vmov.f32 s1, s4
-; CHECK-NEXT:    vmov.f32 s2, s17
-; CHECK-NEXT:    vmov.f32 s3, s5
-; CHECK-NEXT:    vmov.f32 s4, s18
-; CHECK-NEXT:    vmov.f32 s5, s6
-; CHECK-NEXT:    vmov.f32 s6, s19
-; CHECK-NEXT:    vmov.f32 s16, s8
-; CHECK-NEXT:    vmov.f32 s17, s10
-; CHECK-NEXT:    vmov.f32 s28, s25
-; CHECK-NEXT:    vmov.f32 s29, s27
-; CHECK-NEXT:    vmov.f32 s8, s9
-; CHECK-NEXT:    vmov.f32 s9, s11
-; CHECK-NEXT:    vmov.f32 s25, s26
-; CHECK-NEXT:    vmov.f32 s18, s12
-; CHECK-NEXT:    vmov.f32 s19, s14
-; CHECK-NEXT:    vmov.f32 s30, s21
-; CHECK-NEXT:    vmov.f32 s31, s23
-; CHECK-NEXT:    vmov.f32 s10, s13
-; CHECK-NEXT:    vadd.i32 q4, q7, q4
-; CHECK-NEXT:    vmov.f32 s11, s15
-; CHECK-NEXT:    vmov.f32 s26, s20
-; CHECK-NEXT:    vmov.f32 s27, s22
-; CHECK-NEXT:    vsub.i32 q3, q6, q2
-; CHECK-NEXT:    vmov.f32 s9, s16
-; CHECK-NEXT:    vmov.f32 s11, s17
-; CHECK-NEXT:    vmov.f32 s17, s18
-; CHECK-NEXT:    vmov.f32 s16, s14
-; CHECK-NEXT:    vmov.f32 s18, s15
-; CHECK-NEXT:    vmov.f32 s8, s12
-; CHECK-NEXT:    vmov.f32 s10, s13
-; CHECK-NEXT:    vmov q3, q4
+; CHECK-NEXT:    vcadd.i32 q5, q0, q1, #90
+; CHECK-NEXT:    vldrw.u32 q0, [r0]
+; CHECK-NEXT:    add r0, sp, #112
+; CHECK-NEXT:    vmov q1, q5
+; CHECK-NEXT:    vcadd.i32 q6, q0, q2, #90
+; CHECK-NEXT:    vldrw.u32 q0, [r0]
+; CHECK-NEXT:    vmov q2, q6
+; CHECK-NEXT:    vcadd.i32 q7, q0, q3, #90
+; CHECK-NEXT:    vmov q0, q4
+; CHECK-NEXT:    vmov q3, q7
 ; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13, d14, d15}
 ; CHECK-NEXT:    bx lr
 entry:
   %a.real = shufflevector <16 x i32> %a, <16 x i32> zeroinitializer, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
   %a.imag = shufflevector <16 x i32> %a, <16 x i32> zeroinitializer, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
   %b.real = shufflevector <16 x i32> %b, <16 x i32> zeroinitializer, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
   %b.imag = shufflevector <16 x i32> %b, <16 x i32> zeroinitializer, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
   %0 = sub <8 x i32> %b.real, %a.imag
   %1 = add <8 x i32> %b.imag, %a.real
   %interleaved.vec = shufflevector <8 x i32> %0, <8 x i32> %1, <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
   ret <16 x i32> %interleaved.vec
 }
 
 ; Expected to transform
 define arm_aapcs_vfpcc <32 x i32> @complex_add_v32i32(<32 x i32> %a, <32 x i32> %b) {
 ; CHECK-LABEL: complex_add_v32i32:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13, d14, d15}
-; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13, d14, d15}
-; CHECK-NEXT:    .pad #32
-; CHECK-NEXT:    sub sp, #32
-; CHECK-NEXT:    add r1, sp, #144
-; CHECK-NEXT:    vstrw.32 q1, [sp, #16] @ 16-byte Spill
-; CHECK-NEXT:    vldrw.u32 q5, [r1]
-; CHECK-NEXT:    add r1, sp, #128
-; CHECK-NEXT:    vldrw.u32 q7, [r1]
-; CHECK-NEXT:    add r1, sp, #272
-; CHECK-NEXT:    vldrw.u32 q1, [r1]
-; CHECK-NEXT:    add r1, sp, #256
-; CHECK-NEXT:    vstrw.32 q0, [sp] @ 16-byte Spill
-; CHECK-NEXT:    vldrw.u32 q0, [r1]
-; CHECK-NEXT:    vmov.f32 s16, s28
-; CHECK-NEXT:    add r1, sp, #112
-; CHECK-NEXT:    vmov.f32 s24, s1
-; CHECK-NEXT:    vmov.f32 s25, s3
-; CHECK-NEXT:    vmov.f32 s17, s30
-; CHECK-NEXT:    vmov.f32 s18, s20
-; CHECK-NEXT:    vmov.f32 s19, s22
-; CHECK-NEXT:    vmov.f32 s26, s5
-; CHECK-NEXT:    vmov.f32 s27, s7
-; CHECK-NEXT:    vadd.i32 q4, q6, q4
-; CHECK-NEXT:    vmov.f32 s1, s2
-; CHECK-NEXT:    vmov.f32 s24, s29
-; CHECK-NEXT:    vmov.f32 s25, s31
-; CHECK-NEXT:    vmov.f32 s26, s21
-; CHECK-NEXT:    vmov.f32 s27, s23
-; CHECK-NEXT:    vldrw.u32 q5, [r1]
+; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13}
+; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13}
 ; CHECK-NEXT:    add r1, sp, #96
-; CHECK-NEXT:    vmov.f32 s2, s4
-; CHECK-NEXT:    vmov.f32 s3, s6
-; CHECK-NEXT:    vldrw.u32 q7, [r1]
-; CHECK-NEXT:    vsub.i32 q0, q0, q6
-; CHECK-NEXT:    add r1, sp, #240
-; CHECK-NEXT:    vldrw.u32 q6, [r1]
-; CHECK-NEXT:    vmov.f32 s5, s18
-; CHECK-NEXT:    vmov.f32 s7, s19
+; CHECK-NEXT:    vldrw.u32 q4, [r1]
 ; CHECK-NEXT:    add r1, sp, #224
-; CHECK-NEXT:    vmov.f32 s4, s2
-; CHECK-NEXT:    vmov.f32 s6, s3
-; CHECK-NEXT:    vstrw.32 q1, [r0, #112]
-; CHECK-NEXT:    vmov.f32 s4, s0
-; CHECK-NEXT:    vmov.f32 s6, s1
-; CHECK-NEXT:    vldrw.u32 q0, [r1]
-; CHECK-NEXT:    vmov.f32 s5, s16
+; CHECK-NEXT:    vldrw.u32 q5, [r1]
+; CHECK-NEXT:    add r1, sp, #80
+; CHECK-NEXT:    vcadd.i32 q6, q5, q4, #90
+; CHECK-NEXT:    vldrw.u32 q4, [r1]
 ; CHECK-NEXT:    add r1, sp, #208
-; CHECK-NEXT:    vmov.f32 s7, s17
-; CHECK-NEXT:    vstrw.32 q1, [r0, #96]
-; CHECK-NEXT:    vmov.f32 s16, s1
-; CHECK-NEXT:    vmov.f32 s17, s3
-; CHECK-NEXT:    vmov.f32 s4, s28
-; CHECK-NEXT:    vmov.f32 s5, s30
-; CHECK-NEXT:    vmov.f32 s6, s20
-; CHECK-NEXT:    vmov.f32 s7, s22
-; CHECK-NEXT:    vmov.f32 s18, s25
-; CHECK-NEXT:    vmov.f32 s19, s27
-; CHECK-NEXT:    vadd.i32 q4, q4, q1
-; CHECK-NEXT:    vmov.f32 s1, s2
-; CHECK-NEXT:    vmov.f32 s4, s29
-; CHECK-NEXT:    vmov.f32 s5, s31
-; CHECK-NEXT:    vmov.f32 s6, s21
-; CHECK-NEXT:    vmov.f32 s7, s23
+; CHECK-NEXT:    vstrw.32 q6, [r0, #112]
 ; CHECK-NEXT:    vldrw.u32 q5, [r1]
-; CHECK-NEXT:    vmov.f32 s2, s24
+; CHECK-NEXT:    add r1, sp, #64
+; CHECK-NEXT:    vcadd.i32 q6, q5, q4, #90
+; CHECK-NEXT:    vldrw.u32 q4, [r1]
 ; CHECK-NEXT:    add r1, sp, #192
-; CHECK-NEXT:    vmov.f32 s3, s26
-; CHECK-NEXT:    vsub.i32 q0, q0, q1
-; CHECK-NEXT:    vmov.f32 s5, s18
-; CHECK-NEXT:    vmov.f32 s7, s19
-; CHECK-NEXT:    vmov.f32 s4, s2
-; CHECK-NEXT:    vmov.f32 s6, s3
-; CHECK-NEXT:    vstrw.32 q1, [r0, #80]
-; CHECK-NEXT:    vmov.f32 s4, s0
