[AMDGPU] gfx11 Remove SDWA from shuffle_vector ISel

[Joe_Nash]

Summary

gfx11 does not have SDWA

llvm/lib/Target/AMDGPU/AMDGPUInstructionSelector.cpp

bool AMDGPUInstructionSelector::selectG_SHUFFLE_VECTOR(
   const LLT V2S16 = LLT::fixed_vector(2, 16);
   if (MRI->getType(DstReg) != V2S16 || MRI->getType(Src0Reg) != V2S16)
     return false;
 
   if (!AMDGPU::isLegalVOP3PShuffleMask(ShufMask))
     return false;
 
   assert(ShufMask.size() == 2);
-  assert(STI.hasSDWA() && "no target has VOP3P but not SDWA");
 
   MachineBasicBlock *MBB = MI.getParent();
   const DebugLoc &DL = MI.getDebugLoc();
 
   const RegisterBank *DstRB = RBI.getRegBank(DstReg, *MRI, TRI);
   const bool IsVALU = DstRB->getID() == AMDGPU::VGPRRegBankID;
   const TargetRegisterClass &RC = IsVALU ?
     AMDGPU::VGPR_32RegClass : AMDGPU::SReg_32RegClass;
if (IsVALU) {
         .addReg(SrcVec);
     } else {
       BuildMI(*MBB, MI, DL, TII.get(AMDGPU::S_LSHL_B32), DstReg)
         .addReg(SrcVec)
         .addImm(16);
     }
   } else if (Mask[0] == 0 && Mask[1] == 0) {
     if (IsVALU) {
+      if (STI.hasSDWA()) {
         // Write low half of the register into the high half.
         MachineInstr *MovSDWA =
             BuildMI(*MBB, MI, DL, TII.get(AMDGPU::V_MOV_B32_sdwa), DstReg)
                 .addImm(0)                             // $src0_modifiers
                 .addReg(SrcVec)                        // $src0
                 .addImm(0)                             // $clamp
                 .addImm(AMDGPU::SDWA::WORD_1)          // $dst_sel
                 .addImm(AMDGPU::SDWA::UNUSED_PRESERVE) // $dst_unused
                 .addImm(AMDGPU::SDWA::WORD_0)          // $src0_sel
                 .addReg(SrcVec, RegState::Implicit);
         MovSDWA->tieOperands(0, MovSDWA->getNumOperands() - 1);
       } else {
+        Register TmpReg = MRI->createVirtualRegister(&AMDGPU::VGPR_32RegClass);
+        BuildMI(*MBB, MI, DL, TII.get(AMDGPU::V_AND_B32_e32), TmpReg)
+            .addImm(0xFFFF)
+            .addReg(SrcVec);
+        BuildMI(*MBB, MI, DL, TII.get(AMDGPU::V_LSHL_OR_B32_e64), DstReg)
+            .addReg(TmpReg)
+            .addImm(16)
+            .addReg(TmpReg);
+      }
+    } else {
       BuildMI(*MBB, MI, DL, TII.get(AMDGPU::S_PACK_LL_B32_B16), DstReg)
         .addReg(SrcVec)
         .addReg(SrcVec);
     }
   } else if (Mask[0] == 1 && Mask[1] == 1) {
     if (IsVALU) {
+      if (STI.hasSDWA()) {
         // Write high half of the register into the low half.
         MachineInstr *MovSDWA =
             BuildMI(*MBB, MI, DL, TII.get(AMDGPU::V_MOV_B32_sdwa), DstReg)
                 .addImm(0)                             // $src0_modifiers
                 .addReg(SrcVec)                        // $src0
                 .addImm(0)                             // $clamp
                 .addImm(AMDGPU::SDWA::WORD_0)          // $dst_sel
                 .addImm(AMDGPU::SDWA::UNUSED_PRESERVE) // $dst_unused
                 .addImm(AMDGPU::SDWA::WORD_1)          // $src0_sel
                 .addReg(SrcVec, RegState::Implicit);
         MovSDWA->tieOperands(0, MovSDWA->getNumOperands() - 1);
       } else {
+        Register TmpReg = MRI->createVirtualRegister(&AMDGPU::VGPR_32RegClass);
+        BuildMI(*MBB, MI, DL, TII.get(AMDGPU::V_LSHRREV_B32_e64), TmpReg)
+            .addImm(16)
+            .addReg(SrcVec);
+        BuildMI(*MBB, MI, DL, TII.get(AMDGPU::V_LSHL_OR_B32_e64), DstReg)
+            .addReg(TmpReg)
+            .addImm(16)
+            .addReg(TmpReg);
+      }
+    } else {
       BuildMI(*MBB, MI, DL, TII.get(AMDGPU::S_PACK_HH_B32_B16), DstReg)
         .addReg(SrcVec)
         .addReg(SrcVec);
     }
   } else if (Mask[0] == 1 && Mask[1] == 0) {
     if (IsVALU) {
       BuildMI(*MBB, MI, DL, TII.get(AMDGPU::V_ALIGNBIT_B32_e64), DstReg)
         .addReg(SrcVec)
         .addReg(SrcVec)
         .addImm(16);
     } else {
+      if (STI.hasSPackHL()) {
+        BuildMI(*MBB, MI, DL, TII.get(AMDGPU::S_PACK_HL_B32_B16), DstReg)
+            .addReg(SrcVec)
+            .addReg(SrcVec);
+      } else {
         Register TmpReg = MRI->createVirtualRegister(&AMDGPU::SReg_32RegClass);
         BuildMI(*MBB, MI, DL, TII.get(AMDGPU::S_LSHR_B32), TmpReg)
             .addReg(SrcVec)
             .addImm(16);
         BuildMI(*MBB, MI, DL, TII.get(AMDGPU::S_PACK_LL_B32_B16), DstReg)
             .addReg(TmpReg)
             .addReg(SrcVec);
       }
+    }
   } else
     llvm_unreachable("all shuffle masks should be handled");
 
   MI.eraseFromParent();
   return true;
 }
 
 bool AMDGPUInstructionSelector::selectAMDGPU_BUFFER_ATOMIC_FADD(

llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-shuffle-vector.v2s16.mir

 # NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
 # RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx900 -run-pass=instruction-select -o - %s | FileCheck -check-prefix=GFX9 %s
 # RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 -run-pass=instruction-select -o - %s | FileCheck -check-prefix=GFX9 %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1100 -run-pass=instruction-select -o - %s | FileCheck -check-prefix=GFX11 %s
 
