Index: lib/Target/AMDGPU/AMDGPUISelDAGToDAG.cpp
===================================================================
--- lib/Target/AMDGPU/AMDGPUISelDAGToDAG.cpp
+++ lib/Target/AMDGPU/AMDGPUISelDAGToDAG.cpp
@@ -135,6 +135,8 @@
 
   void SelectADD_SUB_I64(SDNode *N);
   void SelectDIV_SCALE(SDNode *N);
+  void SelectFMA_W_CHAIN(SDNode *N);
+  void SelectFMUL_W_CHAIN(SDNode *N);
 
   SDNode *getS_BFE(unsigned Opcode, const SDLoc &DL, SDValue Val,
                    uint32_t Offset, uint32_t Width);
@@ -296,6 +298,15 @@
     SelectADD_SUB_I64(N);
     return;
   }
+  case AMDGPUISD::FMUL_W_CHAIN: {
+    SelectFMUL_W_CHAIN(N);
+    return;
+  }
+  case AMDGPUISD::FMA_W_CHAIN: {
+    SelectFMA_W_CHAIN(N);
+    return;
+  }
+
   case ISD::SCALAR_TO_VECTOR:
   case AMDGPUISD::BUILD_VERTICAL_VECTOR:
   case ISD::BUILD_VECTOR: {
@@ -653,6 +664,33 @@
   CurDAG->RemoveDeadNode(N);
 }
 
+void AMDGPUDAGToDAGISel::SelectFMA_W_CHAIN(SDNode *N) {
+  SDLoc SL(N);
+  //  src0_modifiers, src0,  src1_modifiers, src1, src2_modifiers, src2, clamp, omod
+  SDValue Ops[10];
+
+  SelectVOP3Mods0(N->getOperand(1), Ops[1], Ops[0], Ops[6], Ops[7]);
+  SelectVOP3Mods(N->getOperand(2), Ops[3], Ops[2]);
+  SelectVOP3Mods(N->getOperand(3), Ops[5], Ops[4]);
+  Ops[8] = N->getOperand(0);
+  Ops[9] = N->getOperand(4);
+
+  CurDAG->SelectNodeTo(N, AMDGPU::V_FMA_F32, N->getVTList(), Ops);
+}
+
+void AMDGPUDAGToDAGISel::SelectFMUL_W_CHAIN(SDNode *N) {
+  SDLoc SL(N);
+  //	src0_modifiers, src0,  src1_modifiers, src1, clamp, omod
+  SDValue Ops[8];
+
+  SelectVOP3Mods0(N->getOperand(1), Ops[1], Ops[0], Ops[4], Ops[5]);
+  SelectVOP3Mods(N->getOperand(2), Ops[3], Ops[2]);
+  Ops[6] = N->getOperand(0);
+  Ops[7] = N->getOperand(3);
+
+  CurDAG->SelectNodeTo(N, AMDGPU::V_MUL_F32_e64, N->getVTList(), Ops);
+}
+
 // We need to handle this here because tablegen doesn't support matching
 // instructions with multiple outputs.
 void AMDGPUDAGToDAGISel::SelectDIV_SCALE(SDNode *N) {
Index: lib/Target/AMDGPU/AMDGPUISelLowering.h
===================================================================
--- lib/Target/AMDGPU/AMDGPUISelLowering.h
+++ lib/Target/AMDGPU/AMDGPUISelLowering.h
@@ -230,6 +230,10 @@
   // This is SETCC with the full mask result which is used for a compare with a
   // result bit per item in the wavefront.
   SETCC,
+  SETREG,
+  // FP ops with input and output chain.
+  FMA_W_CHAIN,
+  FMUL_W_CHAIN,
 
