[GlobalISel] Handle div-by-pow2

This patch adds similar handling of div-by-pow2 as in `SelectionDAG`.

llvm/include/llvm/CodeGen/GlobalISel/CombinerHelper.h

@@ -673,6 +673,14 @@ public:
   bool matchSDivByConst(MachineInstr &MI);
   void applySDivByConst(MachineInstr &MI);
 
+  /// Given an G_SDIV \p MI expressing a signed divided by a pow2 constant,
+  /// return expressions that implements it by shifting.
+  bool matchDivByPow2(MachineInstr &MI, bool IsSigned);
+  void applySDivByPow2(MachineInstr &MI);
+  /// Given an G_UDIV \p MI expressing an unsigned divided by a pow2 constant,
+  /// return expressions that implements it by shifting.
+  void applyUDivByPow2(MachineInstr &MI);
+
   // G_UMULH x, (1 << c)) -> x >> (bitwidth - c)
   bool matchUMulHToLShr(MachineInstr &MI);
   void applyUMulHToLShr(MachineInstr &MI);

llvm/include/llvm/Target/GlobalISel/Combine.td

@@ -179,6 +179,7 @@ def FmArcp      : MIFlagEnum<"FmArcp">;
 def FmContract  : MIFlagEnum<"FmContract">;
 def FmAfn       : MIFlagEnum<"FmAfn">;
 def FmReassoc   : MIFlagEnum<"FmReassoc">;
+def IsExact     : MIFlagEnum<"IsExact">;
 
 def MIFlags;
 // def not; -> Already defined as a SDNode
@@ -1036,7 +1037,20 @@ def sdiv_by_const : GICombineRule<
    [{ return Helper.matchSDivByConst(*${root}); }]),
   (apply [{ Helper.applySDivByConst(*${root}); }])>;
 
-def intdiv_combines : GICombineGroup<[udiv_by_const, sdiv_by_const]>;
+def sdiv_by_pow2 : GICombineRule<
+  (defs root:$root),
+  (match (G_SDIV $dst, $x, $y, (MIFlags (not IsExact))):$root,
+   [{ return Helper.matchDivByPow2(*${root}, /*IsSigned=*/true); }]),
+  (apply [{ Helper.applySDivByPow2(*${root}); }])>;
+
+def udiv_by_pow2 : GICombineRule<
+  (defs root:$root),
+  (match (G_UDIV $dst, $x, $y, (MIFlags (not IsExact))):$root,
+   [{ return Helper.matchDivByPow2(*${root}, /*IsSigned=*/false); }]),
+  (apply [{ Helper.applyUDivByPow2(*${root}); }])>;
+
+def intdiv_combines : GICombineGroup<[udiv_by_const, sdiv_by_const,
+                                      sdiv_by_pow2, udiv_by_pow2]>;
 
 def reassoc_ptradd : GICombineRule<
   (defs root:$root, build_fn_matchinfo:$matchinfo),

llvm/lib/CodeGen/GlobalISel/CombinerHelper.cpp

@@ -5270,6 +5270,144 @@ MachineInstr *CombinerHelper::buildSDivUsingMul(MachineInstr &MI) {
   return MIB.buildMul(Ty, Res, Factor);
 }
 
+bool CombinerHelper::matchDivByPow2(MachineInstr &MI, bool IsSigned) {
+  assert((MI.getOpcode() == TargetOpcode::G_SDIV ||
+          MI.getOpcode() == TargetOpcode::G_UDIV) &&
+         "Expected SDIV or UDIV");
+  auto &Div = cast<GenericMachineInstr>(MI);
+  Register RHS = Div.getReg(2);
+  auto MatchPow2 = [&](const Constant *C) {
+    auto *CI = dyn_cast<ConstantInt>(C);
+    return CI && (CI->getValue().isPowerOf2() ||
+                  (IsSigned && CI->getValue().isNegatedPowerOf2()));
+  };
+  return matchUnaryPredicate(MRI, RHS, MatchPow2, /*AllowUndefs=*/false);
+}
+
+void CombinerHelper::applySDivByPow2(MachineInstr &MI) {
+  assert(MI.getOpcode() == TargetOpcode::G_SDIV && "Expected SDIV");
+  auto &SDiv = cast<GenericMachineInstr>(MI);
+  Register Dst = SDiv.getReg(0);
+  Register LHS = SDiv.getReg(1);
+  Register RHS = SDiv.getReg(2);
+  LLT Ty = MRI.getType(Dst);
+  LLT ShiftAmtTy = getTargetLowering().getPreferredShiftAmountTy(Ty);
+
+  Builder.setInstrAndDebugLoc(MI);
+
+  // Effectively we want to lower G_SDIV %lhs, %rhs, where %rhs is a power of 2,
+  // to the following version:
+  //
+  // %c1 = G_CTTZ %rhs
+  // %inexact = G_SUB $bitwidth, %c1
+  // %sign = %G_ASHR %lhs, $(bitwidth - 1)
+  // %srl = G_SHR %sign, %inexact
+  // %add = G_ADD %lhs, %srl
+  // %sra = G_ASHR %add, %c1
+  // %sra = G_SELECT, %isoneorallones, %lhs, %sra
+  // %zero = G_CONSTANT $0
+  // %neg = G_NEG %sra
+  // %isneg = G_ICMP SLT %lhs, %zero
+  // %res = G_SELECT %isneg, %neg, %sra
+  //
+  // When %rhs is a constant integer, or a splat vector, we can check its value
+  // at compile time such that the first two G_ICMP conditional statements, as
+  // well as the corresponding non-taken branches, can be eliminated. This can
+  // generate compact code even w/o any constant folding afterwards. When $rhs
+  // is not a splat vector, we have to generate those checks via instructions.
+
+  unsigned Bitwidth = Ty.getScalarSizeInBits();
+  auto Zero = Builder.buildConstant(Ty, 0);
+
+  // TODO: It is not necessary to have this specialized version. We need it *for
+  // now* because the folding/combine can't handle it. Remove this large
+  // conditional statement once we can properly fold the two G_ICMP.
+  if (auto RHSC = getConstantOrConstantSplatVector(RHS)) {
+    // Special case: (sdiv X, 1) -> X
+    if (RHSC->isOne()) {
+      replaceSingleDefInstWithReg(MI, LHS);
+      return;
+    }
+    // Special Case: (sdiv X, -1) -> 0-X
+    if (RHSC->isAllOnes()) {
+      auto Neg = Builder.buildNeg(Ty, LHS);
+      replaceSingleDefInstWithReg(MI, Neg->getOperand(0).getReg());
+      return;
+    }
+
+    unsigned TrailingZeros = RHSC->countTrailingZeros();
+    auto C1 = Builder.buildConstant(ShiftAmtTy, TrailingZeros);
+    auto Inexact = Builder.buildConstant(ShiftAmtTy, Bitwidth - TrailingZeros);
+    auto Sign = Builder.buildAShr(
+        Ty, LHS, Builder.buildConstant(ShiftAmtTy, Bitwidth - 1));
+    // Add (LHS < 0) ? abs2 - 1 : 0;
+    auto Lshr = Builder.buildLShr(Ty, Sign, Inexact);
+    auto Add = Builder.buildAdd(Ty, LHS, Lshr);
+    auto Shr = Builder.buildAShr(Ty, Add, C1);
+
+    // If dividing by a positive value, we're done. Otherwise, the result must
+    // be negated.
+    auto Res = RHSC->isNegative() ? Builder.buildNeg(Ty, Shr) : Shr;
+    replaceSingleDefInstWithReg(MI, Res->getOperand(0).getReg());
+    return;
+  }
+
+  // RHS is not a splat vector. Build the above version with instructions.
+  auto Bits = Builder.buildConstant(ShiftAmtTy, Bitwidth);
+  auto C1 = Builder.buildCTTZ(Ty, RHS);
+  C1 = Builder.buildZExtOrTrunc(ShiftAmtTy, C1);
+  auto Inexact = Builder.buildSub(ShiftAmtTy, Bits, C1);
+  auto Sign = Builder.buildAShr(
+      Ty, LHS, Builder.buildConstant(ShiftAmtTy, Bitwidth - 1));
+
+  // Add (LHS < 0) ? abs2 - 1 : 0;
+  auto Shl = Builder.buildShl(Ty, Sign, Inexact);
+  auto Add = Builder.buildAdd(Ty, LHS, Shl);
+  auto Shr = Builder.buildAShr(Ty, Add, C1);
+
+  LLT CCVT = LLT::vector(Ty.getElementCount(), 1);
+
+  auto One = Builder.buildConstant(Ty, 1);
+  auto AllOnes =
+      Builder.buildConstant(Ty, APInt::getAllOnes(Ty.getScalarSizeInBits()));
+  auto IsOne = Builder.buildICmp(CmpInst::Predicate::ICMP_EQ, CCVT, RHS, One);
+  auto IsAllOnes =
+      Builder.buildICmp(CmpInst::Predicate::ICMP_EQ, CCVT, RHS, AllOnes);
+  auto IsOneOrAllOnes = Builder.buildOr(CCVT, IsOne, IsAllOnes);
+  Shr = Builder.buildSelect(Ty, IsOneOrAllOnes, LHS, Shr);
+
+  // If dividing by a positive value, we're done. Otherwise, the result must
+  // be negated.
+  auto Neg = Builder.buildNeg(Ty, Shr);
+  auto IsNeg = Builder.buildICmp(CmpInst::Predicate::ICMP_SLT, CCVT, LHS, Zero);
+  Builder.buildSelect(MI.getOperand(0).getReg(), IsNeg, Neg, Shr);
+}
+
+void CombinerHelper::applyUDivByPow2(MachineInstr &MI) {
+  assert(MI.getOpcode() == TargetOpcode::G_UDIV && "Expected UDIV");
+  auto &UDiv = cast<GenericMachineInstr>(MI);
+  Register Dst = UDiv.getReg(0);
+  Register LHS = UDiv.getReg(1);
+  Register RHS = UDiv.getReg(2);
+  LLT Ty = MRI.getType(Dst);
+  LLT ShiftAmtTy = getTargetLowering().getPreferredShiftAmountTy(Ty);
+
+  Builder.setInstrAndDebugLoc(MI);
+
+  // TODO: It is not necessary to have this specialized version. We need it *for
+  // now* because the folding/combine can't handle CTTZ.
+  if (auto RHSC = getConstantOrConstantSplatVector(RHS)) {
+    auto C1 = Builder.buildConstant(ShiftAmtTy, RHSC->countTrailingZeros());
+    auto Res = Builder.buildLShr(Ty, LHS, C1);
+    replaceSingleDefInstWithReg(MI, Res->getOperand(0).getReg());
+    return;
+  }
+
+  auto C1 = Builder.buildCTTZ(Ty, RHS);
+  C1 = Builder.buildZExtOrTrunc(ShiftAmtTy, C1);
+  Builder.buildLShr(MI.getOperand(0).getReg(), LHS, C1);
+}
+
 bool CombinerHelper::matchUMulHToLShr(MachineInstr &MI) {
   assert(MI.getOpcode() == TargetOpcode::G_UMULH);
   Register RHS = MI.getOperand(2).getReg();

llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.sbfe.ll

@@ -670,36 +670,19 @@ define amdgpu_kernel void @bfe_sext_in_reg_i24(ptr addrspace(1) %out, ptr addrsp
 define amdgpu_kernel void @simplify_demanded_bfe_sdiv(ptr addrspace(1) %out, ptr addrspace(1) %in) #0 {
 ; GFX6-LABEL: simplify_demanded_bfe_sdiv:
 ; GFX6:       ; %bb.0:
-; GFX6-NEXT:    v_rcp_iflag_f32_e32 v0, 2.0
+; GFX6-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
-; GFX6-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x0
-; GFX6-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
-; GFX6-NEXT:    v_cvt_u32_f32_e32 v0, v0
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX6-NEXT:    s_load_dword s0, s[6:7], 0x0
+; GFX6-NEXT:    s_load_dword s3, s[2:3], 0x0
-; GFX6-NEXT:    s_mov_b32 s6, -1
+; GFX6-NEXT:    s_mov_b32 s2, -1
-; GFX6-NEXT:    s_mov_b32 s7, 0xf000
-; GFX6-NEXT:    v_mul_lo_u32 v1, v0, -2
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX6-NEXT:    s_bfe_i32 s0, s0, 0x100001
+; GFX6-NEXT:    s_bfe_i32 s3, s3, 0x100001
-; GFX6-NEXT:    s_ashr_i32 s2, s0, 31
+; GFX6-NEXT:    s_ashr_i32 s4, s3, 31
-; GFX6-NEXT:    v_mul_hi_u32 v1, v0, v1
+; GFX6-NEXT:    s_lshr_b32 s4, s4, 31
-; GFX6-NEXT:    s_add_i32 s0, s0, s2
+; GFX6-NEXT:    s_add_i32 s3, s3, s4
-; GFX6-NEXT:    s_xor_b32 s0, s0, s2
+; GFX6-NEXT:    s_ashr_i32 s3, s3, 1
-; GFX6-NEXT:    v_add_i32_e32 v0, vcc, v0, v1
+; GFX6-NEXT:    v_mov_b32_e32 v0, s3
-; GFX6-NEXT:    v_mul_hi_u32 v0, s0, v0
+; GFX6-NEXT:    s_mov_b32 s3, 0xf000
-; GFX6-NEXT:    v_lshlrev_b32_e32 v1, 1, v0
+; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
-; GFX6-NEXT:    v_add_i32_e32 v2, vcc, 1, v0
-; GFX6-NEXT:    v_sub_i32_e32 v1, vcc, s0, v1
-; GFX6-NEXT:    v_cmp_le_u32_e32 vcc, 2, v1
-; GFX6-NEXT:    v_cndmask_b32_e32 v0, v0, v2, vcc
-; GFX6-NEXT:    v_subrev_i32_e64 v2, s[0:1], 2, v1
-; GFX6-NEXT:    v_cndmask_b32_e32 v1, v1, v2, vcc
-; GFX6-NEXT:    v_add_i32_e32 v2, vcc, 1, v0
-; GFX6-NEXT:    v_cmp_le_u32_e32 vcc, 2, v1
-; GFX6-NEXT:    v_cndmask_b32_e32 v0, v0, v2, vcc
-; GFX6-NEXT:    v_xor_b32_e32 v0, s2, v0
-; GFX6-NEXT:    v_subrev_i32_e32 v0, vcc, s2, v0
-; GFX6-NEXT:    buffer_store_dword v0, off, s[4:7], 0
 ; GFX6-NEXT:    s_endpgm
   %src = load i32, ptr addrspace(1) %in, align 4
   %bfe = call i32 @llvm.amdgcn.sbfe.i32(i32 %src, i32 1, i32 16)

llvm/test/CodeGen/AMDGPU/GlobalISel/sdiv.i32.ll

@@ -279,125 +279,27 @@ define i32 @v_sdiv_i32_pow2k_denom(i32 %num) {
 ; CHECK:       ; %bb.0:
 ; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; CHECK-NEXT:    v_ashrrev_i32_e32 v1, 31, v0
-; CHECK-NEXT:    v_rcp_iflag_f32_e32 v2, 0x45800000
+; CHECK-NEXT:    v_lshrrev_b32_e32 v1, 20, v1
-; CHECK-NEXT:    v_mov_b32_e32 v3, 0xfffff000
-; CHECK-NEXT:    v_mov_b32_e32 v4, 0x1000
 ; CHECK-NEXT:    v_add_i32_e32 v0, vcc, v0, v1
-; CHECK-NEXT:    v_mul_f32_e32 v2, 0x4f7ffffe, v2
+; CHECK-NEXT:    v_ashrrev_i32_e32 v0, 12, v0
-; CHECK-NEXT:    v_xor_b32_e32 v0, v0, v1
-; CHECK-NEXT:    v_cvt_u32_f32_e32 v2, v2
-; CHECK-NEXT:    v_mul_lo_u32 v3, v2, v3
-; CHECK-NEXT:    v_mul_hi_u32 v3, v2, v3
-; CHECK-NEXT:    v_add_i32_e32 v2, vcc, v2, v3
-; CHECK-NEXT:    v_mul_hi_u32 v2, v0, v2
-; CHECK-NEXT:    v_lshlrev_b32_e32 v3, 12, v2
-; CHECK-NEXT:    v_add_i32_e32 v5, vcc, 1, v2
-; CHECK-NEXT:    v_sub_i32_e32 v0, vcc, v0, v3
-; CHECK-NEXT:    v_cmp_ge_u32_e64 s[4:5], v0, v4
-; CHECK-NEXT:    v_cndmask_b32_e64 v2, v2, v5, s[4:5]
-; CHECK-NEXT:    v_subrev_i32_e32 v3, vcc, 0x1000, v0
-; CHECK-NEXT:    v_cndmask_b32_e64 v0, v0, v3, s[4:5]
-; CHECK-NEXT:    v_add_i32_e32 v3, vcc, 1, v2
-; CHECK-NEXT:    v_cmp_ge_u32_e32 vcc, v0, v4
-; CHECK-NEXT:    v_cndmask_b32_e32 v0, v2, v3, vcc
-; CHECK-NEXT:    v_xor_b32_e32 v0, v0, v1
-; CHECK-NEXT:    v_sub_i32_e32 v0, vcc, v0, v1
 ; CHECK-NEXT:    s_setpc_b64 s[30:31]
   %result = sdiv i32 %num, 4096
   ret i32 %result
 }
 