 ---
 name: v_shufflevector_v2s16_v2s16_u_u
 tracksRegLiveness: true
 legalized:       true
 regBankSelected: true
 
 body: |
   bb.0:
     liveins: $vgpr0, $vgpr1
 
     ; GFX9-LABEL: name: v_shufflevector_v2s16_v2s16_u_u
     ; GFX9: liveins: $vgpr0, $vgpr1
     ; GFX9: [[DEF:%[0-9]+]]:vgpr_32 = IMPLICIT_DEF
     ; GFX9: $vgpr0 = COPY [[DEF]]
+    ; GFX11-LABEL: name: v_shufflevector_v2s16_v2s16_u_u
+    ; GFX11: liveins: $vgpr0, $vgpr1
+    ; GFX11: [[DEF:%[0-9]+]]:vgpr_32 = IMPLICIT_DEF
+    ; GFX11: $vgpr0 = COPY [[DEF]]
     %0:vgpr(<2 x s16>) = COPY $vgpr0
     %1:vgpr(<2 x s16>) = COPY $vgpr1
     %2:vgpr(<2 x s16>) = G_SHUFFLE_VECTOR %0, %1, shufflemask(undef, undef)
     $vgpr0 = COPY %2
 
 ...
 
 ---
 name: v_shufflevector_v2s16_v2s16_0_u
 tracksRegLiveness: true
 legalized:       true
 regBankSelected: true
 
 body: |
   bb.0:
     liveins: $vgpr0, $vgpr1
 
     ; GFX9-LABEL: name: v_shufflevector_v2s16_v2s16_0_u
     ; GFX9: liveins: $vgpr0, $vgpr1
     ; GFX9: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
     ; GFX9: $vgpr0 = COPY [[COPY]]
+    ; GFX11-LABEL: name: v_shufflevector_v2s16_v2s16_0_u
+    ; GFX11: liveins: $vgpr0, $vgpr1
+    ; GFX11: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+    ; GFX11: $vgpr0 = COPY [[COPY]]
     %0:vgpr(<2 x s16>) = COPY $vgpr0
     %1:vgpr(<2 x s16>) = COPY $vgpr1
     %2:vgpr(<2 x s16>) = G_SHUFFLE_VECTOR %0, %1, shufflemask(0, undef)
     $vgpr0 = COPY %2
 
 ...
 
 ---
 name: v_shufflevector_v2s16_v2s16_u_0
 tracksRegLiveness: true
 legalized:       true
 regBankSelected: true
 
 body: |
   bb.0:
     liveins: $vgpr0, $vgpr1
 
     ; GFX9-LABEL: name: v_shufflevector_v2s16_v2s16_u_0
     ; GFX9: liveins: $vgpr0, $vgpr1
     ; GFX9: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
     ; GFX9: [[V_LSHLREV_B32_e64_:%[0-9]+]]:vgpr_32 = V_LSHLREV_B32_e64 16, [[COPY]], implicit $exec
     ; GFX9: $vgpr0 = COPY [[V_LSHLREV_B32_e64_]]
+    ; GFX11-LABEL: name: v_shufflevector_v2s16_v2s16_u_0
+    ; GFX11: liveins: $vgpr0, $vgpr1
+    ; GFX11: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+    ; GFX11: [[V_LSHLREV_B32_e64_:%[0-9]+]]:vgpr_32 = V_LSHLREV_B32_e64 16, [[COPY]], implicit $exec
+    ; GFX11: $vgpr0 = COPY [[V_LSHLREV_B32_e64_]]
     %0:vgpr(<2 x s16>) = COPY $vgpr0
     %1:vgpr(<2 x s16>) = COPY $vgpr1
     %2:vgpr(<2 x s16>) = G_SHUFFLE_VECTOR %0, %1, shufflemask(undef, 0)
     $vgpr0 = COPY %2
 
 ...
 
 ---
 name: v_shufflevector_v2s16_v2s16_1_u
 tracksRegLiveness: true
 legalized:       true
 regBankSelected: true
 
 body: |
   bb.0:
     liveins: $vgpr0, $vgpr1
 
     ; GFX9-LABEL: name: v_shufflevector_v2s16_v2s16_1_u
     ; GFX9: liveins: $vgpr0, $vgpr1
     ; GFX9: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
     ; GFX9: [[V_LSHRREV_B32_e64_:%[0-9]+]]:vgpr_32 = V_LSHRREV_B32_e64 16, [[COPY]], implicit $exec
     ; GFX9: $vgpr0 = COPY [[V_LSHRREV_B32_e64_]]
+    ; GFX11-LABEL: name: v_shufflevector_v2s16_v2s16_1_u
+    ; GFX11: liveins: $vgpr0, $vgpr1
+    ; GFX11: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+    ; GFX11: [[V_LSHRREV_B32_e64_:%[0-9]+]]:vgpr_32 = V_LSHRREV_B32_e64 16, [[COPY]], implicit $exec
+    ; GFX11: $vgpr0 = COPY [[V_LSHRREV_B32_e64_]]
     %0:vgpr(<2 x s16>) = COPY $vgpr0
     %1:vgpr(<2 x s16>) = COPY $vgpr1
     %2:vgpr(<2 x s16>) = G_SHUFFLE_VECTOR %0, %1, shufflemask(1, undef)
     $vgpr0 = COPY %2
 
 ...
 
 ---
 name: v_shufflevector_v2s16_v2s16_u_1
 tracksRegLiveness: true
 legalized:       true
 regBankSelected: true
 
 body: |
   bb.0:
     liveins: $vgpr0, $vgpr1
 
     ; GFX9-LABEL: name: v_shufflevector_v2s16_v2s16_u_1
     ; GFX9: liveins: $vgpr0, $vgpr1
     ; GFX9: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
     ; GFX9: $vgpr0 = COPY [[COPY]]
+    ; GFX11-LABEL: name: v_shufflevector_v2s16_v2s16_u_1
+    ; GFX11: liveins: $vgpr0, $vgpr1
+    ; GFX11: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+    ; GFX11: $vgpr0 = COPY [[COPY]]
     %0:vgpr(<2 x s16>) = COPY $vgpr0
     %1:vgpr(<2 x s16>) = COPY $vgpr1
     %2:vgpr(<2 x s16>) = G_SHUFFLE_VECTOR %0, %1, shufflemask(undef, 1)
     $vgpr0 = COPY %2
 
 ...
 