   // SIN_HW, COS_HW - f32 for SI, 1 ULP max error, valid from -100 pi to 100 pi.
   // Denormals handled on some parts.
Index: lib/Target/AMDGPU/AMDGPUISelLowering.cpp
===================================================================
--- lib/Target/AMDGPU/AMDGPUISelLowering.cpp
+++ lib/Target/AMDGPU/AMDGPUISelLowering.cpp
@@ -2953,6 +2953,9 @@
   NODE_NAME_CASE(DWORDADDR)
   NODE_NAME_CASE(FRACT)
   NODE_NAME_CASE(SETCC)
+  NODE_NAME_CASE(SETREG)
+  NODE_NAME_CASE(FMA_W_CHAIN)
+  NODE_NAME_CASE(FMUL_W_CHAIN)
   NODE_NAME_CASE(CLAMP)
   NODE_NAME_CASE(COS_HW)
   NODE_NAME_CASE(SIN_HW)
Index: lib/Target/AMDGPU/AMDGPUInstrInfo.td
===================================================================
--- lib/Target/AMDGPU/AMDGPUInstrInfo.td
+++ lib/Target/AMDGPU/AMDGPUInstrInfo.td
@@ -150,6 +150,19 @@
 
 def AMDGPUsetcc : SDNode<"AMDGPUISD::SETCC", AMDGPUSetCCOp>;
 
+def AMDGPUSetRegOp :  SDTypeProfile<0, 2, [
+  SDTCisInt<0>, SDTCisInt<1>
+]>;
+
+def AMDGPUsetreg : SDNode<"AMDGPUISD::SETREG", AMDGPUSetRegOp, [
+  SDNPHasChain, SDNPSideEffect, SDNPOptInGlue, SDNPOutGlue]>;
+
+def AMDGPUfma : SDNode<"AMDGPUISD::FMA_W_CHAIN", SDTFPTernaryOp, [
+   SDNPHasChain, SDNPOptInGlue, SDNPOutGlue]>;
+
+def AMDGPUmul : SDNode<"AMDGPUISD::FMUL_W_CHAIN", SDTFPBinOp, [
+  SDNPHasChain, SDNPOptInGlue, SDNPOutGlue]>;
+
 def AMDGPUcvt_f32_ubyte0 : SDNode<"AMDGPUISD::CVT_F32_UBYTE0",
   SDTIntToFPOp, []>;
 def AMDGPUcvt_f32_ubyte1 : SDNode<"AMDGPUISD::CVT_F32_UBYTE1",
Index: lib/Target/AMDGPU/SIFoldOperands.cpp
===================================================================
--- lib/Target/AMDGPU/SIFoldOperands.cpp
+++ lib/Target/AMDGPU/SIFoldOperands.cpp
@@ -173,6 +173,16 @@
       MI->setDesc(TII->get(Opc));
     }
 
+    // Special case for s_setreg_b32
+    if (Opc == AMDGPU::S_SETREG_B32) {
+      MI->setDesc(TII->get(AMDGPU::S_SETREG_IMM32_B32));
+      bool FoldAsImm32 = tryAddToFoldList(FoldList, MI, OpNo, OpToFold, TII);
+      if (!FoldAsImm32) {
+        MI->setDesc(TII->get(Opc));
+      }
+      return FoldAsImm32;
+    }
+
     // If we are already folding into another operand of MI, then
     // we can't commute the instruction, otherwise we risk making the
     // other fold illegal.
Index: lib/Target/AMDGPU/SIISelLowering.cpp
===================================================================
--- lib/Target/AMDGPU/SIISelLowering.cpp
+++ lib/Target/AMDGPU/SIISelLowering.cpp
@@ -21,6 +21,7 @@
 #include "AMDGPU.h"
 #include "AMDGPUIntrinsicInfo.h"
 #include "AMDGPUSubtarget.h"
+#include "SIDefines.h"
 #include "SIISelLowering.h"
 #include "SIInstrInfo.h"
 #include "SIMachineFunctionInfo.h"
@@ -2858,6 +2859,47 @@
   return SDValue();
 }
 