-; CHECK-NEXT:    vmov.f32 s6, s1
-; CHECK-NEXT:    vldrw.u32 q0, [r1]
-; CHECK-NEXT:    vmov.f32 s5, s16
+; CHECK-NEXT:    vstrw.32 q6, [r0, #96]
+; CHECK-NEXT:    vldrw.u32 q5, [r1]
+; CHECK-NEXT:    add r1, sp, #48
+; CHECK-NEXT:    vcadd.i32 q6, q5, q4, #90
+; CHECK-NEXT:    vldrw.u32 q4, [r1]
 ; CHECK-NEXT:    add r1, sp, #176
-; CHECK-NEXT:    vmov.f32 s7, s17
-; CHECK-NEXT:    vstrw.32 q1, [r0, #64]
-; CHECK-NEXT:    vmov.f32 s16, s1
-; CHECK-NEXT:    vmov.f32 s17, s3
-; CHECK-NEXT:    vmov.f32 s4, s8
-; CHECK-NEXT:    vmov.f32 s5, s10
-; CHECK-NEXT:    vmov.f32 s6, s12
-; CHECK-NEXT:    vmov.f32 s7, s14
-; CHECK-NEXT:    vmov.f32 s18, s21
-; CHECK-NEXT:    vmov.f32 s19, s23
-; CHECK-NEXT:    vadd.i32 q4, q4, q1
-; CHECK-NEXT:    vmov.f32 s1, s2
-; CHECK-NEXT:    vmov.f32 s4, s9
-; CHECK-NEXT:    vmov.f32 s5, s11
-; CHECK-NEXT:    vmov.f32 s6, s13
-; CHECK-NEXT:    vmov.f32 s7, s15
-; CHECK-NEXT:    vldrw.u32 q3, [r1]
-; CHECK-NEXT:    vmov.f32 s2, s20
+; CHECK-NEXT:    vstrw.32 q6, [r0, #80]
+; CHECK-NEXT:    vldrw.u32 q5, [r1]
 ; CHECK-NEXT:    add r1, sp, #160
-; CHECK-NEXT:    vmov.f32 s3, s22
-; CHECK-NEXT:    vsub.i32 q0, q0, q1
-; CHECK-NEXT:    vmov.f32 s5, s18
-; CHECK-NEXT:    vmov.f32 s4, s2
-; CHECK-NEXT:    vmov.f32 s6, s3
-; CHECK-NEXT:    vmov.f32 s7, s19
-; CHECK-NEXT:    vstrw.32 q1, [r0, #48]
-; CHECK-NEXT:    vmov.f32 s4, s0
-; CHECK-NEXT:    vmov.f32 s6, s1
-; CHECK-NEXT:    vldrw.u32 q0, [r1]
-; CHECK-NEXT:    vmov.f32 s5, s16
-; CHECK-NEXT:    vmov.f32 s7, s17
-; CHECK-NEXT:    vstrw.32 q1, [r0, #32]
-; CHECK-NEXT:    vmov.f32 s8, s1
-; CHECK-NEXT:    vldrw.u32 q5, [sp] @ 16-byte Reload
-; CHECK-NEXT:    vldrw.u32 q4, [sp, #16] @ 16-byte Reload
-; CHECK-NEXT:    vmov.f32 s9, s3
-; CHECK-NEXT:    vmov.f32 s4, s20
-; CHECK-NEXT:    vmov.f32 s5, s22
-; CHECK-NEXT:    vmov.f32 s6, s16
-; CHECK-NEXT:    vmov.f32 s7, s18
-; CHECK-NEXT:    vmov.f32 s10, s13
-; CHECK-NEXT:    vmov.f32 s11, s15
-; CHECK-NEXT:    vadd.i32 q2, q2, q1
-; CHECK-NEXT:    vmov.f32 s1, s2
-; CHECK-NEXT:    vmov.f32 s4, s21
-; CHECK-NEXT:    vmov.f32 s5, s23
-; CHECK-NEXT:    vmov.f32 s6, s17
-; CHECK-NEXT:    vmov.f32 s7, s19
-; CHECK-NEXT:    vmov.f32 s2, s12
-; CHECK-NEXT:    vmov.f32 s3, s14
-; CHECK-NEXT:    vsub.i32 q0, q0, q1
-; CHECK-NEXT:    vmov.f32 s5, s10
-; CHECK-NEXT:    vmov.f32 s4, s2
-; CHECK-NEXT:    vmov.f32 s6, s3
-; CHECK-NEXT:    vmov.f32 s7, s11
-; CHECK-NEXT:    vstrw.32 q1, [r0, #16]
-; CHECK-NEXT:    vmov.f32 s4, s0
-; CHECK-NEXT:    vmov.f32 s5, s8
-; CHECK-NEXT:    vmov.f32 s6, s1
-; CHECK-NEXT:    vmov.f32 s7, s9
-; CHECK-NEXT:    vstrw.32 q1, [r0]
-; CHECK-NEXT:    add sp, #32
-; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13, d14, d15}
+; CHECK-NEXT:    vcadd.i32 q6, q5, q4, #90
+; CHECK-NEXT:    vldrw.u32 q4, [r1]
+; CHECK-NEXT:    add r1, sp, #144
+; CHECK-NEXT:    vstrw.32 q6, [r0, #64]
+; CHECK-NEXT:    vcadd.i32 q5, q4, q3, #90
+; CHECK-NEXT:    vldrw.u32 q3, [r1]
+; CHECK-NEXT:    add r1, sp, #128
+; CHECK-NEXT:    vstrw.32 q5, [r0, #48]
+; CHECK-NEXT:    vcadd.i32 q4, q3, q2, #90
+; CHECK-NEXT:    vldrw.u32 q2, [r1]
+; CHECK-NEXT:    add r1, sp, #112
+; CHECK-NEXT:    vstrw.32 q4, [r0, #32]
+; CHECK-NEXT:    vcadd.i32 q3, q2, q1, #90
+; CHECK-NEXT:    vldrw.u32 q1, [r1]
+; CHECK-NEXT:    vstrw.32 q3, [r0, #16]
+; CHECK-NEXT:    vcadd.i32 q2, q1, q0, #90
+; CHECK-NEXT:    vstrw.32 q2, [r0]
+; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13}
 ; CHECK-NEXT:    bx lr
 entry:
   %a.real = shufflevector <32 x i32> %a, <32 x i32> zeroinitializer, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30>
   %a.imag = shufflevector <32 x i32> %a, <32 x i32> zeroinitializer, <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
   %b.real = shufflevector <32 x i32> %b, <32 x i32> zeroinitializer, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30>
   %b.imag = shufflevector <32 x i32> %b, <32 x i32> zeroinitializer, <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
   %0 = sub <16 x i32> %b.real, %a.imag
   %1 = add <16 x i32> %b.imag, %a.real
   %interleaved.vec = shufflevector <16 x i32> %0, <16 x i32> %1, <32 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23, i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
   ret <32 x i32> %interleaved.vec
 }

llvm/test/CodeGen/Thumb2/mve-complex-deinterleaving-i64-add.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc < %s --mattr=+mve -o - | FileCheck %s
+
+target triple = "thumbv8.1m.main-none-none-eabi"
+
+
+; Expected to not transform
+define arm_aapcs_vfpcc <2 x i64> @complex_add_v2i64(<2 x i64> %a, <2 x i64> %b) {
+; CHECK-LABEL: complex_add_v2i64:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    .save {r7, lr}
+; CHECK-NEXT:    push {r7, lr}
+; CHECK-NEXT:    vmov r0, r1, d0
+; CHECK-NEXT:    vmov r2, r3, d3
+; CHECK-NEXT:    adds.w lr, r2, r0
+; CHECK-NEXT:    adc.w r12, r3, r1
+; CHECK-NEXT:    vmov r2, r3, d1
+; CHECK-NEXT:    vmov r1, r0, d2
+; CHECK-NEXT:    subs r1, r1, r2
+; CHECK-NEXT:    vmov q0[2], q0[0], r1, lr
+; CHECK-NEXT:    sbcs r0, r3
+; CHECK-NEXT:    vmov q0[3], q0[1], r0, r12
+; CHECK-NEXT:    pop {r7, pc}
+entry:
+  %a.real = shufflevector <2 x i64> %a, <2 x i64> zeroinitializer, <1 x i32> <i32 0>
+  %a.imag = shufflevector <2 x i64> %a, <2 x i64> zeroinitializer, <1 x i32> <i32 1>
+  %b.real = shufflevector <2 x i64> %b, <2 x i64> zeroinitializer, <1 x i32> <i32 0>
+  %b.imag = shufflevector <2 x i64> %b, <2 x i64> zeroinitializer, <1 x i32> <i32 1>
+  %0 = sub <1 x i64> %b.real, %a.imag
+  %1 = add <1 x i64> %b.imag, %a.real
+  %interleaved.vec = shufflevector <1 x i64> %0, <1 x i64> %1, <2 x i32> <i32 0, i32 1>