 
 ---
 name: v_shufflevector_v2s16_v2s16_2_u
 tracksRegLiveness: true
 legalized:       true
 regBankSelected: true
 
 body: |
   bb.0:
     liveins: $vgpr0, $vgpr1
 
     ; GFX9-LABEL: name: v_shufflevector_v2s16_v2s16_2_u
     ; GFX9: liveins: $vgpr0, $vgpr1
     ; GFX9: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr1
     ; GFX9: $vgpr0 = COPY [[COPY]]
+    ; GFX11-LABEL: name: v_shufflevector_v2s16_v2s16_2_u
+    ; GFX11: liveins: $vgpr0, $vgpr1
+    ; GFX11: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr1
+    ; GFX11: $vgpr0 = COPY [[COPY]]
     %0:vgpr(<2 x s16>) = COPY $vgpr0
     %1:vgpr(<2 x s16>) = COPY $vgpr1
     %2:vgpr(<2 x s16>) = G_SHUFFLE_VECTOR %0, %1, shufflemask(2, undef)
     $vgpr0 = COPY %2
 
 ...
 
 ---
 name: v_shufflevector_v2s16_v2s16_u_2
 tracksRegLiveness: true
 legalized:       true
 regBankSelected: true
 
 body: |
   bb.0:
     liveins: $vgpr0, $vgpr1
 
     ; GFX9-LABEL: name: v_shufflevector_v2s16_v2s16_u_2
     ; GFX9: liveins: $vgpr0, $vgpr1
     ; GFX9: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr1
     ; GFX9: [[V_LSHLREV_B32_e64_:%[0-9]+]]:vgpr_32 = V_LSHLREV_B32_e64 16, [[COPY]], implicit $exec
     ; GFX9: $vgpr0 = COPY [[V_LSHLREV_B32_e64_]]
+    ; GFX11-LABEL: name: v_shufflevector_v2s16_v2s16_u_2
+    ; GFX11: liveins: $vgpr0, $vgpr1
+    ; GFX11: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr1
+    ; GFX11: [[V_LSHLREV_B32_e64_:%[0-9]+]]:vgpr_32 = V_LSHLREV_B32_e64 16, [[COPY]], implicit $exec
+    ; GFX11: $vgpr0 = COPY [[V_LSHLREV_B32_e64_]]
     %0:vgpr(<2 x s16>) = COPY $vgpr0
     %1:vgpr(<2 x s16>) = COPY $vgpr1
     %2:vgpr(<2 x s16>) = G_SHUFFLE_VECTOR %0, %1, shufflemask(undef, 2)
     $vgpr0 = COPY %2
 
 ...
 
 ---
 name: v_shufflevector_v2s16_v2s16_3_u
 tracksRegLiveness: true
 legalized:       true
 regBankSelected: true
 
 body: |
   bb.0:
     liveins: $vgpr0, $vgpr1
 
     ; GFX9-LABEL: name: v_shufflevector_v2s16_v2s16_3_u
     ; GFX9: liveins: $vgpr0, $vgpr1
     ; GFX9: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr1
     ; GFX9: [[V_LSHRREV_B32_e64_:%[0-9]+]]:vgpr_32 = V_LSHRREV_B32_e64 16, [[COPY]], implicit $exec
     ; GFX9: $vgpr0 = COPY [[V_LSHRREV_B32_e64_]]
+    ; GFX11-LABEL: name: v_shufflevector_v2s16_v2s16_3_u
+    ; GFX11: liveins: $vgpr0, $vgpr1
+    ; GFX11: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr1
+    ; GFX11: [[V_LSHRREV_B32_e64_:%[0-9]+]]:vgpr_32 = V_LSHRREV_B32_e64 16, [[COPY]], implicit $exec
+    ; GFX11: $vgpr0 = COPY [[V_LSHRREV_B32_e64_]]
     %0:vgpr(<2 x s16>) = COPY $vgpr0
     %1:vgpr(<2 x s16>) = COPY $vgpr1
     %2:vgpr(<2 x s16>) = G_SHUFFLE_VECTOR %0, %1, shufflemask(3, undef)
     $vgpr0 = COPY %2
 
 ...
 
 ---
 name: v_shufflevector_v2s16_v2s16_u_3
 tracksRegLiveness: true
 legalized:       true
 regBankSelected: true
 
 body: |
   bb.0:
     liveins: $vgpr0, $vgpr1
 
     ; GFX9-LABEL: name: v_shufflevector_v2s16_v2s16_u_3
     ; GFX9: liveins: $vgpr0, $vgpr1
     ; GFX9: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr1
     ; GFX9: $vgpr0 = COPY [[COPY]]
+    ; GFX11-LABEL: name: v_shufflevector_v2s16_v2s16_u_3
+    ; GFX11: liveins: $vgpr0, $vgpr1
+    ; GFX11: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr1
+    ; GFX11: $vgpr0 = COPY [[COPY]]
     %0:vgpr(<2 x s16>) = COPY $vgpr0
     %1:vgpr(<2 x s16>) = COPY $vgpr1
     %2:vgpr(<2 x s16>) = G_SHUFFLE_VECTOR %0, %1, shufflemask(undef, 3)
     $vgpr0 = COPY %2
 
 ...
 
 ---
 name: v_shufflevector_v2s16_v2s16_0_0
 tracksRegLiveness: true
 legalized:       true
 regBankSelected: true
 
 body: |
   bb.0:
     liveins: $vgpr0, $vgpr1
 
     ; GFX9-LABEL: name: v_shufflevector_v2s16_v2s16_0_0
     ; GFX9: liveins: $vgpr0, $vgpr1
     ; GFX9: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
     ; GFX9: [[V_MOV_B32_sdwa:%[0-9]+]]:vgpr_32 = V_MOV_B32_sdwa 0, [[COPY]], 0, 5, 2, 4, implicit $exec, implicit [[COPY]](tied-def 0)
     ; GFX9: $vgpr0 = COPY [[V_MOV_B32_sdwa]]
+    ; GFX11-LABEL: name: v_shufflevector_v2s16_v2s16_0_0
+    ; GFX11: liveins: $vgpr0, $vgpr1
+    ; GFX11: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+    ; GFX11: [[V_AND_B32_e32_:%[0-9]+]]:vgpr_32 = V_AND_B32_e32 65535, [[COPY]], implicit $exec
+    ; GFX11: [[V_LSHL_OR_B32_e64_:%[0-9]+]]:vgpr_32 = V_LSHL_OR_B32_e64 [[V_AND_B32_e32_]], 16, [[V_AND_B32_e32_]], implicit $exec
+    ; GFX11: $vgpr0 = COPY [[V_LSHL_OR_B32_e64_]]
     %0:vgpr(<2 x s16>) = COPY $vgpr0
     %1:vgpr(<2 x s16>) = COPY $vgpr1
     %2:vgpr(<2 x s16>) = G_SHUFFLE_VECTOR %0, %1, shufflemask(0, 0)
     $vgpr0 = COPY %2
 
 ...
 