+static SDValue getFPBinOp(SelectionDAG &DAG, unsigned Opcode, const SDLoc &SL,
+                          EVT VT, SDValue A, SDValue B, SDValue GlueChain) {
+  if (GlueChain->getNumValues() <= 1) {
+    return DAG.getNode(Opcode, SL, VT, A, B);
+  }
+
+  assert(GlueChain->getNumValues() == 3);
+
+  SDVTList VTList = DAG.getVTList(VT, MVT::Other, MVT::Glue);
+  switch (Opcode) {
+  default: llvm_unreachable("no chain equivalent for opcode");
+  case ISD::FMUL:
+    Opcode = AMDGPUISD::FMUL_W_CHAIN;
+    break;
+  }
+
+  return DAG.getNode(Opcode, SL, VTList, GlueChain.getValue(1), A, B,
+                     GlueChain.getValue(2));
+}
+
+static SDValue getFPTernOp(SelectionDAG &DAG, unsigned Opcode, const SDLoc &SL,
+                           EVT VT, SDValue A, SDValue B, SDValue C,
+                           SDValue GlueChain) {
+  if (GlueChain->getNumValues() <= 1) {
+    return DAG.getNode(Opcode, SL, VT, A, B, C);
+  }
+
+  assert(GlueChain->getNumValues() == 3);
+
+  SDVTList VTList = DAG.getVTList(VT, MVT::Other, MVT::Glue);
+  switch (Opcode) {
+  default: llvm_unreachable("no chain equivalent for opcode");
+  case ISD::FMA:
+    Opcode = AMDGPUISD::FMA_W_CHAIN;
+    break;
+  }
+
+  return DAG.getNode(Opcode, SL, VTList, GlueChain.getValue(1), A, B, C,
+                     GlueChain.getValue(2));
+}
+
 // Faster 2.5 ULP division that does not support denormals.
 SDValue SITargetLowering::lowerFDIV_FAST(SDValue Op, SelectionDAG &DAG) const {
   SDLoc SL(Op);
@@ -2904,25 +2946,73 @@
 
   SDVTList ScaleVT = DAG.getVTList(MVT::f32, MVT::i1);
 
-  SDValue DenominatorScaled = DAG.getNode(AMDGPUISD::DIV_SCALE, SL, ScaleVT, RHS, RHS, LHS);
-  SDValue NumeratorScaled = DAG.getNode(AMDGPUISD::DIV_SCALE, SL, ScaleVT, LHS, RHS, LHS);
+  SDValue DenominatorScaled = DAG.getNode(AMDGPUISD::DIV_SCALE, SL, ScaleVT,
+                                          RHS, RHS, LHS);
+  SDValue NumeratorScaled = DAG.getNode(AMDGPUISD::DIV_SCALE, SL, ScaleVT,
+                                        LHS, RHS, LHS);
 
   // Denominator is scaled to not be denormal, so using rcp is ok.
-  SDValue ApproxRcp = DAG.getNode(AMDGPUISD::RCP, SL, MVT::f32, DenominatorScaled);
+  SDValue ApproxRcp = DAG.getNode(AMDGPUISD::RCP, SL, MVT::f32,
+                                  DenominatorScaled);
+  SDValue NegDivScale0 = DAG.getNode(ISD::FNEG, SL, MVT::f32,
+                                     DenominatorScaled);
+
+  const unsigned Denorm32Reg = AMDGPU::Hwreg::ID_MODE |
+                               (4 << AMDGPU::Hwreg::OFFSET_SHIFT_) |
+                               (1 << AMDGPU::Hwreg::WIDTH_M1_SHIFT_);
+
+  const SDValue BitField = DAG.getTargetConstant(Denorm32Reg, SL, MVT::i16);
+
+  if (!Subtarget->hasFP32Denormals()) {
+    SDVTList BindParamVTs = DAG.getVTList(MVT::Other, MVT::Glue);
+    const SDValue EnableDenormValue = DAG.getConstant(FP_DENORM_FLUSH_NONE,
+                                                      SL, MVT::i32);
+    SDValue EnableDenorm = DAG.getNode(AMDGPUISD::SETREG, SL, BindParamVTs,
+                                       DAG.getEntryNode(),
+                                       EnableDenormValue, BitField);
+    SDValue Ops[3] = {
+      NegDivScale0,
+      EnableDenorm.getValue(0),
+      EnableDenorm.getValue(1)
+    };
+
+    NegDivScale0 = DAG.getMergeValues(Ops, SL);
+  }
+
+  SDValue Fma0 = getFPTernOp(DAG, ISD::FMA, SL, MVT::f32, NegDivScale0,
+                             ApproxRcp, One, NegDivScale0);
+
+  SDValue Fma1 = getFPTernOp(DAG, ISD::FMA, SL, MVT::f32, Fma0, ApproxRcp,
+                             ApproxRcp, Fma0);
 
