[AMDGPU] gfx11 CodeGen for new DPP instructions

[Joe_Nash]

Summary

Modifies the GCNDPPCombine pass to enable DPP formation for the new DPP
instruction in gfx11, namely VOP3 encoded instructions with DPP and VOPC
with DPP.

Depends on D128656

[rampitec]

llvm/lib/Target/AMDGPU/GCNDPPCombine.cpp

145

No need to pass HasVOP3DPP, you have ST available here and it is a trivial getter.

147

You could search for DPP4 opcode only is HasVOP3DPP.

[arsenm]

llvm/lib/Target/AMDGPU/GCNDPPCombine.cpp

146–149

Bad auto

[Joe_Nash]

Removed HasVOP3DPP parameter. explicit type instead of auto. Check ST->hasVOP3DPP before calling getDPPOp64

[rampitec]

llvm/lib/Target/AMDGPU/GCNDPPCombine.cpp

157

You can call getDPPOp64() here, after early exit for DPP32.

[Joe_Nash]

call getDPPOp64 after early exit for DPP32

llvm/lib/Target/AMDGPU/GCNDPPCombine.cpp

if (!hasNoImmOrEqual(MI, AMDGPU::OpName::src0_modifiers, 0, Mask) ||
       !hasNoImmOrEqual(MI, AMDGPU::OpName::clamp, 0) ||
       !hasNoImmOrEqual(MI, AMDGPU::OpName::omod, 0)) {
     LLVM_DEBUG(dbgs() << "  Inst has non-default modifiers\n");
     return false;
   }
   return true;
 }
 
 int GCNDPPCombine::getDPPOp(unsigned Op, bool IsShrinkable) const {
-  auto DPP32 = AMDGPU::getDPPOp32(Op);
+  int DPP32 = AMDGPU::getDPPOp32(Op);
+  int DPP64 = -1;
+  if (ST->hasVOP3DPP())
+    DPP64 = AMDGPU::getDPPOp64(Op);
   if (IsShrinkable) {
     assert(DPP32 == -1);
-    auto E32 = AMDGPU::getVOPe32(Op);
+    int E32 = AMDGPU::getVOPe32(Op);
     DPP32 = (E32 == -1) ? -1 : AMDGPU::getDPPOp32(E32);
   }
-  return (DPP32 == -1 || TII->pseudoToMCOpcode(DPP32) == -1) ? -1 : DPP32;
+  if (DPP32 != -1 && TII->pseudoToMCOpcode(DPP32) != -1)
+    return DPP32;
+  if (DPP64 != -1 && TII->pseudoToMCOpcode(DPP64) != -1)
+    return DPP64;
+  return -1;
 }
 
 // tracks the register operand definition and returns:
 //   1. immediate operand used to initialize the register if found
 //   2. nullptr if the register operand is undef
 //   3. the operand itself otherwise
 MachineOperand *GCNDPPCombine::getOldOpndValue(MachineOperand &OldOpnd) const {
   auto *Def = getVRegSubRegDef(getRegSubRegPair(OldOpnd), *MRI);
MachineInstr *GCNDPPCombine::createDPPInst(MachineInstr &OrigMI,
                                            MachineInstr &MovMI,
                                            RegSubRegPair CombOldVGPR,
                                            bool CombBCZ,
                                            bool IsShrinkable) const {
   assert(MovMI.getOpcode() == AMDGPU::V_MOV_B32_dpp ||
          MovMI.getOpcode() == AMDGPU::V_MOV_B64_dpp ||
          MovMI.getOpcode() == AMDGPU::V_MOV_B64_DPP_PSEUDO);
 
+  bool HasVOP3DPP = ST->hasVOP3DPP();
   auto OrigOp = OrigMI.getOpcode();
   auto DPPOp = getDPPOp(OrigOp, IsShrinkable);
   if (DPPOp == -1) {
     LLVM_DEBUG(dbgs() << "  failed: no DPP opcode\n");
     return nullptr;
   }
 
   auto DPPInst = BuildMI(*OrigMI.getParent(), OrigMI,
                          OrigMI.getDebugLoc(), TII->get(DPPOp))
     .setMIFlags(OrigMI.getFlags());
 
   bool Fail = false;
   do {
-    auto *Dst = TII->getNamedOperand(OrigMI, AMDGPU::OpName::vdst);
-    assert(Dst);
+    int NumOperands = 0;
+    if (auto *Dst = TII->getNamedOperand(OrigMI, AMDGPU::OpName::vdst)) {
       DPPInst.add(*Dst);
-    int NumOperands = 1;
+      ++NumOperands;
+    }
+    if (auto *SDst = TII->getNamedOperand(OrigMI, AMDGPU::OpName::sdst)) {
+      if (TII->isOperandLegal(*DPPInst.getInstr(), NumOperands, SDst)) {
+        DPPInst.add(*SDst);
+        ++NumOperands;
+      }
+      // If we shrunk a 64bit vop3b to 32bits, just ignore the sdst
+    }
 
     const int OldIdx = AMDGPU::getNamedOperandIdx(DPPOp, AMDGPU::OpName::old);
     if (OldIdx != -1) {
       assert(OldIdx == NumOperands);
       assert(isOfRegClass(
           CombOldVGPR,
           *MRI->getRegClass(
               TII->getNamedOperand(MovMI, AMDGPU::OpName::vdst)->getReg()),
if (OldIdx != -1) {
       Fail = true;
       break;
     }
 
     if (auto *Mod0 = TII->getNamedOperand(OrigMI,
                                           AMDGPU::OpName::src0_modifiers)) {
       assert(NumOperands == AMDGPU::getNamedOperandIdx(DPPOp,
                                           AMDGPU::OpName::src0_modifiers));
-      assert(0LL == (Mod0->getImm() & ~(SISrcMods::ABS | SISrcMods::NEG)));
+      assert(HasVOP3DPP ||
+             (0LL == (Mod0->getImm() & ~(SISrcMods::ABS | SISrcMods::NEG))));
       DPPInst.addImm(Mod0->getImm());
       ++NumOperands;
     } else if (AMDGPU::getNamedOperandIdx(DPPOp,
                    AMDGPU::OpName::src0_modifiers) != -1) {
       DPPInst.addImm(0);
       ++NumOperands;
     }
     auto *Src0 = TII->getNamedOperand(MovMI, AMDGPU::OpName::src0);
     assert(Src0);
     if (!TII->isOperandLegal(*DPPInst.getInstr(), NumOperands, Src0)) {
       LLVM_DEBUG(dbgs() << "  failed: src0 is illegal\n");
       Fail = true;
       break;
     }
     DPPInst.add(*Src0);
     DPPInst->getOperand(NumOperands).setIsKill(false);
     ++NumOperands;
 
     if (auto *Mod1 = TII->getNamedOperand(OrigMI,
                                           AMDGPU::OpName::src1_modifiers)) {
       assert(NumOperands == AMDGPU::getNamedOperandIdx(DPPOp,
                                           AMDGPU::OpName::src1_modifiers));
-      assert(0LL == (Mod1->getImm() & ~(SISrcMods::ABS | SISrcMods::NEG)));
+      assert(HasVOP3DPP ||
+             (0LL == (Mod1->getImm() & ~(SISrcMods::ABS | SISrcMods::NEG))));
       DPPInst.addImm(Mod1->getImm());
       ++NumOperands;
     } else if (AMDGPU::getNamedOperandIdx(DPPOp,
                    AMDGPU::OpName::src1_modifiers) != -1) {
       DPPInst.addImm(0);
       ++NumOperands;
     }
-    if (auto *Src1 = TII->getNamedOperand(OrigMI, AMDGPU::OpName::src1)) {
+    auto *Src1 = TII->getNamedOperand(OrigMI, AMDGPU::OpName::src1);
+    if (Src1) {
       if (!TII->isOperandLegal(*DPPInst.getInstr(), NumOperands, Src1)) {
         LLVM_DEBUG(dbgs() << "  failed: src1 is illegal\n");
         Fail = true;
         break;
       }
       DPPInst.add(*Src1);
       ++NumOperands;
     }
-
-    if (auto *Src2 = TII->getNamedOperand(OrigMI, AMDGPU::OpName::src2)) {
+    if (auto *Mod2 =
+            TII->getNamedOperand(OrigMI, AMDGPU::OpName::src2_modifiers)) {
+      assert(NumOperands ==
+             AMDGPU::getNamedOperandIdx(DPPOp, AMDGPU::OpName::src2_modifiers));
+      assert(HasVOP3DPP ||
+             (0LL == (Mod2->getImm() & ~(SISrcMods::ABS | SISrcMods::NEG))));
+      DPPInst.addImm(Mod2->getImm());
+      ++NumOperands;
+    }
+    auto *Src2 = TII->getNamedOperand(OrigMI, AMDGPU::OpName::src2);
+    if (Src2) {
       if (!TII->getNamedOperand(*DPPInst.getInstr(), AMDGPU::OpName::src2) ||
           !TII->isOperandLegal(*DPPInst.getInstr(), NumOperands, Src2)) {
         LLVM_DEBUG(dbgs() << "  failed: src2 is illegal\n");
         Fail = true;
         break;
       }
       DPPInst.add(*Src2);
+      ++NumOperands;
+    }
+    if (HasVOP3DPP) {
+      auto *ClampOpr = TII->getNamedOperand(OrigMI, AMDGPU::OpName::clamp);
+      if (ClampOpr &&
+          AMDGPU::getNamedOperandIdx(DPPOp, AMDGPU::OpName::clamp) != -1) {
+        DPPInst.addImm(ClampOpr->getImm());
+      }
+      auto *VdstInOpr = TII->getNamedOperand(OrigMI, AMDGPU::OpName::vdst_in);
+      if (VdstInOpr &&
+          AMDGPU::getNamedOperandIdx(DPPOp, AMDGPU::OpName::vdst_in) != -1) {
+        DPPInst.add(*VdstInOpr);
+      }
+      auto *OmodOpr = TII->getNamedOperand(OrigMI, AMDGPU::OpName::omod);
+      if (OmodOpr &&
+          AMDGPU::getNamedOperandIdx(DPPOp, AMDGPU::OpName::omod) != -1) {
+        DPPInst.addImm(OmodOpr->getImm());
+      }
+      // Validate OP_SEL has to be set to all 0 and OP_SEL_HI has to be set to
+      // all 1.
+      if (auto *OpSelOpr =
+              TII->getNamedOperand(OrigMI, AMDGPU::OpName::op_sel)) {
+        auto OpSel = OpSelOpr->getImm();
+        if (OpSel != 0) {
+          LLVM_DEBUG(dbgs() << "  failed: op_sel must be zero\n");
+          Fail = true;
+          break;
+        }
+        if (AMDGPU::getNamedOperandIdx(DPPOp, AMDGPU::OpName::op_sel) != -1)
+          DPPInst.addImm(OpSel);
+      }
+      if (auto *OpSelHiOpr =
+              TII->getNamedOperand(OrigMI, AMDGPU::OpName::op_sel_hi)) {
+        auto OpSelHi = OpSelHiOpr->getImm();
+        // Only vop3p has op_sel_hi, and all vop3p have 3 operands, so check
+        // the bitmask for 3 op_sel_hi bits set
+        assert(Src2 && "Expected vop3p with 3 operands");
+        if (OpSelHi != 7) {
+          LLVM_DEBUG(dbgs() << "  failed: op_sel_hi must be all set to one\n");
+          Fail = true;
+          break;
+        }
+        if (AMDGPU::getNamedOperandIdx(DPPOp, AMDGPU::OpName::op_sel_hi) != -1)
+          DPPInst.addImm(OpSelHi);
+      }
+      auto *NegOpr = TII->getNamedOperand(OrigMI, AMDGPU::OpName::neg_lo);
+      if (NegOpr &&
+          AMDGPU::getNamedOperandIdx(DPPOp, AMDGPU::OpName::neg_lo) != -1) {
+        DPPInst.addImm(NegOpr->getImm());
+      }
+      auto *NegHiOpr = TII->getNamedOperand(OrigMI, AMDGPU::OpName::neg_hi);
+      if (NegHiOpr &&
+          AMDGPU::getNamedOperandIdx(DPPOp, AMDGPU::OpName::neg_hi) != -1) {
+        DPPInst.addImm(NegHiOpr->getImm());
+      }
     }
-
     DPPInst.add(*TII->getNamedOperand(MovMI, AMDGPU::OpName::dpp_ctrl));
     DPPInst.add(*TII->getNamedOperand(MovMI, AMDGPU::OpName::row_mask));
     DPPInst.add(*TII->getNamedOperand(MovMI, AMDGPU::OpName::bank_mask));
     DPPInst.addImm(CombBCZ ? 1 : 0);
   } while (false);
 
   if (Fail) {
     DPPInst.getInstr()->eraseFromParent();
if (OrigOp == AMDGPU::REG_SEQUENCE) {
         if (Op.getSubReg() == FwdSubReg)
           Uses.push_back(&Op);
       }
       RegSeqWithOpNos[&OrigMI].push_back(OpNo);
       continue;
     }
 
     bool IsShrinkable = isShrinkable(OrigMI);
-    if (!(IsShrinkable || TII->isVOP1(OrigOp) || TII->isVOP2(OrigOp))) {
-      LLVM_DEBUG(dbgs() << "  failed: not VOP1/2/3\n");
+    if (!(IsShrinkable ||
+          ((TII->isVOP3P(OrigOp) || TII->isVOPC(OrigOp) ||
+            TII->isVOP3(OrigOp)) &&
+           ST->hasVOP3DPP()) ||
+          TII->isVOP1(OrigOp) || TII->isVOP2(OrigOp))) {
+      LLVM_DEBUG(dbgs() << "  failed: not VOP1/2/3/3P/C\n");
+      break;
+    }
+    if (OrigMI.modifiesRegister(AMDGPU::EXEC, ST->getRegisterInfo())) {
+      LLVM_DEBUG(dbgs() << "  failed: can't combine v_cmpx\n");
       break;
     }
 
     auto *Src0 = TII->getNamedOperand(OrigMI, AMDGPU::OpName::src0);
     auto *Src1 = TII->getNamedOperand(OrigMI, AMDGPU::OpName::src1);
     if (Use != Src0 && !(Use == Src1 && OrigMI.isCommutable())) { // [1]
       LLVM_DEBUG(dbgs() << "  failed: no suitable operands\n");
       break;
     }
 