 ---
 name: v_shufflevector_v2s16_v2s16_0_1
 tracksRegLiveness: true
 legalized:       true
 regBankSelected: true
 
 body: |
   bb.0:
     liveins: $vgpr0, $vgpr1
 
     ; GFX9-LABEL: name: v_shufflevector_v2s16_v2s16_0_1
     ; GFX9: liveins: $vgpr0, $vgpr1
     ; GFX9: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
     ; GFX9: $vgpr0 = COPY [[COPY]]
+    ; GFX11-LABEL: name: v_shufflevector_v2s16_v2s16_0_1
+    ; GFX11: liveins: $vgpr0, $vgpr1
+    ; GFX11: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+    ; GFX11: $vgpr0 = COPY [[COPY]]
     %0:vgpr(<2 x s16>) = COPY $vgpr0
     %1:vgpr(<2 x s16>) = COPY $vgpr1
     %2:vgpr(<2 x s16>) = G_SHUFFLE_VECTOR %0, %1, shufflemask(0, 1)
     $vgpr0 = COPY %2
 
 ...
 
 ---
 name: v_shufflevector_v2s16_v2s16_1_0
 tracksRegLiveness: true
 legalized:       true
 regBankSelected: true
 
 body: |
   bb.0:
     liveins: $vgpr0, $vgpr1
 
     ; GFX9-LABEL: name: v_shufflevector_v2s16_v2s16_1_0
     ; GFX9: liveins: $vgpr0, $vgpr1
     ; GFX9: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
     ; GFX9: [[V_ALIGNBIT_B32_e64_:%[0-9]+]]:vgpr_32 = V_ALIGNBIT_B32_e64 [[COPY]], [[COPY]], 16, implicit $exec
     ; GFX9: $vgpr0 = COPY [[V_ALIGNBIT_B32_e64_]]
+    ; GFX11-LABEL: name: v_shufflevector_v2s16_v2s16_1_0
+    ; GFX11: liveins: $vgpr0, $vgpr1
+    ; GFX11: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+    ; GFX11: [[V_ALIGNBIT_B32_e64_:%[0-9]+]]:vgpr_32 = V_ALIGNBIT_B32_e64 [[COPY]], [[COPY]], 16, implicit $exec
+    ; GFX11: $vgpr0 = COPY [[V_ALIGNBIT_B32_e64_]]
     %0:vgpr(<2 x s16>) = COPY $vgpr0
     %1:vgpr(<2 x s16>) = COPY $vgpr1
     %2:vgpr(<2 x s16>) = G_SHUFFLE_VECTOR %0, %1, shufflemask(1, 0)
     $vgpr0 = COPY %2
 
 ...
 
 ---
 name: v_shufflevector_v2s16_v2s16_1_1
 tracksRegLiveness: true
 legalized:       true
 regBankSelected: true
 
 body: |
   bb.0:
     liveins: $vgpr0, $vgpr1
 
     ; GFX9-LABEL: name: v_shufflevector_v2s16_v2s16_1_1
     ; GFX9: liveins: $vgpr0, $vgpr1
     ; GFX9: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
     ; GFX9: [[V_MOV_B32_sdwa:%[0-9]+]]:vgpr_32 = V_MOV_B32_sdwa 0, [[COPY]], 0, 4, 2, 5, implicit $exec, implicit [[COPY]](tied-def 0)
     ; GFX9: $vgpr0 = COPY [[V_MOV_B32_sdwa]]
+    ; GFX11-LABEL: name: v_shufflevector_v2s16_v2s16_1_1
+    ; GFX11: liveins: $vgpr0, $vgpr1
+    ; GFX11: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+    ; GFX11: [[V_LSHRREV_B32_e64_:%[0-9]+]]:vgpr_32 = V_LSHRREV_B32_e64 16, [[COPY]], implicit $exec
+    ; GFX11: [[V_LSHL_OR_B32_e64_:%[0-9]+]]:vgpr_32 = V_LSHL_OR_B32_e64 [[V_LSHRREV_B32_e64_]], 16, [[V_LSHRREV_B32_e64_]], implicit $exec
+    ; GFX11: $vgpr0 = COPY [[V_LSHL_OR_B32_e64_]]
     %0:vgpr(<2 x s16>) = COPY $vgpr0
     %1:vgpr(<2 x s16>) = COPY $vgpr1
     %2:vgpr(<2 x s16>) = G_SHUFFLE_VECTOR %0, %1, shufflemask(1, 1)
     $vgpr0 = COPY %2
 
 ...
 
 ---
 name: v_shufflevector_v2s16_v2s16_2_2
 tracksRegLiveness: true
 legalized:       true
 regBankSelected: true
 
 body: |
   bb.0:
     liveins: $vgpr0, $vgpr1
 
     ; GFX9-LABEL: name: v_shufflevector_v2s16_v2s16_2_2
     ; GFX9: liveins: $vgpr0, $vgpr1
     ; GFX9: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr1
     ; GFX9: [[V_MOV_B32_sdwa:%[0-9]+]]:vgpr_32 = V_MOV_B32_sdwa 0, [[COPY]], 0, 5, 2, 4, implicit $exec, implicit [[COPY]](tied-def 0)
     ; GFX9: $vgpr0 = COPY [[V_MOV_B32_sdwa]]
+    ; GFX11-LABEL: name: v_shufflevector_v2s16_v2s16_2_2
+    ; GFX11: liveins: $vgpr0, $vgpr1
+    ; GFX11: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr1
+    ; GFX11: [[V_AND_B32_e32_:%[0-9]+]]:vgpr_32 = V_AND_B32_e32 65535, [[COPY]], implicit $exec
+    ; GFX11: [[V_LSHL_OR_B32_e64_:%[0-9]+]]:vgpr_32 = V_LSHL_OR_B32_e64 [[V_AND_B32_e32_]], 16, [[V_AND_B32_e32_]], implicit $exec
+    ; GFX11: $vgpr0 = COPY [[V_LSHL_OR_B32_e64_]]
     %0:vgpr(<2 x s16>) = COPY $vgpr0
     %1:vgpr(<2 x s16>) = COPY $vgpr1
     %2:vgpr(<2 x s16>) = G_SHUFFLE_VECTOR %0, %1, shufflemask(2, 2)
     $vgpr0 = COPY %2
 
 ...
 
 ---
 name: v_shufflevector_v2s16_v2s16_2_3
 tracksRegLiveness: true
 legalized:       true
 regBankSelected: true
 
 body: |
   bb.0:
     liveins: $vgpr0, $vgpr1
 
     ; GFX9-LABEL: name: v_shufflevector_v2s16_v2s16_2_3
     ; GFX9: liveins: $vgpr0, $vgpr1
     ; GFX9: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr1
     ; GFX9: $vgpr0 = COPY [[COPY]]
+    ; GFX11-LABEL: name: v_shufflevector_v2s16_v2s16_2_3
+    ; GFX11: liveins: $vgpr0, $vgpr1
+    ; GFX11: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr1
+    ; GFX11: $vgpr0 = COPY [[COPY]]
     %0:vgpr(<2 x s16>) = COPY $vgpr0
     %1:vgpr(<2 x s16>) = COPY $vgpr1
     %2:vgpr(<2 x s16>) = G_SHUFFLE_VECTOR %0, %1, shufflemask(2, 3)
     $vgpr0 = COPY %2
 
 ...
 
