[ARM] Favour PL/MI over GE/LT when possible

The arm condition codes for GE is N==V (and for LT is N!=V). If the source of
flags cannot set V (overflow), such as a cmp against #0, then we can use the
simpler PL and MI conditions that only check N. As these PL/MI conditions are
simpler than GE/LT, other passes like the peephole optimiser can have a better
time optimising away the redundant CMPs.

The exception is the VSEL instruction, which cannot take the PL code, so there
the transform favours GE.

Differential Revision: https://reviews.llvm.org/D64160

llvm-svn: 365117

1 files changed, 19 insertions, 0 deletions

diff --git a/llvm/lib/Target/ARM/ARMISelLowering.cpp b/llvm/lib/Target/ARM/ARMISelLowering.cpp
index 28d014c297e..ea8e84d0afa 100644
--- a/llvm/lib/Target/ARM/ARMISelLowering.cpp
+++ b/llvm/lib/Target/ARM/ARMISelLowering.cpp
@@ -4017,6 +4017,22 @@ SDValue ARMTargetLowering::getARMCmp(SDValue LHS, SDValue RHS, ISD::CondCode CC,
   }
 
   ARMCC::CondCodes CondCode = IntCCToARMCC(CC);
+
+  // If the RHS is a constant zero then the V (overflow) flag will never be
+  // set. This can allow us to simplify GE to PL or LT to MI, which can be
+  // simpler for other passes (like the peephole optimiser) to deal with.
+  if (isNullConstant(RHS)) {
+    switch (CondCode) {
+      default: break;
+      case ARMCC::GE:
+        CondCode = ARMCC::PL;
+        break;
+      case ARMCC::LT:
+        CondCode = ARMCC::MI;
+        break;
+    }
+  }
+
   ARMISD::NodeType CompareType;
   switch (CondCode) {
   default:
@@ -4645,6 +4661,9 @@ SDValue ARMTargetLowering::LowerSELECT_CC(SDValue Op, SelectionDAG &DAG) const {
     SDValue ARMcc;
     SDValue CCR = DAG.getRegister(ARM::CPSR, MVT::i32);
     SDValue Cmp = getARMCmp(LHS, RHS, CC, ARMcc, DAG, dl);
+    // Choose GE over PL, which vsel does now support
+    if (cast<ConstantSDNode>(ARMcc)->getZExtValue() == ARMCC::PL)
+      ARMcc = DAG.getConstant(ARMCC::GE, dl, MVT::i32);
     return getCMOV(dl, VT, FalseVal, TrueVal, ARMcc, CCR, Cmp, DAG);
   }