Index: lib/CodeGen/SelectionDAG/DAGCombiner.cpp
===================================================================
--- lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+++ lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -6860,13 +6860,25 @@
   if (N1CFP && N1CFP->isExactlyValue(1.0))
     return N0;
 
-  // If allowed, fold (fmul (fmul x, c1), c2) -> (fmul x, (fmul c1, c2))
-  if (DAG.getTarget().Options.UnsafeFPMath &&
-      N1CFP && N0.getOpcode() == ISD::FMUL &&
-      N0.getNode()->hasOneUse() && isConstOrConstSplatFP(N0.getOperand(1))) {
-    return DAG.getNode(ISD::FMUL, SDLoc(N), VT, N0.getOperand(0),
-                       DAG.getNode(ISD::FMUL, SDLoc(N), VT,
-                                   N0.getOperand(1), N1));
+  if (DAG.getTarget().Options.UnsafeFPMath) {
+    // If allowed, fold (fmul (fmul x, c1), c2) -> (fmul x, (fmul c1, c2))
+    if (N1CFP && N0.getOpcode() == ISD::FMUL &&
+        N0.getNode()->hasOneUse() && isConstOrConstSplatFP(N0.getOperand(1))) {
+      SDLoc SL(N);
+      SDValue MulConsts = DAG.getNode(ISD::FMUL, SL, VT, N0.getOperand(1), N1);
+      return DAG.getNode(ISD::FMUL, SL, VT, N0.getOperand(0), MulConsts);
+    }
+
+    // If allowed, fold (fmul (fadd x, x), c) -> (fmul x, (fmul 2.0, c))
+    // Undo the fmul 2.0, x -> fadd x, x transformation, since if it occurs
+    // during an early run of DAGCombiner can prevent folding with fmuls
+    // inserted during lowering.
+    if (N0.getOpcode() == ISD::FADD && N0.getOperand(0) == N0.getOperand(1)) {
+      SDLoc SL(N);
+      const SDValue Two = DAG.getConstantFP(2.0, VT);
+      SDValue MulConsts = DAG.getNode(ISD::FMUL, SL, VT, Two, N1);
+      return DAG.getNode(ISD::FMUL, SDLoc(N), VT, N0.getOperand(0), MulConsts);
+    }
   }
 
   // fold (fmul X, 2.0) -> (fadd X, X)
Index: test/CodeGen/R600/llvm.sin.ll
===================================================================
--- test/CodeGen/R600/llvm.sin.ll
+++ test/CodeGen/R600/llvm.sin.ll
@@ -1,53 +1,84 @@
-;RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
-;RUN: llc -march=r600 -mcpu=SI < %s | FileCheck -check-prefix=SI -check-prefix=SI-SAFE -check-prefix=FUNC %s
-;RUN: llc -march=r600 -mcpu=SI -enable-unsafe-fp-math < %s | FileCheck -check-prefix=SI -check-prefix=SI-UNSAFE -check-prefix=FUNC %s
-
-;FUNC-LABEL: test
-;EG: MULADD_IEEE *
-;EG: FRACT *
-;EG: ADD *
-;EG: SIN * T{{[0-9]+\.[XYZW], PV\.[XYZW]}}
-;EG-NOT: SIN
-;SI: V_MUL_F32
-;SI: V_FRACT_F32
-;SI: V_SIN_F32
-;SI-NOT: V_SIN_F32
-
-define void @test(float addrspace(1)* %out, float %x) #1 {
+; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
+; RUN: llc -march=r600 -mcpu=SI < %s | FileCheck -check-prefix=SI -check-prefix=SI-SAFE -check-prefix=FUNC %s
+; RUN: llc -march=r600 -mcpu=SI -enable-unsafe-fp-math < %s | FileCheck -check-prefix=SI -check-prefix=SI-UNSAFE -check-prefix=FUNC %s
+
+; FUNC-LABEL: sin_f32
+; EG: MULADD_IEEE *
+; EG: FRACT *
+; EG: ADD *
+; EG: SIN * T{{[0-9]+\.[XYZW], PV\.[XYZW]}}
+; EG-NOT: SIN
+; SI: V_MUL_F32
+; SI: V_FRACT_F32
+; SI: V_SIN_F32
+; SI-NOT: V_SIN_F32
+
+define void @sin_f32(float addrspace(1)* %out, float %x) #1 {
    %sin = call float @llvm.sin.f32(float %x)
    store float %sin, float addrspace(1)* %out
    ret void
 }
 
