replica
/
hanchenye-llvm-project
2
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Micro-optimization: 

If we have this situation:

    jCC  L1
    jmp  L2
L1:
  ...
L2:
  ...

We can get a small performance boost by emitting this instead:

    jnCC L2
L1:
  ...
L2:
  ...

This testcase shows an example of this:

float func(float x, float y) {
    double product = (double)x * y;
    if (product == 0.0)
        return product;
    return product - 1.0;
}

llvm-svn: 101075
This commit is contained in:
Bill Wendling 2010-04-12 22:19:57 +00:00
parent 3cdcc3f728
commit b02bbe416f
2 changed files with 84 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

47
llvm/lib/Target/X86/X86InstrInfo.cpp
View File

@ -1684,6 +1684,7 @@ bool X86InstrInfo::AnalyzeBranch(MachineBasicBlock &MBB,
// Start from the bottom of the block and work up, examining the
// terminator instructions.
MachineBasicBlock::iterator I = MBB.end();
MachineBasicBlock::iterator UnCondBrIter = MBB.end();
while (I != MBB.begin()) {
--I;
if (I->isDebugValue())
@ -1701,6 +1702,8 @@ bool X86InstrInfo::AnalyzeBranch(MachineBasicBlock &MBB,
// Handle unconditional branches.
if (I->getOpcode() == X86::JMP_4) {
UnCondBrIter = I;
if (!AllowModify) {
TBB = I->getOperand(0).getMBB();
continue;
@ -1718,10 +1721,11 @@ bool X86InstrInfo::AnalyzeBranch(MachineBasicBlock &MBB,
TBB = 0;
I->eraseFromParent();
I = MBB.end();
UnCondBrIter = MBB.end();
continue;
}
// TBB is used to indicate the unconditinal destination.
// TBB is used to indicate the unconditional destination.
TBB = I->getOperand(0).getMBB();
continue;
}
@ -1733,7 +1737,48 @@ bool X86InstrInfo::AnalyzeBranch(MachineBasicBlock &MBB,
// Working from the bottom, handle the first conditional branch.
if (Cond.empty()) {
MachineBasicBlock *TargetBB = I->getOperand(0).getMBB();
if (AllowModify && UnCondBrIter != MBB.end() &&
MBB.isLayoutSuccessor(TargetBB)) {
// If we can modify the code and it ends in something like:
//
// jCC L1
// jmp L2
// L1:
// ...
// L2:
//
// Then we can change this to:
//
// jnCC L2
// L1:
// ...
// L2:
//
// Which is a bit more efficient.
// We conditionally jump to the fall-through block.
BranchCode = GetOppositeBranchCondition(BranchCode);
unsigned JNCC = GetCondBranchFromCond(BranchCode);
MachineBasicBlock::iterator OldInst = I;
--I;
BuildMI(MBB, UnCondBrIter, MBB.findDebugLoc(I), get(JNCC))
.addMBB(UnCondBrIter->getOperand(0).getMBB());
BuildMI(MBB, UnCondBrIter, MBB.findDebugLoc(I), get(X86::JMP_4))
.addMBB(TargetBB);
MBB.addSuccessor(TargetBB);
OldInst->eraseFromParent();
UnCondBrIter->eraseFromParent();
// Restart the analysis.
UnCondBrIter = MBB.end();
I = MBB.end();
continue;
}
FBB = TBB;
TBB = TargetBB;
TBB = I->getOperand(0).getMBB();
Cond.push_back(MachineOperand::CreateImm(BranchCode));
continue;

38
llvm/test/CodeGen/X86/brcond.ll
View File

@ -67,3 +67,41 @@ return: ; preds = %entry
; CHECK-NEXT: orl 8(%esp), %eax
; CHECK-NEXT: je LBB3_2
}
; <rdar://problem/7598384>:
;
; jCC L1
; jmp L2
; L1:
; ...
; L2:
; ...
;
; to:
;
; jnCC L2
; L1:
; ...
; L2:
; ...
define float @test4(float %x, float %y) nounwind readnone optsize ssp {
entry:
%0 = fpext float %x to double ; <double> [#uses=1]
%1 = fpext float %y to double ; <double> [#uses=1]
%2 = fmul double %0, %1 ; <double> [#uses=3]
%3 = fcmp oeq double %2, 0.000000e+00 ; <i1> [#uses=1]
br i1 %3, label %bb2, label %bb1
; CHECK: jne
; CHECK-NOT: jmp
; CHECK-NEXT: jnp
bb1: ; preds = %entry
%4 = fadd double %2, -1.000000e+00 ; <double> [#uses=1]
br label %bb2
bb2: ; preds = %entry, %bb1
%.0.in = phi double [ %4, %bb1 ], [ %2, %entry ] ; <double> [#uses=1]
%.0 = fptrunc double %.0.in to float ; <float> [#uses=1]
ret float %.0
}