-  SDValue NegDivScale0 = DAG.getNode(ISD::FNEG, SL, MVT::f32, DenominatorScaled);
+  SDValue Mul = getFPBinOp(DAG, ISD::FMUL, SL, MVT::f32, NumeratorScaled,
+                           Fma1, Fma1);
 
-  SDValue Fma0 = DAG.getNode(ISD::FMA, SL, MVT::f32, NegDivScale0, ApproxRcp, One);
-  SDValue Fma1 = DAG.getNode(ISD::FMA, SL, MVT::f32, Fma0, ApproxRcp, ApproxRcp);
+  SDValue Fma2 = getFPTernOp(DAG, ISD::FMA, SL, MVT::f32, NegDivScale0, Mul,
+                             NumeratorScaled, Mul);
 
-  SDValue Mul = DAG.getNode(ISD::FMUL, SL, MVT::f32, NumeratorScaled, Fma1);
+  SDValue Fma3 = getFPTernOp(DAG, ISD::FMA,SL, MVT::f32, Fma2, Fma1, Mul, Fma2);
 
-  SDValue Fma2 = DAG.getNode(ISD::FMA, SL, MVT::f32, NegDivScale0, Mul, NumeratorScaled);
-  SDValue Fma3 = DAG.getNode(ISD::FMA, SL, MVT::f32, Fma2, Fma1, Mul);
-  SDValue Fma4 = DAG.getNode(ISD::FMA, SL, MVT::f32, NegDivScale0, Fma3, NumeratorScaled);
+  SDValue Fma4 = getFPTernOp(DAG, ISD::FMA, SL, MVT::f32, NegDivScale0, Fma3,
+                             NumeratorScaled, Fma3);
+
+  if (!Subtarget->hasFP32Denormals()) {
+    const SDValue DisableDenormValue =
+        DAG.getConstant(FP_DENORM_FLUSH_IN_FLUSH_OUT, SL, MVT::i32);
+    SDValue DisableDenorm = DAG.getNode(AMDGPUISD::SETREG, SL, MVT::Other,
+                                        Fma4.getValue(1),
+                                        DisableDenormValue,
+                                        BitField,
+                                        Fma4.getValue(2));
+
+    SDValue OutputChain = DAG.getNode(ISD::TokenFactor, SL, MVT::Other,
+                                      DisableDenorm, DAG.getRoot());
+    DAG.setRoot(OutputChain);
+  }
 
   SDValue Scale = NumeratorScaled.getValue(1);
-  SDValue Fmas = DAG.getNode(AMDGPUISD::DIV_FMAS, SL, MVT::f32, Fma4, Fma1, Fma3, Scale);
+  SDValue Fmas = DAG.getNode(AMDGPUISD::DIV_FMAS, SL, MVT::f32,
+                             Fma4, Fma1, Fma3, Scale);
 
   return DAG.getNode(AMDGPUISD::DIV_FIXUP, SL, MVT::f32, Fmas, RHS, LHS);
 }
Index: lib/Target/AMDGPU/SIInstrInfo.cpp
===================================================================
--- lib/Target/AMDGPU/SIInstrInfo.cpp
+++ lib/Target/AMDGPU/SIInstrInfo.cpp
@@ -1669,6 +1669,8 @@
   // boundaries prevents incorrect movements of such instructions.
   return TargetInstrInfo::isSchedulingBoundary(MI, MBB, MF) ||
          MI.modifiesRegister(AMDGPU::EXEC, &RI) ||
+         MI.getOpcode() == AMDGPU::S_SETREG_IMM32_B32 ||
+         MI.getOpcode() == AMDGPU::S_SETREG_B32 ||
          changesVGPRIndexingMode(MI);
 }
 
