diff options
Diffstat (limited to 'llvm/lib/Target/ARM/ARMISelLowering.cpp')
| -rw-r--r-- | llvm/lib/Target/ARM/ARMISelLowering.cpp | 25 |
1 files changed, 25 insertions, 0 deletions
diff --git a/llvm/lib/Target/ARM/ARMISelLowering.cpp b/llvm/lib/Target/ARM/ARMISelLowering.cpp index a87770125fd..e372bdb4f58 100644 --- a/llvm/lib/Target/ARM/ARMISelLowering.cpp +++ b/llvm/lib/Target/ARM/ARMISelLowering.cpp @@ -1490,6 +1490,7 @@ const char *ARMTargetLowering::getTargetNodeName(unsigned Opcode) const { case ARMISD::ADDE: return "ARMISD::ADDE"; case ARMISD::SUBC: return "ARMISD::SUBC"; case ARMISD::SUBE: return "ARMISD::SUBE"; + case ARMISD::LSLS: return "ARMISD::LSLS"; case ARMISD::VMOVRRD: return "ARMISD::VMOVRRD"; case ARMISD::VMOVDRR: return "ARMISD::VMOVDRR"; @@ -4110,6 +4111,30 @@ SDValue ARMTargetLowering::getARMCmp(SDValue LHS, SDValue RHS, ISD::CondCode CC, } } } + + // The specific comparison "(x<<c) > 0x80000000U" can be optimized to a + // single "lsls x, c+1". The shift sets the "C" and "Z" flags the same + // way a cmp would. + // FIXME: Add support for ARM/Thumb2; this would need isel patterns, and + // some tweaks to the heuristics for the previous and->shift transform. + // FIXME: Optimize cases where the LHS isn't a shift. + if (Subtarget->isThumb1Only() && LHS->getOpcode() == ISD::SHL && + isa<ConstantSDNode>(RHS) && + cast<ConstantSDNode>(RHS)->getZExtValue() == 0x80000000U && + CC == ISD::SETUGT && isa<ConstantSDNode>(LHS.getOperand(1)) && + cast<ConstantSDNode>(LHS.getOperand(1))->getZExtValue() < 31) { + unsigned ShiftAmt = + cast<ConstantSDNode>(LHS.getOperand(1))->getZExtValue() + 1; + SDValue Shift = DAG.getNode(ARMISD::LSLS, dl, + DAG.getVTList(MVT::i32, MVT::i32), + LHS.getOperand(0), + DAG.getConstant(ShiftAmt, dl, MVT::i32)); + SDValue Chain = DAG.getCopyToReg(DAG.getEntryNode(), dl, ARM::CPSR, + Shift.getValue(1), SDValue()); + ARMcc = DAG.getConstant(ARMCC::HI, dl, MVT::i32); + return Chain.getValue(1); + } + ARMCC::CondCodes CondCode = IntCCToARMCC(CC); // If the RHS is a constant zero then the V (overflow) flag will never be |