+    auto *Src2 = TII->getNamedOperand(OrigMI, AMDGPU::OpName::src2);
     assert(Src0 && "Src1 without Src0?");
-    if (Src1 && Src1->isIdenticalTo(*Src0)) {
-      assert(Src1->isReg());
+    if ((Use == Src0 && ((Src1 && Src1->isIdenticalTo(*Src0)) ||
+                         (Src2 && Src2->isIdenticalTo(*Src0)))) ||
+        (Use == Src1 && (Src1->isIdenticalTo(*Src0) ||
+                         (Src2 && Src2->isIdenticalTo(*Src1))))) {
       LLVM_DEBUG(
           dbgs()
           << "  " << OrigMI
           << "  failed: DPP register is used more than once per instruction\n");
       break;
     }
 
     LLVM_DEBUG(dbgs() << "  combining: " << OrigMI);

llvm/lib/Target/AMDGPU/SIInstrInfo.h

namespace AMDGPU {
 
   LLVM_READONLY
   int getSDWAOp(uint16_t Opcode);
 
   LLVM_READONLY
   int getDPPOp32(uint16_t Opcode);
 
   LLVM_READONLY
+  int getDPPOp64(uint16_t Opcode);
+
+  LLVM_READONLY
   int getBasicFromSDWAOp(uint16_t Opcode);
 
   LLVM_READONLY
   int getCommuteRev(uint16_t Opcode);
 
   LLVM_READONLY
   int getCommuteOrig(uint16_t Opcode);
 

llvm/lib/Target/AMDGPU/SIInstrInfo.td

class getAsmVOP3DPPBase <int NumSrcArgs, bit HasDst, bit HasClamp,
                        bit HasOpSel, bit HasOMod, bit IsVOP3P,
                        bit HasModifiers, bit Src0HasMods,
                        bit Src1HasMods, bit Src2HasMods, ValueType DstVT = i32> {
   string dst = !if(HasDst,
                    !if(!eq(DstVT.Size, 1),
                        "$sdst",
                        "$vdst"),
                     ""); // use $sdst for VOPC
-  string isrc0 = !if(!eq(NumSrcArgs, 1), "$src0", "$src0,");
-  string isrc1 = !if(!eq(NumSrcArgs, 1), "",
-                     !if(!eq(NumSrcArgs, 2), " $src1",
-                                             " $src1,"));
-  string isrc2 = !if(!eq(NumSrcArgs, 3), " $src2", "");
+  string src0nomods = !if(!eq(NumSrcArgs, 1), "$src0", "$src0,");
+  string src1nomods = !if(!eq(NumSrcArgs, 1), "",
+  string src2nomods = !if(!eq(NumSrcArgs, 3), " $src2", "");
 
-  string fsrc0 = !if(!eq(NumSrcArgs, 1), "$src0_modifiers", "$src0_modifiers,");
-  string fsrc1 = !if(!eq(NumSrcArgs, 1), "",
-                     !if(!eq(NumSrcArgs, 2), " $src1_modifiers",
-                                             " $src1_modifiers,"));
-  string fsrc2 = !if(!eq(NumSrcArgs, 3), " $src2_modifiers", "");
+  string src0mods = !if(!eq(NumSrcArgs, 1), "$src0_modifiers", "$src0_modifiers,");
+  string src1mods = !if(!eq(NumSrcArgs, 1), "",
+  string src2mods = !if(!eq(NumSrcArgs, 3), " $src2_modifiers", "");
 
-  string src0 = !if(Src0HasMods, fsrc0, isrc0);
-  string src1 = !if(Src1HasMods, fsrc1, isrc1);
-  string src2 = !if(Src2HasMods, fsrc2, isrc2);
+  string src0 = !if(Src0HasMods, src0mods, src0nomods);
+  string src1 = !if(Src1HasMods, src1mods, src1nomods);
+  string src2 = !if(Src2HasMods, src2mods, src2nomods);
   string opsel = !if(HasOpSel, "$op_sel", "");
   string 3PMods = !if(IsVOP3P,
                       !if(HasOpSel, "$op_sel_hi", "")
                         #!if(HasModifiers, "$neg_lo$neg_hi", ""),
                       "");
   string clamp = !if(HasClamp, "$clamp", "");
   string omod = !if(HasOMod, "$omod", "");
 
field string AsmVOP3OpSel = getAsmVOP3OpSel<NumSrcArgs,
                                               HasSrc2FloatMods>.ret;
   field string AsmDPP = !if(HasExtDPP,
                             getAsmDPP<HasDst, NumSrcArgs, HasModifiers, DstVT>.ret, "");
   field string AsmDPP16 = getAsmDPP16<HasDst, NumSrcArgs, HasModifiers, DstVT>.ret;
   // DPP8 encoding has no fields for modifiers, and it is enforced by setting
   // the asm operand name via this HasModifiers flag
   field string AsmDPP8 = getAsmDPP8<HasDst, NumSrcArgs, 0 /*HasModifiers*/, DstVT>.ret;
   field string AsmVOP3DPPBase = getAsmVOP3DPPBase<NumSrcArgs, HasDst, HasClamp,
-   HasOpSel, HasOMod, IsVOP3P, HasModifiers, HasSrc0FloatMods, HasSrc1FloatMods,
-   HasSrc2FloatMods, DstVT >.ret;
+   HasOpSel, HasOMod, IsVOP3P, HasModifiers, HasModifiers, HasModifiers,
+   HasModifiers, DstVT>.ret;
   field string AsmVOP3DPP = getAsmVOP3DPP<AsmVOP3DPPBase>.ret;
   field string AsmVOP3DPP16 = getAsmVOP3DPP16<AsmVOP3DPPBase>.ret;
   field string AsmVOP3DPP8 = getAsmVOP3DPP8<AsmVOP3DPPBase>.ret;
   field string AsmSDWA = getAsmSDWA<HasDst, NumSrcArgs, DstVT>.ret;
   field string AsmSDWA9 = getAsmSDWA9<HasDst, HasSDWAOMod, NumSrcArgs, DstVT>.ret;
   field string AsmVOPDX = getAsmVOPDPart<NumSrcArgs, "X">.ret;
   field string AsmVOPDY = getAsmVOPDPart<NumSrcArgs, "Y">.ret;
   field string TieRegDPP = "$old";

 def getDPPOp32 : InstrMapping {
   let FilterClass = "VOP";
   let RowFields = ["OpName"];
   let ColFields = ["AsmVariantName"];
   let KeyCol = ["Default"];
   let ValueCols = [["DPP"]];
 }
 
+def getDPPOp64 : InstrMapping {
+  let FilterClass = "VOP";
+  let RowFields = ["OpName"];
+  let ColFields = ["AsmVariantName"];
+  let KeyCol = ["VOP3"];
+  let ValueCols = [["VOP3_DPP"]];
+}
+
 // Maps an commuted opcode to its original version
 def getCommuteOrig : InstrMapping {
   let FilterClass = "Commutable_REV";
   let RowFields = ["RevOp"];
   let ColFields = ["IsOrig"];
   let KeyCol = ["0"];
   let ValueCols = [["1"]];
 }

llvm/lib/Target/AMDGPU/VOPInstructions.td

 
 class VOP3_Profile_Base<VOPProfile P, VOP3Features Features = VOP3_REGULAR> : VOPProfile<P.ArgVT> {
 
   let HasClamp = !if(Features.HasClamp, 1, P.HasClamp);
   let HasOpSel = !if(Features.HasOpSel, 1, P.HasOpSel);
   let IsMAI    = !if(Features.IsMAI,    1, P.IsMAI);
   let IsPacked = !if(Features.IsPacked, 1, P.IsPacked);
 
-  let HasModifiers = !if(Features.IsMAI, 0, !or(Features.IsPacked, P.HasModifiers));
+  let HasModifiers =
+      !if (Features.IsMAI, 0,
+           !or(Features.IsPacked, Features.HasOpSel, P.HasModifiers));
 }
 
 class VOP3_Profile<VOPProfile P, VOP3Features Features = VOP3_REGULAR> : VOP3_Profile_Base<P, Features> {
   let IsSingle = 1;
 
 }
 
 // consistently gives instructions a _e64 suffix

llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.mov.dpp.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -global-isel -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=GFX8 %s
 ; RUN: llc -global-isel -march=amdgcn -mcpu=gfx1010 -verify-machineinstrs -show-mc-encoding < %s | FileCheck -check-prefix=GFX10 %s
+; RUN: llc -global-isel -march=amdgcn -mcpu=gfx1100 -amdgpu-enable-delay-alu=0 -verify-machineinstrs -show-mc-encoding < %s | FileCheck -check-prefix=GFX11 %s
 
 ; FIXME: Merge with DAG test
 
 define amdgpu_kernel void @dpp_test(i32 addrspace(1)* %out, i32 %in) {
 ; GFX8-LABEL: dpp_test:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_load_dword s2, s[0:1], 0x2c
 ; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24

 ; GFX10-NEXT:    s_load_dword s4, s[0:1], 0x2c ; encoding: [0x00,0x01,0x00,0xf4,0x2c,0x00,0x00,0xfa]
 ; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24 ; encoding: [0x80,0x00,0x04,0xf4,0x24,0x00,0x00,0xfa]
 ; GFX10-NEXT:    v_mov_b32_e32 v1, 0 ; encoding: [0x80,0x02,0x02,0x7e]
 ; GFX10-NEXT:    s_waitcnt lgkmcnt(0) ; encoding: [0x7f,0xc0,0x8c,0xbf]
 ; GFX10-NEXT:    v_mov_b32_e32 v0, s4 ; encoding: [0x04,0x02,0x00,0x7e]
 ; GFX10-NEXT:    v_mov_b32_dpp v0, v0 quad_perm:[1,0,0,0] row_mask:0x1 bank_mask:0x1 bound_ctrl:1 ; encoding: [0xfa,0x02,0x00,0x7e,0x00,0x01,0x08,0x11]
 ; GFX10-NEXT:    global_store_dword v1, v0, s[2:3] ; encoding: [0x00,0x80,0x70,0xdc,0x01,0x00,0x02,0x00]
 ; GFX10-NEXT:    s_endpgm ; encoding: [0x00,0x00,0x81,0xbf]
+;
+; GFX11-LABEL: dpp_test:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_clause 0x1 ; encoding: [0x01,0x00,0x85,0xbf]
+; GFX11-NEXT:    s_load_b32 s2, s[0:1], 0x2c ; encoding: [0x80,0x00,0x00,0xf4,0x2c,0x00,0x00,0xf8]
+; GFX11-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24 ; encoding: [0x00,0x00,0x04,0xf4,0x24,0x00,0x00,0xf8]
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0) ; encoding: [0x07,0xfc,0x89,0xbf]
+; GFX11-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s2 ; encoding: [0x80,0x00,0x10,0xca,0x02,0x00,0x00,0x01]
+; GFX11-NEXT:    v_mov_b32_dpp v0, v0 quad_perm:[1,0,0,0] row_mask:0x1 bank_mask:0x1 bound_ctrl:1 ; encoding: [0xfa,0x02,0x00,0x7e,0x00,0x01,0x08,0x11]
+; GFX11-NEXT:    global_store_b32 v1, v0, s[0:1] ; encoding: [0x00,0x00,0x6a,0xdc,0x01,0x00,0x00,0x00]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS) ; encoding: [0x03,0x00,0xb6,0xbf]
+; GFX11-NEXT:    s_endpgm ; encoding: [0x00,0x00,0xb0,0xbf]
   %tmp0 = call i32 @llvm.amdgcn.mov.dpp.i32(i32 %in, i32 1, i32 1, i32 1, i1 true) #0
   store i32 %tmp0, i32 addrspace(1)* %out
   ret void
 }
 define amdgpu_kernel void @mov_dpp64_test(i64 addrspace(1)* %out, i64 %in1) {
 ; GFX8-LABEL: mov_dpp64_test:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24

 ; GFX10-NEXT:    v_mov_b32_e32 v2, 0 ; encoding: [0x80,0x02,0x04,0x7e]
 ; GFX10-NEXT:    s_waitcnt lgkmcnt(0) ; encoding: [0x7f,0xc0,0x8c,0xbf]
 ; GFX10-NEXT:    v_mov_b32_e32 v0, s2 ; encoding: [0x02,0x02,0x00,0x7e]
 ; GFX10-NEXT:    v_mov_b32_e32 v1, s3 ; encoding: [0x03,0x02,0x02,0x7e]
 ; GFX10-NEXT:    v_mov_b32_dpp v0, v0 quad_perm:[1,0,0,0] row_mask:0x1 bank_mask:0x1 ; encoding: [0xfa,0x02,0x00,0x7e,0x00,0x01,0x00,0x11]
 ; GFX10-NEXT:    v_mov_b32_dpp v1, v1 quad_perm:[1,0,0,0] row_mask:0x1 bank_mask:0x1 ; encoding: [0xfa,0x02,0x02,0x7e,0x01,0x01,0x00,0x11]
 ; GFX10-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1] ; encoding: [0x00,0x80,0x74,0xdc,0x02,0x00,0x00,0x00]
 ; GFX10-NEXT:    s_endpgm ; encoding: [0x00,0x00,0x81,0xbf]
+;
+; GFX11-LABEL: mov_dpp64_test:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24 ; encoding: [0x00,0x00,0x08,0xf4,0x24,0x00,0x00,0xf8]
+; GFX11-NEXT:    v_mov_b32_e32 v2, 0 ; encoding: [0x80,0x02,0x04,0x7e]
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0) ; encoding: [0x07,0xfc,0x89,0xbf]
+; GFX11-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3 ; encoding: [0x02,0x00,0x10,0xca,0x03,0x00,0x00,0x00]
+; GFX11-NEXT:    v_mov_b32_dpp v0, v0 quad_perm:[1,0,0,0] row_mask:0x1 bank_mask:0x1 ; encoding: [0xfa,0x02,0x00,0x7e,0x00,0x01,0x00,0x11]
+; GFX11-NEXT:    v_mov_b32_dpp v1, v1 quad_perm:[1,0,0,0] row_mask:0x1 bank_mask:0x1 ; encoding: [0xfa,0x02,0x02,0x7e,0x01,0x01,0x00,0x11]
+; GFX11-NEXT:    global_store_b64 v2, v[0:1], s[0:1] ; encoding: [0x00,0x00,0x6e,0xdc,0x02,0x00,0x00,0x00]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS) ; encoding: [0x03,0x00,0xb6,0xbf]
+; GFX11-NEXT:    s_endpgm ; encoding: [0x00,0x00,0xb0,0xbf]
   %tmp0 = call i64 @llvm.amdgcn.mov.dpp.i64(i64 %in1, i32 1, i32 1, i32 1, i1 false) #0
   store i64 %tmp0, i64 addrspace(1)* %out
   ret void
 }
 
 declare i32 @llvm.amdgcn.mov.dpp.i32(i32, i32 immarg, i32 immarg, i32 immarg, i1 immarg) #0
 declare i64 @llvm.amdgcn.mov.dpp.i64(i64, i32 immarg, i32 immarg, i32 immarg, i1 immarg) #0
 
 attributes #0 = { convergent nounwind readnone }

llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.update.dpp.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -global-isel -march=amdgcn -mcpu=tonga -amdgpu-dpp-combine=false -verify-machineinstrs < %s | FileCheck -check-prefix=GFX8 %s
 ; RUN: llc -global-isel -march=amdgcn -mcpu=gfx1010 -amdgpu-dpp-combine=false -verify-machineinstrs < %s | FileCheck -check-prefix=GFX10 %s
+; RUN: llc -global-isel -march=amdgcn -mcpu=gfx1100 -amdgpu-enable-delay-alu=0 -amdgpu-dpp-combine=false -verify-machineinstrs < %s | FileCheck -check-prefix=GFX11 %s
 
 define amdgpu_kernel void @dpp_test(i32 addrspace(1)* %out, i32 %in1, i32 %in2) {
 ; GFX8-LABEL: dpp_test:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x2c
 ; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s2

