Fix the order of the operands in the llvm.fma intrinsic patterns for ARM,

<rdar://problem/11325085>.

llvm-svn: 155724

3 files changed, 28 insertions, 27 deletions

diff --git a/llvm/lib/Target/ARM/ARMInstrNEON.td b/llvm/lib/Target/ARM/ARMInstrNEON.td
index 31dd843fe89..6132989b636 100644
--- a/llvm/lib/Target/ARM/ARMInstrNEON.td
+++ b/llvm/lib/Target/ARM/ARMInstrNEON.td
@@ -4136,16 +4136,16 @@ def  VFMSfq   : N3VQMulOp<0, 0, 0b10, 0b1100, 1, IIC_VFMACQ, "vfms", "f32",
                 Requires<[HasVFP4,UseFusedMAC]>;
 
 // Match @llvm.fma.* intrinsics
-def : Pat<(v2f32 (fma DPR:$src1, DPR:$Vn, DPR:$Vm)),
+def : Pat<(v2f32 (fma DPR:$Vn, DPR:$Vm, DPR:$src1)),
           (VFMAfd DPR:$src1, DPR:$Vn, DPR:$Vm)>,
           Requires<[HasVFP4]>;
-def : Pat<(v4f32 (fma QPR:$src1, QPR:$Vn, QPR:$Vm)),
+def : Pat<(v4f32 (fma QPR:$Vn, QPR:$Vm, QPR:$src1)),
           (VFMAfq QPR:$src1, QPR:$Vn, QPR:$Vm)>,
           Requires<[HasVFP4]>;
-def : Pat<(v2f32 (fma (fneg DPR:$src1), DPR:$Vn, DPR:$Vm)),
+def : Pat<(v2f32 (fma (fneg DPR:$Vn), DPR:$Vm, DPR:$src1)),
           (VFMSfd DPR:$src1, DPR:$Vn, DPR:$Vm)>,
       Requires<[HasVFP4]>;
-def : Pat<(v4f32 (fma (fneg QPR:$src1), QPR:$Vn, QPR:$Vm)),
+def : Pat<(v4f32 (fma (fneg QPR:$Vn), QPR:$Vm, QPR:$src1)),
           (VFMSfq QPR:$src1, QPR:$Vn, QPR:$Vm)>,
       Requires<[HasVFP4]>;
 
diff --git a/llvm/lib/Target/ARM/ARMInstrVFP.td b/llvm/lib/Target/ARM/ARMInstrVFP.td
index f447e304b49..dccbffa4c9f 100644
--- a/llvm/lib/Target/ARM/ARMInstrVFP.td
+++ b/llvm/lib/Target/ARM/ARMInstrVFP.td
@@ -1089,10 +1089,11 @@ def : Pat<(fadd_mlx SPR:$dstin, (fmul_su SPR:$a, SPR:$b)),
           Requires<[HasVFP4,DontUseNEONForFP,UseFusedMAC]>;
 
 // Match @llvm.fma.* intrinsics
-def : Pat<(f64 (fma DPR:$Ddin, DPR:$Dn, DPR:$Dm)),
+// (fma x, y, z) -> (vfms z, x, y)
+def : Pat<(f64 (fma DPR:$Dn, DPR:$Dm, DPR:$Ddin)),
           (VFMAD DPR:$Ddin, DPR:$Dn, DPR:$Dm)>,
       Requires<[HasVFP4]>;
-def : Pat<(f32 (fma SPR:$Sdin, SPR:$Sn, SPR:$Sm)),
+def : Pat<(f32 (fma SPR:$Sn, SPR:$Sm, SPR:$Sdin)),
           (VFMAS SPR:$Sdin, SPR:$Sn, SPR:$Sm)>,
       Requires<[HasVFP4]>;
 
@@ -1123,18 +1124,18 @@ def : Pat<(fsub_mlx SPR:$dstin, (fmul_su SPR:$a, SPR:$b)),
           Requires<[HasVFP4,DontUseNEONForFP,UseFusedMAC]>;
 