Index: lib/Target/AMDGPU/SOPInstructions.td
===================================================================
--- lib/Target/AMDGPU/SOPInstructions.td
+++ lib/Target/AMDGPU/SOPInstructions.td
@@ -590,10 +590,13 @@
 >;
 }
 
+let hasSideEffects = 1 in {
+
 def S_SETREG_B32 : SOPK_Pseudo <
   "s_setreg_b32",
   (outs), (ins SReg_32:$sdst, hwreg:$simm16),
-  "$simm16, $sdst"
+  "$simm16, $sdst",
+  [(AMDGPUsetreg i32:$sdst, (i16 timm:$simm16))]
 >;
 
 // FIXME: Not on SI?
@@ -607,6 +610,7 @@
   let has_sdst = 0;
 }
 
+} // End hasSideEffects = 1
 
 //===----------------------------------------------------------------------===//
 // SOPC Instructions
Index: test/CodeGen/AMDGPU/fdiv.ll
===================================================================
--- test/CodeGen/AMDGPU/fdiv.ll
+++ test/CodeGen/AMDGPU/fdiv.ll
@@ -1,4 +1,6 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
+; RUN: llc -march=amdgcn -mcpu=tahiti -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=FUNC %s
+; RUN: llc -march=amdgcn -mcpu=hawaii -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=FUNC %s
+; RUN: llc -march=amdgcn -mcpu=fiji -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=FUNC %s
 ; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=R600 -check-prefix=FUNC %s
 
 ; These tests check that fdiv is expanded correctly and also test that the
@@ -11,18 +13,20 @@
 ; R600-DAG: RECIP_IEEE * T{{[0-9]+\.[XYZW]}}, KC0[2].W
 ; R600-DAG: MUL_IEEE {{\** *}}T{{[0-9]+\.[XYZW]}}, KC0[2].Z, PS
 
-; SI: v_div_scale_f32
-; SI-DAG: v_div_scale_f32
-
-; SI-DAG: v_rcp_f32
-; SI: v_fma_f32
-; SI: v_fma_f32
-; SI: v_mul_f32
-; SI: v_fma_f32
-; SI: v_fma_f32
-; SI: v_fma_f32
-; SI: v_div_fmas_f32
-; SI: v_div_fixup_f32
+; GCN: v_div_scale_f32 [[NUM_SCALE:v[0-9]+]]
+; GCN-DAG: v_div_scale_f32 [[DEN_SCALE:v[0-9]+]]
+; GCN-DAG: v_rcp_f32_e32 [[NUM_RCP:v[0-9]+]], [[NUM_SCALE]]
+
+; GCN: s_setreg_imm32_b32 hwreg(HW_REG_MODE, 4, 2), 3
+; GCN: v_fma_f32 [[A:v[0-9]+]], -[[NUM_SCALE]], [[NUM_RCP]], 1.0
+; GCN: v_fma_f32 [[B:v[0-9]+]], [[A]], [[NUM_RCP]], [[NUM_RCP]]
+; GCN: v_mul_f32_e32 [[C:v[0-9]+]], [[B]], [[DEN_SCALE]]
+; GCN: v_fma_f32 [[D:v[0-9]+]], -[[NUM_SCALE]], [[C]], [[DEN_SCALE]]
+; GCN: v_fma_f32 [[E:v[0-9]+]], [[D]], [[B]], [[C]]
+; GCN: v_fma_f32 [[F:v[0-9]+]], -[[NUM_SCALE]], [[E]], [[DEN_SCALE]]
+; GCN: s_setreg_imm32_b32 hwreg(HW_REG_MODE, 4, 2), 0
+; GCN: v_div_fmas_f32 [[FMAS:v[0-9]+]], [[F]], [[B]], [[E]]
+; GCN: v_div_fixup_f32 v{{[0-9]+}}, [[FMAS]],
 define void @fdiv_f32(float addrspace(1)* %out, float %a, float %b) #0 {
 entry:
   %fdiv = fdiv float %a, %b
@@ -30,12 +34,37 @@
   ret void
 }
 
