diff options
author | Daniel Sanders <daniel.sanders@imgtec.com> | 2014-03-10 15:01:57 +0000 |
---|---|---|
committer | Daniel Sanders <daniel.sanders@imgtec.com> | 2014-03-10 15:01:57 +0000 |
commit | e1820a6a4e5cfd1a41bd40ce6c21635d04289fec (patch) | |
tree | 19b87bbe2f524a4e358ee507bca26f4129e538a5 /lib/Target/Mips/MipsSEInstrInfo.cpp | |
parent | 7d4afb2711974795e655688c1f7964cd60f3cb96 (diff) | |
download | llvm-e1820a6a4e5cfd1a41bd40ce6c21635d04289fec.tar.gz llvm-e1820a6a4e5cfd1a41bd40ce6c21635d04289fec.tar.bz2 llvm-e1820a6a4e5cfd1a41bd40ce6c21635d04289fec.tar.xz |
[mips][fp64] Add an implicit def to MFHC1 claiming that it reads the lower 32-bits of 64-bit FPR
Summary:
This is a white lie to workaround a widespread bug in the -mfp64
implementation.
The problem is that none of the 32-bit fpu ops mention the fact that they
clobber the upper 32-bits of the 64-bit FPR. This allows MFHC1 to be
scheduled on the wrong side of most 32-bit FPU ops. Fixing that requires a
major overhaul of the FPU implementation which can't be done right now due to
time constraints.
MFHC1 is one of two affected instructions. These instructions are the only
FPU instructions that don't read or write the lower 32-bits. We therefore
pretend that it reads the bottom 32-bits to artificially create a dependency and
prevent the scheduler changing the behaviour of the code.
The other instruction is MTHC1 which will be fixed once I've have found a failing
test case for it.
The testcase is test-suite/SingleSource/UnitTests/Vector/simple.c when
given TARGET_CFLAGS="-mips32r2 -mfp64 -mmsa".
Reviewers: jacksprat, matheusalmeida
Reviewed By: jacksprat
Differential Revision: http://llvm-reviews.chandlerc.com/D2966
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@203464 91177308-0d34-0410-b5e6-96231b3b80d8
Diffstat (limited to 'lib/Target/Mips/MipsSEInstrInfo.cpp')
-rw-r--r-- | lib/Target/Mips/MipsSEInstrInfo.cpp | 18 |
1 files changed, 15 insertions, 3 deletions
diff --git a/lib/Target/Mips/MipsSEInstrInfo.cpp b/lib/Target/Mips/MipsSEInstrInfo.cpp index 96aaf26d00..a7ade1f47b 100644 --- a/lib/Target/Mips/MipsSEInstrInfo.cpp +++ b/lib/Target/Mips/MipsSEInstrInfo.cpp @@ -506,9 +506,21 @@ void MipsSEInstrInfo::expandExtractElementF64(MachineBasicBlock &MBB, unsigned SubIdx = N ? Mips::sub_hi : Mips::sub_lo; unsigned SubReg = getRegisterInfo().getSubReg(SrcReg, SubIdx); - if (SubIdx == Mips::sub_hi && FP64) - BuildMI(MBB, I, dl, get(Mips::MFHC1), DstReg).addReg(SubReg); - else + if (SubIdx == Mips::sub_hi && FP64) { + // FIXME: The .addReg(SrcReg, RegState::Implicit) is a white lie used to + // temporarily work around a widespread bug in the -mfp64 support. + // The problem is that none of the 32-bit fpu ops mention the fact + // that they clobber the upper 32-bits of the 64-bit FPR. Fixing that + // requires a major overhaul of the FPU implementation which can't + // be done right now due to time constraints. + // MFHC1 is the only instruction that is affected since it is the + // only instruction that doesn't read the lower 32-bits. We therefore + // pretend that it reads the bottom 32-bits to artificially create a + // dependency and prevent the scheduler changing the behaviour of the + // code. + BuildMI(MBB, I, dl, get(Mips::MFHC1), DstReg).addReg(SubReg).addReg( + SrcReg, RegState::Implicit); + } else BuildMI(MBB, I, dl, get(Mips::MFC1), DstReg).addReg(SubReg); } |