 // Match @llvm.fma.* intrinsics
-// (fma (fneg x), y, z) -> (vfms x, y, z)
-def : Pat<(f64 (fma (fneg DPR:$Ddin), DPR:$Dn, DPR:$Dm)),
+// (fma (fneg x), y, z) -> (vfms z, x, y)
+def : Pat<(f64 (fma (fneg DPR:$Dn), DPR:$Dm, DPR:$Ddin)),
           (VFMSD DPR:$Ddin, DPR:$Dn, DPR:$Dm)>,
       Requires<[HasVFP4]>;
-def : Pat<(f32 (fma (fneg SPR:$Sdin), SPR:$Sn, SPR:$Sm)),
+def : Pat<(f32 (fma (fneg SPR:$Sn), SPR:$Sm, SPR:$Sdin)),
           (VFMSS SPR:$Sdin, SPR:$Sn, SPR:$Sm)>,
       Requires<[HasVFP4]>;
-// (fneg (fma x, (fneg y), z) -> (vfms x, y, z)
-def : Pat<(fneg (f64 (fma DPR:$Ddin, (fneg DPR:$Dn), DPR:$Dm))),
+// (fma x, (fneg y), z) -> (vfms z, x, y)
+def : Pat<(f64 (fma DPR:$Dn, (fneg DPR:$Dm), DPR:$Ddin)),
           (VFMSD DPR:$Ddin, DPR:$Dn, DPR:$Dm)>,
       Requires<[HasVFP4]>;
-def : Pat<(fneg (f32 (fma SPR:$Sdin, (fneg SPR:$Sn), SPR:$Sm))),
+def : Pat<(f32 (fma SPR:$Sn, (fneg SPR:$Sm), SPR:$Sdin)),
           (VFMSS SPR:$Sdin, SPR:$Sn, SPR:$Sm)>,
       Requires<[HasVFP4]>;
 
@@ -1165,18 +1166,18 @@ def : Pat<(fsub_mlx (fneg (fmul_su SPR:$a, SPR:$b)), SPR:$dstin),
           Requires<[HasVFP4,DontUseNEONForFP,UseFusedMAC]>;
 
 // Match @llvm.fma.* intrinsics
-// (fneg (fma x, y, z)) -> (vfnma x, y, z)
-def : Pat<(fneg (fma (f64 DPR:$Ddin), (f64 DPR:$Dn), (f64 DPR:$Dm))),
+// (fneg (fma x, y, z)) -> (vfnma z, x, y)
+def : Pat<(fneg (fma (f64 DPR:$Dn), (f64 DPR:$Dm), (f64 DPR:$Ddin))),
           (VFNMAD DPR:$Ddin, DPR:$Dn, DPR:$Dm)>,
       Requires<[HasVFP4]>;
-def : Pat<(fneg (fma (f32 SPR:$Sdin), (f32 SPR:$Sn), (f32 SPR:$Sm))),
+def : Pat<(fneg (fma (f32 SPR:$Sn), (f32 SPR:$Sm), (f32 SPR:$Sdin))),
           (VFNMAS SPR:$Sdin, SPR:$Sn, SPR:$Sm)>,
       Requires<[HasVFP4]>;
-// (fma (fneg x), y, (fneg z)) -> (vfnma x, y, z)
-def : Pat<(f64 (fma (fneg DPR:$Ddin), DPR:$Dn, (fneg DPR:$Dm))),
+// (fma (fneg x), y, (fneg z)) -> (vfnma z, x, y)
+def : Pat<(f64 (fma (fneg DPR:$Dn), DPR:$Dm, (fneg DPR:$Ddin))),
           (VFNMAD DPR:$Ddin, DPR:$Dn, DPR:$Dm)>,
       Requires<[HasVFP4]>;
-def : Pat<(f32 (fma (fneg SPR:$Sdin), SPR:$Sn, (fneg SPR:$Sm))),
+def : Pat<(f32 (fma (fneg SPR:$Sn), SPR:$Sm, (fneg SPR:$Sdin))),
           (VFNMAS SPR:$Sdin, SPR:$Sn, SPR:$Sm)>,
       Requires<[HasVFP4]>;
 