 ; GFX10-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
 ; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX10-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-NEXT:    v_mov_b32_dpp v0, v1 quad_perm:[1,0,0,0] row_mask:0x1 bank_mask:0x1
 ; GFX10-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: dpp_test:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_clause 0x1
+; GFX11-NEXT:    s_load_b64 s[2:3], s[0:1], 0x2c
+; GFX11-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3
+; GFX11-NEXT:    v_mov_b32_dpp v0, v1 quad_perm:[1,0,0,0] row_mask:0x1 bank_mask:0x1
+; GFX11-NEXT:    v_mov_b32_e32 v1, 0
+; GFX11-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
   %tmp0 = call i32 @llvm.amdgcn.update.dpp.i32(i32 %in1, i32 %in2, i32 1, i32 1, i32 1, i1 false)
   store i32 %tmp0, i32 addrspace(1)* %out
   ret void
 }
 define amdgpu_kernel void @update_dpp64_test(i64 addrspace(1)* %arg, i64 %in1, i64 %in2) {
 ; GFX8-LABEL: update_dpp64_test:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24

 ; GFX10-NEXT:    global_load_dwordx2 v[0:1], v4, s[0:1]
 ; GFX10-NEXT:    v_mov_b32_e32 v2, s2
 ; GFX10-NEXT:    v_mov_b32_e32 v3, s3
 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    v_mov_b32_dpp v2, v0 quad_perm:[1,0,0,0] row_mask:0x1 bank_mask:0x1
 ; GFX10-NEXT:    v_mov_b32_dpp v3, v1 quad_perm:[1,0,0,0] row_mask:0x1 bank_mask:0x1
 ; GFX10-NEXT:    global_store_dwordx2 v4, v[2:3], s[0:1]
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: update_dpp64_test:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-NEXT:    v_lshlrev_b32_e32 v4, 3, v0
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3
+; GFX11-NEXT:    global_load_b64 v[0:1], v4, s[0:1]
+; GFX11-NEXT:    s_waitcnt vmcnt(0)
+; GFX11-NEXT:    v_mov_b32_dpp v2, v0 quad_perm:[1,0,0,0] row_mask:0x1 bank_mask:0x1
+; GFX11-NEXT:    v_mov_b32_dpp v3, v1 quad_perm:[1,0,0,0] row_mask:0x1 bank_mask:0x1
+; GFX11-NEXT:    global_store_b64 v4, v[2:3], s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
   %id = tail call i32 @llvm.amdgcn.workitem.id.x()
   %gep = getelementptr inbounds i64, i64 addrspace(1)* %arg, i32 %id
   %load = load i64, i64 addrspace(1)* %gep
   %tmp0 = call i64 @llvm.amdgcn.update.dpp.i64(i64 %in1, i64 %load, i32 1, i32 1, i32 1, i1 false) #1
   store i64 %tmp0, i64 addrspace(1)* %gep
   ret void
 }
 
 declare i32 @llvm.amdgcn.workitem.id.x() #0
 declare i32 @llvm.amdgcn.update.dpp.i32(i32, i32, i32 immarg, i32 immarg, i32 immarg, i1 immarg) #1
 declare i64 @llvm.amdgcn.update.dpp.i64(i64, i64, i32 immarg, i32 immarg, i32 immarg, i1 immarg) #1
 
 attributes #0 = { nounwind readnone speculatable }
 attributes #1 = { convergent nounwind readnone }

llvm/test/CodeGen/AMDGPU/dpp64_combine.ll

 ; RUN: llc -march=amdgcn -mcpu=gfx90a -verify-machineinstrs < %s | FileCheck %s -check-prefixes=GCN,DPP64,GFX90A
 ; RUN: llc -march=amdgcn -mcpu=gfx940 -verify-machineinstrs < %s | FileCheck %s -check-prefixes=GCN,DPP64,DPPMOV64
-; RUN: llc -march=amdgcn -mcpu=gfx1010 -verify-machineinstrs < %s | FileCheck %s -check-prefixes=GCN,DPP32,GFX10
+; RUN: llc -march=amdgcn -mcpu=gfx1010 -verify-machineinstrs < %s | FileCheck %s -check-prefixes=GCN,DPP32,GFX10PLUS
+; RUN: llc -march=amdgcn -mcpu=gfx1100 -verify-machineinstrs < %s | FileCheck %s -check-prefixes=GCN,DPP32,GFX10PLUS
 
 ; GCN-LABEL: {{^}}dpp64_ceil:
-; GCN:           global_load_dwordx2 [[V:v\[[0-9:]+\]]],
+; GCN:           global_load_{{dwordx2|b64}} [[V:v\[[0-9:]+\]]],
 ; DPP64:         v_ceil_f64_dpp [[V]], [[V]] row_newbcast:1 row_mask:0xf bank_mask:0xf bound_ctrl:1{{$}}
 ; DPP32-COUNT-2: v_mov_b32_dpp v{{[0-9]+}}, v{{[0-9]+}} row_share:1 row_mask:0xf bank_mask:0xf bound_ctrl:1{{$}}
 define amdgpu_kernel void @dpp64_ceil(i64 addrspace(1)* %arg, i64 %in1) {
   %id = tail call i32 @llvm.amdgcn.workitem.id.x()
   %gep = getelementptr inbounds i64, i64 addrspace(1)* %arg, i32 %id
   %load = load i64, i64 addrspace(1)* %gep
   %tmp0 = call i64 @llvm.amdgcn.update.dpp.i64(i64 %in1, i64 %load, i32 337, i32 15, i32 15, i1 1) #0
   %tmp1 = bitcast i64 %tmp0 to double
   %round = tail call double @llvm.ceil.f64(double %tmp1)
   %tmp2 = bitcast double %round to i64
   store i64 %tmp2, i64 addrspace(1)* %gep
   ret void
 }
 
 ; GCN-LABEL: {{^}}dpp64_rcp:
-; GCN:           global_load_dwordx2 [[V:v\[[0-9:]+\]]],
+; GCN:           global_load_{{dwordx2|b64}} [[V:v\[[0-9:]+\]]],
 ; DPP64:         v_rcp_f64_dpp [[V]], [[V]] row_newbcast:1 row_mask:0xf bank_mask:0xf bound_ctrl:1{{$}}
 ; DPP32-COUNT-2: v_mov_b32_dpp v{{[0-9]+}}, v{{[0-9]+}} row_share:1 row_mask:0xf bank_mask:0xf bound_ctrl:1{{$}}
 define amdgpu_kernel void @dpp64_rcp(i64 addrspace(1)* %arg, i64 %in1) {
   %id = tail call i32 @llvm.amdgcn.workitem.id.x()
   %gep = getelementptr inbounds i64, i64 addrspace(1)* %arg, i32 %id
   %load = load i64, i64 addrspace(1)* %gep
   %tmp0 = call i64 @llvm.amdgcn.update.dpp.i64(i64 %in1, i64 %load, i32 337, i32 15, i32 15, i1 1) #0
   %tmp1 = bitcast i64 %tmp0 to double
define amdgpu_kernel void @dpp64_rcp_unsupported_ctl(i64 addrspace(1)* %arg, i64 %in1) {
   %tmp1 = bitcast i64 %tmp0 to double
   %rcp = fdiv fast double 1.0, %tmp1
   %tmp2 = bitcast double %rcp to i64
   store i64 %tmp2, i64 addrspace(1)* %gep
   ret void
 }
 
 ; GCN-LABEL: {{^}}dpp64_div:
-; GCN:            global_load_dwordx2 [[V:v\[[0-9:]+\]]],
-; DPPMOV64:       v_mov_b64_dpp v[{{[0-9:]+}}], [[V]] row_newbcast:1 row_mask:0xf bank_mask:0xf bound_ctrl:1{{$}}
-; GFX90A-COUNT-2: v_mov_b32_dpp v{{[0-9]+}}, v{{[0-9]+}} row_newbcast:1 row_mask:0xf bank_mask:0xf bound_ctrl:1{{$}}
-; GFX10-COUNT-2:  v_mov_b32_dpp v{{[0-9]+}}, v{{[0-9]+}} row_share:1 row_mask:0xf bank_mask:0xf bound_ctrl:1{{$}}
-; GCN:            v_div_scale_f64
-; GCN:            v_rcp_f64_e32
+; GCN:               global_load_{{dwordx2|b64}} [[V:v\[[0-9:]+\]]],
+; DPPMOV64:          v_mov_b64_dpp v[{{[0-9:]+}}], [[V]] row_newbcast:1 row_mask:0xf bank_mask:0xf bound_ctrl:1{{$}}
+; GFX90A-COUNT-2:    v_mov_b32_dpp v{{[0-9]+}}, v{{[0-9]+}} row_newbcast:1 row_mask:0xf bank_mask:0xf bound_ctrl:1{{$}}
+; GFX10PLUS-COUNT-2: v_mov_b32_dpp v{{[0-9]+}}, v{{[0-9]+}} row_share:1 row_mask:0xf bank_mask:0xf bound_ctrl:1{{$}}
+; GCN:               v_div_scale_f64
+; GCN:               v_rcp_f64_e32
 define amdgpu_kernel void @dpp64_div(i64 addrspace(1)* %arg, i64 %in1) {
   %id = tail call i32 @llvm.amdgcn.workitem.id.x()
   %gep = getelementptr inbounds i64, i64 addrspace(1)* %arg, i32 %id
   %load = load i64, i64 addrspace(1)* %gep
   %tmp0 = call i64 @llvm.amdgcn.update.dpp.i64(i64 %in1, i64 %load, i32 337, i32 15, i32 15, i1 1) #0
   %tmp1 = bitcast i64 %tmp0 to double
   %rcp = fdiv double 15.0, %tmp1
   %tmp2 = bitcast double %rcp to i64

llvm/test/CodeGen/AMDGPU/dpp_combine.ll

 ; RUN: llc -march=amdgcn -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck %s -check-prefix=GCN
 ; RUN: llc -march=amdgcn -mcpu=gfx1010 -verify-machineinstrs < %s | FileCheck %s -check-prefix=GCN
+; RUN: llc -march=amdgcn -mcpu=gfx1100 -verify-machineinstrs < %s | FileCheck %s -check-prefix=GCN
 
 ; GCN-LABEL: {{^}}dpp_add:
-; GCN: global_load_dword [[V:v[0-9]+]],
+; GCN: global_load_{{dword|b32}} [[V:v[0-9]+]],
 ; GCN: v_add_{{(nc_)?}}u32_dpp [[V]], [[V]], [[V]] quad_perm:[1,0,0,0] row_mask:0xf bank_mask:0xf bound_ctrl:1{{$}}
 define amdgpu_kernel void @dpp_add(i32 addrspace(1)* %arg) {
   %id = tail call i32 @llvm.amdgcn.workitem.id.x()
   %gep = getelementptr inbounds i32, i32 addrspace(1)* %arg, i32 %id
   %load = load i32, i32 addrspace(1)* %gep
   %tmp0 = call i32 @llvm.amdgcn.update.dpp.i32(i32 %load, i32 %load, i32 1, i32 15, i32 15, i1 1) #0
   %add = add i32 %tmp0, %load
   store i32 %add, i32 addrspace(1)* %gep
   ret void
 }
 
 ; GCN-LABEL: {{^}}dpp_ceil:
-; GCN: global_load_dword [[V:v[0-9]+]],
+; GCN: global_load_{{dword|b32}} [[V:v[0-9]+]],
 ; GCN: v_ceil_f32_dpp [[V]], [[V]] quad_perm:[1,0,0,0] row_mask:0xf bank_mask:0xf bound_ctrl:1{{$}}
 define amdgpu_kernel void @dpp_ceil(i32 addrspace(1)* %arg) {
   %id = tail call i32 @llvm.amdgcn.workitem.id.x()
   %gep = getelementptr inbounds i32, i32 addrspace(1)* %arg, i32 %id
   %load = load i32, i32 addrspace(1)* %gep
   %tmp0 = call i32 @llvm.amdgcn.update.dpp.i32(i32 %load, i32 %load, i32 1, i32 15, i32 15, i1 1) #0
   %tmp1 = bitcast i32 %tmp0 to float
   %round = tail call float @llvm.ceil.f32(float %tmp1)
   %tmp2 = bitcast float %round to i32
   store i32 %tmp2, i32 addrspace(1)* %gep
   ret void
 }
 