+  ret <2 x i64> %interleaved.vec
+}
+
+; Expected to not transform
+define arm_aapcs_vfpcc <4 x i64> @complex_add_v4i64(<4 x i64> %a, <4 x i64> %b) {
+; CHECK-LABEL: complex_add_v4i64:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    .save {r7, lr}
+; CHECK-NEXT:    push {r7, lr}
+; CHECK-NEXT:    .vsave {d8, d9}
+; CHECK-NEXT:    vpush {d8, d9}
+; CHECK-NEXT:    vmov q4, q1
+; CHECK-NEXT:    vmov r2, r3, d7
+; CHECK-NEXT:    vmov r0, r1, d8
+; CHECK-NEXT:    adds.w lr, r2, r0
+; CHECK-NEXT:    adc.w r12, r3, r1
+; CHECK-NEXT:    vmov r2, r3, d0
+; CHECK-NEXT:    vmov r1, r0, d5
+; CHECK-NEXT:    adds r1, r1, r2
+; CHECK-NEXT:    adcs r0, r3
+; CHECK-NEXT:    vmov q1[2], q1[0], r1, lr
+; CHECK-NEXT:    vmov q1[3], q1[1], r0, r12
+; CHECK-NEXT:    vmov r0, r1, d9
+; CHECK-NEXT:    vmov r2, r3, d6
+; CHECK-NEXT:    subs.w lr, r2, r0
+; CHECK-NEXT:    sbc.w r12, r3, r1
+; CHECK-NEXT:    vmov r2, r3, d1
+; CHECK-NEXT:    vmov r1, r0, d4
+; CHECK-NEXT:    vmov.f32 s2, s4
+; CHECK-NEXT:    vmov.f32 s3, s5
+; CHECK-NEXT:    subs r1, r1, r2
+; CHECK-NEXT:    vmov q2[2], q2[0], r1, lr
+; CHECK-NEXT:    sbcs r0, r3
+; CHECK-NEXT:    vmov q2[3], q2[1], r0, r12
+; CHECK-NEXT:    vmov.f32 s0, s8
+; CHECK-NEXT:    vmov.f32 s4, s10
+; CHECK-NEXT:    vmov.f32 s1, s9
+; CHECK-NEXT:    vmov.f32 s5, s11
+; CHECK-NEXT:    vpop {d8, d9}
+; CHECK-NEXT:    pop {r7, pc}
+entry:
+  %a.real = shufflevector <4 x i64> %a, <4 x i64> zeroinitializer, <2 x i32> <i32 0, i32 2>
+  %a.imag = shufflevector <4 x i64> %a, <4 x i64> zeroinitializer, <2 x i32> <i32 1, i32 3>
+  %b.real = shufflevector <4 x i64> %b, <4 x i64> zeroinitializer, <2 x i32> <i32 0, i32 2>
+  %b.imag = shufflevector <4 x i64> %b, <4 x i64> zeroinitializer, <2 x i32> <i32 1, i32 3>
+  %0 = sub <2 x i64> %b.real, %a.imag
+  %1 = add <2 x i64> %b.imag, %a.real
+  %interleaved.vec = shufflevector <2 x i64> %0, <2 x i64> %1, <4 x i32> <i32 0, i32 2, i32 1, i32 3>
+  ret <4 x i64> %interleaved.vec
+}
+
+; Expected to not transform
+define arm_aapcs_vfpcc <8 x i64> @complex_add_v8i64(<8 x i64> %a, <8 x i64> %b) {
+; CHECK-LABEL: complex_add_v8i64:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    .save {r7, lr}
+; CHECK-NEXT:    push {r7, lr}
+; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13}
+; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13}
+; CHECK-NEXT:    add r2, sp, #72
+; CHECK-NEXT:    vmov q4, q1
+; CHECK-NEXT:    vldrw.u32 q5, [r2]
+; CHECK-NEXT:    vmov r0, r1, d8
+; CHECK-NEXT:    vmov r2, r3, d11
+; CHECK-NEXT:    adds.w lr, r2, r0
+; CHECK-NEXT:    adc.w r12, r3, r1
+; CHECK-NEXT:    add r1, sp, #56
+; CHECK-NEXT:    vldrw.u32 q6, [r1]
+; CHECK-NEXT:    vmov r2, r3, d0
+; CHECK-NEXT:    vmov r1, r0, d13
+; CHECK-NEXT:    adds r1, r1, r2
+; CHECK-NEXT:    adcs r0, r3
+; CHECK-NEXT:    vmov q1[2], q1[0], r1, lr
+; CHECK-NEXT:    vmov q1[3], q1[1], r0, r12
+; CHECK-NEXT:    vmov r0, r1, d9
+; CHECK-NEXT:    vmov r2, r3, d10
+; CHECK-NEXT:    subs.w lr, r2, r0
+; CHECK-NEXT:    sbc.w r12, r3, r1
+; CHECK-NEXT:    vmov r2, r3, d1
+; CHECK-NEXT:    vmov r1, r0, d12
+; CHECK-NEXT:    vmov.f32 s2, s4
+; CHECK-NEXT:    vmov.f32 s3, s5
+; CHECK-NEXT:    subs r1, r1, r2
+; CHECK-NEXT:    add r2, sp, #104
+; CHECK-NEXT:    vldrw.u32 q5, [r2]
+; CHECK-NEXT:    sbcs r0, r3
+; CHECK-NEXT:    vmov q4[2], q4[0], r1, lr
+; CHECK-NEXT:    vmov q4[3], q4[1], r0, r12
+; CHECK-NEXT:    vmov r0, r1, d6
+; CHECK-NEXT:    vmov r2, r3, d11
+; CHECK-NEXT:    vmov.f32 s0, s16
+; CHECK-NEXT:    vmov.f32 s4, s18
+; CHECK-NEXT:    vmov.f32 s1, s17
+; CHECK-NEXT:    vmov.f32 s5, s19
+; CHECK-NEXT:    adds.w lr, r2, r0
+; CHECK-NEXT:    adc.w r12, r3, r1
+; CHECK-NEXT:    add r1, sp, #88
+; CHECK-NEXT:    vldrw.u32 q6, [r1]
+; CHECK-NEXT:    vmov r2, r3, d4
+; CHECK-NEXT:    vmov r1, r0, d13
+; CHECK-NEXT:    adds r1, r1, r2
+; CHECK-NEXT:    adcs r0, r3
+; CHECK-NEXT:    vmov q4[2], q4[0], r1, lr
+; CHECK-NEXT:    vmov q4[3], q4[1], r0, r12
+; CHECK-NEXT:    vmov r0, r1, d7
+; CHECK-NEXT:    vmov r2, r3, d10
+; CHECK-NEXT:    subs.w lr, r2, r0
+; CHECK-NEXT:    sbc.w r12, r3, r1
+; CHECK-NEXT:    vmov r2, r3, d5
+; CHECK-NEXT:    vmov r1, r0, d12
+; CHECK-NEXT:    vmov.f32 s10, s16
+; CHECK-NEXT:    vmov.f32 s11, s17
+; CHECK-NEXT:    subs r1, r1, r2
+; CHECK-NEXT:    vmov q3[2], q3[0], r1, lr
+; CHECK-NEXT:    sbcs r0, r3
+; CHECK-NEXT:    vmov q3[3], q3[1], r0, r12
+; CHECK-NEXT:    vmov.f32 s16, s14
+; CHECK-NEXT:    vmov.f32 s8, s12
+; CHECK-NEXT:    vmov.f32 s17, s15
+; CHECK-NEXT:    vmov.f32 s9, s13
+; CHECK-NEXT:    vmov q3, q4
+; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13}
+; CHECK-NEXT:    pop {r7, pc}
+entry:
+  %a.real = shufflevector <8 x i64> %a, <8 x i64> zeroinitializer, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+  %a.imag = shufflevector <8 x i64> %a, <8 x i64> zeroinitializer, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+  %b.real = shufflevector <8 x i64> %b, <8 x i64> zeroinitializer, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+  %b.imag = shufflevector <8 x i64> %b, <8 x i64> zeroinitializer, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+  %0 = sub <4 x i64> %b.real, %a.imag
+  %1 = add <4 x i64> %b.imag, %a.real
+  %interleaved.vec = shufflevector <4 x i64> %0, <4 x i64> %1, <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
+  ret <8 x i64> %interleaved.vec
+}
+
+; Expected to not transform
+define arm_aapcs_vfpcc <16 x i64> @complex_add_v16i64(<16 x i64> %a, <16 x i64> %b) {
+; CHECK-LABEL: complex_add_v16i64:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    .save {r4, lr}
+; CHECK-NEXT:    push {r4, lr}
+; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13, d14, d15}
+; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13, d14, d15}
+; CHECK-NEXT:    .pad #16
+; CHECK-NEXT:    sub sp, #16
+; CHECK-NEXT:    add r1, sp, #136
+; CHECK-NEXT:    add r3, sp, #264
+; CHECK-NEXT:    vldrw.u32 q5, [r1]
+; CHECK-NEXT:    vldrw.u32 q6, [r3]
+; CHECK-NEXT:    vstrw.32 q0, [sp] @ 16-byte Spill
+; CHECK-NEXT:    vmov r1, r12, d10
+; CHECK-NEXT:    vmov r3, r2, d13
+; CHECK-NEXT:    adds.w lr, r3, r1
+; CHECK-NEXT:    add r3, sp, #120
+; CHECK-NEXT:    add r1, sp, #248
+; CHECK-NEXT:    vldrw.u32 q7, [r3]
+; CHECK-NEXT:    vldrw.u32 q0, [r1]
+; CHECK-NEXT:    adc.w r12, r12, r2
+; CHECK-NEXT:    vmov r3, r2, d14
+; CHECK-NEXT:    vmov r1, r4, d1
+; CHECK-NEXT:    adds r1, r1, r3
+; CHECK-NEXT:    vmov q4[2], q4[0], r1, lr
+; CHECK-NEXT:    adc.w r1, r4, r2
+; CHECK-NEXT:    vmov q4[3], q4[1], r1, r12
+; CHECK-NEXT:    vmov r1, r2, d11
+; CHECK-NEXT:    vmov r3, r4, d12
+; CHECK-NEXT:    vmov.f32 s22, s18
+; CHECK-NEXT:    vmov.f32 s23, s19
+; CHECK-NEXT:    subs.w lr, r3, r1
+; CHECK-NEXT:    sbc.w r12, r4, r2
+; CHECK-NEXT:    vmov r3, r4, d15
+; CHECK-NEXT:    vmov r2, r1, d0
+; CHECK-NEXT:    subs r2, r2, r3
+; CHECK-NEXT:    add r3, sp, #232
+; CHECK-NEXT:    vmov q0[2], q0[0], r2, lr
+; CHECK-NEXT:    sbcs r1, r4
+; CHECK-NEXT:    vmov q0[3], q0[1], r1, r12
+; CHECK-NEXT:    add r1, sp, #104
+; CHECK-NEXT:    vmov.f32 s20, s2
+; CHECK-NEXT:    vldrw.u32 q6, [r3]
+; CHECK-NEXT:    vmov.f32 s21, s3
+; CHECK-NEXT:    vstrw.32 q5, [r0, #112]
+; CHECK-NEXT:    vldrw.u32 q5, [r1]
+; CHECK-NEXT:    vmov r3, r4, d13
+; CHECK-NEXT:    vmov r1, r2, d10
+; CHECK-NEXT:    vmov.f32 s2, s16
+; CHECK-NEXT:    vmov.f32 s3, s17
+; CHECK-NEXT:    vstrw.32 q0, [r0, #96]
+; CHECK-NEXT:    adds.w lr, r3, r1
+; CHECK-NEXT:    add r3, sp, #88
+; CHECK-NEXT:    adc.w r12, r4, r2
+; CHECK-NEXT:    add r2, sp, #216
+; CHECK-NEXT:    vldrw.u32 q7, [r3]
+; CHECK-NEXT:    vldrw.u32 q0, [r2]
+; CHECK-NEXT:    vmov r3, r4, d14
+; CHECK-NEXT:    vmov r2, r1, d1
+; CHECK-NEXT:    adds r2, r2, r3
+; CHECK-NEXT:    adcs r1, r4
+; CHECK-NEXT:    vmov q4[2], q4[0], r2, lr
+; CHECK-NEXT:    vmov q4[3], q4[1], r1, r12
+; CHECK-NEXT:    vmov r1, r2, d11
+; CHECK-NEXT:    vmov r3, r4, d12
+; CHECK-NEXT:    vmov.f32 s22, s18
+; CHECK-NEXT:    vmov.f32 s23, s19
+; CHECK-NEXT:    subs.w lr, r3, r1
+; CHECK-NEXT:    sbc.w r12, r4, r2
+; CHECK-NEXT:    vmov r3, r4, d15
+; CHECK-NEXT:    vmov r2, r1, d0
+; CHECK-NEXT:    subs r2, r2, r3
+; CHECK-NEXT:    add r3, sp, #200
+; CHECK-NEXT:    vmov q0[2], q0[0], r2, lr
+; CHECK-NEXT:    sbcs r1, r4
+; CHECK-NEXT:    vmov q0[3], q0[1], r1, r12
+; CHECK-NEXT:    vmov r1, r2, d6
+; CHECK-NEXT:    vmov.f32 s20, s2
+; CHECK-NEXT:    vmov.f32 s21, s3
+; CHECK-NEXT:    vstrw.32 q5, [r0, #80]
+; CHECK-NEXT:    vldrw.u32 q5, [r3]
+; CHECK-NEXT:    vmov.f32 s2, s16
+; CHECK-NEXT:    vmov r3, r4, d11
+; CHECK-NEXT:    vmov.f32 s3, s17
+; CHECK-NEXT:    vstrw.32 q0, [r0, #64]
+; CHECK-NEXT:    adds.w lr, r3, r1
+; CHECK-NEXT:    adc.w r12, r4, r2
+; CHECK-NEXT:    add r2, sp, #184
+; CHECK-NEXT:    vldrw.u32 q0, [r2]
+; CHECK-NEXT:    vmov r3, r4, d4
+; CHECK-NEXT:    vmov r2, r1, d1
+; CHECK-NEXT:    adds r2, r2, r3
+; CHECK-NEXT:    adcs r1, r4
+; CHECK-NEXT:    vmov q4[2], q4[0], r2, lr
+; CHECK-NEXT:    vmov q4[3], q4[1], r1, r12
+; CHECK-NEXT:    vmov r1, r2, d7
+; CHECK-NEXT:    vmov r3, r4, d10
+; CHECK-NEXT:    subs.w lr, r3, r1
+; CHECK-NEXT:    sbc.w r12, r4, r2
+; CHECK-NEXT:    vmov r3, r4, d5
+; CHECK-NEXT:    vmov r2, r1, d0
+; CHECK-NEXT:    vmov.f32 s10, s18
+; CHECK-NEXT:    vmov.f32 s11, s19
+; CHECK-NEXT:    subs r2, r2, r3
+; CHECK-NEXT:    add r3, sp, #168
+; CHECK-NEXT:    vldrw.u32 q3, [r3]
+; CHECK-NEXT:    sbcs r1, r4
+; CHECK-NEXT:    vmov q0[2], q0[0], r2, lr
+; CHECK-NEXT:    vmov q0[3], q0[1], r1, r12
+; CHECK-NEXT:    vmov r1, r2, d2
+; CHECK-NEXT:    vmov r3, r4, d7
+; CHECK-NEXT:    vmov.f32 s8, s2
+; CHECK-NEXT:    vmov.f32 s2, s16
+; CHECK-NEXT:    vmov.f32 s9, s3
+; CHECK-NEXT:    vmov.f32 s3, s17
+; CHECK-NEXT:    vstrw.32 q2, [r0, #48]
+; CHECK-NEXT:    vstrw.32 q0, [r0, #32]
+; CHECK-NEXT:    vldrw.u32 q4, [sp] @ 16-byte Reload
+; CHECK-NEXT:    adds.w lr, r3, r1
+; CHECK-NEXT:    adc.w r12, r4, r2
+; CHECK-NEXT:    add r2, sp, #152
+; CHECK-NEXT:    vldrw.u32 q0, [r2]
+; CHECK-NEXT:    vmov r3, r4, d8
+; CHECK-NEXT:    vmov r2, r1, d1
+; CHECK-NEXT:    adds r2, r2, r3
+; CHECK-NEXT:    adcs r1, r4
+; CHECK-NEXT:    vmov q2[2], q2[0], r2, lr
+; CHECK-NEXT:    vmov q2[3], q2[1], r1, r12
+; CHECK-NEXT:    vmov r1, r2, d3
+; CHECK-NEXT:    vmov r3, r4, d6
+; CHECK-NEXT:    vmov.f32 s6, s10
+; CHECK-NEXT:    vmov.f32 s7, s11
+; CHECK-NEXT:    subs.w lr, r3, r1
+; CHECK-NEXT:    sbc.w r12, r4, r2
+; CHECK-NEXT:    vmov r3, r4, d9
+; CHECK-NEXT:    vmov r2, r1, d0
+; CHECK-NEXT:    subs r2, r2, r3
+; CHECK-NEXT:    vmov q0[2], q0[0], r2, lr
+; CHECK-NEXT:    sbcs r1, r4
+; CHECK-NEXT:    vmov q0[3], q0[1], r1, r12
+; CHECK-NEXT:    vmov.f32 s4, s2
+; CHECK-NEXT:    vmov.f32 s2, s8
+; CHECK-NEXT:    vmov.f32 s5, s3
+; CHECK-NEXT:    vmov.f32 s3, s9
+; CHECK-NEXT:    vstrw.32 q1, [r0, #16]
+; CHECK-NEXT:    vstrw.32 q0, [r0]
+; CHECK-NEXT:    add sp, #16
+; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13, d14, d15}
+; CHECK-NEXT:    pop {r4, pc}
+entry:
+  %a.real = shufflevector <16 x i64> %a, <16 x i64> zeroinitializer, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
+  %a.imag = shufflevector <16 x i64> %a, <16 x i64> zeroinitializer, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
