diff --git a/lib/Target/ARM/ARMInstrVFP.td b/lib/Target/ARM/ARMInstrVFP.td
index dccbffa..4e2cda4 100644
--- a/lib/Target/ARM/ARMInstrVFP.td
+++ b/lib/Target/ARM/ARMInstrVFP.td
@@ -1207,6 +1207,14 @@
           Requires<[HasVFP4,DontUseNEONForFP,UseFusedMAC]>;
 
 // Match @llvm.fma.* intrinsics
+
+// (fma x, y, (fneg z)) -> (vfnms z, x, y))
+def : Pat<(f64 (fma DPR:$Dn, DPR:$Dm, (fneg DPR:$Ddin))),
+          (VFNMSD DPR:$Ddin, DPR:$Dn, DPR:$Dm)>,
+      Requires<[HasVFP4]>;
+def : Pat<(f32 (fma SPR:$Sn, SPR:$Sm, (fneg SPR:$Sdin))),
+          (VFNMSS SPR:$Sdin, SPR:$Sn, SPR:$Sm)>,
+      Requires<[HasVFP4]>;
 // (fneg (fma (fneg x), y, z)) -> (vfnms z, x, y)
 def : Pat<(fneg (f64 (fma (fneg DPR:$Dn), DPR:$Dm, DPR:$Ddin))),
           (VFNMSD DPR:$Ddin, DPR:$Dn, DPR:$Dm)>,
diff --git a/test/CodeGen/ARM/fusedMAC.ll b/test/CodeGen/ARM/fusedMAC.ll
index 61e7d7b..0cc1cdd 100644
--- a/test/CodeGen/ARM/fusedMAC.ll
+++ b/test/CodeGen/ARM/fusedMAC.ll
@@ -141,6 +141,15 @@
   ret double %tmp2
 }
 
+define float @test_fnms_f32(float %a, float %b, float* %c) nounwind readnone ssp {
+; CHECK: test_fnms_f32
+; CHECK: vfnms.f32
+  %tmp1 = load float* %c, align 4
+  %tmp2 = fsub float -0.0, %tmp1
+  %tmp3 = tail call float @llvm.fma.f32(float %a, float %b, float %tmp2) nounwind readnone
+  ret float %tmp3 
+}
+
 define double @test_fnms_f64(double %a, double %b, double %c) nounwind readnone ssp {
 entry:
 ; CHECK: test_fnms_f64