+; FUNC-LABEL: {{^}}fdiv_f32_denormals:
+; R600-DAG: RECIP_IEEE * T{{[0-9]+\.[XYZW]}}, KC0[2].W
+; R600-DAG: MUL_IEEE {{\** *}}T{{[0-9]+\.[XYZW]}}, KC0[2].Z, PS
+
+; GCN: v_div_scale_f32 [[NUM_SCALE:v[0-9]+]]
+; GCN-DAG: v_div_scale_f32 [[DEN_SCALE:v[0-9]+]]
+; GCN-DAG: v_rcp_f32_e32 [[NUM_RCP:v[0-9]+]], [[NUM_SCALE]]
+
+; GCN-NOT: s_setreg
+; GCN: v_fma_f32 [[A:v[0-9]+]], -[[NUM_SCALE]], [[NUM_RCP]], 1.0
+; GCN: v_fma_f32 [[B:v[0-9]+]], [[A]], [[NUM_RCP]], [[NUM_RCP]]
+; GCN: v_mul_f32_e32 [[C:v[0-9]+]], [[B]], [[DEN_SCALE]]
+; GCN: v_fma_f32 [[D:v[0-9]+]], -[[NUM_SCALE]], [[C]], [[DEN_SCALE]]
+; GCN: v_fma_f32 [[E:v[0-9]+]], [[D]], [[B]], [[C]]
+; GCN: v_fma_f32 [[F:v[0-9]+]], -[[NUM_SCALE]], [[E]], [[DEN_SCALE]]
+; GCN-NOT: s_setreg
+; GCN: v_div_fmas_f32 [[FMAS:v[0-9]+]], [[F]], [[B]], [[E]]
+; GCN: v_div_fixup_f32 v{{[0-9]+}}, [[FMAS]],
+define void @fdiv_f32_denormals(float addrspace(1)* %out, float %a, float %b) #2 {
+entry:
+  %fdiv = fdiv float %a, %b
+  store float %fdiv, float addrspace(1)* %out
+  ret void
+}
+
 ; FUNC-LABEL: {{^}}fdiv_25ulp_f32:
-; SI: v_cndmask_b32
-; SI: v_mul_f32
-; SI: v_rcp_f32
-; SI: v_mul_f32
-; SI: v_mul_f32
+; GCN: v_cndmask_b32
+; GCN: v_mul_f32
+; GCN: v_rcp_f32
+; GCN: v_mul_f32
+; GCN: v_mul_f32
 define void @fdiv_25ulp_f32(float addrspace(1)* %out, float %a, float %b) #0 {
 entry:
   %fdiv = fdiv float %a, %b, !fpmath !0
@@ -45,9 +74,9 @@
 
 ; Use correct fdiv
 ; FUNC-LABEL: {{^}}fdiv_25ulp_denormals_f32:
-; SI: v_fma_f32
-; SI: v_div_fmas_f32
-; SI: v_div_fixup_f32
+; GCN: v_fma_f32
+; GCN: v_div_fmas_f32
+; GCN: v_div_fixup_f32
 define void @fdiv_25ulp_denormals_f32(float addrspace(1)* %out, float %a, float %b) #2 {
 entry:
   %fdiv = fdiv float %a, %b, !fpmath !0
@@ -56,10 +85,10 @@
 }
 
