1
0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-24 19:52:54 +01:00

Optimize the "bit test" code path for switch lowering in the

case where the bit mask has exactly one bit.

llvm-svn: 106716
This commit is contained in:
Dan Gohman 2010-06-24 02:06:24 +00:00
parent 1c9d50ed92
commit 7953fd5824
2 changed files with 77 additions and 14 deletions

View File

@ -1557,9 +1557,20 @@ void SelectionDAGBuilder::visitBitTestCase(MachineBasicBlock* NextMBB,
unsigned Reg, unsigned Reg,
BitTestCase &B, BitTestCase &B,
MachineBasicBlock *SwitchBB) { MachineBasicBlock *SwitchBB) {
// Make desired shift
SDValue ShiftOp = DAG.getCopyFromReg(getControlRoot(), getCurDebugLoc(), Reg, SDValue ShiftOp = DAG.getCopyFromReg(getControlRoot(), getCurDebugLoc(), Reg,
TLI.getPointerTy()); TLI.getPointerTy());
SDValue Cmp;
if (CountPopulation_64(B.Mask) == 1) {
// Testing for a single bit; just compare the shift count with what it
// would need to be to shift a 1 bit in that position.
Cmp = DAG.getSetCC(getCurDebugLoc(),
TLI.getSetCCResultType(ShiftOp.getValueType()),
ShiftOp,
DAG.getConstant(CountTrailingZeros_64(B.Mask),
TLI.getPointerTy()),
ISD::SETEQ);
} else {
// Make desired shift
SDValue SwitchVal = DAG.getNode(ISD::SHL, getCurDebugLoc(), SDValue SwitchVal = DAG.getNode(ISD::SHL, getCurDebugLoc(),
TLI.getPointerTy(), TLI.getPointerTy(),
DAG.getConstant(1, TLI.getPointerTy()), DAG.getConstant(1, TLI.getPointerTy()),
@ -1569,17 +1580,18 @@ void SelectionDAGBuilder::visitBitTestCase(MachineBasicBlock* NextMBB,
SDValue AndOp = DAG.getNode(ISD::AND, getCurDebugLoc(), SDValue AndOp = DAG.getNode(ISD::AND, getCurDebugLoc(),
TLI.getPointerTy(), SwitchVal, TLI.getPointerTy(), SwitchVal,
DAG.getConstant(B.Mask, TLI.getPointerTy())); DAG.getConstant(B.Mask, TLI.getPointerTy()));
SDValue AndCmp = DAG.getSetCC(getCurDebugLoc(), Cmp = DAG.getSetCC(getCurDebugLoc(),
TLI.getSetCCResultType(AndOp.getValueType()), TLI.getSetCCResultType(AndOp.getValueType()),
AndOp, DAG.getConstant(0, TLI.getPointerTy()), AndOp, DAG.getConstant(0, TLI.getPointerTy()),
ISD::SETNE); ISD::SETNE);
}
SwitchBB->addSuccessor(B.TargetBB); SwitchBB->addSuccessor(B.TargetBB);
SwitchBB->addSuccessor(NextMBB); SwitchBB->addSuccessor(NextMBB);
SDValue BrAnd = DAG.getNode(ISD::BRCOND, getCurDebugLoc(), SDValue BrAnd = DAG.getNode(ISD::BRCOND, getCurDebugLoc(),
MVT::Other, getControlRoot(), MVT::Other, getControlRoot(),
AndCmp, DAG.getBasicBlock(B.TargetBB)); Cmp, DAG.getBasicBlock(B.TargetBB));
// Set NextBlock to be the MBB immediately after the current one, if any. // Set NextBlock to be the MBB immediately after the current one, if any.
// This is used to avoid emitting unnecessary branches to the next block. // This is used to avoid emitting unnecessary branches to the next block.

View File

@ -0,0 +1,51 @@
; RUN: llc -march=x86-64 -asm-verbose=false < %s | FileCheck %s
; This switch should use bit tests, and the third bit test case is just
; testing for one possible value, so it doesn't need a bt.
; CHECK: movabsq $2305843009482129440, %r
; CHECK-NEXT: btq %rax, %r
; CHECK-NEXT: jb
; CHECK-NEXT: movl $671088640, %e
; CHECK-NEXT: btq %rax, %r
; CHECK-NEXT: jb
; CHECK-NEXT: testq %rax, %r
; CHECK-NEXT: j
define void @test(i8* %l) nounwind {
entry:
%l.addr = alloca i8*, align 8 ; <i8**> [#uses=2]
store i8* %l, i8** %l.addr
%tmp = load i8** %l.addr ; <i8*> [#uses=1]
%tmp1 = load i8* %tmp ; <i8> [#uses=1]
%conv = sext i8 %tmp1 to i32 ; <i32> [#uses=1]
switch i32 %conv, label %sw.default [
i32 62, label %sw.bb
i32 60, label %sw.bb
i32 38, label %sw.bb2
i32 94, label %sw.bb2
i32 61, label %sw.bb2
i32 33, label %sw.bb4
]
sw.bb: ; preds = %entry, %entry
call void @foo(i32 0)
br label %sw.epilog
sw.bb2: ; preds = %entry, %entry, %entry
call void @foo(i32 1)
br label %sw.epilog
sw.bb4: ; preds = %entry
call void @foo(i32 3)
br label %sw.epilog
sw.default: ; preds = %entry
call void @foo(i32 97)
br label %sw.epilog
sw.epilog: ; preds = %sw.default, %sw.bb4, %sw.bb2, %sw.bb
ret void
}
declare void @foo(i32)