diff options
| -rw-r--r-- | lib/Target/ARM/ARMInstrVFP.td | 8 | ||||
| -rw-r--r-- | test/CodeGen/ARM/fusedMAC.ll | 31 |
2 files changed, 33 insertions, 6 deletions
diff --git a/lib/Target/ARM/ARMInstrVFP.td b/lib/Target/ARM/ARMInstrVFP.td index a9ad4fa4ec..1b0d58784d 100644 --- a/lib/Target/ARM/ARMInstrVFP.td +++ b/lib/Target/ARM/ARMInstrVFP.td @@ -1140,6 +1140,14 @@ def : Pat<(fsub_mlx (fneg (fmul_su SPR:$a, SPR:$b)), SPR:$dstin), (VFNMAS SPR:$dstin, SPR:$a, SPR:$b)>, Requires<[HasVFP4,DontUseNEONForFP,FPContractions]>; +// Match @llvm.fma.* intrinsics +def : Pat<(fneg (fma (f64 DPR:$Ddin), (f64 DPR:$Dn), (f64 DPR:$Dm))), + (VFNMAD DPR:$Ddin, DPR:$Dn, DPR:$Dm)>, + Requires<[HasVFP4]>; +def : Pat<(fneg (fma (f32 SPR:$Sdin), (f32 SPR:$Sn), (f32 SPR:$Sm))), + (VFNMAS SPR:$Sdin, SPR:$Sn, SPR:$Sm)>, + Requires<[HasVFP4]>; + def VFNMSD : ADbI<0b11101, 0b01, 0, 0, (outs DPR:$Dd), (ins DPR:$Ddin, DPR:$Dn, DPR:$Dm), IIC_fpFMAC64, "vfnms", ".f64\t$Dd, $Dn, $Dm", diff --git a/test/CodeGen/ARM/fusedMAC.ll b/test/CodeGen/ARM/fusedMAC.ll index 381eb437e0..b5398c7a58 100644 --- a/test/CodeGen/ARM/fusedMAC.ll +++ b/test/CodeGen/ARM/fusedMAC.ll @@ -99,30 +99,49 @@ define <4 x float> @fusedMACTest12(<4 x float> %a, <4 x float> %b) { ret <4 x float> %sub } -define float @test_f32(float %a, float %b, float %c) nounwind readnone ssp { +define float @test_fma_f32(float %a, float %b, float %c) nounwind readnone ssp { entry: -; CHECK: test_f32 +; CHECK: test_fma_f32 ; CHECK: vfma.f32 %call = tail call float @llvm.fma.f32(float %a, float %b, float %c) nounwind readnone ret float %call } -define double @test_f64(double %a, double %b, double %c) nounwind readnone ssp { +define double @test_fma_f64(double %a, double %b, double %c) nounwind readnone ssp { entry: -; CHECK: test_f64 +; CHECK: test_fma_f64 ; CHECK: vfma.f64 %call = tail call double @llvm.fma.f64(double %a, double %b, double %c) nounwind readnone ret double %call } -define <2 x float> @test_v2f32(<2 x float> %a, <2 x float> %b, <2 x float> %c) nounwind readnone ssp { +define <2 x float> @test_fma_v2f32(<2 x float> %a, <2 x float> %b, <2 x float> %c) nounwind readnone ssp { entry: -; CHECK: test_v2f32 +; CHECK: test_fma_v2f32 ; CHECK: vfma.f32 %0 = tail call <2 x float> @llvm.fma.v2f32(<2 x float> %a, <2 x float> %b, <2 x float> %c) nounwind ret <2 x float> %0 } +define float @test_fnma_f32(float %a, float %b, float %c) nounwind readnone ssp { +entry: +; CHECK: test_fnma_f32 +; CHECK: vfnma.f32 + %call = tail call float @llvm.fma.f32(float %a, float %b, float %c) nounwind readnone + %tmp1 = fsub float -0.0, %call + %tmp2 = fsub float %tmp1, %c + ret float %tmp2 +} + +define double @test_fnma_f64(double %a, double %b, double %c) nounwind readnone ssp { +entry: +; CHECK: test_fnma_f64 +; CHECK: vfnma.f64 + %call = tail call double @llvm.fma.f64(double %a, double %b, double %c) nounwind readnone + %tmp = fsub double -0.0, %call + ret double %tmp +} + declare float @llvm.fma.f32(float, float, float) nounwind readnone declare double @llvm.fma.f64(double, double, double) nounwind readnone declare <2 x float> @llvm.fma.v2f32(<2 x float>, <2 x float>, <2 x float>) nounwind readnone |