 ; GCN-LABEL: {{^}}dpp_fadd:
-; GCN: global_load_dword [[V:v[0-9]+]],
+; GCN: global_load_{{dword|b32}} [[V:v[0-9]+]],
 ; GCN: v_add_f32_dpp [[V]], [[V]], [[V]] quad_perm:[1,0,0,0] row_mask:0xf bank_mask:0xf bound_ctrl:1{{$}}
 define amdgpu_kernel void @dpp_fadd(i32 addrspace(1)* %arg) {
   %id = tail call i32 @llvm.amdgcn.workitem.id.x()
   %gep = getelementptr inbounds i32, i32 addrspace(1)* %arg, i32 %id
   %load = load i32, i32 addrspace(1)* %gep
   %tmp0 = call i32 @llvm.amdgcn.update.dpp.i32(i32 %load, i32 %load, i32 1, i32 15, i32 15, i1 1) #0
   %tmp1 = bitcast i32 %tmp0 to float
   %t = bitcast i32 %load to float

llvm/test/CodeGen/AMDGPU/dpp_combine_gfx11.mir

+# RUN: llc -march=amdgcn -mcpu=gfx1100 -run-pass=gcn-dpp-combine -verify-machineinstrs -o - %s | FileCheck %s -check-prefix=GCN
+
+---
+
+# GCN-label: name: vop3
+# GCN: %6:vgpr_32, %7:sreg_32_xm0_xexec = V_SUBBREV_U32_e64_dpp %3, %0, %1, %5, 1, 1, 15, 15, 1, implicit $exec
+# GCN: %8:vgpr_32 = V_CVT_PK_U8_F32_e64_dpp %3, 4, %0, 2, %2, 2, %1, 1, 1, 15, 15, 1, implicit $mode, implicit $exec
+# GCN: %10:vgpr_32 = V_MED3_F32_e64 0, %9, 0, %0, 0, 12345678, 0, 0, implicit $mode, implicit $exec
+# GCN: %12:vgpr_32 = V_MED3_F32_e64 0, %11, 0, 2, 0, %7, 0, 0, implicit $mode, implicit $exec
+name: vop3
+tracksRegLiveness: true
+body:             |
+  bb.0:
+    liveins: $vgpr0, $vgpr1, $vgpr2
+
+    %0:vgpr_32 = COPY $vgpr0
+    %1:vgpr_32 = COPY $vgpr1
+    %2:vgpr_32 = COPY $vgpr2
+    %3:vgpr_32 = IMPLICIT_DEF
+    %4:vgpr_32 = V_MOV_B32_dpp %3, %0, 1, 15, 15, 1, implicit $exec
+
+    %5:sreg_32_xm0_xexec = IMPLICIT_DEF
+    %6:vgpr_32, %7:sreg_32_xm0_xexec = V_SUBBREV_U32_e64 %4, %1, %5, 1, implicit $exec
+
+    %8:vgpr_32 = V_CVT_PK_U8_F32_e64 4, %4, 2, %2, 2, %1, 1, implicit $mode, implicit $exec
+
+    ; should not be combined because src2 literal is illegal
+    %9:vgpr_32 = V_MOV_B32_dpp %3, %1, 1, 15, 15, 1, implicit $exec
+    %10:vgpr_32 = V_MED3_F32_e64 0, %9, 0, %0, 0, 12345678, 0, 0, implicit $mode, implicit $exec
+
+    ; should not be combined because src1 imm is illegal
+    %11:vgpr_32 = V_MOV_B32_dpp %3, %1, 1, 15, 15, 1, implicit $exec
+    %12:vgpr_32 = V_MED3_F32_e64 0, %11, 0, 2, 0, %7, 0, 0, implicit $mode, implicit $exec
+...
+
+# Regression test for src_modifiers on base u16 opcode
+# GCN-label: name: vop3_u16
+# GCN: %5:vgpr_32 = V_ADD_NC_U16_e64_dpp %3, 0, %1, 0, %3, 0, 0, 1, 15, 15, 1, implicit $exec
+# GCN: %7:vgpr_32 = V_ADD_NC_U16_e64_dpp %3, 4, %5, 8, %5, 0, 0, 1, 15, 15, 1, implicit $exec
+name: vop3_u16
+tracksRegLiveness: true
+body:             |
+  bb.0:
+    liveins: $vgpr0, $vgpr1, $vgpr2
+
+    %0:vgpr_32 = COPY $vgpr0
+    %1:vgpr_32 = COPY $vgpr1
+    %2:vgpr_32 = COPY $vgpr2
+    %3:vgpr_32 = IMPLICIT_DEF
+    %4:vgpr_32 = V_MOV_B32_dpp %3, %1, 1, 15, 15, 1, implicit $exec
+    %5:vgpr_32 = V_ADD_NC_U16_e64 0, %4, 0, %3, 0, 0, implicit $exec
+    %6:vgpr_32 = V_MOV_B32_dpp %3, %5, 1, 15, 15, 1, implicit $exec
+    %7:vgpr_32 = V_ADD_NC_U16_e64 4, %6, 8, %5, 0, 0, implicit $exec
+...
+
+name: vop3p
+tracksRegLiveness: true
+body:             |
+  bb.0:
+    liveins: $vgpr0, $vgpr1, $vgpr2
+
+    ; GCN-LABEL: name: vop3p
+    ; GCN: liveins: $vgpr0, $vgpr1, $vgpr2
+    ; GCN: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+    ; GCN: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr1
+    ; GCN: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; GCN: [[DEF:%[0-9]+]]:vgpr_32 = IMPLICIT_DEF
+    ; GCN: [[V_MOV_B32_dpp:%[0-9]+]]:vgpr_32 = V_MOV_B32_dpp [[DEF]], [[COPY1]], 1, 15, 15, 1, implicit $exec
+    ; GCN: [[V_DOT2_F32_F16_:%[0-9]+]]:vgpr_32 = V_DOT2_F32_F16 0, [[V_MOV_B32_dpp]], 0, [[COPY]], 0, [[COPY2]], 0, 5, 0, 0, 0, implicit $mode, implicit $exec
+    ; GCN: [[V_MOV_B32_dpp1:%[0-9]+]]:vgpr_32 = V_MOV_B32_dpp [[DEF]], [[COPY1]], 1, 15, 15, 1, implicit $exec
+    ; GCN: [[V_DOT2_F32_F16_1:%[0-9]+]]:vgpr_32 = V_DOT2_F32_F16 0, [[V_MOV_B32_dpp1]], 0, [[COPY]], 0, [[COPY2]], 0, 0, 4, 0, 0, implicit $mode, implicit $exec
+    ; GCN: [[V_DOT2_F32_F16_dpp:%[0-9]+]]:vgpr_32 = V_DOT2_F32_F16_dpp [[DEF]], 10, [[COPY1]], 8, [[COPY]], 13, [[COPY2]], 1, 0, 7, 4, 5, 1, 15, 15, 1, implicit $mode, implicit $exec
+    ; GCN: [[V_FMA_MIX_F32_dpp:%[0-9]+]]:vgpr_32 = V_FMA_MIX_F32_dpp [[DEF]], 8, [[COPY1]], 8, [[COPY]], 8, [[COPY2]], 1, 0, 7, 1, 15, 15, 1, implicit $mode, implicit $exec
+    ; GCN: [[V_FMA_MIXLO_F16_dpp:%[0-9]+]]:vgpr_32 = V_FMA_MIXLO_F16_dpp [[DEF]], 8, [[COPY1]], 8, [[COPY]], 8, [[COPY2]], 0, [[COPY2]], 0, 7, 1, 15, 15, 1, implicit $mode, implicit $exec
+    ; GCN: [[V_FMA_MIXHI_F16_dpp:%[0-9]+]]:vgpr_32 = V_FMA_MIXHI_F16_dpp [[DEF]], 8, [[COPY1]], 8, [[COPY]], 8, [[COPY2]], 1, [[COPY]], 0, 7, 1, 15, 15, 1, implicit $mode, implicit $exec
+    %0:vgpr_32 = COPY $vgpr0
+    %1:vgpr_32 = COPY $vgpr1
+    %2:vgpr_32 = COPY $vgpr2
+    %3:vgpr_32 = IMPLICIT_DEF
+
+    ; this should not be combined because op_sel is not zero
+    %4:vgpr_32 = V_MOV_B32_dpp %3, %1, 1, 15, 15, 1, implicit $exec
+    %5:vgpr_32 = V_DOT2_F32_F16 0, %4, 0, %0, 0, %2, 0, 5, 0, 0, 0, implicit $mode, implicit $exec
+
+    ; this should not be combined because op_sel_hi is not all set
+    %6:vgpr_32 = V_MOV_B32_dpp %3, %1, 1, 15, 15, 1, implicit $exec
+    %7:vgpr_32 = V_DOT2_F32_F16 0, %6, 0, %0, 0, %2, 0, 0, 4, 0, 0, implicit $mode, implicit $exec
+
+    %8:vgpr_32 = V_MOV_B32_dpp %3, %1, 1, 15, 15, 1, implicit $exec
+    %9:vgpr_32 = V_DOT2_F32_F16 10, %8, 8, %0, 13, %2, 1, 0, 7, 4, 5, implicit $mode, implicit $exec
+
+    %10:vgpr_32 = V_MOV_B32_dpp %3, %1, 1, 15, 15, 1, implicit $exec
+    %11:vgpr_32 = V_FMA_MIX_F32 8, %10, 8, %0, 8, %2, 1, 0, 7, implicit $mode, implicit $exec
+
+    %12:vgpr_32 = V_MOV_B32_dpp %3, %1, 1, 15, 15, 1, implicit $exec
+    %13:vgpr_32 = V_FMA_MIXLO_F16 8, %12, 8, %0, 8, %2, 0, %2, 0, 7, implicit $mode, implicit $exec
+
+    %14:vgpr_32 = V_MOV_B32_dpp %3, %1, 1, 15, 15, 1, implicit $exec
+    %15:vgpr_32 = V_FMA_MIXHI_F16 8, %14, 8, %0, 8, %2, 1, %0, 0, 7, implicit $mode, implicit $exec
+
+...
+
+# when the DPP source isn't a src0 operand the operation should be commuted if possible
+# GCN-LABEL: name: dpp_commute_shrink
+# GCN: %4:vgpr_32 = V_MUL_U32_U24_dpp %1, %0, %1, 1, 14, 15, 0, implicit $exec
+# GCN: %7:vgpr_32 = V_AND_B32_dpp %1, %0, %1, 1, 15, 14, 0, implicit $exec
+# GCN: %10:vgpr_32 = V_MAX_I32_dpp %1, %0, %1, 1, 14, 15, 0, implicit $exec
+# GCN: %13:vgpr_32 = V_MIN_I32_dpp %1, %0, %1, 1, 15, 14, 0, implicit $exec
+# GCN: %16:vgpr_32 = V_SUBREV_U32_dpp %1, %0, %1, 1, 14, 15, 0, implicit $exec
+name: dpp_commute_shrink
+tracksRegLiveness: true
+body:             |
+  bb.0:
+    liveins: $vgpr0, $vgpr1
+
+    %0:vgpr_32 = COPY $vgpr0
+    %1:vgpr_32 = COPY $vgpr1
+
+    %2:vgpr_32 = V_MOV_B32_e32 1, implicit $exec
+    %3:vgpr_32 = V_MOV_B32_dpp %2, %0, 1, 14, 15, 0, implicit $exec
+    %4:vgpr_32 = V_MUL_U32_U24_e64 %1, %3, 0, implicit $exec
+
+    %5:vgpr_32 = V_MOV_B32_e32 4294967295, implicit $exec
+    %6:vgpr_32 = V_MOV_B32_dpp %5, %0, 1, 15, 14, 0, implicit $exec
+    %7:vgpr_32 = V_AND_B32_e64 %1, %6, implicit $exec
+
+    %8:vgpr_32 = V_MOV_B32_e32 -2147483648, implicit $exec
+    %9:vgpr_32 = V_MOV_B32_dpp %8, %0, 1, 14, 15, 0, implicit $exec
+    %10:vgpr_32 = V_MAX_I32_e64 %1, %9, implicit $exec
+
+    %11:vgpr_32 = V_MOV_B32_e32 2147483647, implicit $exec
+    %12:vgpr_32 = V_MOV_B32_dpp %11, %0, 1, 15, 14, 0, implicit $exec
+    %13:vgpr_32 = V_MIN_I32_e64 %1, %12, implicit $exec
+
+    %14:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    %15:vgpr_32 = V_MOV_B32_dpp %14, %0, 1, 14, 15, 0, implicit $exec
+    %16:vgpr_32 = V_SUB_U32_e64 %1, %15, 0, implicit $exec
+
+...
+
+# do not combine, dpp arg used twice
+# GCN-label: name: dpp_arg_twice
+# GCN: %4:vgpr_32 = V_FMA_F32_e64 1, %1, 2, %3, 2, %3, 1, 2, implicit $mode, implicit $exec
+# GCN: %6:vgpr_32 = V_FMA_F32_e64 2, %5, 2, %1, 2, %5, 1, 2, implicit $mode, implicit $exec
+# GCN: %8:vgpr_32 = V_FMA_F32_e64 2, %7, 2, %7, 2, %1, 1, 2, implicit $mode, implicit $exec
+name: dpp_arg_twice
+tracksRegLiveness: true
+body:             |
+  bb.0:
+    liveins: $vgpr0, $vgpr1
+
+    %0:vgpr_32 = COPY $vgpr0
+    %1:vgpr_32 = COPY $vgpr1
+    %2:vgpr_32 = IMPLICIT_DEF
+
+    %3:vgpr_32 = V_MOV_B32_dpp %2, %0, 1, 15, 15, 1, implicit $exec
+    %4:vgpr_32 = V_FMA_F32_e64 1, %1, 2, %3, 2, %3, 1, 2, implicit $mode, implicit $exec
+
+    %5:vgpr_32 = V_MOV_B32_dpp %2, %0, 1, 15, 15, 1, implicit $exec
+    %6:vgpr_32 = V_FMA_F32_e64 2, %5, 2, %1, 2, %5, 1, 2, implicit $mode, implicit $exec
+
+    %7:vgpr_32 = V_MOV_B32_dpp %2, %0, 1, 15, 15, 1, implicit $exec
+    %8:vgpr_32 = V_FMA_F32_e64 2, %7, 2, %7, 2, %1, 1, 2, implicit $mode, implicit $exec
+
+...
+
+# when the dpp source isn't a src0 operand the operation should be commuted if possible
+# GCN-label: name: dpp_commute_e64
