diff options
| author | Evan Cheng <evan.cheng@apple.com> | 2012-04-11 00:13:00 +0000 |
|---|---|---|
| committer | Evan Cheng <evan.cheng@apple.com> | 2012-04-11 00:13:00 +0000 |
| commit | 82509e5c62a99912c636b22e227b810eaf6eda78 (patch) | |
| tree | 4c7f36ce001590abdda1adea175a40a34ae0eb7c /lib/Target/ARM/ARMInstrNEON.td | |
| parent | 71fbed45d9f4e2e886afc7f22c058087e7872dc6 (diff) | |
| download | llvm-82509e5c62a99912c636b22e227b810eaf6eda78.tar.gz llvm-82509e5c62a99912c636b22e227b810eaf6eda78.tar.bz2 llvm-82509e5c62a99912c636b22e227b810eaf6eda78.tar.xz | |
Fix a number of problems with ARM fused multiply add/subtract instructions.
1. The new instruction itinerary entries are not properly described.
2. The asm parser can't handle vfms and vfnms.
3. There were no assembler, disassembler test cases.
4. HasNEON2 has the wrong assembler predicate.
rdar://10139676
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@154456 91177308-0d34-0410-b5e6-96231b3b80d8
Diffstat (limited to 'lib/Target/ARM/ARMInstrNEON.td')
| -rw-r--r-- | lib/Target/ARM/ARMInstrNEON.td | 9 |
1 files changed, 4 insertions, 5 deletions
diff --git a/lib/Target/ARM/ARMInstrNEON.td b/lib/Target/ARM/ARMInstrNEON.td index 99dbb95431..501cc8f4db 100644 --- a/lib/Target/ARM/ARMInstrNEON.td +++ b/lib/Target/ARM/ARMInstrNEON.td @@ -4115,7 +4115,6 @@ defm VQDMLSL : N3VLInt3_HS<0, 1, 0b1011, 0, IIC_VMACi16D, IIC_VMACi32D, "vqdmlsl", "s", int_arm_neon_vqdmlsl>; defm VQDMLSLsl: N3VLInt3SL_HS<0, 0b111, "vqdmlsl", "s", int_arm_neon_vqdmlsl>; - // Fused Vector Multiply-Accumulate and Fused Multiply-Subtract Operations. def VFMAfd : N3VDMulOp<0, 0, 0b00, 0b1100, 1, IIC_VFMACD, "vfma", "f32", v2f32, fmul_su, fadd_mlx>, @@ -4136,10 +4135,10 @@ def VFMSfq : N3VQMulOp<0, 0, 0b10, 0b1100, 1, IIC_VFMACQ, "vfms", "f32", // Match @llvm.fma.* intrinsics def : Pat<(fma (v2f32 DPR:$src1), (v2f32 DPR:$Vn), (v2f32 DPR:$Vm)), (VFMAfd DPR:$src1, DPR:$Vn, DPR:$Vm)>, - Requires<[HasNEON, HasVFP4]>; + Requires<[HasNEON2]>; def : Pat<(fma (v4f32 QPR:$src1), (v4f32 QPR:$Vn), (v4f32 QPR:$Vm)), (VFMAfq QPR:$src1, QPR:$Vn, QPR:$Vm)>, - Requires<[HasNEON, HasVFP4]>; + Requires<[HasNEON2]>; // Vector Subtract Operations. @@ -5497,9 +5496,9 @@ def : N3VSMulOpPat<fmul, fadd, VMLAfd>, def : N3VSMulOpPat<fmul, fsub, VMLSfd>, Requires<[HasNEON, UseNEONForFP, UseFPVMLx, NoNEON2]>; def : N3VSMulOpPat<fmul, fadd, VFMAfd>, - Requires<[HasNEON2, UseNEONForFP,FPContractions]>; + Requires<[HasNEON2, UseNEONForFP, FPContractions]>; def : N3VSMulOpPat<fmul, fsub, VFMSfd>, - Requires<[HasNEON2, UseNEONForFP,FPContractions]>; + Requires<[HasNEON2, UseNEONForFP, FPContractions]>; def : N2VSPat<fabs, VABSfd>; def : N2VSPat<fneg, VNEGfd>; def : N3VSPat<NEONfmax, VMAXfd>; |