 define <2 x i32> @v_sdiv_v2i32_pow2k_denom(<2 x i32> %num) {
-; GISEL-LABEL: v_sdiv_v2i32_pow2k_denom:
+; CHECK-LABEL: v_sdiv_v2i32_pow2k_denom:
-; GISEL:       ; %bb.0:
+; CHECK:       ; %bb.0:
-; GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GISEL-NEXT:    v_ashrrev_i32_e32 v2, 31, v0
+; CHECK-NEXT:    v_ashrrev_i32_e32 v2, 31, v0
-; GISEL-NEXT:    v_mov_b32_e32 v3, 0x1000
+; CHECK-NEXT:    v_ashrrev_i32_e32 v3, 31, v1
-; GISEL-NEXT:    v_cvt_f32_u32_e32 v4, 0x1000
+; CHECK-NEXT:    v_lshrrev_b32_e32 v2, 20, v2
-; GISEL-NEXT:    v_mov_b32_e32 v5, 0xfffff000
+; CHECK-NEXT:    v_lshrrev_b32_e32 v3, 20, v3
-; GISEL-NEXT:    v_ashrrev_i32_e32 v6, 31, v1
+; CHECK-NEXT:    v_add_i32_e32 v0, vcc, v0, v2
-; GISEL-NEXT:    v_add_i32_e32 v0, vcc, v0, v2
+; CHECK-NEXT:    v_add_i32_e32 v1, vcc, v1, v3
-; GISEL-NEXT:    v_rcp_iflag_f32_e32 v4, v4
+; CHECK-NEXT:    v_ashrrev_i32_e32 v0, 12, v0
-; GISEL-NEXT:    v_add_i32_e32 v1, vcc, v1, v6
+; CHECK-NEXT:    v_ashrrev_i32_e32 v1, 12, v1
-; GISEL-NEXT:    v_xor_b32_e32 v0, v0, v2
+; CHECK-NEXT:    s_setpc_b64 s[30:31]
-; GISEL-NEXT:    v_mul_f32_e32 v4, 0x4f7ffffe, v4
-; GISEL-NEXT:    v_xor_b32_e32 v1, v1, v6
-; GISEL-NEXT:    v_cvt_u32_f32_e32 v4, v4
-; GISEL-NEXT:    v_mul_lo_u32 v5, v4, v5
-; GISEL-NEXT:    v_mul_hi_u32 v5, v4, v5
-; GISEL-NEXT:    v_add_i32_e32 v4, vcc, v4, v5
-; GISEL-NEXT:    v_mul_hi_u32 v5, v0, v4
-; GISEL-NEXT:    v_mul_hi_u32 v4, v1, v4
-; GISEL-NEXT:    v_lshlrev_b32_e32 v7, 12, v5
-; GISEL-NEXT:    v_add_i32_e32 v8, vcc, 1, v5
-; GISEL-NEXT:    v_lshlrev_b32_e32 v9, 12, v4
-; GISEL-NEXT:    v_add_i32_e32 v10, vcc, 1, v4
-; GISEL-NEXT:    v_sub_i32_e32 v0, vcc, v0, v7
-; GISEL-NEXT:    v_sub_i32_e32 v1, vcc, v1, v9
-; GISEL-NEXT:    v_cmp_ge_u32_e64 s[4:5], v0, v3
-; GISEL-NEXT:    v_cndmask_b32_e64 v5, v5, v8, s[4:5]
-; GISEL-NEXT:    v_sub_i32_e32 v7, vcc, v0, v3
-; GISEL-NEXT:    v_cmp_ge_u32_e64 s[6:7], v1, v3
-; GISEL-NEXT:    v_cndmask_b32_e64 v4, v4, v10, s[6:7]
-; GISEL-NEXT:    v_subrev_i32_e32 v8, vcc, 0x1000, v1
-; GISEL-NEXT:    v_cndmask_b32_e64 v0, v0, v7, s[4:5]
-; GISEL-NEXT:    v_add_i32_e32 v7, vcc, 1, v5
-; GISEL-NEXT:    v_cndmask_b32_e64 v1, v1, v8, s[6:7]
-; GISEL-NEXT:    v_add_i32_e32 v8, vcc, 1, v4
-; GISEL-NEXT:    v_cmp_ge_u32_e32 vcc, v0, v3
-; GISEL-NEXT:    v_cndmask_b32_e32 v0, v5, v7, vcc
-; GISEL-NEXT:    v_cmp_ge_u32_e32 vcc, v1, v3
-; GISEL-NEXT:    v_cndmask_b32_e32 v1, v4, v8, vcc
-; GISEL-NEXT:    v_xor_b32_e32 v0, v0, v2
-; GISEL-NEXT:    v_xor_b32_e32 v1, v1, v6
-; GISEL-NEXT:    v_sub_i32_e32 v0, vcc, v0, v2
-; GISEL-NEXT:    v_sub_i32_e32 v1, vcc, v1, v6
-; GISEL-NEXT:    s_setpc_b64 s[30:31]
-;
-; CGP-LABEL: v_sdiv_v2i32_pow2k_denom:
-; CGP:       ; %bb.0:
-; CGP-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; CGP-NEXT:    v_ashrrev_i32_e32 v2, 31, v0
-; CGP-NEXT:    v_rcp_iflag_f32_e32 v3, 0x45800000
-; CGP-NEXT:    v_mov_b32_e32 v4, 0xfffff000
-; CGP-NEXT:    v_mov_b32_e32 v5, 0x1000
-; CGP-NEXT:    v_ashrrev_i32_e32 v6, 31, v1
-; CGP-NEXT:    v_add_i32_e32 v0, vcc, v0, v2
-; CGP-NEXT:    v_mul_f32_e32 v3, 0x4f7ffffe, v3
-; CGP-NEXT:    v_add_i32_e32 v1, vcc, v1, v6
-; CGP-NEXT:    v_xor_b32_e32 v0, v0, v2
-; CGP-NEXT:    v_cvt_u32_f32_e32 v3, v3
-; CGP-NEXT:    v_xor_b32_e32 v1, v1, v6
-; CGP-NEXT:    v_mul_lo_u32 v4, v3, v4
-; CGP-NEXT:    v_mul_hi_u32 v4, v3, v4
-; CGP-NEXT:    v_add_i32_e32 v3, vcc, v3, v4
-; CGP-NEXT:    v_mul_hi_u32 v4, v0, v3
-; CGP-NEXT:    v_mul_hi_u32 v3, v1, v3
-; CGP-NEXT:    v_lshlrev_b32_e32 v7, 12, v4
-; CGP-NEXT:    v_add_i32_e32 v8, vcc, 1, v4
-; CGP-NEXT:    v_lshlrev_b32_e32 v9, 12, v3
-; CGP-NEXT:    v_add_i32_e32 v10, vcc, 1, v3
-; CGP-NEXT:    v_sub_i32_e32 v0, vcc, v0, v7
-; CGP-NEXT:    v_sub_i32_e32 v1, vcc, v1, v9
-; CGP-NEXT:    v_cmp_ge_u32_e64 s[4:5], v0, v5
-; CGP-NEXT:    v_cndmask_b32_e64 v4, v4, v8, s[4:5]
-; CGP-NEXT:    v_sub_i32_e32 v7, vcc, v0, v5
-; CGP-NEXT:    v_cmp_ge_u32_e64 s[6:7], v1, v5
-; CGP-NEXT:    v_cndmask_b32_e64 v3, v3, v10, s[6:7]
-; CGP-NEXT:    v_subrev_i32_e32 v8, vcc, 0x1000, v1
-; CGP-NEXT:    v_cndmask_b32_e64 v0, v0, v7, s[4:5]
-; CGP-NEXT:    v_add_i32_e32 v7, vcc, 1, v4
-; CGP-NEXT:    v_cndmask_b32_e64 v1, v1, v8, s[6:7]
-; CGP-NEXT:    v_add_i32_e32 v8, vcc, 1, v3
-; CGP-NEXT:    v_cmp_ge_u32_e32 vcc, v0, v5
-; CGP-NEXT:    v_cndmask_b32_e32 v0, v4, v7, vcc
-; CGP-NEXT:    v_cmp_ge_u32_e32 vcc, v1, v5
-; CGP-NEXT:    v_cndmask_b32_e32 v1, v3, v8, vcc
-; CGP-NEXT:    v_xor_b32_e32 v0, v0, v2
-; CGP-NEXT:    v_xor_b32_e32 v1, v1, v6
-; CGP-NEXT:    v_sub_i32_e32 v0, vcc, v0, v2
-; CGP-NEXT:    v_sub_i32_e32 v1, vcc, v1, v6
-; CGP-NEXT:    s_setpc_b64 s[30:31]
   %result = sdiv <2 x i32> %num, <i32 4096, i32 4096>
   ret <2 x i32> %result
 }
@@ -884,3 +786,13 @@ define <2 x i32> @v_sdiv_v2i32_24bit(<2 x i32> %num, <2 x i32> %den) {
   %result = sdiv <2 x i32> %num.mask, %den.mask
   ret <2 x i32> %result
 }
+
+define i32 @v_sdiv_i32_exact(i32 %num) {
+; CHECK-LABEL: v_sdiv_i32_exact:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    v_ashrrev_i32_e32 v0, 12, v0
+; CHECK-NEXT:    s_setpc_b64 s[30:31]
+  %result = sdiv exact i32 %num, 4096
+  ret i32 %result
+}