+# GCN: %4:vgpr_32  = V_MUL_U32_U24_e64_dpp %1, %0, %1, 1, 1, 14, 15, 0, implicit $exec
+# GCN: %7:vgpr_32 = V_FMA_F32_e64_dpp %5, 2, %0, 1, %1, 2, %1, 1, 2, 1, 15, 15, 1, implicit $mode, implicit $exec
+# GCN: %10:vgpr_32 = V_SUBREV_U32_e64_dpp %1, %0, %1, 1, 1, 14, 15, 0, implicit $exec
+# GCN: %13:vgpr_32, %14:sreg_32_xm0_xexec = V_ADD_CO_U32_e64_dpp %1, %0, %1, 0, 1, 14, 15, 0, implicit $exec
+# GCN: %17:vgpr_32, %18:sreg_32_xm0_xexec = V_ADD_CO_U32_e64 5, %16, 0, implicit $exec
+name: dpp_commute_e64
+tracksRegLiveness: true
+body:             |
+  bb.0:
+    liveins: $vgpr0, $vgpr1
+
+    %0:vgpr_32 = COPY $vgpr0
+    %1:vgpr_32 = COPY $vgpr1
+
+    %2:vgpr_32 = V_MOV_B32_e32 1, implicit $exec
+    %3:vgpr_32 = V_MOV_B32_dpp %2, %0, 1, 14, 15, 0, implicit $exec
+    %4:vgpr_32 = V_MUL_U32_U24_e64 %1, %3, 1, implicit $exec
+
+    %5:vgpr_32 = IMPLICIT_DEF
+    %6:vgpr_32 = V_MOV_B32_dpp %5, %0, 1, 15, 15, 1, implicit $exec
+    %7:vgpr_32 = V_FMA_F32_e64 1, %1, 2, %6, 2, %1, 1, 2, implicit $mode, implicit $exec
+
+    %8:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    %9:vgpr_32 = V_MOV_B32_dpp %8, %0, 1, 14, 15, 0, implicit $exec
+    %10:vgpr_32 = V_SUB_U32_e64 %1, %9, 1, implicit $exec
+
+    %11:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    %12:vgpr_32 = V_MOV_B32_dpp %11, %0, 1, 14, 15, 0, implicit $exec
+    %13:vgpr_32, %14:sreg_32_xm0_xexec = V_ADD_CO_U32_e64 %1, %12, 0, implicit $exec
+
+    ; this cannot be combined because immediate as src0 isn't commutable
+    %15:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    %16:vgpr_32 = V_MOV_B32_dpp %15, %0, 1, 14, 15, 0, implicit $exec
+    %17:vgpr_32, %18:sreg_32_xm0_xexec = V_ADD_CO_U32_e64 5, %16, 0, implicit $exec
+...
+
+---
+
+# check for floating point modifiers
+# GCN-LABEL: name: add_f32_e64
+# GCN:   %4:vgpr_32 = V_ADD_F32_e64_dpp %2, 0, %1, 0, %0, 0, 1, 1, 15, 15, 1, implicit $mode, implicit $exec
+# GCN:   %6:vgpr_32 = V_ADD_F32_dpp %2, 0, %1, 0, %0, 1, 15, 15, 1, implicit $mode, implicit $exec
+# GCN:   %8:vgpr_32 = V_ADD_F32_dpp %2, 1, %1, 2, %0, 1, 15, 15, 1, implicit $mode, implicit $exec
+# GCN:   %10:vgpr_32 = V_ADD_F32_e64_dpp %2, 4, %1, 8, %0, 0, 0, 1, 15, 15, 1, implicit $mode, implicit $exec
+
+name: add_f32_e64
+tracksRegLiveness: true
+body:             |
+  bb.0:
+    liveins: $vgpr0, $vgpr1
+
+    %0:vgpr_32 = COPY $vgpr0
+    %1:vgpr_32 = COPY $vgpr1
+    %2:vgpr_32 = IMPLICIT_DEF
+
+    ; this should be combined as e64
+    %3:vgpr_32 = V_MOV_B32_dpp undef %2, %1, 1, 15, 15, 1, implicit $exec
+    %4:vgpr_32 = V_ADD_F32_e64 0, %3, 0, %0, 0, 1, implicit $mode, implicit $exec
+
+    ; this should be combined and shrunk as all modifiers are default
+    %5:vgpr_32 = V_MOV_B32_dpp undef %2, %1, 1, 15, 15, 1, implicit $exec
+    %6:vgpr_32 = V_ADD_F32_e64 0, %5, 0, %0, 0, 0, implicit $mode, implicit $exec
+
+    ; this should be combined and shrunk as modifiers other than abs|neg are default
+    %7:vgpr_32 = V_MOV_B32_dpp undef %2, %1, 1, 15, 15, 1, implicit $exec
+    %8:vgpr_32 = V_ADD_F32_e64 1, %7, 2, %0, 0, 0, implicit $mode, implicit $exec
+
+    ; this should be combined as e64
+    %9:vgpr_32 = V_MOV_B32_dpp undef %2, %1, 1, 15, 15, 1, implicit $exec
+    %10:vgpr_32 = V_ADD_F32_e64 4, %9, 8, %0, 0, 0, implicit $mode, implicit $exec
+...
+
+# check for e64 modifiers
+# GCN-LABEL: name: add_u32_e64
+# GCN: %4:vgpr_32 = V_ADD_U32_dpp %2, %0, %1, 1, 15, 15, 1, implicit $exec
+# GCN: %6:vgpr_32 = V_ADD_U32_e64_dpp %2, %0, %1, 1, 1, 15, 15, 1, implicit $exec
+
+name: add_u32_e64
+tracksRegLiveness: true
+body:             |
+  bb.0:
+    liveins: $vgpr0, $vgpr1
+
+    %0:vgpr_32 = COPY $vgpr0
+    %1:vgpr_32 = COPY $vgpr1
+    %2:vgpr_32 = IMPLICIT_DEF
+
+    ; this should be combined  and shrunk as all modifiers are default
+    %3:vgpr_32 = V_MOV_B32_dpp undef %2, %0, 1, 15, 15, 1, implicit $exec
+    %4:vgpr_32 = V_ADD_U32_e64 %3, %1, 0, implicit $exec
+
+    ; this should be combined as _e64
+    %5:vgpr_32 = V_MOV_B32_dpp undef %2, %0, 1, 15, 15, 1, implicit $exec
+    %6:vgpr_32 = V_ADD_U32_e64 %5, %1, 1, implicit $exec
+...
+
+# tests on sequences of dpp consumers
+# GCN-LABEL: name: dpp_seq
+# GCN: %4:vgpr_32 = V_ADD_U32_dpp %1, %0, %1, 1, 14, 15, 0,  implicit $exec
+# GCN: %5:vgpr_32 = V_SUBREV_U32_dpp %1, %0, %1, 1, 14, 15, 0,  implicit $exec
+# GCN: %6:vgpr_32 = V_OR_B32_dpp %1, %0, %1, 1, 14, 15, 0, implicit $exec
+# broken sequence:
+# GCN: %7:vgpr_32 = V_MOV_B32_dpp %2, %0, 1, 14, 15, 0, implicit $exec
+
+name: dpp_seq
+tracksRegLiveness: true
+body: |
+  bb.0:
+    liveins: $vgpr0, $vgpr1
+    %0:vgpr_32 = COPY $vgpr0
+    %1:vgpr_32 = COPY $vgpr1
+    %2:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+
+    %3:vgpr_32 = V_MOV_B32_dpp %2, %0, 1, 14, 15, 0, implicit $exec
+    %4:vgpr_32 = V_ADD_U32_e32 %3, %1,  implicit $exec
+    %5:vgpr_32 = V_SUB_U32_e32 %1, %3,  implicit $exec
+    %6:vgpr_32 = V_OR_B32_e32 %3, %1, implicit $exec
+
+    %7:vgpr_32 = V_MOV_B32_dpp %2, %0, 1, 14, 15, 0, implicit $exec
+    %8:vgpr_32 = V_ADD_U32_e32 %7, %1,  implicit $exec
+    ; this breaks the sequence
+    %9:vgpr_32 = V_SUB_U32_e32 5, %7,  implicit $exec
+...
+
+# tests on sequences of dpp consumers followed by control flow
+# GCN-LABEL: name: dpp_seq_cf
+# GCN: %4:vgpr_32 = V_ADD_U32_dpp %1, %0, %1, 1, 14, 15, 0, implicit $exec
+# GCN: %5:vgpr_32 = V_SUBREV_U32_dpp %1, %0, %1, 1, 14, 15, 0, implicit $exec
+# GCN: %6:vgpr_32 = V_OR_B32_dpp %1, %0, %1, 1, 14, 15, 0, implicit $exec
+
+name: dpp_seq_cf
+tracksRegLiveness: true
+body: |
+  bb.0:
+    successors: %bb.1, %bb.2
+    liveins: $vgpr0, $vgpr1
+    %0:vgpr_32 = COPY $vgpr0
+    %1:vgpr_32 = COPY $vgpr1
+    %2:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+
+    %3:vgpr_32 = V_MOV_B32_dpp %2, %0, 1, 14, 15, 0, implicit $exec
+    %4:vgpr_32 = V_ADD_U32_e32 %3, %1, implicit $exec
+    %5:vgpr_32 = V_SUB_U32_e32 %1, %3, implicit $exec
+    %6:vgpr_32 = V_OR_B32_e32 %3, %1, implicit $exec
+
+    %7:sreg_32 = V_CMP_EQ_U32_e64 %5, %6, implicit $exec
+    %8:sreg_32 = SI_IF %7, %bb.2, implicit-def dead $exec, implicit-def dead $scc, implicit $exec
+    S_BRANCH %bb.1
+
+  bb.1:
+    successors: %bb.2
+
+  bb.2:
+    SI_END_CF %8, implicit-def dead $exec, implicit-def dead $scc, implicit $exec
+...
+
+# GCN-LABEL: name: old_in_diff_bb
+# GCN: %4:vgpr_32 = V_ADD_U32_dpp %0, %1, %0, 1, 1, 1, 0, implicit $exec
+
+name: old_in_diff_bb
+tracksRegLiveness: true
+body: |
+  bb.0:
+    successors: %bb.1
+    liveins: $vgpr0, $vgpr1
+
+    %0:vgpr_32 = COPY $vgpr0
+    %1:vgpr_32 = COPY $vgpr1
+    %2:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    S_BRANCH %bb.1
+
+  bb.1:
+    %3:vgpr_32 = V_MOV_B32_dpp %2, %1, 1, 1, 1, 0, implicit $exec
+    %4:vgpr_32 = V_ADD_U32_e32 %3, %0, implicit $exec
+...
+
+# old reg def is in diff BB but bound_ctrl:1 - can combine
+# GCN-LABEL: name: old_in_diff_bb_bctrl_zero
+# GCN: %4:vgpr_32 = V_ADD_U32_dpp {{%[0-9]}}, %0, %1, 1, 15, 15, 1, implicit $exec
+
+name: old_in_diff_bb_bctrl_zero
+tracksRegLiveness: true
+body: |
+  bb.0:
+    successors: %bb.1
+    liveins: $vgpr0, $vgpr1
+
+    %0:vgpr_32 = COPY $vgpr0
+    %1:vgpr_32 = COPY $vgpr1
+    %2:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    S_BRANCH %bb.1
+
+  bb.1:
+    %3:vgpr_32 = V_MOV_B32_dpp %2, %0, 1, 15, 15, 1, implicit $exec
+    %4:vgpr_32 = V_ADD_U32_e32 %3, %1, implicit $exec
+...
+
+# EXEC mask changed between def and use - cannot combine
+# GCN-LABEL: name: exec_changed
+# GCN: %3:vgpr_32 = V_MOV_B32_dpp %2, %0, 1, 15, 15, 1, implicit $exec
+
+name: exec_changed
+tracksRegLiveness: true
+body: |
+  bb.0:
+    liveins: $vgpr0, $vgpr1
+
+    %0:vgpr_32 = COPY $vgpr0
+    %1:vgpr_32 = COPY $vgpr1
+    %2:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    %3:vgpr_32 = V_MOV_B32_dpp %2, %0, 1, 15, 15, 1, implicit $exec
+    %4:vgpr_32 = V_ADD_U32_e32 %3, %1, implicit $exec
+    %5:sreg_64 = COPY $exec, implicit-def $exec
+    %6:vgpr_32 = V_ADD_U32_e32 %3, %1, implicit $exec
+...
+
+# test if $old definition is correctly tracked through subreg manipulation pseudos
+
+# GCN-LABEL: name: mul_old_subreg
+# GCN: %7:vgpr_32 = V_MUL_I32_I24_dpp %0.sub1, %1, %0.sub1, 1, 1, 1, 0, implicit $exec
+
+name:            mul_old_subreg
+tracksRegLiveness: true
+body: |
+  bb.0:
+    liveins: $vgpr0, $vgpr1
+
+    %0:vreg_64 = COPY $vgpr0
+    %1:vgpr_32 = COPY $vgpr1
+    %2:vgpr_32 = V_MOV_B32_e32 1, implicit $exec
+    %3:vgpr_32 = V_MOV_B32_e32 42, implicit $exec
+    %4:vreg_64 = REG_SEQUENCE %2, %subreg.sub0, %3, %subreg.sub1
+    %5:vreg_64 = INSERT_SUBREG %4, %1, %subreg.sub1 ; %5.sub0 is taken from %4
+    %6:vgpr_32 = V_MOV_B32_dpp %5.sub0, %1, 1, 1, 1, 0, implicit $exec
+    %7:vgpr_32 = V_MUL_I32_I24_e32 %6, %0.sub1, implicit $exec
+...
+
+# GCN-LABEL: name: add_old_subreg
+# GCN: %5:vgpr_32 = V_ADD_U32_dpp %0.sub1, %1, %0.sub1, 1, 1, 1, 0, implicit $exec
