From 54f1326e1f7c971a67f31582ebe54e6d94794f40 Mon Sep 17 00:00:00 2001 From: Juergen Ributzka Date: Wed, 11 Jun 2014 23:10:58 +0000 Subject: [PATCH] [FastIsel][X86] Add support for lowering the first 8 floating-point arguments. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210719 91177308-0d34-0410-b5e6-96231b3b80d8 --- lib/Target/X86/X86FastISel.cpp | 55 ++++++++++++++++++++++------------ 1 file changed, 36 insertions(+), 19 deletions(-) diff --git a/lib/Target/X86/X86FastISel.cpp b/lib/Target/X86/X86FastISel.cpp index 2b6b378059b..23919849648 100644 --- a/lib/Target/X86/X86FastISel.cpp +++ b/lib/Target/X86/X86FastISel.cpp @@ -1896,31 +1896,39 @@ bool X86FastISel::FastLowerArguments() { return false; // Only handle simple cases. i.e. Up to 6 i32/i64 scalar arguments. + unsigned GPRCnt = 0; + unsigned FPRCnt = 0; unsigned Idx = 1; - for (Function::const_arg_iterator I = F->arg_begin(), E = F->arg_end(); - I != E; ++I, ++Idx) { - if (Idx > 6) - return false; - + for (auto const &Arg : F->args()) { if (F->getAttributes().hasAttribute(Idx, Attribute::ByVal) || F->getAttributes().hasAttribute(Idx, Attribute::InReg) || F->getAttributes().hasAttribute(Idx, Attribute::StructRet) || F->getAttributes().hasAttribute(Idx, Attribute::Nest)) return false; - Type *ArgTy = I->getType(); + Type *ArgTy = Arg.getType(); if (ArgTy->isStructTy() || ArgTy->isArrayTy() || ArgTy->isVectorTy()) return false; EVT ArgVT = TLI.getValueType(ArgTy); if (!ArgVT.isSimple()) return false; switch (ArgVT.getSimpleVT().SimpleTy) { + default: return false; case MVT::i32: case MVT::i64: + ++GPRCnt; break; - default: - return false; + case MVT::f32: + case MVT::f64: + ++FPRCnt; + break; } + + if (GPRCnt > 6) + return false; + + if (FPRCnt > 8) + return false; } static const MCPhysReg GPR32ArgRegs[] = { @@ -1929,24 +1937,33 @@ bool X86FastISel::FastLowerArguments() { static const MCPhysReg GPR64ArgRegs[] = { X86::RDI, X86::RSI, X86::RDX, X86::RCX, X86::R8 , X86::R9 }; + static const MCPhysReg XMMArgRegs[] = { + X86::XMM0, X86::XMM1, X86::XMM2, X86::XMM3, + X86::XMM4, X86::XMM5, X86::XMM6, X86::XMM7 + }; - Idx = 0; - const TargetRegisterClass *RC32 = TLI.getRegClassFor(MVT::i32); - const TargetRegisterClass *RC64 = TLI.getRegClassFor(MVT::i64); - for (Function::const_arg_iterator I = F->arg_begin(), E = F->arg_end(); - I != E; ++I, ++Idx) { - bool is32Bit = TLI.getValueType(I->getType()) == MVT::i32; - const TargetRegisterClass *RC = is32Bit ? RC32 : RC64; - unsigned SrcReg = is32Bit ? GPR32ArgRegs[Idx] : GPR64ArgRegs[Idx]; + unsigned GPRIdx = 0; + unsigned FPRIdx = 0; + for (auto const &Arg : F->args()) { + MVT VT = TLI.getSimpleValueType(Arg.getType()); + const TargetRegisterClass *RC = TLI.getRegClassFor(VT); + unsigned SrcReg; + switch (VT.SimpleTy) { + default: llvm_unreachable("Unexpected value type."); + case MVT::i32: SrcReg = GPR32ArgRegs[GPRIdx++]; break; + case MVT::i64: SrcReg = GPR64ArgRegs[GPRIdx++]; break; + case MVT::f32: // fall-through + case MVT::f64: SrcReg = XMMArgRegs[FPRIdx++]; break; + } unsigned DstReg = FuncInfo.MF->addLiveIn(SrcReg, RC); // FIXME: Unfortunately it's necessary to emit a copy from the livein copy. // Without this, EmitLiveInCopies may eliminate the livein if its only // use is a bitcast (which isn't turned into an instruction). unsigned ResultReg = createResultReg(RC); BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, - TII.get(TargetOpcode::COPY), - ResultReg).addReg(DstReg, getKillRegState(true)); - UpdateValueMap(I, ResultReg); + TII.get(TargetOpcode::COPY), ResultReg) + .addReg(DstReg, getKillRegState(true)); + UpdateValueMap(&Arg, ResultReg); } return true; }