Index: llvm/lib/Target/AMDGPU/AMDGPUInstructionSelector.h
===================================================================
--- llvm/lib/Target/AMDGPU/AMDGPUInstructionSelector.h
+++ llvm/lib/Target/AMDGPU/AMDGPUInstructionSelector.h
@@ -107,6 +107,7 @@
   bool selectInterpP1F16(MachineInstr &MI) const;
   bool selectDivScale(MachineInstr &MI) const;
   bool selectIntrinsicIcmp(MachineInstr &MI) const;
+  bool selectBallot(MachineInstr &I) const;
   bool selectG_INTRINSIC(MachineInstr &I) const;
 
   bool selectEndCfIntrinsic(MachineInstr &MI) const;
Index: llvm/lib/Target/AMDGPU/AMDGPUInstructionSelector.cpp
===================================================================
--- llvm/lib/Target/AMDGPU/AMDGPUInstructionSelector.cpp
+++ llvm/lib/Target/AMDGPU/AMDGPUInstructionSelector.cpp
@@ -891,6 +891,8 @@
     return selectDivScale(I);
   case Intrinsic::amdgcn_icmp:
     return selectIntrinsicIcmp(I);
+  case Intrinsic::amdgcn_ballot:
+    return selectBallot(I);
   default:
     return selectImpl(I, *CoverageInfo);
   }
@@ -1039,6 +1041,41 @@
   return Ret;
 }
 
+bool AMDGPUInstructionSelector::selectBallot(MachineInstr &I) const {
+  MachineBasicBlock *BB = I.getParent();
+  const DebugLoc &DL = I.getDebugLoc();
+  Register DstReg = I.getOperand(0).getReg();
+  unsigned Size = MRI->getType(DstReg).getSizeInBits();
+  const bool Is64 = Size == 64;
+
+  if (Size != STI.getWavefrontSize())
+    return false;
+
+  MachineInstr *Value = MRI->getVRegDef(I.getOperand(2).getReg());
+  if (Value->getOpcode() == AMDGPU::COPY)
+    Value = MRI->getVRegDef(Value->getOperand(1).getReg());
+
+  if (Value->getOpcode() == AMDGPU::G_CONSTANT) {
+    const APInt &Val = Value->getOperand(1).getCImm()->getValue();
+
+    if (Val.isNullValue()) {
+      unsigned Opcode = Is64 ? AMDGPU::V_MOV_B32_e64 : AMDGPU::V_MOV_B32_e32;
+      BuildMI(*BB, &I, DL, TII.get(Opcode), DstReg).addImm(0);
+    } else if (Val.isAllOnesValue()) {
+      unsigned SrcReg = Is64 ? AMDGPU::EXEC : AMDGPU::EXEC_LO;
+      BuildMI(*BB, &I, DL, TII.get(AMDGPU::COPY), DstReg).addReg(SrcReg);
+    } else
+      return false;
+
+  } else {
+    Register SrcReg = I.getOperand(2).getReg();
+    BuildMI(*BB, &I, DL, TII.get(AMDGPU::COPY), DstReg).addReg(SrcReg);
+  }
+
+  I.eraseFromParent();
+  return true;
+}
+
 bool AMDGPUInstructionSelector::selectEndCfIntrinsic(MachineInstr &MI) const {
   // FIXME: Manually selecting to avoid dealiing with the SReg_1 trick
   // SelectionDAG uses for wave32 vs wave64.