+
+name:            add_old_subreg
+tracksRegLiveness: true
+body: |
+  bb.0:
+    liveins: $vgpr0, $vgpr1
+
+    %0:vreg_64 = COPY $vgpr0
+    %1:vgpr_32 = COPY $vgpr1
+    %2:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    %3:vreg_64 = INSERT_SUBREG %0, %2, %subreg.sub1 ; %3.sub1 is inserted
+    %4:vgpr_32 = V_MOV_B32_dpp %3.sub1, %1, 1, 1, 1, 0, implicit $exec
+    %5:vgpr_32 = V_ADD_U32_e32 %4, %0.sub1, implicit $exec
+...
+
+# GCN-LABEL: name: add_old_subreg_undef
+# GCN: %5:vgpr_32 = V_ADD_U32_dpp undef %3.sub1, %1, %0.sub1, 1, 15, 15, 1, implicit $exec
+
+name:            add_old_subreg_undef
+tracksRegLiveness: true
+body: |
+  bb.0:
+    liveins: $vgpr0, $vgpr1
+
+    %0:vreg_64 = COPY $vgpr0
+    %1:vgpr_32 = COPY $vgpr1
+    %2:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
+    %3:vreg_64 = REG_SEQUENCE %2, %subreg.sub0 ; %3.sub1 is undef
+    %4:vgpr_32 = V_MOV_B32_dpp %3.sub1, %1, 1, 15, 15, 1, implicit $exec
+    %5:vgpr_32 = V_ADD_U32_e32 %4, %0.sub1, implicit $exec
+...
+
+# Test instruction which does not have modifiers in VOP1 form but does in DPP form.
+# GCN-LABEL: name: dpp_vop1
+# GCN: %3:vgpr_32 = V_CEIL_F32_dpp %0, 0, undef %2:vgpr_32, 1, 15, 15, 1, implicit $mode, implicit $exec
+name: dpp_vop1
+tracksRegLiveness: true
+body: |
+  bb.0:
+    %1:vgpr_32 = IMPLICIT_DEF
+    %2:vgpr_32 = V_MOV_B32_dpp %1:vgpr_32, undef %0:vgpr_32, 1, 15, 15, 1, implicit $exec
+    %3:vgpr_32 = V_CEIL_F32_e32 %2, implicit $mode, implicit $exec
+...
+
+# Test instruction which does not have modifiers in VOP2 form but does in DPP form.
+# GCN-LABEL: name: dpp_min
+# GCN: %3:vgpr_32 = V_MIN_F32_dpp %0, 0, undef %2:vgpr_32, 0, undef %4:vgpr_32, 1, 15, 15, 1, implicit $mode, implicit $exec
+name: dpp_min
+tracksRegLiveness: true
+body: |
+  bb.0:
+    %1:vgpr_32 = IMPLICIT_DEF
+    %2:vgpr_32 = V_MOV_B32_dpp %1:vgpr_32, undef %0:vgpr_32, 1, 15, 15, 1, implicit $exec
+    %4:vgpr_32 = V_MIN_F32_e32 %2, undef %3:vgpr_32, implicit $mode, implicit $exec
+...
+
+# Test an undef old operand
+# GCN-LABEL: name: dpp_undef_old
+# GCN: %3:vgpr_32 = V_CEIL_F32_dpp undef %1:vgpr_32, 0, undef %2:vgpr_32, 1, 15, 15, 1, implicit $mode, implicit $exec
+name: dpp_undef_old
+tracksRegLiveness: true
+body: |
+  bb.0:
+    %2:vgpr_32 = V_MOV_B32_dpp undef %1:vgpr_32, undef %0:vgpr_32, 1, 15, 15, 1, implicit $exec
+    %3:vgpr_32 = V_CEIL_F32_e32 %2, implicit $mode, implicit $exec
+...
+
+# Do not combine a dpp mov which writes a physreg.
+# GCN-LABEL: name: phys_dpp_mov_dst
+# GCN: $vgpr0 = V_MOV_B32_dpp undef %0:vgpr_32, undef %1:vgpr_32, 1, 15, 15, 1, implicit $exec
+# GCN: %2:vgpr_32 = V_CEIL_F32_e32 $vgpr0, implicit $mode, implicit $exec
+name: phys_dpp_mov_dst
+tracksRegLiveness: true
+body: |
+  bb.0:
+    $vgpr0 = V_MOV_B32_dpp undef %1:vgpr_32, undef %0:vgpr_32, 1, 15, 15, 1, implicit $exec
+    %2:vgpr_32 = V_CEIL_F32_e32 $vgpr0, implicit $mode, implicit $exec
+...
+
+# Do not combine a dpp mov which reads a physreg.
+# GCN-LABEL: name: phys_dpp_mov_old_src
+# GCN: %0:vgpr_32 = V_MOV_B32_dpp undef $vgpr0, undef %1:vgpr_32, 1, 15, 15, 1, implicit $exec
+# GCN: %2:vgpr_32 = V_CEIL_F32_e32 %0, implicit $mode, implicit $exec
+name: phys_dpp_mov_old_src
+tracksRegLiveness: true
+body: |
+  bb.0:
+    %1:vgpr_32 = V_MOV_B32_dpp undef $vgpr0, undef %0:vgpr_32, 1, 15, 15, 1, implicit $exec
+    %2:vgpr_32 = V_CEIL_F32_e32 %1, implicit $mode, implicit $exec
+...
+
+# Do not combine a dpp mov which reads a physreg.
+# GCN-LABEL: name: phys_dpp_mov_src
+# GCN: %0:vgpr_32 = V_MOV_B32_dpp undef %1:vgpr_32, undef $vgpr0, 1, 15, 15, 1, implicit $exec
+# GCN: %2:vgpr_32 = V_CEIL_F32_e32 %0, implicit $mode, implicit $exec
+name: phys_dpp_mov_src
+tracksRegLiveness: true
+body: |
+  bb.0:
+    %1:vgpr_32 = V_MOV_B32_dpp undef %0:vgpr_32, undef $vgpr0, 1, 15, 15, 1, implicit $exec
+    %2:vgpr_32 = V_CEIL_F32_e32 %1, implicit $mode, implicit $exec
+...
+
+# GCN-LABEL: name: dpp_reg_sequence_both_combined
+# GCN: %0:vreg_64 = COPY $vgpr0_vgpr1
+# GCN: %1:vreg_64 = COPY $vgpr2_vgpr3
+# GCN: %2:vgpr_32 = V_MOV_B32_e32 5, implicit $exec
+# GCN: %9:vgpr_32 = IMPLICIT_DEF
+# GCN: %8:vgpr_32 = IMPLICIT_DEF
+# GCN: %6:vgpr_32 = V_ADD_U32_dpp %9, %1.sub0, %2, 1, 15, 15, 1, implicit $exec
+# GCN: %7:vgpr_32 = V_ADDC_U32_dpp %8, %1.sub1, %2, 1, 15, 15, 1, implicit-def $vcc, implicit $vcc, implicit $exec
+name: dpp_reg_sequence_both_combined
+tracksRegLiveness: true
+body: |
+  bb.0:
+    liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+
+    %0:vreg_64 = COPY $vgpr0_vgpr1
+    %1:vreg_64 = COPY $vgpr2_vgpr3
+    %5:vgpr_32 = V_MOV_B32_e32 5, implicit $exec
+    %2:vgpr_32 = V_MOV_B32_dpp %0.sub0, %1.sub0, 1, 15, 15, 1, implicit $exec
+    %3:vgpr_32 = V_MOV_B32_dpp %0.sub1, %1.sub1, 1, 15, 15, 1, implicit $exec
+    %4:vreg_64 = REG_SEQUENCE %2, %subreg.sub0, %3, %subreg.sub1
+    %6:vgpr_32 = V_ADD_U32_e32 %4.sub0, %5, implicit $exec
+    %7:vgpr_32 = V_ADDC_U32_e32 %4.sub1, %5, implicit-def $vcc, implicit $vcc, implicit $exec
+...
+
+# GCN-LABEL: name: dpp_reg_sequence_first_combined
+# GCN: %0:vreg_64 = COPY $vgpr0_vgpr1
+# GCN: %1:vreg_64 = COPY $vgpr2_vgpr3
+# GCN: %2:vgpr_32 = V_MOV_B32_e32 5, implicit $exec
+# GCN: %8:vgpr_32 = IMPLICIT_DEF
+# GCN: %4:vgpr_32 = V_MOV_B32_dpp %0.sub1, %1.sub1, 1, 1, 1, 1, implicit $exec
+# GCN: %5:vreg_64 = REG_SEQUENCE undef %3:vgpr_32, %subreg.sub0, %4, %subreg.sub1
+# GCN: %6:vgpr_32 = V_ADD_U32_dpp %8, %1.sub0, %2, 1, 15, 15, 1, implicit $exec
+# GCN: %7:vgpr_32 = V_ADDC_U32_e32 %5.sub1, %2, implicit-def $vcc, implicit $vcc, implicit $exec
+name: dpp_reg_sequence_first_combined
+tracksRegLiveness: true
+body: |
+  bb.0:
+    liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+
+    %0:vreg_64 = COPY $vgpr0_vgpr1
+    %1:vreg_64 = COPY $vgpr2_vgpr3
+    %5:vgpr_32 = V_MOV_B32_e32 5, implicit $exec
+    %2:vgpr_32 = V_MOV_B32_dpp %0.sub0, %1.sub0, 1, 15, 15, 1, implicit $exec
+    %3:vgpr_32 = V_MOV_B32_dpp %0.sub1, %1.sub1, 1, 1, 1, 1, implicit $exec
+    %4:vreg_64 = REG_SEQUENCE %2, %subreg.sub0, %3, %subreg.sub1
+    %6:vgpr_32 = V_ADD_U32_e32 %4.sub0, %5, implicit $exec
+    %7:vgpr_32 = V_ADDC_U32_e32 %4.sub1, %5, implicit-def $vcc, implicit $vcc, implicit $exec
+...
+
+# GCN-LABEL: name: dpp_reg_sequence_second_combined
+# GCN: %0:vreg_64 = COPY $vgpr0_vgpr1
+# GCN: %1:vreg_64 = COPY $vgpr2_vgpr3
+# GCN: %2:vgpr_32 = V_MOV_B32_e32 5, implicit $exec
+# GCN: %3:vgpr_32 = V_MOV_B32_dpp %0.sub0, %1.sub0, 1, 1, 1, 1, implicit $exec
+# GCN: %8:vgpr_32 = IMPLICIT_DEF
+# GCN: %5:vreg_64 = REG_SEQUENCE %3, %subreg.sub0, undef %4:vgpr_32, %subreg.sub1
+# GCN: %6:vgpr_32 = V_ADD_U32_e32 %5.sub0, %2, implicit $exec
+# GCN: %7:vgpr_32 = V_ADDC_U32_dpp %8, %1.sub1, %2, 1, 15, 15, 1, implicit-def $vcc, implicit $vcc, implicit $exec
+name: dpp_reg_sequence_second_combined
+tracksRegLiveness: true
+body: |
+  bb.0:
+    liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+
+    %0:vreg_64 = COPY $vgpr0_vgpr1
+    %1:vreg_64 = COPY $vgpr2_vgpr3
+    %5:vgpr_32 = V_MOV_B32_e32 5, implicit $exec
+    %2:vgpr_32 = V_MOV_B32_dpp %0.sub0, %1.sub0, 1, 1, 1, 1, implicit $exec
+    %3:vgpr_32 = V_MOV_B32_dpp %0.sub1, %1.sub1, 1, 15, 15, 1, implicit $exec
+    %4:vreg_64 = REG_SEQUENCE %2, %subreg.sub0, %3, %subreg.sub1
+    %6:vgpr_32 = V_ADD_U32_e32 %4.sub0, %5, implicit $exec
+    %7:vgpr_32 = V_ADDC_U32_e32 %4.sub1, %5, implicit-def $vcc, implicit $vcc, implicit $exec
+...
+
+# GCN-LABEL: name: dpp_reg_sequence_none_combined
+# GCN: %0:vreg_64 = COPY $vgpr0_vgpr1
+# GCN: %1:vreg_64 = COPY $vgpr2_vgpr3
+# GCN: %2:vgpr_32 = V_MOV_B32_e32 5, implicit $exec
+# GCN: %3:vgpr_32 = V_MOV_B32_dpp %0.sub0, %1.sub0, 1, 1, 1, 1, implicit $exec
+# GCN: %4:vgpr_32 = V_MOV_B32_dpp %0.sub1, %1.sub1, 1, 1, 1, 1, implicit $exec
+# GCN: %5:vreg_64 = REG_SEQUENCE %3, %subreg.sub0, %4, %subreg.sub1
+# GCN: %6:vgpr_32 = V_ADD_U32_e32 %5.sub0, %2, implicit $exec
+# GCN: %7:vgpr_32 = V_ADDC_U32_e32 %5.sub1, %2, implicit-def $vcc, implicit $vcc, implicit $exec
+name: dpp_reg_sequence_none_combined
+tracksRegLiveness: true
+body: |
+  bb.0:
+    liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+
+    %0:vreg_64 = COPY $vgpr0_vgpr1
+    %1:vreg_64 = COPY $vgpr2_vgpr3
+    %5:vgpr_32 = V_MOV_B32_e32 5, implicit $exec
+    %2:vgpr_32 = V_MOV_B32_dpp %0.sub0, %1.sub0, 1, 1, 1, 1, implicit $exec
+    %3:vgpr_32 = V_MOV_B32_dpp %0.sub1, %1.sub1, 1, 1, 1, 1, implicit $exec
+    %4:vreg_64 = REG_SEQUENCE %2, %subreg.sub0, %3, %subreg.sub1
+    %6:vgpr_32 = V_ADD_U32_e32 %4.sub0, %5, implicit $exec
+    %7:vgpr_32 = V_ADDC_U32_e32 %4.sub1, %5, implicit-def $vcc, implicit $vcc, implicit $exec
+...
+
+# GCN-LABEL: name: dpp_reg_sequence_exec_changed
+# GCN:   %0:vreg_64 = COPY $vgpr0_vgpr1
+# GCN:   %1:vreg_64 = COPY $vgpr2_vgpr3
+# GCN:   %2:vgpr_32 = V_MOV_B32_e32 5, implicit $exec