llvm/test/CodeGen/AMDGPU/GlobalISel/sdiv.i64.ll

@@ -999,126 +999,11 @@ define i64 @v_sdiv_i64_pow2k_denom(i64 %num) {
 ; CHECK-LABEL: v_sdiv_i64_pow2k_denom:
 ; CHECK:       ; %bb.0:
 ; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; CHECK-NEXT:    v_cvt_f32_u32_e32 v2, 0x1000
+; CHECK-NEXT:    v_ashrrev_i32_e32 v2, 31, v1
-; CHECK-NEXT:    v_cvt_f32_ubyte0_e32 v3, 0
+; CHECK-NEXT:    v_lshrrev_b32_e32 v2, 20, v2
-; CHECK-NEXT:    v_mov_b32_e32 v6, 0xfffff000
+; CHECK-NEXT:    v_add_i32_e32 v0, vcc, v0, v2
-; CHECK-NEXT:    v_mac_f32_e32 v2, 0x4f800000, v3
+; CHECK-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
-; CHECK-NEXT:    v_rcp_iflag_f32_e32 v2, v2
+; CHECK-NEXT:    v_ashr_i64 v[0:1], v[0:1], 12
-; CHECK-NEXT:    v_mul_f32_e32 v2, 0x5f7ffffc, v2
-; CHECK-NEXT:    v_mul_f32_e32 v3, 0x2f800000, v2
-; CHECK-NEXT:    v_trunc_f32_e32 v4, v3
-; CHECK-NEXT:    v_mac_f32_e32 v2, 0xcf800000, v4
-; CHECK-NEXT:    v_cvt_u32_f32_e32 v5, v2
-; CHECK-NEXT:    v_cvt_u32_f32_e32 v7, v4
-; CHECK-NEXT:    v_mad_u64_u32 v[2:3], s[4:5], v6, v5, 0
-; CHECK-NEXT:    v_mad_u64_u32 v[3:4], s[4:5], v6, v7, v[3:4]
-; CHECK-NEXT:    v_mul_hi_u32 v8, v5, v2
-; CHECK-NEXT:    v_mad_u64_u32 v[3:4], s[4:5], -1, v5, v[3:4]
-; CHECK-NEXT:    v_mul_lo_u32 v4, v7, v2
-; CHECK-NEXT:    v_mul_hi_u32 v2, v7, v2
-; CHECK-NEXT:    v_mul_lo_u32 v9, v5, v3
-; CHECK-NEXT:    v_mul_lo_u32 v10, v7, v3
-; CHECK-NEXT:    v_mul_hi_u32 v11, v5, v3
-; CHECK-NEXT:    v_mul_hi_u32 v3, v7, v3
-; CHECK-NEXT:    v_add_i32_e32 v4, vcc, v4, v9
-; CHECK-NEXT:    v_cndmask_b32_e64 v9, 0, 1, vcc
-; CHECK-NEXT:    v_add_i32_e32 v2, vcc, v10, v2
-; CHECK-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
-; CHECK-NEXT:    v_add_i32_e32 v4, vcc, v4, v8
-; CHECK-NEXT:    v_cndmask_b32_e64 v4, 0, 1, vcc
-; CHECK-NEXT:    v_add_i32_e32 v4, vcc, v9, v4
-; CHECK-NEXT:    v_add_i32_e32 v2, vcc, v2, v11
-; CHECK-NEXT:    v_cndmask_b32_e64 v8, 0, 1, vcc
-; CHECK-NEXT:    v_add_i32_e32 v8, vcc, v10, v8
-; CHECK-NEXT:    v_add_i32_e32 v2, vcc, v2, v4
-; CHECK-NEXT:    v_cndmask_b32_e64 v4, 0, 1, vcc
-; CHECK-NEXT:    v_add_i32_e32 v4, vcc, v8, v4
-; CHECK-NEXT:    v_add_i32_e32 v3, vcc, v3, v4
-; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v5, v2
-; CHECK-NEXT:    v_addc_u32_e32 v7, vcc, v7, v3, vcc
-; CHECK-NEXT:    v_mad_u64_u32 v[2:3], s[4:5], v6, v5, 0
-; CHECK-NEXT:    v_mad_u64_u32 v[3:4], s[4:5], v6, v7, v[3:4]
-; CHECK-NEXT:    v_ashrrev_i32_e32 v6, 31, v1
-; CHECK-NEXT:    v_add_i32_e32 v0, vcc, v0, v6
-; CHECK-NEXT:    v_mad_u64_u32 v[3:4], s[4:5], -1, v5, v[3:4]
-; CHECK-NEXT:    v_addc_u32_e32 v1, vcc, v1, v6, vcc
-; CHECK-NEXT:    v_xor_b32_e32 v4, v0, v6
-; CHECK-NEXT:    v_mul_lo_u32 v0, v7, v2
-; CHECK-NEXT:    v_mul_lo_u32 v8, v5, v3
-; CHECK-NEXT:    v_xor_b32_e32 v9, v1, v6
-; CHECK-NEXT:    v_mul_hi_u32 v1, v5, v2
-; CHECK-NEXT:    v_mul_hi_u32 v2, v7, v2
-; CHECK-NEXT:    v_add_i32_e32 v0, vcc, v0, v8
-; CHECK-NEXT:    v_cndmask_b32_e64 v8, 0, 1, vcc
-; CHECK-NEXT:    v_add_i32_e32 v0, vcc, v0, v1
-; CHECK-NEXT:    v_cndmask_b32_e64 v0, 0, 1, vcc
-; CHECK-NEXT:    v_mul_lo_u32 v1, v7, v3
-; CHECK-NEXT:    v_add_i32_e32 v0, vcc, v8, v0
-; CHECK-NEXT:    v_mul_hi_u32 v8, v5, v3
-; CHECK-NEXT:    v_add_i32_e32 v1, vcc, v1, v2
-; CHECK-NEXT:    v_cndmask_b32_e64 v2, 0, 1, vcc
-; CHECK-NEXT:    v_add_i32_e32 v1, vcc, v1, v8
-; CHECK-NEXT:    v_cndmask_b32_e64 v8, 0, 1, vcc
-; CHECK-NEXT:    v_add_i32_e32 v2, vcc, v2, v8
-; CHECK-NEXT:    v_mul_hi_u32 v3, v7, v3
-; CHECK-NEXT:    v_add_i32_e32 v0, vcc, v1, v0
-; CHECK-NEXT:    v_cndmask_b32_e64 v1, 0, 1, vcc
-; CHECK-NEXT:    v_add_i32_e32 v1, vcc, v2, v1
-; CHECK-NEXT:    v_add_i32_e32 v1, vcc, v3, v1
-; CHECK-NEXT:    v_add_i32_e32 v0, vcc, v5, v0
-; CHECK-NEXT:    v_addc_u32_e32 v1, vcc, v7, v1, vcc
-; CHECK-NEXT:    v_mul_lo_u32 v2, v9, v0
-; CHECK-NEXT:    v_mul_lo_u32 v3, v4, v1
-; CHECK-NEXT:    v_mul_hi_u32 v7, v4, v0
-; CHECK-NEXT:    v_mul_hi_u32 v0, v9, v0
-; CHECK-NEXT:    v_mov_b32_e32 v5, 0x1000
-; CHECK-NEXT:    v_add_i32_e32 v2, vcc, v2, v3
-; CHECK-NEXT:    v_cndmask_b32_e64 v3, 0, 1, vcc
-; CHECK-NEXT:    v_add_i32_e32 v2, vcc, v2, v7
-; CHECK-NEXT:    v_cndmask_b32_e64 v2, 0, 1, vcc
-; CHECK-NEXT:    v_mul_lo_u32 v7, v9, v1
-; CHECK-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
-; CHECK-NEXT:    v_mul_hi_u32 v3, v4, v1
-; CHECK-NEXT:    v_add_i32_e32 v0, vcc, v7, v0
-; CHECK-NEXT:    v_cndmask_b32_e64 v7, 0, 1, vcc
-; CHECK-NEXT:    v_add_i32_e32 v0, vcc, v0, v3
-; CHECK-NEXT:    v_cndmask_b32_e64 v3, 0, 1, vcc
-; CHECK-NEXT:    v_add_i32_e32 v3, vcc, v7, v3
-; CHECK-NEXT:    v_add_i32_e32 v7, vcc, v0, v2
-; CHECK-NEXT:    v_mul_hi_u32 v8, v9, v1
-; CHECK-NEXT:    v_mad_u64_u32 v[0:1], s[4:5], v5, v7, 0
-; CHECK-NEXT:    v_cndmask_b32_e64 v2, 0, 1, vcc
-; CHECK-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
-; CHECK-NEXT:    v_add_i32_e32 v3, vcc, v8, v2
-; CHECK-NEXT:    v_mad_u64_u32 v[1:2], s[4:5], v5, v3, v[1:2]
-; CHECK-NEXT:    v_sub_i32_e32 v0, vcc, v4, v0
-; CHECK-NEXT:    v_subb_u32_e64 v2, s[4:5], v9, v1, vcc
-; CHECK-NEXT:    v_sub_i32_e64 v1, s[4:5], v9, v1
-; CHECK-NEXT:    v_subbrev_u32_e32 v1, vcc, 0, v1, vcc
-; CHECK-NEXT:    v_cmp_ge_u32_e64 s[4:5], v0, v5
-; CHECK-NEXT:    v_sub_i32_e32 v0, vcc, v0, v5
-; CHECK-NEXT:    v_cndmask_b32_e64 v4, 0, -1, s[4:5]
-; CHECK-NEXT:    v_cmp_eq_u32_e64 s[4:5], 0, v2
-; CHECK-NEXT:    v_subbrev_u32_e32 v1, vcc, 0, v1, vcc
-; CHECK-NEXT:    v_cndmask_b32_e64 v2, -1, v4, s[4:5]
-; CHECK-NEXT:    v_add_i32_e32 v4, vcc, 1, v7
-; CHECK-NEXT:    v_addc_u32_e32 v8, vcc, 0, v3, vcc
-; CHECK-NEXT:    v_cmp_ge_u32_e32 vcc, v0, v5
-; CHECK-NEXT:    v_cndmask_b32_e64 v0, 0, -1, vcc
-; CHECK-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v1
-; CHECK-NEXT:    v_cndmask_b32_e32 v0, -1, v0, vcc
-; CHECK-NEXT:    v_add_i32_e32 v1, vcc, 1, v4
-; CHECK-NEXT:    v_addc_u32_e32 v5, vcc, 0, v8, vcc
-; CHECK-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v0
-; CHECK-NEXT:    v_cndmask_b32_e32 v0, v4, v1, vcc
-; CHECK-NEXT:    v_cndmask_b32_e32 v1, v8, v5, vcc
-; CHECK-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v2
-; CHECK-NEXT:    v_cndmask_b32_e32 v0, v7, v0, vcc