Index: llvm/lib/Target/AMDGPU/AMDGPURegisterBankInfo.cpp
===================================================================
--- llvm/lib/Target/AMDGPU/AMDGPURegisterBankInfo.cpp
+++ llvm/lib/Target/AMDGPU/AMDGPURegisterBankInfo.cpp
@@ -2989,6 +2989,7 @@
       constrainOpWithReadfirstlane(MI, MRI, 3); // Index
       return;
     }
+    case Intrinsic::amdgcn_ballot:
     case Intrinsic::amdgcn_interp_p1:
     case Intrinsic::amdgcn_interp_p2:
     case Intrinsic::amdgcn_interp_mov:
@@ -4158,6 +4159,13 @@
       OpdsMapping[M0Idx] = AMDGPU::getValueMapping(M0Bank, 32);
       break;
     }
+    case Intrinsic::amdgcn_ballot: {
+      unsigned DstSize = MRI.getType(MI.getOperand(0).getReg()).getSizeInBits();
+      unsigned SrcSize = MRI.getType(MI.getOperand(2).getReg()).getSizeInBits();
+      OpdsMapping[0] = AMDGPU::getValueMapping(AMDGPU::VGPRRegBankID, DstSize);
+      OpdsMapping[2] = AMDGPU::getValueMapping(AMDGPU::VCCRegBankID, SrcSize);
+      break;
+    }
     }
     break;
   }
Index: llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.ballot.i32.ll
===================================================================
--- /dev/null
+++ llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.ballot.i32.ll
@@ -0,0 +1,93 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -march=amdgcn -mcpu=gfx1010 -mattr=+wavefrontsize32,-wavefrontsize64 -global-isel < %s | FileCheck %s
+
+declare i32 @llvm.amdgcn.ballot.i32(i1)
+
+; Test ballot(0)
+
+define i32 @test0() {
+; CHECK-LABEL: test0:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    s_waitcnt_vscnt null, 0x0
+; CHECK-NEXT:    v_mov_b32_e32 v0, 0
+; CHECK-NEXT:    ; implicit-def: $vcc_hi
+; CHECK-NEXT:    s_setpc_b64 s[30:31]
+  %ballot = call i32 @llvm.amdgcn.ballot.i32(i1 0)
+  ret i32 %ballot
+}
+
+; Test ballot(1)
+
+define i32 @test1() {
+; CHECK-LABEL: test1:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    s_waitcnt_vscnt null, 0x0
+; CHECK-NEXT:    v_mov_b32_e32 v0, exec_lo
+; CHECK-NEXT:    ; implicit-def: $vcc_hi
+; CHECK-NEXT:    s_setpc_b64 s[30:31]
+  %ballot = call i32 @llvm.amdgcn.ballot.i32(i1 1)
+  ret i32 %ballot
+}
+
+; Test ballot of a non-comparison operation
+
+define i32 @test2(i32 %x) {
+; CHECK-LABEL: test2:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    s_waitcnt_vscnt null, 0x0
+; CHECK-NEXT:    v_and_b32_e32 v0, 1, v0
+; CHECK-NEXT:    ; implicit-def: $vcc_hi
+; CHECK-NEXT:    v_cmp_ne_u32_e64 s4, 0, v0
+; CHECK-NEXT:    v_mov_b32_e32 v0, s4
+; CHECK-NEXT:    s_setpc_b64 s[30:31]
+  %trunc = trunc i32 %x to i1
+  %ballot = call i32 @llvm.amdgcn.ballot.i32(i1 %trunc)
+  ret i32 %ballot
+}
+
+; Test ballot of comparisons
+
+define i32 @test3(i32 %x, i32 %y) {
+; CHECK-LABEL: test3:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    s_waitcnt_vscnt null, 0x0
+; CHECK-NEXT:    v_cmp_eq_u32_e64 s4, v0, v1
+; CHECK-NEXT:    ; implicit-def: $vcc_hi
+; CHECK-NEXT:    v_mov_b32_e32 v0, s4
+; CHECK-NEXT:    s_setpc_b64 s[30:31]
+  %cmp = icmp eq i32 %x, %y
+  %ballot = call i32 @llvm.amdgcn.ballot.i32(i1 %cmp)
+  ret i32 %ballot
+}
+
+define i32 @test4(i32 %x) {
+; CHECK-LABEL: test4:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    s_waitcnt_vscnt null, 0x0
+; CHECK-NEXT:    v_cmp_le_i32_e64 s4, 0x63, v0