@@ -1206,18 +1207,18 @@ def : Pat<(fsub_mlx (fmul_su SPR:$a, SPR:$b), SPR:$dstin),
           Requires<[HasVFP4,DontUseNEONForFP,UseFusedMAC]>;
 
 // Match @llvm.fma.* intrinsics
-// (fneg (fma (fneg x), y, z)) -> (vnfms x, y, z)
-def : Pat<(fneg (f64 (fma (fneg DPR:$Ddin), DPR:$Dn, DPR:$Dm))),
+// (fneg (fma (fneg x), y, z)) -> (vfnms z, x, y)
+def : Pat<(fneg (f64 (fma (fneg DPR:$Dn), DPR:$Dm, DPR:$Ddin))),
           (VFNMSD DPR:$Ddin, DPR:$Dn, DPR:$Dm)>,
       Requires<[HasVFP4]>;
-def : Pat<(fneg (f32 (fma (fneg SPR:$Sdin), SPR:$Sn, SPR:$Sm))),
+def : Pat<(fneg (f32 (fma (fneg SPR:$Sn), SPR:$Sm, SPR:$Sdin))),
           (VFNMSS SPR:$Sdin, SPR:$Sn, SPR:$Sm)>,
       Requires<[HasVFP4]>;
-// (fma x, (fneg y), z) -> (vnfms x, y, z)
-def : Pat<(f64 (fma DPR:$Ddin, (fneg DPR:$Dn), DPR:$Dm)),
+// (fneg (fma x, (fneg y), z) -> (vfnms z, x, y)
+def : Pat<(fneg (f64 (fma DPR:$Dn, (fneg DPR:$Dm), DPR:$Ddin))),
           (VFNMSD DPR:$Ddin, DPR:$Dn, DPR:$Dm)>,
       Requires<[HasVFP4]>;
-def : Pat<(f32 (fma SPR:$Sdin, (fneg SPR:$Sn), SPR:$Sm)),
+def : Pat<(fneg (f32 (fma SPR:$Sn, (fneg SPR:$Sm), SPR:$Sdin))),
           (VFNMSS SPR:$Sdin, SPR:$Sn, SPR:$Sm)>,
       Requires<[HasVFP4]>;
 
diff --git a/llvm/test/CodeGen/ARM/fusedMAC.ll b/llvm/test/CodeGen/ARM/fusedMAC.ll
index 802d1b8b393..1ad7ce18ea1 100644
--- a/llvm/test/CodeGen/ARM/fusedMAC.ll
+++ b/llvm/test/CodeGen/ARM/fusedMAC.ll
@@ -138,8 +138,7 @@ entry:
 ; CHECK: vfms.f64
   %tmp1 = fsub double -0.0, %b
   %tmp2 = tail call double @llvm.fma.f64(double %a, double %tmp1, double %c) nounwind readnone
-  %tmp3 = fsub double -0.0, %tmp2
-  ret double %tmp3
+  ret double %tmp2
 }
 
 define double @test_fnms_f64(double %a, double %b, double %c) nounwind readnone ssp {
@@ -158,7 +157,8 @@ entry:
 ; CHECK: vfnms.f64
   %tmp1 = fsub double -0.0, %b
   %tmp2 = tail call double @llvm.fma.f64(double %a, double %tmp1, double %c) nounwind readnone
-  ret double %tmp2
+  %tmp3 = fsub double -0.0, %tmp2
+  ret double %tmp3
 }
 
 define double @test_fnma_f64(double %a, double %b, double %c) nounwind readnone ssp {