+  %b.real = shufflevector <16 x i64> %b, <16 x i64> zeroinitializer, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
+  %b.imag = shufflevector <16 x i64> %b, <16 x i64> zeroinitializer, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
+  %0 = sub <8 x i64> %b.real, %a.imag
+  %1 = add <8 x i64> %b.imag, %a.real
+  %interleaved.vec = shufflevector <8 x i64> %0, <8 x i64> %1, <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
+  ret <16 x i64> %interleaved.vec
+}
+
+; Expected to not transform
+define arm_aapcs_vfpcc <32 x i64> @complex_add_v32i64(<32 x i64> %a, <32 x i64> %b) {
+; CHECK-LABEL: complex_add_v32i64:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    .save {r4, r5, r6, lr}
+; CHECK-NEXT:    push {r4, r5, r6, lr}
+; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13, d14, d15}
+; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13, d14, d15}
+; CHECK-NEXT:    .pad #16
+; CHECK-NEXT:    sub sp, #16
+; CHECK-NEXT:    add r1, sp, #272
+; CHECK-NEXT:    add r3, sp, #528
+; CHECK-NEXT:    vldrw.u32 q5, [r1]
+; CHECK-NEXT:    vldrw.u32 q6, [r3]
+; CHECK-NEXT:    vstrw.32 q0, [sp] @ 16-byte Spill
+; CHECK-NEXT:    vmov r1, r12, d10
+; CHECK-NEXT:    vmov r3, r2, d13
+; CHECK-NEXT:    adds.w lr, r3, r1
+; CHECK-NEXT:    add r3, sp, #256
+; CHECK-NEXT:    add r1, sp, #512
+; CHECK-NEXT:    vldrw.u32 q7, [r3]
+; CHECK-NEXT:    vldrw.u32 q0, [r1]
+; CHECK-NEXT:    adc.w r12, r12, r2
+; CHECK-NEXT:    vmov r3, r2, d14
+; CHECK-NEXT:    vmov r1, r4, d1
+; CHECK-NEXT:    adds r1, r1, r3
+; CHECK-NEXT:    vmov q4[2], q4[0], r1, lr
+; CHECK-NEXT:    adc.w r1, r4, r2
+; CHECK-NEXT:    vmov q4[3], q4[1], r1, r12
+; CHECK-NEXT:    vmov r1, r2, d11
+; CHECK-NEXT:    vmov r3, r4, d12
+; CHECK-NEXT:    vmov.f32 s22, s18
+; CHECK-NEXT:    vmov.f32 s23, s19
+; CHECK-NEXT:    subs.w lr, r3, r1
+; CHECK-NEXT:    sbc.w r12, r4, r2
+; CHECK-NEXT:    vmov r3, r4, d15
+; CHECK-NEXT:    vmov r2, r1, d0
+; CHECK-NEXT:    subs r2, r2, r3
+; CHECK-NEXT:    add r3, sp, #496
+; CHECK-NEXT:    vmov q0[2], q0[0], r2, lr
+; CHECK-NEXT:    sbcs r1, r4
+; CHECK-NEXT:    vmov q0[3], q0[1], r1, r12
+; CHECK-NEXT:    add r1, sp, #240
+; CHECK-NEXT:    vmov.f32 s20, s2
+; CHECK-NEXT:    vldrw.u32 q6, [r3]
+; CHECK-NEXT:    vmov.f32 s21, s3
+; CHECK-NEXT:    vstrw.32 q5, [r0, #240]
+; CHECK-NEXT:    vldrw.u32 q5, [r1]
+; CHECK-NEXT:    vmov r3, r4, d13
+; CHECK-NEXT:    vmov r1, r2, d10
+; CHECK-NEXT:    vmov.f32 s2, s16
+; CHECK-NEXT:    vmov.f32 s3, s17
+; CHECK-NEXT:    vstrw.32 q0, [r0, #224]
+; CHECK-NEXT:    adds.w lr, r3, r1
+; CHECK-NEXT:    add r3, sp, #224
+; CHECK-NEXT:    adc.w r12, r4, r2
+; CHECK-NEXT:    add r2, sp, #480
+; CHECK-NEXT:    vldrw.u32 q7, [r3]
+; CHECK-NEXT:    vldrw.u32 q0, [r2]
+; CHECK-NEXT:    vmov r3, r4, d14
+; CHECK-NEXT:    vmov r2, r1, d1
+; CHECK-NEXT:    adds r2, r2, r3
+; CHECK-NEXT:    adcs r1, r4
+; CHECK-NEXT:    vmov q4[2], q4[0], r2, lr
+; CHECK-NEXT:    vmov q4[3], q4[1], r1, r12
+; CHECK-NEXT:    vmov r1, r2, d11
+; CHECK-NEXT:    vmov r3, r4, d12
+; CHECK-NEXT:    vmov.f32 s22, s18
+; CHECK-NEXT:    vmov.f32 s23, s19
+; CHECK-NEXT:    subs.w lr, r3, r1
+; CHECK-NEXT:    sbc.w r12, r4, r2
+; CHECK-NEXT:    vmov r3, r4, d15
+; CHECK-NEXT:    vmov r2, r1, d0
+; CHECK-NEXT:    subs r2, r2, r3
+; CHECK-NEXT:    add r3, sp, #464
+; CHECK-NEXT:    vmov q0[2], q0[0], r2, lr
+; CHECK-NEXT:    sbcs r1, r4
+; CHECK-NEXT:    vmov q0[3], q0[1], r1, r12
+; CHECK-NEXT:    add r1, sp, #208
+; CHECK-NEXT:    vmov.f32 s20, s2
+; CHECK-NEXT:    vldrw.u32 q6, [r3]
+; CHECK-NEXT:    vmov.f32 s21, s3
+; CHECK-NEXT:    vstrw.32 q5, [r0, #208]
+; CHECK-NEXT:    vldrw.u32 q5, [r1]
+; CHECK-NEXT:    vmov r3, r4, d13
+; CHECK-NEXT:    vmov r1, r2, d10
+; CHECK-NEXT:    vmov.f32 s2, s16
+; CHECK-NEXT:    vmov.f32 s3, s17
+; CHECK-NEXT:    vstrw.32 q0, [r0, #192]
+; CHECK-NEXT:    adds.w lr, r3, r1
+; CHECK-NEXT:    add r3, sp, #192
+; CHECK-NEXT:    adc.w r12, r4, r2
+; CHECK-NEXT:    add r2, sp, #448
+; CHECK-NEXT:    vldrw.u32 q7, [r3]
+; CHECK-NEXT:    vldrw.u32 q0, [r2]
+; CHECK-NEXT:    vmov r3, r4, d14
+; CHECK-NEXT:    vmov r2, r1, d1
+; CHECK-NEXT:    adds r2, r2, r3
+; CHECK-NEXT:    adcs r1, r4
+; CHECK-NEXT:    vmov q4[2], q4[0], r2, lr
+; CHECK-NEXT:    vmov q4[3], q4[1], r1, r12
+; CHECK-NEXT:    vmov r1, r2, d11
+; CHECK-NEXT:    vmov r3, r4, d12
+; CHECK-NEXT:    vmov.f32 s22, s18
+; CHECK-NEXT:    vmov.f32 s23, s19
+; CHECK-NEXT:    subs.w lr, r3, r1
+; CHECK-NEXT:    sbc.w r12, r4, r2
+; CHECK-NEXT:    vmov r3, r4, d15
+; CHECK-NEXT:    vmov r2, r1, d0
+; CHECK-NEXT:    subs r2, r2, r3
+; CHECK-NEXT:    add r3, sp, #432
+; CHECK-NEXT:    vmov q0[2], q0[0], r2, lr
+; CHECK-NEXT:    sbcs r1, r4
+; CHECK-NEXT:    vmov q0[3], q0[1], r1, r12
+; CHECK-NEXT:    add r1, sp, #176
+; CHECK-NEXT:    vmov.f32 s20, s2
+; CHECK-NEXT:    vldrw.u32 q6, [r3]
+; CHECK-NEXT:    vmov.f32 s21, s3
+; CHECK-NEXT:    vstrw.32 q5, [r0, #176]
+; CHECK-NEXT:    vldrw.u32 q5, [r1]
+; CHECK-NEXT:    vmov r3, r4, d13
+; CHECK-NEXT:    vmov r1, r2, d10
+; CHECK-NEXT:    vmov.f32 s2, s16
+; CHECK-NEXT:    vmov.f32 s3, s17
+; CHECK-NEXT:    vstrw.32 q0, [r0, #160]
+; CHECK-NEXT:    adds.w lr, r3, r1
+; CHECK-NEXT:    add r3, sp, #160
+; CHECK-NEXT:    adc.w r12, r4, r2
+; CHECK-NEXT:    add r2, sp, #416
+; CHECK-NEXT:    vldrw.u32 q7, [r3]
+; CHECK-NEXT:    vldrw.u32 q0, [r2]
+; CHECK-NEXT:    vmov r3, r4, d14
+; CHECK-NEXT:    vmov r2, r1, d1
+; CHECK-NEXT:    adds r2, r2, r3
+; CHECK-NEXT:    adcs r1, r4
+; CHECK-NEXT:    vmov q4[2], q4[0], r2, lr
+; CHECK-NEXT:    vmov q4[3], q4[1], r1, r12
+; CHECK-NEXT:    vmov r1, r2, d11
+; CHECK-NEXT:    vmov r3, r4, d12