 ---
 name: v_shufflevector_v2s16_v2s16_3_2
 tracksRegLiveness: true
 legalized:       true
 regBankSelected: true
 
 body: |
   bb.0:
     liveins: $vgpr0, $vgpr1
 
     ; GFX9-LABEL: name: v_shufflevector_v2s16_v2s16_3_2
     ; GFX9: liveins: $vgpr0, $vgpr1
     ; GFX9: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr1
     ; GFX9: [[V_ALIGNBIT_B32_e64_:%[0-9]+]]:vgpr_32 = V_ALIGNBIT_B32_e64 [[COPY]], [[COPY]], 16, implicit $exec
     ; GFX9: $vgpr0 = COPY [[V_ALIGNBIT_B32_e64_]]
+    ; GFX11-LABEL: name: v_shufflevector_v2s16_v2s16_3_2
+    ; GFX11: liveins: $vgpr0, $vgpr1
+    ; GFX11: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr1
+    ; GFX11: [[V_ALIGNBIT_B32_e64_:%[0-9]+]]:vgpr_32 = V_ALIGNBIT_B32_e64 [[COPY]], [[COPY]], 16, implicit $exec
+    ; GFX11: $vgpr0 = COPY [[V_ALIGNBIT_B32_e64_]]
     %0:vgpr(<2 x s16>) = COPY $vgpr0
     %1:vgpr(<2 x s16>) = COPY $vgpr1
     %2:vgpr(<2 x s16>) = G_SHUFFLE_VECTOR %0, %1, shufflemask(3, 2)
     $vgpr0 = COPY %2
 
 ...
 
 ---
 name: v_shufflevector_v2s16_v2s16_3_3
 tracksRegLiveness: true
 legalized:       true
 regBankSelected: true
 
 body: |
   bb.0:
     liveins: $vgpr0, $vgpr1
 
     ; GFX9-LABEL: name: v_shufflevector_v2s16_v2s16_3_3
     ; GFX9: liveins: $vgpr0, $vgpr1
     ; GFX9: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr1
     ; GFX9: [[V_MOV_B32_sdwa:%[0-9]+]]:vgpr_32 = V_MOV_B32_sdwa 0, [[COPY]], 0, 4, 2, 5, implicit $exec, implicit [[COPY]](tied-def 0)
     ; GFX9: $vgpr0 = COPY [[V_MOV_B32_sdwa]]
+    ; GFX11-LABEL: name: v_shufflevector_v2s16_v2s16_3_3
+    ; GFX11: liveins: $vgpr0, $vgpr1
+    ; GFX11: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr1
+    ; GFX11: [[V_LSHRREV_B32_e64_:%[0-9]+]]:vgpr_32 = V_LSHRREV_B32_e64 16, [[COPY]], implicit $exec
+    ; GFX11: [[V_LSHL_OR_B32_e64_:%[0-9]+]]:vgpr_32 = V_LSHL_OR_B32_e64 [[V_LSHRREV_B32_e64_]], 16, [[V_LSHRREV_B32_e64_]], implicit $exec
+    ; GFX11: $vgpr0 = COPY [[V_LSHL_OR_B32_e64_]]
     %0:vgpr(<2 x s16>) = COPY $vgpr0
     %1:vgpr(<2 x s16>) = COPY $vgpr1
     %2:vgpr(<2 x s16>) = G_SHUFFLE_VECTOR %0, %1, shufflemask(3, 3)
     $vgpr0 = COPY %2
 
 ...
 
 ---
 name: s_shufflevector_v2s16_v2s16_u_u
 tracksRegLiveness: true
 legalized:       true
 regBankSelected: true
 
 body: |
   bb.0:
     liveins: $sgpr0, $sgpr1
 
     ; GFX9-LABEL: name: s_shufflevector_v2s16_v2s16_u_u
     ; GFX9: liveins: $sgpr0, $sgpr1
     ; GFX9: [[DEF:%[0-9]+]]:sreg_32 = IMPLICIT_DEF
     ; GFX9: $sgpr0 = COPY [[DEF]]
+    ; GFX11-LABEL: name: s_shufflevector_v2s16_v2s16_u_u
+    ; GFX11: liveins: $sgpr0, $sgpr1
+    ; GFX11: [[DEF:%[0-9]+]]:sreg_32 = IMPLICIT_DEF
+    ; GFX11: $sgpr0 = COPY [[DEF]]
     %0:sgpr(<2 x s16>) = COPY $sgpr0
     %1:sgpr(<2 x s16>) = COPY $sgpr1
     %2:sgpr(<2 x s16>) = G_SHUFFLE_VECTOR %0, %1, shufflemask(undef, undef)
     $sgpr0 = COPY %2
 
 ...
 
 ---
 name: s_shufflevector_v2s16_v2s16_0_u
 tracksRegLiveness: true
 legalized:       true
 regBankSelected: true
 
 body: |
   bb.0:
     liveins: $sgpr0, $sgpr1
 
     ; GFX9-LABEL: name: s_shufflevector_v2s16_v2s16_0_u
     ; GFX9: liveins: $sgpr0, $sgpr1
     ; GFX9: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr0
     ; GFX9: $sgpr0 = COPY [[COPY]]
+    ; GFX11-LABEL: name: s_shufflevector_v2s16_v2s16_0_u
+    ; GFX11: liveins: $sgpr0, $sgpr1
+    ; GFX11: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr0
+    ; GFX11: $sgpr0 = COPY [[COPY]]
     %0:sgpr(<2 x s16>) = COPY $sgpr0
     %1:sgpr(<2 x s16>) = COPY $sgpr1
     %2:sgpr(<2 x s16>) = G_SHUFFLE_VECTOR %0, %1, shufflemask(0, undef)
     $sgpr0 = COPY %2
 
 ...
 