 ; FUNC-LABEL: {{^}}fdiv_fast_denormals_f32:
-; SI: v_rcp_f32_e32 [[RCP:v[0-9]+]], s{{[0-9]+}}
-; SI: v_mul_f32_e32 [[RESULT:v[0-9]+]], s{{[0-9]+}}, [[RCP]]
-; SI-NOT: [[RESULT]]
-; SI: buffer_store_dword [[RESULT]]
+; GCN: v_rcp_f32_e32 [[RCP:v[0-9]+]], s{{[0-9]+}}
+; GCN: v_mul_f32_e32 [[RESULT:v[0-9]+]], s{{[0-9]+}}, [[RCP]]
+; GCN-NOT: [[RESULT]]
+; GCN: buffer_store_dword [[RESULT]]
 define void @fdiv_fast_denormals_f32(float addrspace(1)* %out, float %a, float %b) #2 {
 entry:
   %fdiv = fdiv fast float %a, %b
@@ -71,10 +100,10 @@
 ; R600-DAG: RECIP_IEEE * T{{[0-9]+\.[XYZW]}}, KC0[2].W
 ; R600-DAG: MUL_IEEE {{\** *}}T{{[0-9]+\.[XYZW]}}, KC0[2].Z, PS
 
-; SI: v_rcp_f32_e32 [[RCP:v[0-9]+]], s{{[0-9]+}}
-; SI: v_mul_f32_e32 [[RESULT:v[0-9]+]], s{{[0-9]+}}, [[RCP]]
-; SI-NOT: [[RESULT]]
-; SI: buffer_store_dword [[RESULT]]
+; GCN: v_rcp_f32_e32 [[RCP:v[0-9]+]], s{{[0-9]+}}
+; GCN: v_mul_f32_e32 [[RESULT:v[0-9]+]], s{{[0-9]+}}, [[RCP]]
+; GCN-NOT: [[RESULT]]
+; GCN: buffer_store_dword [[RESULT]]
 define void @fdiv_f32_fast_math(float addrspace(1)* %out, float %a, float %b) #0 {
 entry:
   %fdiv = fdiv fast float %a, %b
@@ -86,10 +115,10 @@
 ; R600-DAG: RECIP_IEEE * T{{[0-9]+\.[XYZW]}}, KC0[2].W
 ; R600-DAG: MUL_IEEE {{\** *}}T{{[0-9]+\.[XYZW]}}, KC0[2].Z, PS
 
-; SI: v_rcp_f32_e32 [[RCP:v[0-9]+]], s{{[0-9]+}}
-; SI: v_mul_f32_e32 [[RESULT:v[0-9]+]], s{{[0-9]+}}, [[RCP]]
-; SI-NOT: [[RESULT]]
-; SI: buffer_store_dword [[RESULT]]
+; GCN: v_rcp_f32_e32 [[RCP:v[0-9]+]], s{{[0-9]+}}
+; GCN: v_mul_f32_e32 [[RESULT:v[0-9]+]], s{{[0-9]+}}, [[RCP]]
+; GCN-NOT: [[RESULT]]
+; GCN: buffer_store_dword [[RESULT]]
 define void @fdiv_f32_arcp_math(float addrspace(1)* %out, float %a, float %b) #0 {
 entry:
   %fdiv = fdiv arcp float %a, %b
@@ -103,10 +132,10 @@
 ; R600-DAG: MUL_IEEE {{\** *}}T{{[0-9]+\.[XYZW]}}, KC0[3].X, PS
 ; R600-DAG: MUL_IEEE {{\** *}}T{{[0-9]+\.[XYZW]}}, KC0[2].W, PS
 
-; SI: v_div_scale_f32
-; SI: v_div_scale_f32
-; SI: v_div_scale_f32
-; SI: v_div_scale_f32
+; GCN: v_div_scale_f32
+; GCN: v_div_scale_f32
+; GCN: v_div_scale_f32
+; GCN: v_div_scale_f32
 define void @fdiv_v2f32(<2 x float> addrspace(1)* %out, <2 x float> %a, <2 x float> %b) #0 {
 entry:
   %fdiv = fdiv <2 x float> %a, %b
@@ -115,8 +144,8 @@
 }
 
