diff --git a/llvm/include/llvm/CodeGen/GlobalISel/MachineIRBuilder.h b/llvm/include/llvm/CodeGen/GlobalISel/MachineIRBuilder.h
--- a/llvm/include/llvm/CodeGen/GlobalISel/MachineIRBuilder.h
+++ b/llvm/include/llvm/CodeGen/GlobalISel/MachineIRBuilder.h
@@ -1210,6 +1210,12 @@
     return buildInstr(TargetOpcode::G_SMULH, {Dst}, {Src0, Src1}, Flags);
   }
 
+  MachineInstrBuilder buildFMul(const DstOp &Dst, const SrcOp &Src0,
+                                const SrcOp &Src1,
+                                Optional<unsigned> Flags = None) {
+    return buildInstr(TargetOpcode::G_FMUL, {Dst}, {Src0, Src1}, Flags);
+  }
+
   MachineInstrBuilder buildShl(const DstOp &Dst, const SrcOp &Src0,
                                const SrcOp &Src1,
                                Optional<unsigned> Flags = None) {
diff --git a/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.h b/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.h
--- a/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.h
+++ b/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.h
@@ -65,6 +65,9 @@
     MachineInstr &MI, MachineRegisterInfo &MRI, MachineIRBuilder &B,
     AMDGPUFunctionArgInfo::PreloadedValue ArgType) const;
 
+  bool legalizeFDIVFast(MachineInstr &MI, MachineRegisterInfo &MRI,
+                        MachineIRBuilder &B) const;
+
   bool legalizeImplicitArgPtr(MachineInstr &MI, MachineRegisterInfo &MRI,
                               MachineIRBuilder &B) const;
   bool legalizeIntrinsic(MachineInstr &MI, MachineRegisterInfo &MRI,
diff --git a/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp b/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
--- a/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
@@ -1272,6 +1272,45 @@
   return false;
 }
 
+bool AMDGPULegalizerInfo::legalizeFDIVFast(MachineInstr &MI,
+                                           MachineRegisterInfo &MRI,
+                                           MachineIRBuilder &B) const {
+  B.setInstr(MI);
+  Register Res = MI.getOperand(0).getReg();
+  Register LHS = MI.getOperand(2).getReg();
+  Register RHS = MI.getOperand(3).getReg();
+
+  LLT S32 = LLT::scalar(32);
+  LLT S1 = LLT::scalar(1);
+
+  auto Abs = B.buildFAbs(S32, RHS);
+
+  const APFloat C0Val(BitsToFloat(0x6f800000));
+  const APFloat C1Val(BitsToFloat(0x2f800000));
+
+  auto C0 = B.buildFConstant(S32, C0Val);
+  auto C1 = B.buildFConstant(S32, C1Val);
+  auto C2 = B.buildFConstant(S32, 1.0);
+
+  auto CmpRes = B.buildFCmp(CmpInst::FCMP_OGT, S1, Abs, C0);
+  auto Sel = B.buildSelect(S32, CmpRes, C1, C2);
+
+  auto Mul0 = B.buildFMul(S32, RHS, Sel.getReg(0));
+
+  auto RCP = B.buildIntrinsic(Intrinsic::amdgcn_rcp, {S32}, false)
+    .addUse(Mul0.getReg(0));
+
+  auto Mul1 = B.buildFMul(S32, LHS, RCP.getReg(0));
+
+  B.buildInstr(TargetOpcode::G_FMUL)
+    .addDef(Res)
+    .addUse(Sel.getReg(0))
+    .addUse(Mul1.getReg(0));
+
+  MI.eraseFromParent();
+  return true;
+}
+
 bool AMDGPULegalizerInfo::legalizeImplicitArgPtr(MachineInstr &MI,
                                                  MachineRegisterInfo &MRI,
                                                  MachineIRBuilder &B) const {
@@ -1386,6 +1425,8 @@
   case Intrinsic::amdgcn_dispatch_id:
     return legalizePreloadedArgIntrin(MI, MRI, B,
                                       AMDGPUFunctionArgInfo::DISPATCH_ID);
+  case Intrinsic::amdgcn_fdiv_fast:
+    return legalizeFDIVFast(MI, MRI, B);
   default:
     return true;
   }
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-intrinsic-amdgcn-fdiv-fast.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-intrinsic-amdgcn-fdiv-fast.mir
new file mode 100644
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-intrinsic-amdgcn-fdiv-fast.mir
@@ -0,0 +1,28 @@
+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
+# RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -run-pass=legalizer %s -o - | FileCheck %s
+
+---
+name: test_amdgcn_fdiv_fast
+body: |
+  bb.0:
+    liveins: $vgpr0, $vgpr1
+
+    ; CHECK-LABEL: name: test_amdgcn_fdiv_fast
+    ; CHECK: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
+    ; CHECK: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1
+    ; CHECK: [[FABS:%[0-9]+]]:_(s32) = G_FABS [[COPY1]]
+    ; CHECK: [[C:%[0-9]+]]:_(s32) = G_FCONSTANT float 0x45F0000000000000
+    ; CHECK: [[C1:%[0-9]+]]:_(s32) = G_FCONSTANT float 0x3DF0000000000000
+    ; CHECK: [[C2:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.000000e+00
+    ; CHECK: [[FCMP:%[0-9]+]]:_(s1) = G_FCMP floatpred(ogt), [[FABS]](s32), [[C]]
+    ; CHECK: [[SELECT:%[0-9]+]]:_(s32) = G_SELECT [[FCMP]](s1), [[C1]], [[C2]]
+    ; CHECK: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[COPY1]], [[SELECT]]
+    ; CHECK: [[INT:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[FMUL]](s32)
+    ; CHECK: [[FMUL1:%[0-9]+]]:_(s32) = G_FMUL [[COPY]], [[INT]]
+    ; CHECK: [[FMUL2:%[0-9]+]]:_(s32) = G_FMUL [[SELECT]], [[FMUL1]]
+    ; CHECK: $vgpr0 = COPY [[FMUL2]](s32)
+    %0:_(s32) = COPY $vgpr0
+    %1:_(s32) = COPY $vgpr1
+    %2:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.fdiv.fast), %0, %1
+    $vgpr0 = COPY %2
+...