-;FUNC-LABEL: testf
-;SI-UNSAFE: 4.774
-;SI-UNSAFE: V_MUL_F32
-;SI-SAFE: V_MUL_F32
-;SI-SAFE: V_MUL_F32
-;SI: V_FRACT_F32
-;SI: V_SIN_F32
-;SI-NOT: V_SIN_F32
-
-define void @testf(float addrspace(1)* %out, float %x) #1 {
-   %y = fmul float 3.0, %x
+; FUNC-LABEL: @sin_3x_f32
+; SI-UNSAFE-NOT: V_ADD_F32
+; SI-UNSAFE: 4.774648e-01
+; SI-UNSAFE: V_MUL_F32
+; SI-SAFE: V_MUL_F32
+; SI-SAFE: V_MUL_F32
+; SI: V_FRACT_F32
+; SI: V_SIN_F32
+; SI-NOT: V_SIN_F32
+define void @sin_3x_f32(float addrspace(1)* %out, float %x) #1 {
+  %y = fmul float 3.0, %x
+  %sin = call float @llvm.sin.f32(float %y)
+  store float %sin, float addrspace(1)* %out
+  ret void
+}
+
+; FUNC-LABEL: @sin_2x_f32
+; SI-UNSAFE-NOT: V_ADD_F32
+; SI-UNSAFE: 3.183099e-01
+; SI-UNSAFE: V_MUL_F32
+; SI-SAFE: V_ADD_F32
+; SI-SAFE: V_MUL_F32
+; SI: V_FRACT_F32
+; SI: V_SIN_F32
+; SI-NOT: V_SIN_F32
+define void @sin_2x_f32(float addrspace(1)* %out, float %x) #1 {
+  %y = fmul float 2.0, %x
+  %sin = call float @llvm.sin.f32(float %y)
+  store float %sin, float addrspace(1)* %out
+  ret void
+}
+
+; FUNC-LABEL: @test_2sin_f32
+; SI-UNSAFE: 3.183099e-01
+; SI-UNSAFE: V_MUL_F32
+; SI-SAFE: V_ADD_F32
+; SI-SAFE: V_MUL_F32
+; SI: V_FRACT_F32
+; SI: V_SIN_F32
+; SI-NOT: V_SIN_F32
+define void @test_2sin_f32(float addrspace(1)* %out, float %x) #1 {
+   %y = fmul float 2.0, %x
    %sin = call float @llvm.sin.f32(float %y)
    store float %sin, float addrspace(1)* %out
    ret void
 }
 
-;FUNC-LABEL: testv
-;EG: SIN * T{{[0-9]+\.[XYZW], PV\.[XYZW]}}
-;EG: SIN * T{{[0-9]+\.[XYZW], PV\.[XYZW]}}
-;EG: SIN * T{{[0-9]+\.[XYZW], PV\.[XYZW]}}
-;EG: SIN * T{{[0-9]+\.[XYZW], PV\.[XYZW]}}
-;EG-NOT: SIN
-;SI: V_SIN_F32
-;SI: V_SIN_F32
-;SI: V_SIN_F32
-;SI: V_SIN_F32
-;SI-NOT: V_SIN_F32
-
-define void @testv(<4 x float> addrspace(1)* %out, <4 x float> %vx) #1 {
+; FUNC-LABEL: @sin_v4f32
+; EG: SIN * T{{[0-9]+\.[XYZW], PV\.[XYZW]}}
+; EG: SIN * T{{[0-9]+\.[XYZW], PV\.[XYZW]}}
+; EG: SIN * T{{[0-9]+\.[XYZW], PV\.[XYZW]}}
+; EG: SIN * T{{[0-9]+\.[XYZW], PV\.[XYZW]}}
+; EG-NOT: SIN
+; SI: V_SIN_F32
+; SI: V_SIN_F32
+; SI: V_SIN_F32
+; SI: V_SIN_F32
+; SI-NOT: V_SIN_F32
+
+define void @sin_v4f32(<4 x float> addrspace(1)* %out, <4 x float> %vx) #1 {
    %sin = call <4 x float> @llvm.sin.v4f32( <4 x float> %vx)
    store <4 x float> %sin, <4 x float> addrspace(1)* %out
    ret void