-; CHECK-NEXT:    v_cndmask_b32_e32 v1, v3, v1, vcc
-; CHECK-NEXT:    v_xor_b32_e32 v0, v0, v6
-; CHECK-NEXT:    v_xor_b32_e32 v1, v1, v6
-; CHECK-NEXT:    v_sub_i32_e32 v0, vcc, v0, v6
-; CHECK-NEXT:    v_subb_u32_e32 v1, vcc, v1, v6, vcc
 ; CHECK-NEXT:    s_setpc_b64 s[30:31]
   %result = sdiv i64 %num, 4096
   ret i64 %result
@@ -1128,473 +1013,31 @@ define <2 x i64> @v_sdiv_v2i64_pow2k_denom(<2 x i64> %num) {
 ; GISEL-LABEL: v_sdiv_v2i64_pow2k_denom:
 ; GISEL:       ; %bb.0:
 ; GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GISEL-NEXT:    v_cvt_f32_u32_e32 v4, 0x1000
-; GISEL-NEXT:    v_cvt_f32_ubyte0_e32 v5, 0
-; GISEL-NEXT:    s_sub_u32 s6, 0, 0x1000
-; GISEL-NEXT:    s_subb_u32 s7, 0, 0
-; GISEL-NEXT:    v_mac_f32_e32 v4, 0x4f800000, v5
-; GISEL-NEXT:    v_rcp_iflag_f32_e32 v4, v4
-; GISEL-NEXT:    v_mul_f32_e32 v4, 0x5f7ffffc, v4
-; GISEL-NEXT:    v_mul_f32_e32 v5, 0x2f800000, v4
-; GISEL-NEXT:    v_trunc_f32_e32 v7, v5
-; GISEL-NEXT:    v_mac_f32_e32 v4, 0xcf800000, v7
-; GISEL-NEXT:    v_cvt_u32_f32_e32 v6, v4
-; GISEL-NEXT:    v_cvt_u32_f32_e32 v7, v7
-; GISEL-NEXT:    v_mad_u64_u32 v[4:5], s[4:5], s6, v6, 0
-; GISEL-NEXT:    v_mad_u64_u32 v[8:9], s[4:5], s6, v7, v[5:6]
-; GISEL-NEXT:    v_mul_lo_u32 v5, v7, v4
-; GISEL-NEXT:    v_mad_u64_u32 v[8:9], s[4:5], s7, v6, v[8:9]
-; GISEL-NEXT:    v_mul_hi_u32 v9, v6, v4
-; GISEL-NEXT:    v_mul_hi_u32 v4, v7, v4
-; GISEL-NEXT:    v_mul_lo_u32 v10, v6, v8
-; GISEL-NEXT:    v_mul_lo_u32 v11, v7, v8
-; GISEL-NEXT:    v_mul_hi_u32 v12, v6, v8
-; GISEL-NEXT:    v_mul_hi_u32 v8, v7, v8
-; GISEL-NEXT:    v_add_i32_e32 v5, vcc, v5, v10
-; GISEL-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
-; GISEL-NEXT:    v_add_i32_e32 v5, vcc, v5, v9
-; GISEL-NEXT:    v_cndmask_b32_e64 v5, 0, 1, vcc
-; GISEL-NEXT:    v_add_i32_e32 v5, vcc, v10, v5
-; GISEL-NEXT:    v_add_i32_e32 v4, vcc, v11, v4
-; GISEL-NEXT:    v_cndmask_b32_e64 v9, 0, 1, vcc
-; GISEL-NEXT:    v_add_i32_e32 v4, vcc, v4, v12
-; GISEL-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
-; GISEL-NEXT:    v_add_i32_e32 v9, vcc, v9, v10
-; GISEL-NEXT:    v_add_i32_e32 v4, vcc, v4, v5
-; GISEL-NEXT:    v_cndmask_b32_e64 v5, 0, 1, vcc
-; GISEL-NEXT:    v_add_i32_e32 v5, vcc, v9, v5
-; GISEL-NEXT:    v_add_i32_e32 v5, vcc, v8, v5
-; GISEL-NEXT:    v_add_i32_e32 v11, vcc, v6, v4
-; GISEL-NEXT:    v_mad_u64_u32 v[8:9], s[4:5], s6, v11, 0
-; GISEL-NEXT:    v_addc_u32_e32 v5, vcc, v7, v5, vcc
-; GISEL-NEXT:    v_mov_b32_e32 v4, v9
-; GISEL-NEXT:    v_mad_u64_u32 v[9:10], s[4:5], s6, v5, v[4:5]
 ; GISEL-NEXT:    v_ashrrev_i32_e32 v4, 31, v1
+; GISEL-NEXT:    v_lshrrev_b32_e32 v4, 20, v4
+; GISEL-NEXT:    v_ashrrev_i32_e32 v5, 31, v3
 ; GISEL-NEXT:    v_add_i32_e32 v0, vcc, v0, v4
-; GISEL-NEXT:    v_mad_u64_u32 v[9:10], s[4:5], s7, v11, v[9:10]
+; GISEL-NEXT:    v_lshrrev_b32_e32 v5, 20, v5
-; GISEL-NEXT:    v_addc_u32_e32 v1, vcc, v1, v4, vcc
+; GISEL-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
-; GISEL-NEXT:    v_xor_b32_e32 v10, v0, v4
+; GISEL-NEXT:    v_add_i32_e32 v2, vcc, v2, v5
-; GISEL-NEXT:    v_mul_lo_u32 v0, v5, v8
+; GISEL-NEXT:    v_addc_u32_e32 v3, vcc, 0, v3, vcc
-; GISEL-NEXT:    v_mul_lo_u32 v12, v11, v9
+; GISEL-NEXT:    v_ashr_i64 v[0:1], v[0:1], 12
-; GISEL-NEXT:    v_xor_b32_e32 v13, v1, v4
+; GISEL-NEXT:    v_ashr_i64 v[2:3], v[2:3], 12
-; GISEL-NEXT:    v_mul_hi_u32 v1, v11, v8
-; GISEL-NEXT:    v_mul_hi_u32 v8, v5, v8
-; GISEL-NEXT:    v_add_i32_e32 v0, vcc, v0, v12
-; GISEL-NEXT:    v_cndmask_b32_e64 v12, 0, 1, vcc
-; GISEL-NEXT:    v_add_i32_e32 v0, vcc, v0, v1
-; GISEL-NEXT:    v_cndmask_b32_e64 v0, 0, 1, vcc
-; GISEL-NEXT:    v_mul_lo_u32 v1, v5, v9
-; GISEL-NEXT:    v_add_i32_e32 v0, vcc, v12, v0
-; GISEL-NEXT:    v_mul_hi_u32 v12, v11, v9
-; GISEL-NEXT:    v_add_i32_e32 v1, vcc, v1, v8
-; GISEL-NEXT:    v_cndmask_b32_e64 v8, 0, 1, vcc
-; GISEL-NEXT:    v_add_i32_e32 v1, vcc, v1, v12
-; GISEL-NEXT:    v_cndmask_b32_e64 v12, 0, 1, vcc
-; GISEL-NEXT:    v_add_i32_e32 v8, vcc, v8, v12
-; GISEL-NEXT:    v_mul_hi_u32 v9, v5, v9
-; GISEL-NEXT:    v_add_i32_e32 v0, vcc, v1, v0
-; GISEL-NEXT:    v_cndmask_b32_e64 v1, 0, 1, vcc
-; GISEL-NEXT:    v_add_i32_e32 v1, vcc, v8, v1
-; GISEL-NEXT:    v_add_i32_e32 v1, vcc, v9, v1
-; GISEL-NEXT:    v_add_i32_e32 v0, vcc, v11, v0
-; GISEL-NEXT:    v_addc_u32_e32 v1, vcc, v5, v1, vcc
-; GISEL-NEXT:    v_mul_lo_u32 v8, v13, v0
-; GISEL-NEXT:    v_mul_lo_u32 v9, v10, v1
-; GISEL-NEXT:    v_mul_hi_u32 v11, v10, v0
-; GISEL-NEXT:    v_mul_hi_u32 v0, v13, v0
-; GISEL-NEXT:    v_mov_b32_e32 v5, 0x1000
-; GISEL-NEXT:    v_add_i32_e32 v8, vcc, v8, v9
-; GISEL-NEXT:    v_cndmask_b32_e64 v9, 0, 1, vcc
-; GISEL-NEXT:    v_add_i32_e32 v8, vcc, v8, v11
-; GISEL-NEXT:    v_cndmask_b32_e64 v8, 0, 1, vcc
-; GISEL-NEXT:    v_mul_lo_u32 v11, v13, v1
-; GISEL-NEXT:    v_add_i32_e32 v8, vcc, v9, v8
-; GISEL-NEXT:    v_mul_hi_u32 v9, v10, v1
-; GISEL-NEXT:    v_add_i32_e32 v0, vcc, v11, v0
-; GISEL-NEXT:    v_cndmask_b32_e64 v11, 0, 1, vcc
-; GISEL-NEXT:    v_add_i32_e32 v0, vcc, v0, v9
-; GISEL-NEXT:    v_cndmask_b32_e64 v9, 0, 1, vcc
-; GISEL-NEXT:    v_add_i32_e32 v9, vcc, v11, v9
-; GISEL-NEXT:    v_add_i32_e32 v11, vcc, v0, v8
-; GISEL-NEXT:    v_mul_hi_u32 v12, v13, v1
-; GISEL-NEXT:    v_mad_u64_u32 v[0:1], s[4:5], v5, v11, 0
-; GISEL-NEXT:    v_cndmask_b32_e64 v8, 0, 1, vcc
-; GISEL-NEXT:    v_add_i32_e32 v8, vcc, v9, v8
-; GISEL-NEXT:    v_add_i32_e32 v12, vcc, v12, v8
-; GISEL-NEXT:    v_mad_u64_u32 v[8:9], s[4:5], v5, v12, v[1:2]
-; GISEL-NEXT:    v_sub_i32_e32 v0, vcc, v10, v0
-; GISEL-NEXT:    v_mad_u64_u32 v[8:9], s[4:5], 0, v11, v[8:9]
-; GISEL-NEXT:    s_sub_u32 s6, 0, 0x1000
-; GISEL-NEXT:    s_subb_u32 s7, 0, 0
-; GISEL-NEXT:    v_subb_u32_e64 v1, s[4:5], v13, v8, vcc
-; GISEL-NEXT:    v_sub_i32_e64 v8, s[4:5], v13, v8
-; GISEL-NEXT:    v_cmp_ge_u32_e64 s[4:5], v0, v5
-; GISEL-NEXT:    v_cndmask_b32_e64 v9, 0, -1, s[4:5]
-; GISEL-NEXT:    v_cmp_eq_u32_e64 s[4:5], 0, v1
-; GISEL-NEXT:    v_subbrev_u32_e32 v1, vcc, 0, v8, vcc
-; GISEL-NEXT:    v_sub_i32_e32 v8, vcc, v0, v5
-; GISEL-NEXT:    v_cndmask_b32_e64 v10, -1, v9, s[4:5]
-; GISEL-NEXT:    v_subbrev_u32_e32 v9, vcc, 0, v1, vcc
-; GISEL-NEXT:    v_add_i32_e32 v13, vcc, 1, v11