 ---
 name: s_shufflevector_v2s16_v2s16_u_0
 tracksRegLiveness: true
 legalized:       true
 regBankSelected: true
 
 body: |
   bb.0:
     liveins: $sgpr0, $sgpr1
 
     ; GFX9-LABEL: name: s_shufflevector_v2s16_v2s16_u_0
     ; GFX9: liveins: $sgpr0, $sgpr1
     ; GFX9: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr0
     ; GFX9: [[S_LSHL_B32_:%[0-9]+]]:sreg_32 = S_LSHL_B32 [[COPY]], 16, implicit-def $scc
     ; GFX9: $sgpr0 = COPY [[S_LSHL_B32_]]
+    ; GFX11-LABEL: name: s_shufflevector_v2s16_v2s16_u_0
+    ; GFX11: liveins: $sgpr0, $sgpr1
+    ; GFX11: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr0
+    ; GFX11: [[S_LSHL_B32_:%[0-9]+]]:sreg_32 = S_LSHL_B32 [[COPY]], 16, implicit-def $scc
+    ; GFX11: $sgpr0 = COPY [[S_LSHL_B32_]]
     %0:sgpr(<2 x s16>) = COPY $sgpr0
     %1:sgpr(<2 x s16>) = COPY $sgpr1
     %2:sgpr(<2 x s16>) = G_SHUFFLE_VECTOR %0, %1, shufflemask(undef, 0)
     $sgpr0 = COPY %2
 
 ...
 
 ---
 name: s_shufflevector_v2s16_v2s16_1_u
 tracksRegLiveness: true
 legalized:       true
 regBankSelected: true
 
 body: |
   bb.0:
     liveins: $sgpr0, $sgpr1
 
     ; GFX9-LABEL: name: s_shufflevector_v2s16_v2s16_1_u
     ; GFX9: liveins: $sgpr0, $sgpr1
     ; GFX9: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr0
     ; GFX9: [[S_LSHR_B32_:%[0-9]+]]:sreg_32 = S_LSHR_B32 [[COPY]], 16, implicit-def $scc
     ; GFX9: $sgpr0 = COPY [[S_LSHR_B32_]]
+    ; GFX11-LABEL: name: s_shufflevector_v2s16_v2s16_1_u
+    ; GFX11: liveins: $sgpr0, $sgpr1
+    ; GFX11: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr0
+    ; GFX11: [[S_LSHR_B32_:%[0-9]+]]:sreg_32 = S_LSHR_B32 [[COPY]], 16, implicit-def $scc
+    ; GFX11: $sgpr0 = COPY [[S_LSHR_B32_]]
     %0:sgpr(<2 x s16>) = COPY $sgpr0
     %1:sgpr(<2 x s16>) = COPY $sgpr1
     %2:sgpr(<2 x s16>) = G_SHUFFLE_VECTOR %0, %1, shufflemask(1, undef)
     $sgpr0 = COPY %2
 
 ...
 
 ---
 name: s_shufflevector_v2s16_v2s16_u_1
 tracksRegLiveness: true
 legalized:       true
 regBankSelected: true
 
 body: |
   bb.0:
     liveins: $sgpr0, $sgpr1
 
     ; GFX9-LABEL: name: s_shufflevector_v2s16_v2s16_u_1
     ; GFX9: liveins: $sgpr0, $sgpr1
     ; GFX9: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr0
     ; GFX9: $sgpr0 = COPY [[COPY]]
+    ; GFX11-LABEL: name: s_shufflevector_v2s16_v2s16_u_1
+    ; GFX11: liveins: $sgpr0, $sgpr1
+    ; GFX11: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr0
+    ; GFX11: $sgpr0 = COPY [[COPY]]
     %0:sgpr(<2 x s16>) = COPY $sgpr0
     %1:sgpr(<2 x s16>) = COPY $sgpr1
     %2:sgpr(<2 x s16>) = G_SHUFFLE_VECTOR %0, %1, shufflemask(undef, 1)
     $sgpr0 = COPY %2
 
 ...
 
 
 ---
 name: s_shufflevector_v2s16_v2s16_2_u
 tracksRegLiveness: true
 legalized:       true
 regBankSelected: true
 
 body: |
   bb.0:
     liveins: $sgpr0, $sgpr1
 
     ; GFX9-LABEL: name: s_shufflevector_v2s16_v2s16_2_u
     ; GFX9: liveins: $sgpr0, $sgpr1
     ; GFX9: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr1
     ; GFX9: $sgpr0 = COPY [[COPY]]
+    ; GFX11-LABEL: name: s_shufflevector_v2s16_v2s16_2_u
+    ; GFX11: liveins: $sgpr0, $sgpr1
+    ; GFX11: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr1
+    ; GFX11: $sgpr0 = COPY [[COPY]]
     %0:sgpr(<2 x s16>) = COPY $sgpr0
     %1:sgpr(<2 x s16>) = COPY $sgpr1
     %2:sgpr(<2 x s16>) = G_SHUFFLE_VECTOR %0, %1, shufflemask(2, undef)
     $sgpr0 = COPY %2
 
 ...
 
 ---
 name: s_shufflevector_v2s16_v2s16_u_2
 tracksRegLiveness: true
 legalized:       true
 regBankSelected: true
 
 body: |
   bb.0:
     liveins: $sgpr0, $sgpr1
 
     ; GFX9-LABEL: name: s_shufflevector_v2s16_v2s16_u_2
     ; GFX9: liveins: $sgpr0, $sgpr1
     ; GFX9: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr1
     ; GFX9: [[S_LSHL_B32_:%[0-9]+]]:sreg_32 = S_LSHL_B32 [[COPY]], 16, implicit-def $scc
     ; GFX9: $sgpr0 = COPY [[S_LSHL_B32_]]
+    ; GFX11-LABEL: name: s_shufflevector_v2s16_v2s16_u_2
+    ; GFX11: liveins: $sgpr0, $sgpr1
+    ; GFX11: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr1
+    ; GFX11: [[S_LSHL_B32_:%[0-9]+]]:sreg_32 = S_LSHL_B32 [[COPY]], 16, implicit-def $scc
+    ; GFX11: $sgpr0 = COPY [[S_LSHL_B32_]]
     %0:sgpr(<2 x s16>) = COPY $sgpr0
     %1:sgpr(<2 x s16>) = COPY $sgpr1
     %2:sgpr(<2 x s16>) = G_SHUFFLE_VECTOR %0, %1, shufflemask(undef, 2)
     $sgpr0 = COPY %2
 
 ...
 