+; CHECK-NEXT:    vmov.f32 s22, s18
+; CHECK-NEXT:    vmov.f32 s23, s19
+; CHECK-NEXT:    subs.w lr, r3, r1
+; CHECK-NEXT:    sbc.w r12, r4, r2
+; CHECK-NEXT:    vmov r3, r4, d15
+; CHECK-NEXT:    vmov r2, r1, d0
+; CHECK-NEXT:    subs r2, r2, r3
+; CHECK-NEXT:    add r3, sp, #400
+; CHECK-NEXT:    vmov q0[2], q0[0], r2, lr
+; CHECK-NEXT:    sbcs r1, r4
+; CHECK-NEXT:    vmov q0[3], q0[1], r1, r12
+; CHECK-NEXT:    add r1, sp, #144
+; CHECK-NEXT:    vmov.f32 s20, s2
+; CHECK-NEXT:    vldrw.u32 q6, [r3]
+; CHECK-NEXT:    vmov.f32 s21, s3
+; CHECK-NEXT:    vstrw.32 q5, [r0, #144]
+; CHECK-NEXT:    vldrw.u32 q5, [r1]
+; CHECK-NEXT:    vmov r3, r4, d13
+; CHECK-NEXT:    vmov r1, r2, d10
+; CHECK-NEXT:    vmov.f32 s2, s16
+; CHECK-NEXT:    vmov.f32 s3, s17
+; CHECK-NEXT:    vstrw.32 q0, [r0, #128]
+; CHECK-NEXT:    adds.w lr, r3, r1
+; CHECK-NEXT:    add r3, sp, #128
+; CHECK-NEXT:    adc.w r12, r4, r2
+; CHECK-NEXT:    add r2, sp, #384
+; CHECK-NEXT:    vldrw.u32 q7, [r3]
+; CHECK-NEXT:    vldrw.u32 q0, [r2]
+; CHECK-NEXT:    vmov r3, r4, d14
+; CHECK-NEXT:    vmov r2, r1, d1
+; CHECK-NEXT:    adds r2, r2, r3
+; CHECK-NEXT:    adcs r1, r4
+; CHECK-NEXT:    vmov q4[2], q4[0], r2, lr
+; CHECK-NEXT:    vmov q4[3], q4[1], r1, r12
+; CHECK-NEXT:    vmov r1, r2, d11
+; CHECK-NEXT:    vmov r3, r4, d12
+; CHECK-NEXT:    vmov.f32 s22, s18
+; CHECK-NEXT:    vmov.f32 s23, s19
+; CHECK-NEXT:    subs.w lr, r3, r1
+; CHECK-NEXT:    sbc.w r12, r4, r2
+; CHECK-NEXT:    vmov r3, r4, d15
+; CHECK-NEXT:    vmov r2, r1, d0
+; CHECK-NEXT:    subs r2, r2, r3
+; CHECK-NEXT:    add r3, sp, #368
+; CHECK-NEXT:    vmov q0[2], q0[0], r2, lr
+; CHECK-NEXT:    sbcs r1, r4
+; CHECK-NEXT:    vmov q0[3], q0[1], r1, r12
+; CHECK-NEXT:    add r1, sp, #112
+; CHECK-NEXT:    vmov.f32 s20, s2
+; CHECK-NEXT:    vldrw.u32 q6, [r3]
+; CHECK-NEXT:    vmov.f32 s21, s3
+; CHECK-NEXT:    vstrw.32 q5, [r0, #112]
+; CHECK-NEXT:    vldrw.u32 q5, [r1]
+; CHECK-NEXT:    vmov r3, r4, d13
+; CHECK-NEXT:    vmov r1, r2, d10
+; CHECK-NEXT:    vmov.f32 s2, s16
+; CHECK-NEXT:    vmov.f32 s3, s17
+; CHECK-NEXT:    vstrw.32 q0, [r0, #96]
+; CHECK-NEXT:    adds.w lr, r3, r1
+; CHECK-NEXT:    add r3, sp, #96
+; CHECK-NEXT:    adc.w r12, r4, r2
+; CHECK-NEXT:    add r2, sp, #352
+; CHECK-NEXT:    vldrw.u32 q7, [r3]
+; CHECK-NEXT:    vldrw.u32 q0, [r2]
+; CHECK-NEXT:    vmov r3, r4, d14
+; CHECK-NEXT:    vmov r2, r1, d1
+; CHECK-NEXT:    adds r2, r2, r3
+; CHECK-NEXT:    adcs r1, r4
+; CHECK-NEXT:    vmov q4[2], q4[0], r2, lr
+; CHECK-NEXT:    vmov q4[3], q4[1], r1, r12
+; CHECK-NEXT:    vmov r1, r2, d11
+; CHECK-NEXT:    vmov r3, r4, d12
+; CHECK-NEXT:    vmov.f32 s22, s18
+; CHECK-NEXT:    vmov.f32 s23, s19
+; CHECK-NEXT:    subs.w lr, r3, r1
+; CHECK-NEXT:    sbc.w r12, r4, r2
+; CHECK-NEXT:    vmov r3, r4, d15
+; CHECK-NEXT:    vmov r2, r1, d0
+; CHECK-NEXT:    subs r2, r2, r3
+; CHECK-NEXT:    add r3, sp, #336
+; CHECK-NEXT:    vmov q0[2], q0[0], r2, lr
+; CHECK-NEXT:    sbcs r1, r4
+; CHECK-NEXT:    vmov q0[3], q0[1], r1, r12
+; CHECK-NEXT:    vmov r1, r2, d6
+; CHECK-NEXT:    vmov.f32 s20, s2
+; CHECK-NEXT:    vmov.f32 s21, s3
+; CHECK-NEXT:    vstrw.32 q5, [r0, #80]
+; CHECK-NEXT:    vldrw.u32 q5, [r3]
+; CHECK-NEXT:    vmov.f32 s2, s16
+; CHECK-NEXT:    vmov r3, r4, d11
+; CHECK-NEXT:    vmov.f32 s3, s17
+; CHECK-NEXT:    vstrw.32 q0, [r0, #64]
+; CHECK-NEXT:    adds.w lr, r3, r1
+; CHECK-NEXT:    adc.w r12, r4, r2
+; CHECK-NEXT:    add r2, sp, #320
+; CHECK-NEXT:    vldrw.u32 q0, [r2]
+; CHECK-NEXT:    vmov r3, r4, d4
+; CHECK-NEXT:    vmov r2, r1, d1
+; CHECK-NEXT:    adds r2, r2, r3
+; CHECK-NEXT:    adcs r1, r4
+; CHECK-NEXT:    vmov q4[2], q4[0], r2, lr
+; CHECK-NEXT:    vmov q4[3], q4[1], r1, r12
+; CHECK-NEXT:    vmov r1, r2, d7
+; CHECK-NEXT:    vmov r3, r4, d10
+; CHECK-NEXT:    subs.w lr, r3, r1
+; CHECK-NEXT:    sbc.w r12, r4, r2
+; CHECK-NEXT:    vmov r3, r4, d5
+; CHECK-NEXT:    vmov r2, r1, d0
+; CHECK-NEXT:    vmov.f32 s10, s18
+; CHECK-NEXT:    vmov.f32 s11, s19
+; CHECK-NEXT:    subs r2, r2, r3
+; CHECK-NEXT:    add r3, sp, #304
+; CHECK-NEXT:    vldrw.u32 q3, [r3]
+; CHECK-NEXT:    sbcs r1, r4
+; CHECK-NEXT:    vmov q0[2], q0[0], r2, lr
+; CHECK-NEXT:    vmov q0[3], q0[1], r1, r12
+; CHECK-NEXT:    vmov r1, r2, d2
+; CHECK-NEXT:    vmov r3, r4, d7
+; CHECK-NEXT:    vmov.f32 s8, s2
+; CHECK-NEXT:    vmov.f32 s2, s16
+; CHECK-NEXT:    vmov.f32 s9, s3
+; CHECK-NEXT:    vmov.f32 s3, s17
+; CHECK-NEXT:    vstrw.32 q2, [r0, #48]
+; CHECK-NEXT:    vstrw.32 q0, [r0, #32]
+; CHECK-NEXT:    vldrw.u32 q4, [sp] @ 16-byte Reload
+; CHECK-NEXT:    adds.w lr, r3, r1
+; CHECK-NEXT:    adc.w r12, r4, r2
+; CHECK-NEXT:    add r2, sp, #288
+; CHECK-NEXT:    vldrw.u32 q0, [r2]
+; CHECK-NEXT:    vmov r3, r4, d8
+; CHECK-NEXT:    vmov r2, r1, d1
+; CHECK-NEXT:    adds r2, r2, r3
+; CHECK-NEXT:    adcs r1, r4
+; CHECK-NEXT:    vmov q2[2], q2[0], r2, lr
+; CHECK-NEXT:    vmov q2[3], q2[1], r1, r12
+; CHECK-NEXT:    vmov r1, r2, d3
+; CHECK-NEXT:    vmov r3, r4, d6
+; CHECK-NEXT:    vmov.f32 s6, s10
+; CHECK-NEXT:    vmov.f32 s7, s11
+; CHECK-NEXT:    subs.w lr, r3, r1
+; CHECK-NEXT:    sbc.w r12, r4, r2
+; CHECK-NEXT:    vmov r3, r4, d9
+; CHECK-NEXT:    vmov r2, r1, d0
+; CHECK-NEXT:    subs r2, r2, r3
+; CHECK-NEXT:    vmov q0[2], q0[0], r2, lr
+; CHECK-NEXT:    sbcs r1, r4
+; CHECK-NEXT:    vmov q0[3], q0[1], r1, r12
+; CHECK-NEXT:    vmov.f32 s4, s2
+; CHECK-NEXT:    vmov.f32 s2, s8
+; CHECK-NEXT:    vmov.f32 s5, s3
+; CHECK-NEXT:    vmov.f32 s3, s9
+; CHECK-NEXT:    vstrw.32 q1, [r0, #16]
+; CHECK-NEXT:    vstrw.32 q0, [r0]
+; CHECK-NEXT:    add sp, #16
+; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13, d14, d15}
+; CHECK-NEXT:    pop {r4, r5, r6, pc}
+entry:
+  %a.real = shufflevector <32 x i64> %a, <32 x i64> zeroinitializer, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30>
+  %a.imag = shufflevector <32 x i64> %a, <32 x i64> zeroinitializer, <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
+  %b.real = shufflevector <32 x i64> %b, <32 x i64> zeroinitializer, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30>
+  %b.imag = shufflevector <32 x i64> %b, <32 x i64> zeroinitializer, <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
+  %0 = sub <16 x i64> %b.real, %a.imag
+  %1 = add <16 x i64> %b.imag, %a.real
+  %interleaved.vec = shufflevector <16 x i64> %0, <16 x i64> %1, <32 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23, i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
+  ret <32 x i64> %interleaved.vec
+}

llvm/test/CodeGen/Thumb2/mve-complex-deinterleaving-i8-add.ll

entry:
   %interleaved.vec = shufflevector <4 x i8> %0, <4 x i8> %1, <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
   ret <8 x i8> %interleaved.vec
 }
 
 ; Expected to transform
 define arm_aapcs_vfpcc <16 x i8> @complex_add_v16i8(<16 x i8> %a, <16 x i8> %b) {
 ; CHECK-LABEL: complex_add_v16i8:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vrev16.8 q2, q1
-; CHECK-NEXT:    vadd.i16 q2, q2, q0
-; CHECK-NEXT:    vrev16.8 q0, q0
-; CHECK-NEXT:    vsub.i16 q0, q1, q0
-; CHECK-NEXT:    vmovnt.i16 q0, q2
+; CHECK-NEXT:    vcadd.i8 q0, q1, q0, #90
 ; CHECK-NEXT:    bx lr
 entry:
   %a.real = shufflevector <16 x i8> %a, <16 x i8> zeroinitializer, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
   %a.imag = shufflevector <16 x i8> %a, <16 x i8> zeroinitializer, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
   %b.real = shufflevector <16 x i8> %b, <16 x i8> zeroinitializer, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
   %b.imag = shufflevector <16 x i8> %b, <16 x i8> zeroinitializer, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
   %0 = sub <8 x i8> %b.real, %a.imag
   %1 = add <8 x i8> %b.imag, %a.real
   %interleaved.vec = shufflevector <8 x i8> %0, <8 x i8> %1, <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
   ret <16 x i8> %interleaved.vec
 }
 
 ; Expected to transform
 define arm_aapcs_vfpcc <32 x i8> @complex_add_v32i8(<32 x i8> %a, <32 x i8> %b) {
 ; CHECK-LABEL: complex_add_v32i8:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13}
-; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13}
-; CHECK-NEXT:    vmov q4, q0
-; CHECK-NEXT:    vmov.u8 r0, q0[1]
-; CHECK-NEXT:    vmov.8 q0[0], r0
-; CHECK-NEXT:    vmov.u8 r0, q4[3]
-; CHECK-NEXT:    vmov.8 q0[1], r0
-; CHECK-NEXT:    vmov.u8 r0, q4[5]
-; CHECK-NEXT:    vmov.8 q0[2], r0
-; CHECK-NEXT:    vmov.u8 r0, q4[7]
-; CHECK-NEXT:    vmov.8 q0[3], r0
-; CHECK-NEXT:    vmov.u8 r0, q4[9]
-; CHECK-NEXT:    vmov.8 q0[4], r0
-; CHECK-NEXT:    vmov.u8 r0, q4[11]
-; CHECK-NEXT:    vmov.8 q0[5], r0
-; CHECK-NEXT:    vmov.u8 r0, q4[13]
-; CHECK-NEXT:    vmov.8 q0[6], r0
-; CHECK-NEXT:    vmov.u8 r0, q4[15]
-; CHECK-NEXT:    vmov.8 q0[7], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[1]
-; CHECK-NEXT:    vmov.8 q0[8], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[3]
-; CHECK-NEXT:    vmov.8 q0[9], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[5]
-; CHECK-NEXT:    vmov.8 q0[10], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[7]
-; CHECK-NEXT:    vmov.8 q0[11], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[9]
-; CHECK-NEXT:    vmov.8 q0[12], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[11]
-; CHECK-NEXT:    vmov.8 q0[13], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[13]
-; CHECK-NEXT:    vmov.8 q0[14], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[15]
-; CHECK-NEXT:    vmov.8 q0[15], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[0]
-; CHECK-NEXT:    vmov.8 q5[0], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[2]
-; CHECK-NEXT:    vmov.8 q5[1], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[4]
-; CHECK-NEXT:    vmov.8 q5[2], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[6]
-; CHECK-NEXT:    vmov.8 q5[3], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[8]
-; CHECK-NEXT:    vmov.8 q5[4], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[10]
-; CHECK-NEXT:    vmov.8 q5[5], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[12]
-; CHECK-NEXT:    vmov.8 q5[6], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[14]
-; CHECK-NEXT:    vmov.8 q5[7], r0
-; CHECK-NEXT:    vmov.u8 r0, q3[0]
-; CHECK-NEXT:    vmov.8 q5[8], r0
-; CHECK-NEXT:    vmov.u8 r0, q3[2]
-; CHECK-NEXT:    vmov.8 q5[9], r0
-; CHECK-NEXT:    vmov.u8 r0, q3[4]
-; CHECK-NEXT:    vmov.8 q5[10], r0
-; CHECK-NEXT:    vmov.u8 r0, q3[6]
-; CHECK-NEXT:    vmov.8 q5[11], r0
-; CHECK-NEXT:    vmov.u8 r0, q3[8]
-; CHECK-NEXT:    vmov.8 q5[12], r0
-; CHECK-NEXT:    vmov.u8 r0, q3[10]
-; CHECK-NEXT:    vmov.8 q5[13], r0
-; CHECK-NEXT:    vmov.u8 r0, q3[12]
-; CHECK-NEXT:    vmov.8 q5[14], r0
-; CHECK-NEXT:    vmov.u8 r0, q3[14]
-; CHECK-NEXT:    vmov.8 q5[15], r0
-; CHECK-NEXT:    vsub.i8 q5, q5, q0
-; CHECK-NEXT:    vmov.u8 r0, q5[0]
-; CHECK-NEXT:    vmov.8 q0[0], r0
-; CHECK-NEXT:    vmov.u8 r0, q4[0]
-; CHECK-NEXT:    vmov.8 q6[0], r0
-; CHECK-NEXT:    vmov.u8 r0, q4[2]
-; CHECK-NEXT:    vmov.8 q6[1], r0
-; CHECK-NEXT:    vmov.u8 r0, q4[4]
-; CHECK-NEXT:    vmov.8 q6[2], r0
-; CHECK-NEXT:    vmov.u8 r0, q4[6]
-; CHECK-NEXT:    vmov.8 q6[3], r0
-; CHECK-NEXT:    vmov.u8 r0, q4[8]
-; CHECK-NEXT:    vmov.8 q6[4], r0
-; CHECK-NEXT:    vmov.u8 r0, q4[10]