-; GISEL-NEXT:    v_mad_u64_u32 v[0:1], s[4:5], s6, v6, 0
-; GISEL-NEXT:    v_addc_u32_e32 v14, vcc, 0, v12, vcc
-; GISEL-NEXT:    v_cmp_ge_u32_e32 vcc, v8, v5
-; GISEL-NEXT:    v_cndmask_b32_e64 v8, 0, -1, vcc
-; GISEL-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v9
-; GISEL-NEXT:    v_cndmask_b32_e32 v15, -1, v8, vcc
-; GISEL-NEXT:    v_mad_u64_u32 v[8:9], s[4:5], s6, v7, v[1:2]
-; GISEL-NEXT:    v_add_i32_e32 v1, vcc, 1, v13
-; GISEL-NEXT:    v_mad_u64_u32 v[8:9], s[4:5], s7, v6, v[8:9]
-; GISEL-NEXT:    v_addc_u32_e32 v16, vcc, 0, v14, vcc
-; GISEL-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v15
-; GISEL-NEXT:    v_cndmask_b32_e32 v9, v13, v1, vcc
-; GISEL-NEXT:    v_mul_lo_u32 v1, v7, v0
-; GISEL-NEXT:    v_mul_lo_u32 v13, v6, v8
-; GISEL-NEXT:    v_mul_hi_u32 v15, v6, v0
-; GISEL-NEXT:    v_cndmask_b32_e32 v14, v14, v16, vcc
-; GISEL-NEXT:    v_mul_hi_u32 v0, v7, v0
-; GISEL-NEXT:    v_add_i32_e32 v1, vcc, v1, v13
-; GISEL-NEXT:    v_cndmask_b32_e64 v13, 0, 1, vcc
-; GISEL-NEXT:    v_add_i32_e32 v1, vcc, v1, v15
-; GISEL-NEXT:    v_cndmask_b32_e64 v1, 0, 1, vcc
-; GISEL-NEXT:    v_mul_lo_u32 v15, v7, v8
-; GISEL-NEXT:    v_add_i32_e32 v1, vcc, v13, v1
-; GISEL-NEXT:    v_mul_hi_u32 v13, v6, v8
-; GISEL-NEXT:    v_add_i32_e32 v0, vcc, v15, v0
-; GISEL-NEXT:    v_cndmask_b32_e64 v15, 0, 1, vcc
-; GISEL-NEXT:    v_add_i32_e32 v0, vcc, v0, v13
-; GISEL-NEXT:    v_cndmask_b32_e64 v13, 0, 1, vcc
-; GISEL-NEXT:    v_add_i32_e32 v13, vcc, v15, v13
-; GISEL-NEXT:    v_mul_hi_u32 v8, v7, v8
-; GISEL-NEXT:    v_add_i32_e32 v0, vcc, v0, v1
-; GISEL-NEXT:    v_cndmask_b32_e64 v1, 0, 1, vcc
-; GISEL-NEXT:    v_add_i32_e32 v1, vcc, v13, v1
-; GISEL-NEXT:    v_add_i32_e32 v1, vcc, v8, v1
-; GISEL-NEXT:    v_add_i32_e32 v8, vcc, v6, v0
-; GISEL-NEXT:    v_addc_u32_e32 v13, vcc, v7, v1, vcc
-; GISEL-NEXT:    v_mad_u64_u32 v[0:1], s[4:5], s6, v8, 0
-; GISEL-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v10
-; GISEL-NEXT:    v_cndmask_b32_e32 v9, v11, v9, vcc
-; GISEL-NEXT:    v_mad_u64_u32 v[6:7], s[4:5], s6, v13, v[1:2]
-; GISEL-NEXT:    v_xor_b32_e32 v1, v9, v4
-; GISEL-NEXT:    v_ashrrev_i32_e32 v9, 31, v3
-; GISEL-NEXT:    v_mad_u64_u32 v[6:7], s[4:5], s7, v8, v[6:7]
-; GISEL-NEXT:    v_cndmask_b32_e32 v10, v12, v14, vcc
-; GISEL-NEXT:    v_add_i32_e32 v2, vcc, v2, v9
-; GISEL-NEXT:    v_addc_u32_e32 v3, vcc, v3, v9, vcc
-; GISEL-NEXT:    v_xor_b32_e32 v11, v2, v9
-; GISEL-NEXT:    v_mul_lo_u32 v2, v13, v0
-; GISEL-NEXT:    v_mul_lo_u32 v7, v8, v6
-; GISEL-NEXT:    v_xor_b32_e32 v12, v3, v9
-; GISEL-NEXT:    v_mul_hi_u32 v3, v8, v0
-; GISEL-NEXT:    v_mul_hi_u32 v0, v13, v0
-; GISEL-NEXT:    v_add_i32_e32 v2, vcc, v2, v7
-; GISEL-NEXT:    v_cndmask_b32_e64 v7, 0, 1, vcc
-; GISEL-NEXT:    v_add_i32_e32 v2, vcc, v2, v3
-; GISEL-NEXT:    v_cndmask_b32_e64 v2, 0, 1, vcc
-; GISEL-NEXT:    v_mul_lo_u32 v3, v13, v6
-; GISEL-NEXT:    v_add_i32_e32 v2, vcc, v7, v2
-; GISEL-NEXT:    v_mul_hi_u32 v7, v8, v6
-; GISEL-NEXT:    v_add_i32_e32 v0, vcc, v3, v0
-; GISEL-NEXT:    v_cndmask_b32_e64 v3, 0, 1, vcc
-; GISEL-NEXT:    v_add_i32_e32 v0, vcc, v0, v7
-; GISEL-NEXT:    v_cndmask_b32_e64 v7, 0, 1, vcc
-; GISEL-NEXT:    v_add_i32_e32 v3, vcc, v3, v7
-; GISEL-NEXT:    v_mul_hi_u32 v6, v13, v6
-; GISEL-NEXT:    v_add_i32_e32 v0, vcc, v0, v2
-; GISEL-NEXT:    v_cndmask_b32_e64 v2, 0, 1, vcc
-; GISEL-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
-; GISEL-NEXT:    v_add_i32_e32 v2, vcc, v6, v2
-; GISEL-NEXT:    v_add_i32_e32 v0, vcc, v8, v0
-; GISEL-NEXT:    v_addc_u32_e32 v2, vcc, v13, v2, vcc
-; GISEL-NEXT:    v_mul_lo_u32 v3, v12, v0
-; GISEL-NEXT:    v_mul_lo_u32 v6, v11, v2
-; GISEL-NEXT:    v_mul_hi_u32 v7, v11, v0
-; GISEL-NEXT:    v_mul_hi_u32 v0, v12, v0
-; GISEL-NEXT:    v_xor_b32_e32 v8, v10, v4
-; GISEL-NEXT:    v_add_i32_e32 v3, vcc, v3, v6
-; GISEL-NEXT:    v_cndmask_b32_e64 v6, 0, 1, vcc
-; GISEL-NEXT:    v_add_i32_e32 v3, vcc, v3, v7
-; GISEL-NEXT:    v_cndmask_b32_e64 v3, 0, 1, vcc
-; GISEL-NEXT:    v_mul_lo_u32 v7, v12, v2
-; GISEL-NEXT:    v_add_i32_e32 v3, vcc, v6, v3
-; GISEL-NEXT:    v_mul_hi_u32 v6, v11, v2
-; GISEL-NEXT:    v_add_i32_e32 v0, vcc, v7, v0
-; GISEL-NEXT:    v_cndmask_b32_e64 v7, 0, 1, vcc
-; GISEL-NEXT:    v_add_i32_e32 v0, vcc, v0, v6
-; GISEL-NEXT:    v_cndmask_b32_e64 v6, 0, 1, vcc
-; GISEL-NEXT:    v_add_i32_e32 v6, vcc, v7, v6
-; GISEL-NEXT:    v_add_i32_e32 v10, vcc, v0, v3
-; GISEL-NEXT:    v_mul_hi_u32 v7, v12, v2
-; GISEL-NEXT:    v_mad_u64_u32 v[2:3], s[4:5], v5, v10, 0
-; GISEL-NEXT:    v_cndmask_b32_e64 v0, 0, 1, vcc
-; GISEL-NEXT:    v_add_i32_e32 v0, vcc, v6, v0
-; GISEL-NEXT:    v_add_i32_e32 v13, vcc, v7, v0
-; GISEL-NEXT:    v_mov_b32_e32 v0, v3
-; GISEL-NEXT:    v_mad_u64_u32 v[6:7], s[4:5], v5, v13, v[0:1]
-; GISEL-NEXT:    v_sub_i32_e32 v0, vcc, v1, v4
-; GISEL-NEXT:    v_subb_u32_e32 v1, vcc, v8, v4, vcc
-; GISEL-NEXT:    v_mad_u64_u32 v[3:4], s[4:5], 0, v10, v[6:7]
-; GISEL-NEXT:    v_sub_i32_e32 v2, vcc, v11, v2
-; GISEL-NEXT:    v_subb_u32_e64 v4, s[4:5], v12, v3, vcc
-; GISEL-NEXT:    v_sub_i32_e64 v3, s[4:5], v12, v3
-; GISEL-NEXT:    v_subbrev_u32_e32 v3, vcc, 0, v3, vcc
-; GISEL-NEXT:    v_cmp_ge_u32_e64 s[4:5], v2, v5
-; GISEL-NEXT:    v_sub_i32_e32 v2, vcc, v2, v5
-; GISEL-NEXT:    v_cndmask_b32_e64 v6, 0, -1, s[4:5]
-; GISEL-NEXT:    v_cmp_eq_u32_e64 s[4:5], 0, v4
-; GISEL-NEXT:    v_subbrev_u32_e32 v3, vcc, 0, v3, vcc
-; GISEL-NEXT:    v_cndmask_b32_e64 v4, -1, v6, s[4:5]
-; GISEL-NEXT:    v_add_i32_e32 v6, vcc, 1, v10
-; GISEL-NEXT:    v_addc_u32_e32 v7, vcc, 0, v13, vcc
-; GISEL-NEXT:    v_cmp_ge_u32_e32 vcc, v2, v5
-; GISEL-NEXT:    v_cndmask_b32_e64 v2, 0, -1, vcc
-; GISEL-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v3
-; GISEL-NEXT:    v_cndmask_b32_e32 v2, -1, v2, vcc
-; GISEL-NEXT:    v_add_i32_e32 v3, vcc, 1, v6
-; GISEL-NEXT:    v_addc_u32_e32 v5, vcc, 0, v7, vcc
-; GISEL-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v2
-; GISEL-NEXT:    v_cndmask_b32_e32 v2, v6, v3, vcc
-; GISEL-NEXT:    v_cndmask_b32_e32 v3, v7, v5, vcc
-; GISEL-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v4
-; GISEL-NEXT:    v_cndmask_b32_e32 v2, v10, v2, vcc
-; GISEL-NEXT:    v_cndmask_b32_e32 v3, v13, v3, vcc
-; GISEL-NEXT:    v_xor_b32_e32 v2, v2, v9