 ---
 name: s_shufflevector_v2s16_v2s16_3_u
 tracksRegLiveness: true
 legalized:       true
 regBankSelected: true
 
 body: |
   bb.0:
     liveins: $sgpr0, $sgpr1
 
     ; GFX9-LABEL: name: s_shufflevector_v2s16_v2s16_3_u
     ; GFX9: liveins: $sgpr0, $sgpr1
     ; GFX9: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr1
     ; GFX9: [[S_LSHR_B32_:%[0-9]+]]:sreg_32 = S_LSHR_B32 [[COPY]], 16, implicit-def $scc
     ; GFX9: $sgpr0 = COPY [[S_LSHR_B32_]]
+    ; GFX11-LABEL: name: s_shufflevector_v2s16_v2s16_3_u
+    ; GFX11: liveins: $sgpr0, $sgpr1
+    ; GFX11: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr1
+    ; GFX11: [[S_LSHR_B32_:%[0-9]+]]:sreg_32 = S_LSHR_B32 [[COPY]], 16, implicit-def $scc
+    ; GFX11: $sgpr0 = COPY [[S_LSHR_B32_]]
     %0:sgpr(<2 x s16>) = COPY $sgpr0
     %1:sgpr(<2 x s16>) = COPY $sgpr1
     %2:sgpr(<2 x s16>) = G_SHUFFLE_VECTOR %0, %1, shufflemask(3, undef)
     $sgpr0 = COPY %2
 
 ...
 
 ---
 name: s_shufflevector_v2s16_v2s16_u_3
 tracksRegLiveness: true
 legalized:       true
 regBankSelected: true
 
 body: |
   bb.0:
     liveins: $sgpr0, $sgpr1
 
     ; GFX9-LABEL: name: s_shufflevector_v2s16_v2s16_u_3
     ; GFX9: liveins: $sgpr0, $sgpr1
     ; GFX9: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr1
     ; GFX9: $sgpr0 = COPY [[COPY]]
+    ; GFX11-LABEL: name: s_shufflevector_v2s16_v2s16_u_3
+    ; GFX11: liveins: $sgpr0, $sgpr1
+    ; GFX11: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr1
+    ; GFX11: $sgpr0 = COPY [[COPY]]
     %0:sgpr(<2 x s16>) = COPY $sgpr0
     %1:sgpr(<2 x s16>) = COPY $sgpr1
     %2:sgpr(<2 x s16>) = G_SHUFFLE_VECTOR %0, %1, shufflemask(undef, 3)
     $sgpr0 = COPY %2
 
 ...
 
 ---
 name: s_shufflevector_v2s16_v2s16_0_0
 tracksRegLiveness: true
 legalized:       true
 regBankSelected: true
 
 body: |
   bb.0:
     liveins: $sgpr0, $sgpr1
 
     ; GFX9-LABEL: name: s_shufflevector_v2s16_v2s16_0_0
     ; GFX9: liveins: $sgpr0, $sgpr1
     ; GFX9: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr0
     ; GFX9: [[S_PACK_LL_B32_B16_:%[0-9]+]]:sreg_32 = S_PACK_LL_B32_B16 [[COPY]], [[COPY]]
     ; GFX9: $sgpr0 = COPY [[S_PACK_LL_B32_B16_]]
+    ; GFX11-LABEL: name: s_shufflevector_v2s16_v2s16_0_0
+    ; GFX11: liveins: $sgpr0, $sgpr1
+    ; GFX11: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr0
+    ; GFX11: [[S_PACK_LL_B32_B16_:%[0-9]+]]:sreg_32 = S_PACK_LL_B32_B16 [[COPY]], [[COPY]]
+    ; GFX11: $sgpr0 = COPY [[S_PACK_LL_B32_B16_]]
     %0:sgpr(<2 x s16>) = COPY $sgpr0
     %1:sgpr(<2 x s16>) = COPY $sgpr1
     %2:sgpr(<2 x s16>) = G_SHUFFLE_VECTOR %0, %1, shufflemask(0, 0)
     $sgpr0 = COPY %2
 
 ...
 
 ---
 name: s_shufflevector_v2s16_v2s16_0_1
 tracksRegLiveness: true
 legalized:       true
 regBankSelected: true
 
 body: |
   bb.0:
     liveins: $sgpr0, $sgpr1
 
     ; GFX9-LABEL: name: s_shufflevector_v2s16_v2s16_0_1
     ; GFX9: liveins: $sgpr0, $sgpr1
     ; GFX9: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr0
     ; GFX9: $sgpr0 = COPY [[COPY]]
+    ; GFX11-LABEL: name: s_shufflevector_v2s16_v2s16_0_1
+    ; GFX11: liveins: $sgpr0, $sgpr1
+    ; GFX11: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr0
+    ; GFX11: $sgpr0 = COPY [[COPY]]
     %0:sgpr(<2 x s16>) = COPY $sgpr0
     %1:sgpr(<2 x s16>) = COPY $sgpr1
     %2:sgpr(<2 x s16>) = G_SHUFFLE_VECTOR %0, %1, shufflemask(0, 1)
     $sgpr0 = COPY %2
 
 ...
 
 ---
 name: s_shufflevector_v2s16_v2s16_1_0
 tracksRegLiveness: true
 legalized:       true
 regBankSelected: true
 
 body: |
   bb.0:
     liveins: $sgpr0, $sgpr1
 
     ; GFX9-LABEL: name: s_shufflevector_v2s16_v2s16_1_0
     ; GFX9: liveins: $sgpr0, $sgpr1
     ; GFX9: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr0
     ; GFX9: [[S_LSHR_B32_:%[0-9]+]]:sreg_32 = S_LSHR_B32 [[COPY]], 16, implicit-def $scc
     ; GFX9: [[S_PACK_LL_B32_B16_:%[0-9]+]]:sreg_32 = S_PACK_LL_B32_B16 [[S_LSHR_B32_]], [[COPY]]
     ; GFX9: $sgpr0 = COPY [[S_PACK_LL_B32_B16_]]
+    ; GFX11-LABEL: name: s_shufflevector_v2s16_v2s16_1_0
+    ; GFX11: liveins: $sgpr0, $sgpr1
+    ; GFX11: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr0
+    ; GFX11: [[S_PACK_HL_B32_B16_:%[0-9]+]]:sreg_32 = S_PACK_HL_B32_B16 [[COPY]], [[COPY]]
+    ; GFX11: $sgpr0 = COPY [[S_PACK_HL_B32_B16_]]
     %0:sgpr(<2 x s16>) = COPY $sgpr0
     %1:sgpr(<2 x s16>) = COPY $sgpr1
     %2:sgpr(<2 x s16>) = G_SHUFFLE_VECTOR %0, %1, shufflemask(1, 0)
     $sgpr0 = COPY %2
 
 ...
 