+; CHECK-NEXT:    ; implicit-def: $vcc_hi
+; CHECK-NEXT:    v_mov_b32_e32 v0, s4
+; CHECK-NEXT:    s_setpc_b64 s[30:31]
+  %cmp = icmp sge i32 %x, 99
+  %ballot = call i32 @llvm.amdgcn.ballot.i32(i1 %cmp)
+  ret i32 %ballot
+}
+
+define i32 @test5(float %x, float %y) {
+; CHECK-LABEL: test5:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    s_waitcnt_vscnt null, 0x0
+; CHECK-NEXT:    v_cmp_gt_f32_e64 s4, v0, v1
+; CHECK-NEXT:    ; implicit-def: $vcc_hi
+; CHECK-NEXT:    v_mov_b32_e32 v0, s4
+; CHECK-NEXT:    s_setpc_b64 s[30:31]
+  %cmp = fcmp ogt float %x, %y
+  %ballot = call i32 @llvm.amdgcn.ballot.i32(i1 %cmp)
+  ret i32 %ballot
+}
Index: llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.ballot.i64.ll
===================================================================
--- /dev/null
+++ llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.ballot.i64.ll
@@ -0,0 +1,88 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -march=amdgcn -mcpu=gfx900 -global-isel < %s | FileCheck %s
+
+declare i64 @llvm.amdgcn.ballot.i64(i1)
+
+; Test ballot(0)
+
+define i64 @test0() {
+; CHECK-LABEL: test0:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    v_mov_b32_e32 v0, 0
+; CHECK-NEXT:    v_mov_b32_e32 v1, 0
+; CHECK-NEXT:    s_setpc_b64 s[30:31]
+  %ballot = call i64 @llvm.amdgcn.ballot.i64(i1 0)
+  ret i64 %ballot
+}
+
+; Test ballot(1)
+
+define i64 @test1() {
+; CHECK-LABEL: test1:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    v_mov_b32_e32 v0, exec_lo
+; CHECK-NEXT:    v_mov_b32_e32 v1, exec_hi
+; CHECK-NEXT:    s_setpc_b64 s[30:31]
+  %ballot = call i64 @llvm.amdgcn.ballot.i64(i1 1)
+  ret i64 %ballot
+}
+
+; Test ballot of a non-comparison operation
+
+define i64 @test2(i32 %x) {
+; CHECK-LABEL: test2:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    v_and_b32_e32 v0, 1, v0
+; CHECK-NEXT:    v_cmp_ne_u32_e64 s[4:5], 0, v0
+; CHECK-NEXT:    v_mov_b32_e32 v0, s4
+; CHECK-NEXT:    v_mov_b32_e32 v1, s5
+; CHECK-NEXT:    s_setpc_b64 s[30:31]
+  %trunc = trunc i32 %x to i1
+  %ballot = call i64 @llvm.amdgcn.ballot.i64(i1 %trunc)
+  ret i64 %ballot
+}
+
+; Test ballot of comparisons
+
+define i64 @test3(i32 %x, i32 %y) {
+; CHECK-LABEL: test3:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    v_cmp_eq_u32_e64 s[4:5], v0, v1
+; CHECK-NEXT:    v_mov_b32_e32 v0, s4
+; CHECK-NEXT:    v_mov_b32_e32 v1, s5
+; CHECK-NEXT:    s_setpc_b64 s[30:31]
+  %cmp = icmp eq i32 %x, %y
+  %ballot = call i64 @llvm.amdgcn.ballot.i64(i1 %cmp)
+  ret i64 %ballot
+}
+
+define i64 @test4(i32 %x) {
+; CHECK-LABEL: test4:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    v_mov_b32_e32 v1, 0x63
+; CHECK-NEXT:    v_cmp_ge_i32_e64 s[4:5], v0, v1
+; CHECK-NEXT:    v_mov_b32_e32 v0, s4
+; CHECK-NEXT:    v_mov_b32_e32 v1, s5
+; CHECK-NEXT:    s_setpc_b64 s[30:31]
+  %cmp = icmp sge i32 %x, 99
+  %ballot = call i64 @llvm.amdgcn.ballot.i64(i1 %cmp)
+  ret i64 %ballot
+}
+
+define i64 @test5(float %x, float %y) {
+; CHECK-LABEL: test5:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    v_cmp_gt_f32_e64 s[4:5], v0, v1
+; CHECK-NEXT:    v_mov_b32_e32 v0, s4
+; CHECK-NEXT:    v_mov_b32_e32 v1, s5
+; CHECK-NEXT:    s_setpc_b64 s[30:31]
+  %cmp = fcmp ogt float %x, %y
+  %ballot = call i64 @llvm.amdgcn.ballot.i64(i1 %cmp)
+  ret i64 %ballot
+}