-; GISEL-NEXT:    v_xor_b32_e32 v3, v3, v9
-; GISEL-NEXT:    v_sub_i32_e32 v2, vcc, v2, v9
-; GISEL-NEXT:    v_subb_u32_e32 v3, vcc, v3, v9, vcc
 ; GISEL-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; CGP-LABEL: v_sdiv_v2i64_pow2k_denom:
 ; CGP:       ; %bb.0:
 ; CGP-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; CGP-NEXT:    v_cvt_f32_u32_e32 v4, 0x1000
+; CGP-NEXT:    v_ashrrev_i32_e32 v4, 31, v1
-; CGP-NEXT:    v_cvt_f32_ubyte0_e32 v5, 0
+; CGP-NEXT:    v_lshrrev_b32_e32 v4, 20, v4
-; CGP-NEXT:    v_mov_b32_e32 v6, 0xfffff000
-; CGP-NEXT:    v_mac_f32_e32 v4, 0x4f800000, v5
-; CGP-NEXT:    v_rcp_iflag_f32_e32 v4, v4
-; CGP-NEXT:    v_mul_f32_e32 v4, 0x5f7ffffc, v4
-; CGP-NEXT:    v_mul_f32_e32 v5, 0x2f800000, v4
-; CGP-NEXT:    v_trunc_f32_e32 v7, v5
-; CGP-NEXT:    v_mac_f32_e32 v4, 0xcf800000, v7
-; CGP-NEXT:    v_cvt_u32_f32_e32 v8, v4
-; CGP-NEXT:    v_cvt_u32_f32_e32 v9, v7
-; CGP-NEXT:    v_mad_u64_u32 v[4:5], s[4:5], v6, v8, 0
-; CGP-NEXT:    v_mov_b32_e32 v7, v5
-; CGP-NEXT:    v_mad_u64_u32 v[10:11], s[4:5], v6, v9, v[7:8]
-; CGP-NEXT:    v_mul_hi_u32 v12, v9, v4
-; CGP-NEXT:    v_mad_u64_u32 v[13:14], s[4:5], -1, v8, v[10:11]
-; CGP-NEXT:    v_mul_lo_u32 v10, v9, v4
-; CGP-NEXT:    v_mul_hi_u32 v11, v8, v4
-; CGP-NEXT:    v_mul_lo_u32 v4, v8, v13
-; CGP-NEXT:    v_mul_lo_u32 v7, v9, v13
-; CGP-NEXT:    v_mul_hi_u32 v14, v8, v13
-; CGP-NEXT:    v_mul_hi_u32 v13, v9, v13
-; CGP-NEXT:    v_add_i32_e32 v4, vcc, v10, v4
-; CGP-NEXT:    v_cndmask_b32_e64 v15, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v4, vcc, v4, v11
-; CGP-NEXT:    v_cndmask_b32_e64 v4, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v4, vcc, v15, v4
-; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v12
-; CGP-NEXT:    v_cndmask_b32_e64 v15, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v14
-; CGP-NEXT:    v_cndmask_b32_e64 v14, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v14, vcc, v15, v14
-; CGP-NEXT:    v_add_i32_e32 v4, vcc, v7, v4
-; CGP-NEXT:    v_cndmask_b32_e64 v7, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v7, vcc, v14, v7
-; CGP-NEXT:    v_add_i32_e32 v7, vcc, v13, v7
-; CGP-NEXT:    v_add_i32_e32 v16, vcc, v8, v4
-; CGP-NEXT:    v_mad_u64_u32 v[13:14], s[4:5], v6, v16, 0
-; CGP-NEXT:    v_addc_u32_e32 v17, vcc, v9, v7, vcc
-; CGP-NEXT:    v_mov_b32_e32 v4, v14
-; CGP-NEXT:    v_mad_u64_u32 v[14:15], s[4:5], v6, v17, v[4:5]
-; CGP-NEXT:    v_ashrrev_i32_e32 v7, 31, v1
-; CGP-NEXT:    v_add_i32_e32 v0, vcc, v0, v7
-; CGP-NEXT:    v_mad_u64_u32 v[14:15], s[4:5], -1, v16, v[14:15]
-; CGP-NEXT:    v_addc_u32_e32 v1, vcc, v1, v7, vcc
-; CGP-NEXT:    v_xor_b32_e32 v15, v0, v7
-; CGP-NEXT:    v_mul_lo_u32 v0, v17, v13
-; CGP-NEXT:    v_mul_lo_u32 v4, v16, v14
-; CGP-NEXT:    v_xor_b32_e32 v18, v1, v7
-; CGP-NEXT:    v_mul_hi_u32 v1, v16, v13
-; CGP-NEXT:    v_mul_hi_u32 v13, v17, v13
 ; CGP-NEXT:    v_add_i32_e32 v0, vcc, v0, v4
-; CGP-NEXT:    v_cndmask_b32_e64 v4, 0, 1, vcc
+; CGP-NEXT:    v_ashrrev_i32_e32 v4, 31, v3
-; CGP-NEXT:    v_add_i32_e32 v0, vcc, v0, v1
+; CGP-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
-; CGP-NEXT:    v_cndmask_b32_e64 v0, 0, 1, vcc
+; CGP-NEXT:    v_lshrrev_b32_e32 v4, 20, v4
-; CGP-NEXT:    v_mul_lo_u32 v1, v17, v14
+; CGP-NEXT:    v_add_i32_e32 v2, vcc, v2, v4
-; CGP-NEXT:    v_add_i32_e32 v0, vcc, v4, v0
+; CGP-NEXT:    v_addc_u32_e32 v3, vcc, 0, v3, vcc
-; CGP-NEXT:    v_mul_hi_u32 v4, v16, v14
+; CGP-NEXT:    v_ashr_i64 v[0:1], v[0:1], 12
-; CGP-NEXT:    v_add_i32_e32 v1, vcc, v1, v13
+; CGP-NEXT:    v_ashr_i64 v[2:3], v[2:3], 12
-; CGP-NEXT:    v_cndmask_b32_e64 v13, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v1, vcc, v1, v4
-; CGP-NEXT:    v_cndmask_b32_e64 v4, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v4, vcc, v13, v4
-; CGP-NEXT:    v_mul_hi_u32 v13, v17, v14
-; CGP-NEXT:    v_add_i32_e32 v0, vcc, v1, v0
-; CGP-NEXT:    v_cndmask_b32_e64 v1, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v1, vcc, v4, v1
-; CGP-NEXT:    v_add_i32_e32 v1, vcc, v13, v1
-; CGP-NEXT:    v_add_i32_e32 v0, vcc, v16, v0
-; CGP-NEXT:    v_addc_u32_e32 v1, vcc, v17, v1, vcc
-; CGP-NEXT:    v_mul_lo_u32 v13, v18, v0
-; CGP-NEXT:    v_mul_lo_u32 v14, v15, v1
-; CGP-NEXT:    v_mul_hi_u32 v16, v15, v0
-; CGP-NEXT:    v_mul_hi_u32 v0, v18, v0
-; CGP-NEXT:    v_mov_b32_e32 v4, 0x1000
-; CGP-NEXT:    v_add_i32_e32 v13, vcc, v13, v14
-; CGP-NEXT:    v_cndmask_b32_e64 v14, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v13, vcc, v13, v16
-; CGP-NEXT:    v_cndmask_b32_e64 v13, 0, 1, vcc
-; CGP-NEXT:    v_mul_lo_u32 v16, v18, v1
-; CGP-NEXT:    v_add_i32_e32 v13, vcc, v14, v13
-; CGP-NEXT:    v_mul_hi_u32 v14, v15, v1
-; CGP-NEXT:    v_add_i32_e32 v0, vcc, v16, v0
-; CGP-NEXT:    v_cndmask_b32_e64 v16, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v0, vcc, v0, v14
-; CGP-NEXT:    v_cndmask_b32_e64 v14, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v14, vcc, v16, v14
-; CGP-NEXT:    v_add_i32_e32 v16, vcc, v0, v13
-; CGP-NEXT:    v_mul_hi_u32 v17, v18, v1
-; CGP-NEXT:    v_mad_u64_u32 v[0:1], s[4:5], v4, v16, 0
-; CGP-NEXT:    v_cndmask_b32_e64 v13, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v13, vcc, v14, v13
-; CGP-NEXT:    v_add_i32_e32 v17, vcc, v17, v13
-; CGP-NEXT:    v_mad_u64_u32 v[13:14], s[4:5], v4, v17, v[1:2]
-; CGP-NEXT:    v_sub_i32_e32 v0, vcc, v15, v0
-; CGP-NEXT:    v_subb_u32_e64 v1, s[4:5], v18, v13, vcc
-; CGP-NEXT:    v_sub_i32_e64 v13, s[4:5], v18, v13
-; CGP-NEXT:    v_cmp_ge_u32_e64 s[4:5], v0, v4
-; CGP-NEXT:    v_cndmask_b32_e64 v14, 0, -1, s[4:5]
-; CGP-NEXT:    v_cmp_eq_u32_e64 s[4:5], 0, v1
-; CGP-NEXT:    v_subbrev_u32_e32 v1, vcc, 0, v13, vcc
-; CGP-NEXT:    v_sub_i32_e32 v0, vcc, v0, v4
-; CGP-NEXT:    v_subbrev_u32_e32 v13, vcc, 0, v1, vcc
-; CGP-NEXT:    v_add_i32_e32 v15, vcc, 1, v16
-; CGP-NEXT:    v_addc_u32_e32 v18, vcc, 0, v17, vcc
-; CGP-NEXT:    v_cmp_ge_u32_e32 vcc, v0, v4
-; CGP-NEXT:    v_mov_b32_e32 v0, v5
-; CGP-NEXT:    v_cndmask_b32_e64 v14, -1, v14, s[4:5]
-; CGP-NEXT:    v_mad_u64_u32 v[0:1], s[4:5], v6, v9, v[0:1]
-; CGP-NEXT:    v_cndmask_b32_e64 v19, 0, -1, vcc
-; CGP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v13