 ; FUNC-LABEL: {{^}}fdiv_ulp25_v2f32:
-; SI: v_cmp_gt_f32
-; SI: v_cmp_gt_f32
+; GCN: v_cmp_gt_f32
+; GCN: v_cmp_gt_f32
 define void @fdiv_ulp25_v2f32(<2 x float> addrspace(1)* %out, <2 x float> %a, <2 x float> %b) #0 {
 entry:
   %fdiv = fdiv arcp <2 x float> %a, %b, !fpmath !0
@@ -130,8 +159,8 @@
 ; R600-DAG: MUL_IEEE {{\** *}}T{{[0-9]+\.[XYZW]}}, KC0[3].X, PS
 ; R600-DAG: MUL_IEEE {{\** *}}T{{[0-9]+\.[XYZW]}}, KC0[2].W, PS
 
-; SI: v_rcp_f32
-; SI: v_rcp_f32
+; GCN: v_rcp_f32
+; GCN: v_rcp_f32
 define void @fdiv_v2f32_fast_math(<2 x float> addrspace(1)* %out, <2 x float> %a, <2 x float> %b) #0 {
 entry:
   %fdiv = fdiv fast <2 x float> %a, %b
@@ -145,8 +174,8 @@
 ; R600-DAG: MUL_IEEE {{\** *}}T{{[0-9]+\.[XYZW]}}, KC0[3].X, PS
 ; R600-DAG: MUL_IEEE {{\** *}}T{{[0-9]+\.[XYZW]}}, KC0[2].W, PS
 
-; SI: v_rcp_f32
-; SI: v_rcp_f32
+; GCN: v_rcp_f32
+; GCN: v_rcp_f32
 define void @fdiv_v2f32_arcp_math(<2 x float> addrspace(1)* %out, <2 x float> %a, <2 x float> %b) #0 {
 entry:
   %fdiv = fdiv arcp <2 x float> %a, %b
@@ -164,10 +193,10 @@
 ; R600-DAG: MUL_IEEE {{\** *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}, PS
 ; R600-DAG: MUL_IEEE {{\** *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}, PS
 
-; SI: v_div_fixup_f32
-; SI: v_div_fixup_f32
-; SI: v_div_fixup_f32
-; SI: v_div_fixup_f32
+; GCN: v_div_fixup_f32
+; GCN: v_div_fixup_f32
+; GCN: v_div_fixup_f32
+; GCN: v_div_fixup_f32
 define void @fdiv_v4f32(<4 x float> addrspace(1)* %out, <4 x float> addrspace(1)* %in) #0 {
   %b_ptr = getelementptr <4 x float>, <4 x float> addrspace(1)* %in, i32 1
   %a = load <4 x float>, <4 x float> addrspace(1) * %in
@@ -187,10 +216,10 @@
 ; R600-DAG: MUL_IEEE {{\** *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}, PS
 ; R600-DAG: MUL_IEEE {{\** *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}, PS
 
-; SI: v_rcp_f32
-; SI: v_rcp_f32
-; SI: v_rcp_f32
-; SI: v_rcp_f32
+; GCN: v_rcp_f32
+; GCN: v_rcp_f32
+; GCN: v_rcp_f32
+; GCN: v_rcp_f32
 define void @fdiv_v4f32_fast_math(<4 x float> addrspace(1)* %out, <4 x float> addrspace(1)* %in) #0 {
   %b_ptr = getelementptr <4 x float>, <4 x float> addrspace(1)* %in, i32 1
   %a = load <4 x float>, <4 x float> addrspace(1) * %in
@@ -210,10 +239,10 @@
 ; R600-DAG: MUL_IEEE {{\** *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}, PS
 ; R600-DAG: MUL_IEEE {{\** *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}, PS
 
-; SI: v_rcp_f32
-; SI: v_rcp_f32
-; SI: v_rcp_f32
-; SI: v_rcp_f32
+; GCN: v_rcp_f32
+; GCN: v_rcp_f32
+; GCN: v_rcp_f32
+; GCN: v_rcp_f32
 define void @fdiv_v4f32_arcp_math(<4 x float> addrspace(1)* %out, <4 x float> addrspace(1)* %in) #0 {
   %b_ptr = getelementptr <4 x float>, <4 x float> addrspace(1)* %in, i32 1
   %a = load <4 x float>, <4 x float> addrspace(1) * %in