-; CHECK-NEXT:    vmov.8 q6[5], r0
-; CHECK-NEXT:    vmov.u8 r0, q4[12]
-; CHECK-NEXT:    vmov.8 q6[6], r0
-; CHECK-NEXT:    vmov.u8 r0, q4[14]
-; CHECK-NEXT:    vmov.8 q6[7], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[0]
-; CHECK-NEXT:    vmov.8 q6[8], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[2]
-; CHECK-NEXT:    vmov.8 q6[9], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[4]
-; CHECK-NEXT:    vmov.8 q6[10], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[6]
-; CHECK-NEXT:    vmov.8 q6[11], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[8]
-; CHECK-NEXT:    vmov.8 q6[12], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[10]
-; CHECK-NEXT:    vmov.8 q6[13], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[12]
-; CHECK-NEXT:    vmov.8 q6[14], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[14]
-; CHECK-NEXT:    vmov.8 q6[15], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[1]
-; CHECK-NEXT:    vmov.8 q1[0], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[3]
-; CHECK-NEXT:    vmov.8 q1[1], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[5]
-; CHECK-NEXT:    vmov.8 q1[2], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[7]
-; CHECK-NEXT:    vmov.8 q1[3], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[9]
-; CHECK-NEXT:    vmov.8 q1[4], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[11]
-; CHECK-NEXT:    vmov.8 q1[5], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[13]
-; CHECK-NEXT:    vmov.8 q1[6], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[15]
-; CHECK-NEXT:    vmov.8 q1[7], r0
-; CHECK-NEXT:    vmov.u8 r0, q3[1]
-; CHECK-NEXT:    vmov.8 q1[8], r0
-; CHECK-NEXT:    vmov.u8 r0, q3[3]
-; CHECK-NEXT:    vmov.8 q1[9], r0
-; CHECK-NEXT:    vmov.u8 r0, q3[5]
-; CHECK-NEXT:    vmov.8 q1[10], r0
-; CHECK-NEXT:    vmov.u8 r0, q3[7]
-; CHECK-NEXT:    vmov.8 q1[11], r0
-; CHECK-NEXT:    vmov.u8 r0, q3[9]
-; CHECK-NEXT:    vmov.8 q1[12], r0
-; CHECK-NEXT:    vmov.u8 r0, q3[11]
-; CHECK-NEXT:    vmov.8 q1[13], r0
-; CHECK-NEXT:    vmov.u8 r0, q3[13]
-; CHECK-NEXT:    vmov.8 q1[14], r0
-; CHECK-NEXT:    vmov.u8 r0, q3[15]
-; CHECK-NEXT:    vmov.8 q1[15], r0
-; CHECK-NEXT:    vadd.i8 q2, q1, q6
-; CHECK-NEXT:    vmov.u8 r0, q2[0]
-; CHECK-NEXT:    vmov.8 q0[1], r0
-; CHECK-NEXT:    vmov.u8 r0, q5[1]
-; CHECK-NEXT:    vmov.8 q0[2], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[1]
-; CHECK-NEXT:    vmov.8 q0[3], r0
-; CHECK-NEXT:    vmov.u8 r0, q5[2]
-; CHECK-NEXT:    vmov.8 q0[4], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[2]
-; CHECK-NEXT:    vmov.8 q0[5], r0
-; CHECK-NEXT:    vmov.u8 r0, q5[3]
-; CHECK-NEXT:    vmov.8 q0[6], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[3]
-; CHECK-NEXT:    vmov.8 q0[7], r0
-; CHECK-NEXT:    vmov.u8 r0, q5[4]
-; CHECK-NEXT:    vmov.8 q0[8], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[4]
-; CHECK-NEXT:    vmov.8 q0[9], r0
-; CHECK-NEXT:    vmov.u8 r0, q5[5]
-; CHECK-NEXT:    vmov.8 q0[10], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[5]
-; CHECK-NEXT:    vmov.8 q0[11], r0
-; CHECK-NEXT:    vmov.u8 r0, q5[6]
-; CHECK-NEXT:    vmov.8 q0[12], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[6]
-; CHECK-NEXT:    vmov.8 q0[13], r0
-; CHECK-NEXT:    vmov.u8 r0, q5[7]
-; CHECK-NEXT:    vmov.8 q0[14], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[7]
-; CHECK-NEXT:    vmov.8 q0[15], r0
-; CHECK-NEXT:    vmov.u8 r0, q5[8]
-; CHECK-NEXT:    vmov.8 q1[0], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[8]
-; CHECK-NEXT:    vmov.8 q1[1], r0
-; CHECK-NEXT:    vmov.u8 r0, q5[9]
-; CHECK-NEXT:    vmov.8 q1[2], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[9]
-; CHECK-NEXT:    vmov.8 q1[3], r0
-; CHECK-NEXT:    vmov.u8 r0, q5[10]
-; CHECK-NEXT:    vmov.8 q1[4], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[10]
-; CHECK-NEXT:    vmov.8 q1[5], r0
-; CHECK-NEXT:    vmov.u8 r0, q5[11]
-; CHECK-NEXT:    vmov.8 q1[6], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[11]
-; CHECK-NEXT:    vmov.8 q1[7], r0
-; CHECK-NEXT:    vmov.u8 r0, q5[12]
-; CHECK-NEXT:    vmov.8 q1[8], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[12]
-; CHECK-NEXT:    vmov.8 q1[9], r0
-; CHECK-NEXT:    vmov.u8 r0, q5[13]
-; CHECK-NEXT:    vmov.8 q1[10], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[13]
-; CHECK-NEXT:    vmov.8 q1[11], r0
-; CHECK-NEXT:    vmov.u8 r0, q5[14]
-; CHECK-NEXT:    vmov.8 q1[12], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[14]
-; CHECK-NEXT:    vmov.8 q1[13], r0
-; CHECK-NEXT:    vmov.u8 r0, q5[15]
-; CHECK-NEXT:    vmov.8 q1[14], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[15]
-; CHECK-NEXT:    vmov.8 q1[15], r0
-; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13}
+; CHECK-NEXT:    vcadd.i8 q0, q2, q0, #90
+; CHECK-NEXT:    vcadd.i8 q1, q3, q1, #90
 ; CHECK-NEXT:    bx lr
 entry:
   %a.real = shufflevector <32 x i8> %a, <32 x i8> zeroinitializer, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30>
   %a.imag = shufflevector <32 x i8> %a, <32 x i8> zeroinitializer, <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
   %b.real = shufflevector <32 x i8> %b, <32 x i8> zeroinitializer, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30>
   %b.imag = shufflevector <32 x i8> %b, <32 x i8> zeroinitializer, <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
   %0 = sub <16 x i8> %b.real, %a.imag
   %1 = add <16 x i8> %b.imag, %a.real
   %interleaved.vec = shufflevector <16 x i8> %0, <16 x i8> %1, <32 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23, i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
   ret <32 x i8> %interleaved.vec
 }