summaryrefslogtreecommitdiff
path: root/lib/CodeGen/SelectionDAG/TargetLowering.cpp
diff options
context:
space:
mode:
authorEvan Cheng <evan.cheng@apple.com>2012-07-17 06:53:39 +0000
committerEvan Cheng <evan.cheng@apple.com>2012-07-17 06:53:39 +0000
commit70e10d3fe4c5df189348f64fce56254a5a32b51c (patch)
tree23550acde05b658d6b1b3ca3594121be0c51c3c1 /lib/CodeGen/SelectionDAG/TargetLowering.cpp
parent06a27cc1aae20e3a2c985dd8dadab3b4ef5d032a (diff)
downloadllvm-70e10d3fe4c5df189348f64fce56254a5a32b51c.tar.gz
llvm-70e10d3fe4c5df189348f64fce56254a5a32b51c.tar.bz2
llvm-70e10d3fe4c5df189348f64fce56254a5a32b51c.tar.xz
This is another case where instcombine demanded bits optimization created
large immediates. Add dag combine logic to recover in case the large immediates doesn't fit in cmp immediate operand field. int foo(unsigned long l) { return (l>> 47) == 1; } we produce %shr.mask = and i64 %l, -140737488355328 %cmp = icmp eq i64 %shr.mask, 140737488355328 %conv = zext i1 %cmp to i32 ret i32 %conv which codegens to movq $0xffff800000000000,%rax andq %rdi,%rax movq $0x0000800000000000,%rcx cmpq %rcx,%rax sete %al movzbl %al,%eax ret TargetLowering::SimplifySetCC would transform (X & -256) == 256 -> (X >> 8) == 1 if the immediate fails the isLegalICmpImmediate() test. For x86, that's immediates which are not a signed 32-bit immediate. Based on a patch by Eli Friedman. PR10328 rdar://9758774 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@160346 91177308-0d34-0410-b5e6-96231b3b80d8
Diffstat (limited to 'lib/CodeGen/SelectionDAG/TargetLowering.cpp')
-rw-r--r--lib/CodeGen/SelectionDAG/TargetLowering.cpp21
1 files changed, 21 insertions, 0 deletions
diff --git a/lib/CodeGen/SelectionDAG/TargetLowering.cpp b/lib/CodeGen/SelectionDAG/TargetLowering.cpp
index 891d401201..9902ed76fa 100644
--- a/lib/CodeGen/SelectionDAG/TargetLowering.cpp
+++ b/lib/CodeGen/SelectionDAG/TargetLowering.cpp
@@ -2322,6 +2322,27 @@ TargetLowering::SimplifySetCC(EVT VT, SDValue N0, SDValue N1,
}
}
}
+
+ if (!isLegalICmpImmediate(C1.getSExtValue())) {
+ // (X & -256) == 256 -> (X >> 8) == 1
+ if ((Cond == ISD::SETEQ || Cond == ISD::SETNE) &&
+ N0.getOpcode() == ISD::AND && N0.hasOneUse()) {
+ if (ConstantSDNode *AndRHS =
+ dyn_cast<ConstantSDNode>(N0.getOperand(1))) {
+ const APInt &AndRHSC = AndRHS->getAPIntValue();
+ if ((-AndRHSC).isPowerOf2() && (AndRHSC & C1) == C1) {
+ unsigned ShiftBits = AndRHSC.countTrailingZeros();
+ EVT ShiftTy = DCI.isBeforeLegalize() ?
+ getPointerTy() : getShiftAmountTy(N0.getValueType());
+ EVT CmpTy = N0.getValueType();
+ SDValue Shift = DAG.getNode(ISD::SRL, dl, CmpTy, N0.getOperand(0),
+ DAG.getConstant(ShiftBits, ShiftTy));
+ SDValue CmpRHS = DAG.getConstant(C1.lshr(ShiftBits), CmpTy);
+ return DAG.getSetCC(dl, VT, Shift, CmpRHS, Cond);
+ }
+ }
+ }
+ }
}
if (isa<ConstantFPSDNode>(N0.getNode())) {