+# GCN:   %3:vgpr_32 = V_MOV_B32_dpp %0.sub0, %1.sub0, 1, 15, 15, 1, implicit $exec
+# GCN:   %4:vgpr_32 = V_MOV_B32_dpp %0.sub1, %1.sub1, 1, 15, 15, 1, implicit $exec
+# GCN:   %5:vreg_64 = REG_SEQUENCE %3, %subreg.sub0, %4, %subreg.sub1
+# GCN:   S_BRANCH %bb.1
+# GCN: bb.1:
+# GCN:   %6:vgpr_32 = V_ADD_U32_e32 %5.sub0, %2, implicit $exec
+# GCN:   %7:vgpr_32 = V_ADDC_U32_e32 %5.sub1, %2, implicit-def $vcc, implicit $vcc, implicit $exec
+name: dpp_reg_sequence_exec_changed
+tracksRegLiveness: true
+body: |
+  bb.0:
+    liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+
+    %0:vreg_64 = COPY $vgpr0_vgpr1
+    %1:vreg_64 = COPY $vgpr2_vgpr3
+    %5:vgpr_32 = V_MOV_B32_e32 5, implicit $exec
+    %2:vgpr_32 = V_MOV_B32_dpp %0.sub0, %1.sub0, 1, 15, 15, 1, implicit $exec
+    %3:vgpr_32 = V_MOV_B32_dpp %0.sub1, %1.sub1, 1, 15, 15, 1, implicit $exec
+    %4:vreg_64 = REG_SEQUENCE %2, %subreg.sub0, %3, %subreg.sub1
+    S_BRANCH %bb.1
+
+  bb.1:
+    %6:vgpr_32 = V_ADD_U32_e32 %4.sub0, %5, implicit $exec
+    %7:vgpr_32 = V_ADDC_U32_e32 %4.sub1, %5, implicit-def $vcc, implicit $vcc, implicit $exec
+...
+
+# GCN-LABEL: name: dpp_reg_sequence_subreg
+# GCN: %0:vreg_64 = COPY $vgpr0_vgpr1
+# GCN: %1:vreg_64 = COPY $vgpr2_vgpr3
+# GCN: %2:vgpr_32 = V_MOV_B32_e32 5, implicit $exec
+# GCN: %3:vgpr_32 = V_MOV_B32_dpp %0.sub0, %1.sub0, 1, 15, 15, 1, implicit $exec
+# GCN: %4:vgpr_32 = V_MOV_B32_dpp %0.sub1, %1.sub1, 1, 15, 15, 1, implicit $exec
+# GCN: %5:vreg_64 = REG_SEQUENCE %3, %subreg.sub0, %4, %subreg.sub1
+# GCN: %6:vreg_64 = REG_SEQUENCE %5.sub0, %subreg.sub0, %5.sub1, %subreg.sub1
+# GCN: %7:vgpr_32 = V_ADD_U32_e32 %6.sub0, %2, implicit $exec
+# GCN: %8:vgpr_32 = V_ADDC_U32_e32 %6.sub1, %2, implicit-def $vcc, implicit $vcc, implicit $exec
+name: dpp_reg_sequence_subreg
+tracksRegLiveness: true
+body: |
+  bb.0:
+    liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+
+    %0:vreg_64 = COPY $vgpr0_vgpr1
+    %1:vreg_64 = COPY $vgpr2_vgpr3
+    %8:vgpr_32 = V_MOV_B32_e32 5, implicit $exec
+    %2:vgpr_32 = V_MOV_B32_dpp %0.sub0, %1.sub0, 1, 15, 15, 1, implicit $exec
+    %3:vgpr_32 = V_MOV_B32_dpp %0.sub1, %1.sub1, 1, 15, 15, 1, implicit $exec
+    %4:vreg_64 = REG_SEQUENCE %2, %subreg.sub0, %3, %subreg.sub1
+    %5:vreg_64 = REG_SEQUENCE %4.sub0, %subreg.sub0, %4.sub1, %subreg.sub1
+    %6:vgpr_32 = V_ADD_U32_e32 %5.sub0, %8, implicit $exec
+    %7:vgpr_32 = V_ADDC_U32_e32 %5.sub1, %8, implicit-def $vcc, implicit $vcc, implicit $exec
+...
+
+# GCN-LABEL: name: dpp_reg_sequence_src2_reject
+#GCN: %2:vgpr_32 = V_MOV_B32_dpp %0.sub0, %1.sub0, 1, 15, 15, 1, implicit $exec
+#GCN: %3:vgpr_32 = V_MOV_B32_dpp %0.sub1, %1.sub1, 1, 15, 15, 1, implicit $exec
+#GCN: %4:vreg_64 = REG_SEQUENCE %2, %subreg.sub0, %3, %subreg.sub1
+#GCN: %5:vgpr_32 = V_MOV_B32_e32 5, implicit $exec
+#GCN: %6:vgpr_32 = V_FMA_F32_e64 2, %4.sub0, 2, %5, 2, %4.sub0, 1, 2, implicit $mode, implicit $exec
+#GCN: %7:vgpr_32 = V_FMA_F32_e64 2, %4.sub0, 2, %5, 2, %4.sub1, 1, 2, implicit $mode, implicit $exec
+name: dpp_reg_sequence_src2_reject
+tracksRegLiveness: true
+body: |
+  bb.0:
+    liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+
+    %0:vreg_64 = COPY $vgpr0_vgpr1
+    %1:vreg_64 = COPY $vgpr2_vgpr3
+    %2:vgpr_32 = V_MOV_B32_dpp %0.sub0, %1.sub0, 1, 15, 15, 1, implicit $exec
+    %3:vgpr_32 = V_MOV_B32_dpp %0.sub1, %1.sub1, 1, 15, 15, 1, implicit $exec
+    %4:vreg_64 = REG_SEQUENCE %2, %subreg.sub0, %3, %subreg.sub1
+    %5:vgpr_32 = V_MOV_B32_e32 5, implicit $exec
+    ; use of dpp arg as src2, reject
+    %6:vgpr_32 = V_FMA_F32_e64 2, %4.sub0, 2, %5, 2, %4.sub0, 1, 2, implicit $mode, implicit $exec
+    ; cannot commute src0 and src2, and %4.sub0 already rejected, reject
+    %7:vgpr_32 = V_FMA_F32_e64 2, %4.sub0, 2, %5, 2, %4.sub1, 1, 2, implicit $mode, implicit $exec
+...
+
+# GCN-LABEL: name: dpp_reg_sequence_src2
+#GCN: %3:vgpr_32 = V_MOV_B32_dpp %0.sub1, %1.sub1, 1, 15, 15, 1, implicit $exec
+#GCN: %4:vreg_64 = REG_SEQUENCE undef %2:vgpr_32, %subreg.sub0, %3, %subreg.sub1
+#GCN: %5:vgpr_32 = V_MOV_B32_e32 5, implicit $exec
+#GCN: %6:vgpr_32 = V_FMA_F32_e64_dpp %8, 2, %1.sub0, 2, %5, 2, %4.sub1, 1, 2, 1, 15, 15, 1, implicit $mode, implicit $exec
+name: dpp_reg_sequence_src2
+tracksRegLiveness: true
+body: |
+  bb.0:
+    liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+
+    %0:vreg_64 = COPY $vgpr0_vgpr1
+    %1:vreg_64 = COPY $vgpr2_vgpr3
+    %2:vgpr_32 = V_MOV_B32_dpp %0.sub0, %1.sub0, 1, 15, 15, 1, implicit $exec
+    %3:vgpr_32 = V_MOV_B32_dpp %0.sub1, %1.sub1, 1, 15, 15, 1, implicit $exec
+    %4:vreg_64 = REG_SEQUENCE %2, %subreg.sub0, %3, %subreg.sub1
+    %5:vgpr_32 = V_MOV_B32_e32 5, implicit $exec
+    %6:vgpr_32 = V_FMA_F32_e64 2, %4.sub0, 2, %5, 2, %4.sub1, 1, 2, implicit $mode, implicit $exec
+...
+
+# GCN-LABEL: name: dpp64_add64_impdef
+# GCN: %3:vgpr_32 = V_ADD_U32_dpp %1.sub0, %0.sub0, undef %4:vgpr_32, 1, 15, 15, 1, implicit $exec
+# GCN: %5:vgpr_32 = V_ADDC_U32_dpp %1.sub1, %0.sub1, undef %4:vgpr_32, 1, 15, 15, 1, implicit-def $vcc, implicit $vcc, implicit $exec
+name: dpp64_add64_impdef
+tracksRegLiveness: true
+body: |
+  bb.0:
+    %0:vreg_64 = IMPLICIT_DEF
+    %1:vreg_64 = IMPLICIT_DEF
+    %2:vreg_64 = V_MOV_B64_DPP_PSEUDO %1:vreg_64, %0:vreg_64, 1, 15, 15, 1, implicit $exec
+    %5:vgpr_32 = V_ADD_U32_e32 %2.sub0, undef %4:vgpr_32, implicit $exec
+    %6:vgpr_32 = V_ADDC_U32_e32 %2.sub1, undef %4, implicit-def $vcc, implicit $vcc, implicit $exec
+...
+
+# GCN-LABEL: name:  dpp64_add64_undef
+# GCN: %3:vgpr_32 = V_ADD_U32_dpp undef %1.sub0:vreg_64, undef %2.sub0:vreg_64, undef %4:vgpr_32, 1, 15, 15, 1, implicit $exec
+# GCN: %5:vgpr_32 = V_ADDC_U32_dpp undef %1.sub1:vreg_64, undef %2.sub1:vreg_64, undef %4:vgpr_32, 1, 15, 15, 1, implicit-def $vcc, implicit $vcc, implicit $exec
+name: dpp64_add64_undef
+tracksRegLiveness: true
+body: |
+  bb.0:
+    %2:vreg_64 = V_MOV_B64_DPP_PSEUDO undef %1:vreg_64, undef %0:vreg_64, 1, 15, 15, 1, implicit $exec
+    %5:vgpr_32 = V_ADD_U32_e32 %2.sub0, undef %4:vgpr_32, implicit $exec
+    %6:vgpr_32 = V_ADDC_U32_e32 %2.sub1, undef %4, implicit-def $vcc, implicit $vcc, implicit $exec
+...
+
+
+# GCN-LABEL: name: cndmask_with_src2
+# GCN: %5:vgpr_32 = V_CNDMASK_B32_e64 0, %3, 0, %1, %4, implicit $exec
+# GCN: %8:vgpr_32 = V_CNDMASK_B32_e64_dpp %2, 4, %1, 0, %1, %7, 1, 15, 15, 1, implicit $exec
+name: cndmask_with_src2
+tracksRegLiveness: true
+body: |
+  bb.0:
+    liveins: $vgpr0, $vgpr1
+    %0:vgpr_32 = COPY $vgpr0
+    %1:vgpr_32 = COPY $vgpr1
+    %2:vgpr_32 = IMPLICIT_DEF
+
+    %3:vgpr_32 = V_MOV_B32_dpp %2, %0, 1, 15, 15, 1, implicit $exec
+    %4:sreg_32_xm0_xexec = IMPLICIT_DEF
+    %5:vgpr_32 = V_CNDMASK_B32_e64 0, %3, 0, %1, %4, implicit $exec
+
+    ; src2 is legal for _e64
+    %6:vgpr_32 = V_MOV_B32_dpp %2, %1, 1, 15, 15, 1, implicit $exec
+    %7:sreg_32_xm0_xexec = IMPLICIT_DEF
+    %8:vgpr_32 = V_CNDMASK_B32_e64 4, %6, 0, %1, %7, implicit $exec
+...
+
+---
+
+# Make sure flags aren't dropped
+# GCN-LABEL: name: flags_add_f32_e64
+# GCN: %4:vgpr_32 = nnan nofpexcept V_ADD_F32_dpp %2, 0, %1, 0, %0, 1, 15, 15, 1, implicit $mode, implicit $exec
+name: flags_add_f32_e64
+tracksRegLiveness: true
+body:             |
+  bb.0:
+    liveins: $vgpr0, $vgpr1
+
+    %0:vgpr_32 = COPY $vgpr0
+    %1:vgpr_32 = COPY $vgpr1
+    %2:vgpr_32 = IMPLICIT_DEF
+
+    %3:vgpr_32 = V_MOV_B32_dpp undef %2, %1, 1, 15, 15, 1, implicit $exec
+    %4:vgpr_32 = nofpexcept nnan V_ADD_F32_e64 0, %3, 0, %0, 0, 0, implicit $mode, implicit $exec
+    S_ENDPGM 0, implicit %4
+
+...
+
+# GCN-LABEL: name: dont_combine_more_than_one_operand
+# GCN: %3:vgpr_32 = V_MAX_F32_e64 0, %2, 0, %2, 0, 0, implicit $mode, implicit $exec
+name: dont_combine_more_than_one_operand
+tracksRegLiveness: true
+body: |
+  bb.0:
+    liveins: $vgpr0, $vgpr1
+    %0:vgpr_32 = COPY $vgpr0
+    %1:vgpr_32 = COPY $vgpr1
+    %2:vgpr_32 = V_MOV_B32_dpp %0, %1, 1, 15, 15, 1, implicit $exec
+    %3:vgpr_32 = V_MAX_F32_e64 0, %2, 0, %2, 0, 0, implicit $mode, implicit $exec
+...
+
+# GCN-LABEL: name: dont_combine_more_than_one_operand_dpp_reg_sequence
+# GCN: %5:vgpr_32 = V_ADD_U32_e32 %4.sub0, %4.sub0, implicit $exec
+# GCN: %6:vgpr_32 = V_ADDC_U32_e32 %4.sub1, %4.sub1, implicit-def $vcc, implicit $vcc, implicit $exec
+name: dont_combine_more_than_one_operand_dpp_reg_sequence
+tracksRegLiveness: true
+body: |
+  bb.0:
+    liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+    %0:vreg_64 = COPY $vgpr0_vgpr1
+    %1:vreg_64 = COPY $vgpr2_vgpr3
+    %2:vgpr_32 = V_MOV_B32_dpp %0.sub0, %1.sub0, 1, 15, 15, 1, implicit $exec
+    %3:vgpr_32 = V_MOV_B32_dpp %0.sub1, %1.sub1, 1, 15, 15, 1, implicit $exec
+    %4:vreg_64 = REG_SEQUENCE %2, %subreg.sub0, %3, %subreg.sub1
+    %5:vgpr_32 = V_ADD_U32_e32 %4.sub0, %4.sub0, implicit $exec
+    %6:vgpr_32 = V_ADDC_U32_e32 %4.sub1, %4.sub1, implicit-def $vcc, implicit $vcc, implicit $exec
+...

llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mov.dpp.ll

 ; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs -show-mc-encoding < %s | FileCheck -check-prefixes=VI,VI-OPT,PREGFX10,PREGFX10-OPT %s
 ; RUN: llc -O0 -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs -show-mc-encoding < %s | FileCheck -check-prefixes=VI,VI-NOOPT,PREGFX10,PREGFX10-NOOPT %s
 ; RUN: llc -march=amdgcn -mcpu=gfx1010 -mattr=-flat-for-global -verify-machineinstrs -show-mc-encoding < %s | FileCheck -check-prefixes=VI,VI-OPT %s
+; RUN: llc -march=amdgcn -mcpu=gfx1100 -mattr=-flat-for-global -amdgpu-enable-vopd=0 -verify-machineinstrs -show-mc-encoding < %s | FileCheck -check-prefixes=VI,VI-OPT %s
 
 ; FIXME: The register allocator / scheduler should be able to avoid these hazards.
 
 ; VI-LABEL: {{^}}dpp_test:
 ; VI: v_mov_b32_e32 v0, s{{[0-9]+}}
 ; VI-NOOPT: v_mov_b32_e32 v1, s{{[0-9]+}}
 ; PREGFX10: s_nop 1
 ; VI-OPT: v_mov_b32_dpp v0, v0 quad_perm:[1,0,0,0] row_mask:0x1 bank_mask:0x1 bound_ctrl:1 ; encoding: [0xfa,0x02,0x00,0x7e,0x00,0x01,0x08,0x11]

llvm/test/CodeGen/AMDGPU/llvm.amdgcn.update.dpp.ll

 ; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -amdgpu-dpp-combine=false -verify-machineinstrs < %s | FileCheck --check-prefixes=GCN,GFX8,GFX8-OPT,GCN-OPT %s
 ; RUN: llc -march=amdgcn -mcpu=tonga -O0 -mattr=-flat-for-global -amdgpu-dpp-combine=false -verify-machineinstrs < %s | FileCheck --check-prefixes=GCN,GFX8,GFX8-NOOPT %s
 ; RUN: llc -march=amdgcn -mcpu=gfx1010 -mattr=-flat-for-global -amdgpu-dpp-combine=false -verify-machineinstrs < %s | FileCheck --check-prefixes=GCN,GFX10,GCN-OPT %s
+; RUN: llc -march=amdgcn -mcpu=gfx1100 -mattr=-flat-for-global -amdgpu-enable-vopd=0 -amdgpu-dpp-combine=false -verify-machineinstrs < %s | FileCheck --check-prefixes=GCN,GFX11,GCN-OPT %s
 
 ; GCN-LABEL: {{^}}dpp_test:
 ; GCN:  v_mov_b32_e32 [[DST:v[0-9]+]], s{{[0-9]+}}
 ; GCN:  v_mov_b32_e32 [[SRC:v[0-9]+]], s{{[0-9]+}}
 ; GFX8: s_nop 1
 ; GCN:  v_mov_b32_dpp [[DST]], [[SRC]] quad_perm:[1,0,0,0] row_mask:0x1 bank_mask:0x1{{$}}
 define amdgpu_kernel void @dpp_test(i32 addrspace(1)* %out, i32 %in1, i32 %in2) {
   %tmp0 = call i32 @llvm.amdgcn.update.dpp.i32(i32 %in1, i32 %in2, i32 1, i32 1, i32 1, i1 0) #0

 define amdgpu_kernel void @dpp_test_bc(i32 addrspace(1)* %out, i32 %in1, i32 %in2) {
   %tmp0 = call i32 @llvm.amdgcn.update.dpp.i32(i32 %in1, i32 %in2, i32 2, i32 1, i32 1, i1 1) #0
   store i32 %tmp0, i32 addrspace(1)* %out
   ret void
 }
 
 
 ; GCN-LABEL: {{^}}dpp_test1:
-; GFX10: v_add_nc_u32_e32 [[REG:v[0-9]+]], v{{[0-9]+}}, v{{[0-9]+}}
+; GFX10,GFX11: v_add_nc_u32_e32 [[REG:v[0-9]+]], v{{[0-9]+}}, v{{[0-9]+}}
 ; GFX8-OPT: v_add_u32_e32 [[REG:v[0-9]+]], vcc, v{{[0-9]+}}, v{{[0-9]+}}
 ; GFX8-NOOPT: v_add_u32_e64 [[REG:v[0-9]+]], s{{\[[0-9]+:[0-9]+\]}}, v{{[0-9]+}}, v{{[0-9]+}}
 ; GFX8-NOOPT: v_mov_b32_e32 v{{[0-9]+}}, 0
 ; GFX8: s_nop 1
 ; GFX8-NEXT: v_mov_b32_dpp {{v[0-9]+}}, [[REG]] quad_perm:[1,0,3,2] row_mask:0xf bank_mask:0xf
 @0 = internal unnamed_addr addrspace(3) global [448 x i32] undef, align 4
 define weak_odr amdgpu_kernel void @dpp_test1(i32* %arg) local_unnamed_addr {
 bb:
   %tmp = tail call i32 @llvm.amdgcn.workitem.id.x()
   %tmp1 = zext i32 %tmp to i64
   %tmp2 = getelementptr inbounds [448 x i32], [448 x i32] addrspace(3)* @0, i32 0, i32 %tmp
   %tmp3 = load i32, i32 addrspace(3)* %tmp2, align 4
   fence syncscope("workgroup-one-as") release
   tail call void @llvm.amdgcn.s.barrier()
   fence syncscope("workgroup-one-as") acquire
   %tmp4 = add nsw i32 %tmp3, %tmp3
   %tmp5 = tail call i32 @llvm.amdgcn.update.dpp.i32(i32 0, i32 %tmp4, i32 177, i32 15, i32 15, i1 zeroext false)
   %tmp6 = add nsw i32 %tmp5, %tmp4
   %tmp7 = getelementptr inbounds i32, i32* %arg, i64 %tmp1
   store i32 %tmp6, i32* %tmp7, align 4
   ret void
 }
 
 ; GCN-LABEL: {{^}}update_dpp64_test:
-; GCN:     load_dwordx2 v[[[SRC_LO:[0-9]+]]:[[SRC_HI:[0-9]+]]]
+; GCN:     load_{{dwordx2|b64}} v[[[SRC_LO:[0-9]+]]:[[SRC_HI:[0-9]+]]]
 ; GCN-DAG: v_mov_b32_dpp v{{[0-9]+}}, v[[SRC_LO]] quad_perm:[1,0,0,0] row_mask:0x1 bank_mask:0x1{{$}}
 ; GCN-DAG: v_mov_b32_dpp v{{[0-9]+}}, v[[SRC_HI]] quad_perm:[1,0,0,0] row_mask:0x1 bank_mask:0x1{{$}}
 define amdgpu_kernel void @update_dpp64_test(i64 addrspace(1)* %arg, i64 %in1, i64 %in2) {
   %id = tail call i32 @llvm.amdgcn.workitem.id.x()
   %gep = getelementptr inbounds i64, i64 addrspace(1)* %arg, i32 %id
   %load = load i64, i64 addrspace(1)* %gep
   %tmp0 = call i64 @llvm.amdgcn.update.dpp.i64(i64 %in1, i64 %load, i32 1, i32 1, i32 1, i1 0) #0
   store i64 %tmp0, i64 addrspace(1)* %gep
   ret void
 }
 
 ; GCN-LABEL: {{^}}update_dpp64_imm_old_test:
 ; GCN-OPT-DAG: v_mov_b32_e32 v[[OLD_LO:[0-9]+]], 0x3afaedd9
-; GCN-OPT-DAG: v_mov_b32_e32 v[[OLD_HI:[0-9]+]], 0x7047
+; GFX8-OPT-DAG,GFX10-DAG: v_mov_b32_e32 v[[OLD_HI:[0-9]+]], 0x7047
+; GFX11-DAG: v_mov_b32_e32 v[[OLD_HI:[0-9]+]], 0x7047
 ; GFX8-NOOPT-DAG: s_mov_b32 s[[SOLD_LO:[0-9]+]], 0x3afaedd9
 ; GFX8-NOOPT-DAG: s_mov_b32 s[[SOLD_HI:[0-9]+]], 0x7047
-; GCN-DAG: load_dwordx2 v[[[SRC_LO:[0-9]+]]:[[SRC_HI:[0-9]+]]]
+; GCN-DAG: load_{{dwordx2|b64}} v[[[SRC_LO:[0-9]+]]:[[SRC_HI:[0-9]+]]]
 ; GCN-OPT-DAG: v_mov_b32_dpp v[[OLD_LO]], v[[SRC_LO]] quad_perm:[1,0,0,0] row_mask:0x1 bank_mask:0x1{{$}}
-; GCN-OPT-DAG: v_mov_b32_dpp v[[OLD_HI]], v[[SRC_HI]] quad_perm:[1,0,0,0] row_mask:0x1 bank_mask:0x1{{$}}
+; GFX8-OPT-DAG,GFX10-DAG,GFX11-DAG: v_mov_b32_dpp v[[OLD_HI]], v[[SRC_HI]] quad_perm:[1,0,0,0] row_mask:0x1 bank_mask:0x1{{$}}
 ; GCN-NOOPT-DAG: v_mov_b32_dpp v{{[0-9]+}}, v[[SRC_LO]] quad_perm:[1,0,0,0] row_mask:0x1 bank_mask:0x1{{$}}
 ; GCN-NOOPT-DAG: v_mov_b32_dpp v{{[0-9]+}}, v[[SRC_HI]] quad_perm:[1,0,0,0] row_mask:0x1 bank_mask:0x1{{$}}
 define amdgpu_kernel void @update_dpp64_imm_old_test(i64 addrspace(1)* %arg, i64 %in2) {
   %id = tail call i32 @llvm.amdgcn.workitem.id.x()
   %gep = getelementptr inbounds i64, i64 addrspace(1)* %arg, i32 %id
   %load = load i64, i64 addrspace(1)* %gep
   %tmp0 = call i64 @llvm.amdgcn.update.dpp.i64(i64 123451234512345, i64 %load, i32 1, i32 1, i32 1, i1 0) #0
   store i64 %tmp0, i64 addrspace(1)* %gep

llvm/test/CodeGen/AMDGPU/vopc_dpp.mir

+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
+# RUN: llc -march=amdgcn -mcpu=gfx1100 -run-pass=gcn-dpp-combine -verify-machineinstrs -o - %s | FileCheck %s -check-prefix=GCN
+
+---
+
+name: vopc
+tracksRegLiveness: true
+body:             |
+  bb.0:
+    liveins: $vgpr0, $vgpr1, $vgpr2
+
+    ; GCN-LABEL: name: vopc
+    ; GCN: liveins: $vgpr0, $vgpr1, $vgpr2
+    ; GCN: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+    ; GCN: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr1
+    ; GCN: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+    ; GCN: [[DEF:%[0-9]+]]:vgpr_32 = IMPLICIT_DEF
+    ; GCN: V_CMP_LT_F32_e32_dpp [[DEF]], 0, [[COPY1]], 0, [[COPY]], 1, 15, 15, 1, implicit-def $vcc, implicit $mode, implicit $exec
+    ; GCN: [[V_MOV_B32_dpp:%[0-9]+]]:vgpr_32 = V_MOV_B32_dpp [[DEF]], [[COPY1]], 1, 15, 15, 1, implicit $exec
+    ; GCN: V_CMPX_EQ_I16_e32 [[V_MOV_B32_dpp]], [[COPY]], implicit-def $exec, implicit-def $vcc, implicit $mode, implicit $exec
+    ; GCN: V_CMP_CLASS_F16_e32_dpp [[DEF]], 0, [[COPY1]], [[COPY]], 1, 15, 15, 1, implicit-def $vcc, implicit $exec
+    ; GCN: [[V_CMP_GE_F16_e64_dpp:%[0-9]+]]:sgpr_32 = V_CMP_GE_F16_e64_dpp [[DEF]], 1, [[COPY1]], 0, [[COPY]], 1, 1, 15, 15, 1, implicit $mode, implicit $exec
+    ; GCN: [[V_MOV_B32_dpp1:%[0-9]+]]:vgpr_32 = V_MOV_B32_dpp [[DEF]], [[COPY1]], 1, 15, 15, 1, implicit $exec
+    ; GCN: V_CMPX_GT_U32_nosdst_e64 [[V_MOV_B32_dpp1]], [[COPY]], implicit-def $exec, implicit $mode, implicit $exec
+    ; GCN: V_CMP_CLASS_F32_e32_dpp [[DEF]], 2, [[COPY1]], [[COPY]], 1, 15, 15, 1, implicit-def $vcc, implicit $exec
+    ; GCN: V_CMP_NGE_F16_e32_dpp [[DEF]], 0, [[COPY1]], 0, [[COPY]], 1, 15, 15, 1, implicit-def $vcc, implicit $mode, implicit $exec
+    ; GCN: [[V_CMP_NGE_F16_e64_dpp:%[0-9]+]]:sgpr_32 = V_CMP_NGE_F16_e64_dpp [[DEF]], 0, [[COPY1]], 0, [[COPY]], 0, 1, 15, 15, 1, implicit $mode, implicit $exec
+    ; GCN: [[S_AND_B32_:%[0-9]+]]:sgpr_32 = S_AND_B32 [[V_CMP_NGE_F16_e64_dpp]], 10101, implicit-def $scc
+    ; GCN: V_CMP_GT_I32_e32_dpp [[DEF]], [[COPY1]], [[COPY]], 1, 15, 15, 1, implicit-def $vcc, implicit $exec
+    %0:vgpr_32 = COPY $vgpr0
+    %1:vgpr_32 = COPY $vgpr1
+    %2:vgpr_32 = COPY $vgpr2
+    %3:vgpr_32 = IMPLICIT_DEF
+
+    %4:vgpr_32 = V_MOV_B32_dpp %3, %1, 1, 15, 15, 1, implicit $exec
+    V_CMP_LT_F32_e32 %4, %0, implicit-def $vcc, implicit $mode, implicit $exec
+
+    ; unsafe to combine cmpx
+    %5:vgpr_32 = V_MOV_B32_dpp %3, %1, 1, 15, 15, 1, implicit $exec
+    V_CMPX_EQ_I16_e32 %5, %0, implicit-def $exec, implicit-def $vcc, implicit $mode, implicit $exec
+
+    %6:vgpr_32 = V_MOV_B32_dpp %3, %1, 1, 15, 15, 1, implicit $exec
+    V_CMP_CLASS_F16_e32 %6, %0, implicit-def $vcc, implicit $mode, implicit $exec
+
+    %7:vgpr_32 = V_MOV_B32_dpp %3, %1, 1, 15, 15, 1, implicit $exec
+    %8:sgpr_32 = V_CMP_GE_F16_e64 1, %7, 0, %0, 1, implicit $mode, implicit $exec
+
+    ; unsafe to combine cmpx
+    %9:vgpr_32 = V_MOV_B32_dpp %3, %1, 1, 15, 15, 1, implicit $exec
+    V_CMPX_GT_U32_nosdst_e64 %9, %0, implicit-def $exec, implicit $mode, implicit $exec
+
+    %11:vgpr_32 = V_MOV_B32_dpp %3, %1, 1, 15, 15, 1, implicit $exec
+    %12:sgpr_32 = V_CMP_CLASS_F32_e64 2, %11, %0, implicit $mode, implicit $exec
+
+    ; shrink
+    %13:vgpr_32 = V_MOV_B32_dpp %3, %1, 1, 15, 15, 1, implicit $exec
+    %14:sgpr_32 = V_CMP_NGE_F16_e64 0, %13, 0, %0, 0, implicit $mode, implicit $exec
+
+    ; do not shrink, sdst used
+    %15:vgpr_32 = V_MOV_B32_dpp %3, %1, 1, 15, 15, 1, implicit $exec
+    %16:sgpr_32 = V_CMP_NGE_F16_e64 0, %15, 0, %0, 0, implicit $mode, implicit $exec
+    %17:sgpr_32 = S_AND_B32 %16, 10101, implicit-def $scc
+
+    ; commute
+    %18:vgpr_32 = V_MOV_B32_dpp %3, %1, 1, 15, 15, 1, implicit $exec
+    V_CMP_LT_I32_e32 %0, %18, implicit-def $vcc, implicit $exec
+
+...