aboutsummaryrefslogtreecommitdiffstats
path: root/lib/Target/X86/X86FloatingPoint.cpp
diff options
context:
space:
mode:
authorLogan Chien <loganchien@google.com>2011-10-20 00:08:13 +0800
committerLogan Chien <loganchien@google.com>2011-10-20 00:09:35 +0800
commit0ebc07a576037e4e36f68bf5cece32740ca120c0 (patch)
treec2e40648043d01498ee25af839a071193561e425 /lib/Target/X86/X86FloatingPoint.cpp
parent62383e889e0b06fd12a6b88311717cd33a1925c4 (diff)
parentcdd8e46bec4e975d00a5abea808d8eb4138515c5 (diff)
downloadexternal_llvm-0ebc07a576037e4e36f68bf5cece32740ca120c0.zip
external_llvm-0ebc07a576037e4e36f68bf5cece32740ca120c0.tar.gz
external_llvm-0ebc07a576037e4e36f68bf5cece32740ca120c0.tar.bz2
Merge with LLVM upstream 2011/10/20 (r142530)
Conflicts: lib/Support/Unix/Host.inc Change-Id: Idc00db3b63912dca6348bddd9f8a1af2a8d5d147
Diffstat (limited to 'lib/Target/X86/X86FloatingPoint.cpp')
-rw-r--r--lib/Target/X86/X86FloatingPoint.cpp34
1 files changed, 34 insertions, 0 deletions
diff --git a/lib/Target/X86/X86FloatingPoint.cpp b/lib/Target/X86/X86FloatingPoint.cpp
index 6eed6ab..e3461c8 100644
--- a/lib/Target/X86/X86FloatingPoint.cpp
+++ b/lib/Target/X86/X86FloatingPoint.cpp
@@ -260,6 +260,21 @@ namespace {
BuildMI(*MBB, I, dl, TII->get(X86::LD_Frr)).addReg(STReg);
}
+ /// duplicatePendingSTBeforeKill - The instruction at I is about to kill
+ /// RegNo. If any PendingST registers still need the RegNo value, duplicate
+ /// them to new scratch registers.
+ void duplicatePendingSTBeforeKill(unsigned RegNo, MachineInstr *I) {
+ for (unsigned i = 0; i != NumPendingSTs; ++i) {
+ if (PendingST[i] != RegNo)
+ continue;
+ unsigned SR = getScratchReg();
+ DEBUG(dbgs() << "Duplicating pending ST" << i
+ << " in FP" << RegNo << " to FP" << SR << '\n');
+ duplicateToTop(RegNo, SR, I);
+ PendingST[i] = SR;
+ }
+ }
+
/// popStackAfter - Pop the current value off of the top of the FP stack
/// after the specified instruction.
void popStackAfter(MachineBasicBlock::iterator &I);
@@ -406,6 +421,10 @@ bool FPS::processBasicBlock(MachineFunction &MF, MachineBasicBlock &BB) {
if (MI->isCopy() && isFPCopy(MI))
FPInstClass = X86II::SpecialFP;
+ if (MI->isImplicitDef() &&
+ X86::RFP80RegClass.contains(MI->getOperand(0).getReg()))
+ FPInstClass = X86II::SpecialFP;
+
if (FPInstClass == X86II::NotFP)
continue; // Efficiently ignore non-fp insts!
@@ -461,6 +480,7 @@ bool FPS::processBasicBlock(MachineFunction &MF, MachineBasicBlock &BB) {
}
dumpStack();
);
+ (void)PrevMI;
Changed = true;
}
@@ -969,6 +989,9 @@ void FPS::handleOneArgFP(MachineBasicBlock::iterator &I) {
unsigned Reg = getFPReg(MI->getOperand(NumOps-1));
bool KillsSrc = MI->killsRegister(X86::FP0+Reg);
+ if (KillsSrc)
+ duplicatePendingSTBeforeKill(Reg, I);
+
// FISTP64m is strange because there isn't a non-popping versions.
// If we have one _and_ we don't want to pop the operand, duplicate the value
// on the stack instead of moving it. This ensure that popping the value is
@@ -1032,6 +1055,7 @@ void FPS::handleOneArgFPRW(MachineBasicBlock::iterator &I) {
bool KillsSrc = MI->killsRegister(X86::FP0+Reg);
if (KillsSrc) {
+ duplicatePendingSTBeforeKill(Reg, I);
// If this is the last use of the source register, just make sure it's on
// the top of the stack.
moveToTop(Reg, I);
@@ -1318,6 +1342,7 @@ void FPS::handleSpecialFP(MachineBasicBlock::iterator &I) {
// When the source is killed, allocate a scratch FP register.
if (KillsSrc) {
+ duplicatePendingSTBeforeKill(SrcFP, I);
unsigned Slot = getSlot(SrcFP);
unsigned SR = getScratchReg();
PendingST[DstST] = SR;
@@ -1369,6 +1394,15 @@ void FPS::handleSpecialFP(MachineBasicBlock::iterator &I) {
break;
}
+ case TargetOpcode::IMPLICIT_DEF: {
+ // All FP registers must be explicitly defined, so load a 0 instead.
+ unsigned Reg = MI->getOperand(0).getReg() - X86::FP0;
+ DEBUG(dbgs() << "Emitting LD_F0 for implicit FP" << Reg << '\n');
+ BuildMI(*MBB, I, MI->getDebugLoc(), TII->get(X86::LD_F0));
+ pushReg(Reg);
+ break;
+ }
+
case X86::FpPOP_RETVAL: {
// The FpPOP_RETVAL instruction is used after calls that return a value on
// the floating point stack. We cannot model this with ST defs since CALL