 ---
 name: s_shufflevector_v2s16_v2s16_1_1
 tracksRegLiveness: true
 legalized:       true
 regBankSelected: true
 
 body: |
   bb.0:
     liveins: $sgpr0, $sgpr1
 
     ; GFX9-LABEL: name: s_shufflevector_v2s16_v2s16_1_1
     ; GFX9: liveins: $sgpr0, $sgpr1
     ; GFX9: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr0
     ; GFX9: [[S_PACK_HH_B32_B16_:%[0-9]+]]:sreg_32 = S_PACK_HH_B32_B16 [[COPY]], [[COPY]]
     ; GFX9: $sgpr0 = COPY [[S_PACK_HH_B32_B16_]]
+    ; GFX11-LABEL: name: s_shufflevector_v2s16_v2s16_1_1
+    ; GFX11: liveins: $sgpr0, $sgpr1
+    ; GFX11: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr0
+    ; GFX11: [[S_PACK_HH_B32_B16_:%[0-9]+]]:sreg_32 = S_PACK_HH_B32_B16 [[COPY]], [[COPY]]
+    ; GFX11: $sgpr0 = COPY [[S_PACK_HH_B32_B16_]]
     %0:sgpr(<2 x s16>) = COPY $sgpr0
     %1:sgpr(<2 x s16>) = COPY $sgpr1
     %2:sgpr(<2 x s16>) = G_SHUFFLE_VECTOR %0, %1, shufflemask(1, 1)
     $sgpr0 = COPY %2
 
 ...
 
 ---
 name: s_shufflevector_v2s16_v2s16_2_2
 tracksRegLiveness: true
 legalized:       true
 regBankSelected: true
 
 body: |
   bb.0:
     liveins: $sgpr0, $sgpr1
 
     ; GFX9-LABEL: name: s_shufflevector_v2s16_v2s16_2_2
     ; GFX9: liveins: $sgpr0, $sgpr1
     ; GFX9: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr1
     ; GFX9: [[S_PACK_LL_B32_B16_:%[0-9]+]]:sreg_32 = S_PACK_LL_B32_B16 [[COPY]], [[COPY]]
     ; GFX9: $sgpr0 = COPY [[S_PACK_LL_B32_B16_]]
+    ; GFX11-LABEL: name: s_shufflevector_v2s16_v2s16_2_2
+    ; GFX11: liveins: $sgpr0, $sgpr1
+    ; GFX11: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr1
+    ; GFX11: [[S_PACK_LL_B32_B16_:%[0-9]+]]:sreg_32 = S_PACK_LL_B32_B16 [[COPY]], [[COPY]]
+    ; GFX11: $sgpr0 = COPY [[S_PACK_LL_B32_B16_]]
     %0:sgpr(<2 x s16>) = COPY $sgpr0
     %1:sgpr(<2 x s16>) = COPY $sgpr1
     %2:sgpr(<2 x s16>) = G_SHUFFLE_VECTOR %0, %1, shufflemask(2, 2)
     $sgpr0 = COPY %2
 
 ...
 
 ---
 name: s_shufflevector_v2s16_v2s16_2_3
 tracksRegLiveness: true
 legalized:       true
 regBankSelected: true
 
 body: |
   bb.0:
     liveins: $sgpr0, $sgpr1
 
     ; GFX9-LABEL: name: s_shufflevector_v2s16_v2s16_2_3
     ; GFX9: liveins: $sgpr0, $sgpr1
     ; GFX9: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr1
     ; GFX9: $sgpr0 = COPY [[COPY]]
+    ; GFX11-LABEL: name: s_shufflevector_v2s16_v2s16_2_3
+    ; GFX11: liveins: $sgpr0, $sgpr1
+    ; GFX11: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr1
+    ; GFX11: $sgpr0 = COPY [[COPY]]
     %0:sgpr(<2 x s16>) = COPY $sgpr0
     %1:sgpr(<2 x s16>) = COPY $sgpr1
     %2:sgpr(<2 x s16>) = G_SHUFFLE_VECTOR %0, %1, shufflemask(2, 3)
     $sgpr0 = COPY %2
 
 ...
 
 ---
 name: s_shufflevector_v2s16_v2s16_3_2
 tracksRegLiveness: true
 legalized:       true
 regBankSelected: true
 
 body: |
   bb.0:
     liveins: $sgpr0, $sgpr1
 
     ; GFX9-LABEL: name: s_shufflevector_v2s16_v2s16_3_2
     ; GFX9: liveins: $sgpr0, $sgpr1
     ; GFX9: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr1
     ; GFX9: [[S_LSHR_B32_:%[0-9]+]]:sreg_32 = S_LSHR_B32 [[COPY]], 16, implicit-def $scc
     ; GFX9: [[S_PACK_LL_B32_B16_:%[0-9]+]]:sreg_32 = S_PACK_LL_B32_B16 [[S_LSHR_B32_]], [[COPY]]
     ; GFX9: $sgpr0 = COPY [[S_PACK_LL_B32_B16_]]
+    ; GFX11-LABEL: name: s_shufflevector_v2s16_v2s16_3_2
+    ; GFX11: liveins: $sgpr0, $sgpr1
+    ; GFX11: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr1
+    ; GFX11: [[S_PACK_HL_B32_B16_:%[0-9]+]]:sreg_32 = S_PACK_HL_B32_B16 [[COPY]], [[COPY]]
+    ; GFX11: $sgpr0 = COPY [[S_PACK_HL_B32_B16_]]
     %0:sgpr(<2 x s16>) = COPY $sgpr0
     %1:sgpr(<2 x s16>) = COPY $sgpr1
     %2:sgpr(<2 x s16>) = G_SHUFFLE_VECTOR %0, %1, shufflemask(3, 2)
     $sgpr0 = COPY %2
 
 ...
 
 ---
 name: s_shufflevector_v2s16_v2s16_3_3
 tracksRegLiveness: true
 legalized:       true
 regBankSelected: true
 
 body: |
   bb.0:
     liveins: $sgpr0, $sgpr1
 
     ; GFX9-LABEL: name: s_shufflevector_v2s16_v2s16_3_3
     ; GFX9: liveins: $sgpr0, $sgpr1
     ; GFX9: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr1
     ; GFX9: [[S_PACK_HH_B32_B16_:%[0-9]+]]:sreg_32 = S_PACK_HH_B32_B16 [[COPY]], [[COPY]]
     ; GFX9: $sgpr0 = COPY [[S_PACK_HH_B32_B16_]]
+    ; GFX11-LABEL: name: s_shufflevector_v2s16_v2s16_3_3
+    ; GFX11: liveins: $sgpr0, $sgpr1
+    ; GFX11: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr1
+    ; GFX11: [[S_PACK_HH_B32_B16_:%[0-9]+]]:sreg_32 = S_PACK_HH_B32_B16 [[COPY]], [[COPY]]
+    ; GFX11: $sgpr0 = COPY [[S_PACK_HH_B32_B16_]]
     %0:sgpr(<2 x s16>) = COPY $sgpr0
     %1:sgpr(<2 x s16>) = COPY $sgpr1
     %2:sgpr(<2 x s16>) = G_SHUFFLE_VECTOR %0, %1, shufflemask(3, 3)
     $sgpr0 = COPY %2
 
 ...