-; CGP-NEXT:    v_mad_u64_u32 v[0:1], s[4:5], -1, v8, v[0:1]
-; CGP-NEXT:    v_cndmask_b32_e32 v5, -1, v19, vcc
-; CGP-NEXT:    v_add_i32_e32 v1, vcc, 1, v15
-; CGP-NEXT:    v_mul_lo_u32 v19, v8, v0
-; CGP-NEXT:    v_addc_u32_e32 v13, vcc, 0, v18, vcc
-; CGP-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v5
-; CGP-NEXT:    v_cndmask_b32_e32 v5, v15, v1, vcc
-; CGP-NEXT:    v_cndmask_b32_e32 v13, v18, v13, vcc
-; CGP-NEXT:    v_add_i32_e32 v1, vcc, v10, v19
-; CGP-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v1, vcc, v1, v11
-; CGP-NEXT:    v_cndmask_b32_e64 v1, 0, 1, vcc
-; CGP-NEXT:    v_mul_lo_u32 v11, v9, v0
-; CGP-NEXT:    v_add_i32_e32 v1, vcc, v10, v1
-; CGP-NEXT:    v_mul_hi_u32 v10, v8, v0
-; CGP-NEXT:    v_add_i32_e32 v11, vcc, v11, v12
-; CGP-NEXT:    v_cndmask_b32_e64 v12, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v10, vcc, v11, v10
-; CGP-NEXT:    v_cndmask_b32_e64 v11, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v11, vcc, v12, v11
-; CGP-NEXT:    v_mul_hi_u32 v0, v9, v0
-; CGP-NEXT:    v_add_i32_e32 v1, vcc, v10, v1
-; CGP-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v10, vcc, v11, v10
-; CGP-NEXT:    v_add_i32_e32 v0, vcc, v0, v10
-; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v1
-; CGP-NEXT:    v_addc_u32_e32 v9, vcc, v9, v0, vcc
-; CGP-NEXT:    v_mad_u64_u32 v[0:1], s[4:5], v6, v8, 0
-; CGP-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v14
-; CGP-NEXT:    v_cndmask_b32_e32 v5, v16, v5, vcc
-; CGP-NEXT:    v_xor_b32_e32 v11, v5, v7
-; CGP-NEXT:    v_mad_u64_u32 v[5:6], s[4:5], v6, v9, v[1:2]
-; CGP-NEXT:    v_cndmask_b32_e32 v10, v17, v13, vcc
-; CGP-NEXT:    v_xor_b32_e32 v1, v10, v7
-; CGP-NEXT:    v_mad_u64_u32 v[5:6], s[4:5], -1, v8, v[5:6]
-; CGP-NEXT:    v_ashrrev_i32_e32 v10, 31, v3
-; CGP-NEXT:    v_add_i32_e32 v2, vcc, v2, v10
-; CGP-NEXT:    v_addc_u32_e32 v3, vcc, v3, v10, vcc
-; CGP-NEXT:    v_xor_b32_e32 v12, v2, v10
-; CGP-NEXT:    v_mul_lo_u32 v2, v9, v0
-; CGP-NEXT:    v_mul_lo_u32 v6, v8, v5
-; CGP-NEXT:    v_xor_b32_e32 v13, v3, v10
-; CGP-NEXT:    v_mul_hi_u32 v3, v8, v0
-; CGP-NEXT:    v_mul_hi_u32 v0, v9, v0
-; CGP-NEXT:    v_add_i32_e32 v2, vcc, v2, v6
-; CGP-NEXT:    v_cndmask_b32_e64 v6, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v2, vcc, v2, v3
-; CGP-NEXT:    v_cndmask_b32_e64 v2, 0, 1, vcc
-; CGP-NEXT:    v_mul_lo_u32 v3, v9, v5
-; CGP-NEXT:    v_add_i32_e32 v2, vcc, v6, v2
-; CGP-NEXT:    v_mul_hi_u32 v6, v8, v5
-; CGP-NEXT:    v_add_i32_e32 v0, vcc, v3, v0
-; CGP-NEXT:    v_cndmask_b32_e64 v3, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v0, vcc, v0, v6
-; CGP-NEXT:    v_cndmask_b32_e64 v6, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v3, vcc, v3, v6
-; CGP-NEXT:    v_mul_hi_u32 v5, v9, v5
-; CGP-NEXT:    v_add_i32_e32 v0, vcc, v0, v2
-; CGP-NEXT:    v_cndmask_b32_e64 v2, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
-; CGP-NEXT:    v_add_i32_e32 v2, vcc, v5, v2
-; CGP-NEXT:    v_add_i32_e32 v3, vcc, v8, v0
-; CGP-NEXT:    v_addc_u32_e32 v2, vcc, v9, v2, vcc
-; CGP-NEXT:    v_mul_lo_u32 v5, v13, v3
-; CGP-NEXT:    v_mul_lo_u32 v6, v12, v2
-; CGP-NEXT:    v_sub_i32_e32 v0, vcc, v11, v7
-; CGP-NEXT:    v_subb_u32_e32 v1, vcc, v1, v7, vcc
-; CGP-NEXT:    v_mul_hi_u32 v7, v12, v3
-; CGP-NEXT:    v_add_i32_e32 v5, vcc, v5, v6
-; CGP-NEXT:    v_cndmask_b32_e64 v6, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v5, vcc, v5, v7
-; CGP-NEXT:    v_cndmask_b32_e64 v5, 0, 1, vcc
-; CGP-NEXT:    v_mul_lo_u32 v7, v13, v2
-; CGP-NEXT:    v_mul_hi_u32 v3, v13, v3
-; CGP-NEXT:    v_add_i32_e32 v5, vcc, v6, v5
-; CGP-NEXT:    v_mul_hi_u32 v6, v12, v2
-; CGP-NEXT:    v_add_i32_e32 v3, vcc, v7, v3
-; CGP-NEXT:    v_cndmask_b32_e64 v7, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v3, vcc, v3, v6
-; CGP-NEXT:    v_cndmask_b32_e64 v6, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v6, vcc, v7, v6
-; CGP-NEXT:    v_add_i32_e32 v7, vcc, v3, v5
-; CGP-NEXT:    v_mul_hi_u32 v8, v13, v2
-; CGP-NEXT:    v_mad_u64_u32 v[2:3], s[4:5], v4, v7, 0
-; CGP-NEXT:    v_cndmask_b32_e64 v5, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v5, vcc, v6, v5
-; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v5
-; CGP-NEXT:    v_mad_u64_u32 v[5:6], s[4:5], v4, v8, v[3:4]
-; CGP-NEXT:    v_sub_i32_e32 v2, vcc, v12, v2
-; CGP-NEXT:    v_subb_u32_e64 v3, s[4:5], v13, v5, vcc
-; CGP-NEXT:    v_sub_i32_e64 v5, s[4:5], v13, v5
-; CGP-NEXT:    v_subbrev_u32_e32 v5, vcc, 0, v5, vcc
-; CGP-NEXT:    v_cmp_ge_u32_e64 s[4:5], v2, v4
-; CGP-NEXT:    v_sub_i32_e32 v2, vcc, v2, v4
-; CGP-NEXT:    v_cndmask_b32_e64 v6, 0, -1, s[4:5]
-; CGP-NEXT:    v_cmp_eq_u32_e64 s[4:5], 0, v3
-; CGP-NEXT:    v_subbrev_u32_e32 v5, vcc, 0, v5, vcc
-; CGP-NEXT:    v_cndmask_b32_e64 v3, -1, v6, s[4:5]
-; CGP-NEXT:    v_add_i32_e32 v6, vcc, 1, v7
-; CGP-NEXT:    v_addc_u32_e32 v9, vcc, 0, v8, vcc
-; CGP-NEXT:    v_cmp_ge_u32_e32 vcc, v2, v4
-; CGP-NEXT:    v_cndmask_b32_e64 v2, 0, -1, vcc
-; CGP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v5
-; CGP-NEXT:    v_cndmask_b32_e32 v2, -1, v2, vcc
-; CGP-NEXT:    v_add_i32_e32 v4, vcc, 1, v6
-; CGP-NEXT:    v_addc_u32_e32 v5, vcc, 0, v9, vcc
-; CGP-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v2
-; CGP-NEXT:    v_cndmask_b32_e32 v2, v6, v4, vcc
-; CGP-NEXT:    v_cndmask_b32_e32 v4, v9, v5, vcc
-; CGP-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v3
-; CGP-NEXT:    v_cndmask_b32_e32 v2, v7, v2, vcc
-; CGP-NEXT:    v_cndmask_b32_e32 v3, v8, v4, vcc
-; CGP-NEXT:    v_xor_b32_e32 v2, v2, v10
-; CGP-NEXT:    v_xor_b32_e32 v3, v3, v10
-; CGP-NEXT:    v_sub_i32_e32 v2, vcc, v2, v10
-; CGP-NEXT:    v_subb_u32_e32 v3, vcc, v3, v10, vcc
 ; CGP-NEXT:    s_setpc_b64 s[30:31]
   %result = sdiv <2 x i64> %num, <i64 4096, i64 4096>
   ret <2 x i64> %result
@@ -3398,3 +2841,13 @@ define <2 x i64> @v_sdiv_v2i64_24bit(<2 x i64> %num, <2 x i64> %den) {
   %result = sdiv <2 x i64> %num.mask, %den.mask
   ret <2 x i64> %result
 }
+
+define i64 @v_sdiv_i64_exact(i64 %num) {
+; CHECK-LABEL: v_sdiv_i64_exact:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    v_ashr_i64 v[0:1], v[0:1], 12
+; CHECK-NEXT:    s_setpc_b64 s[30:31]
+  %result = sdiv exact i64 %num, 4096
+  ret i64 %result
+}