Consider this code snippet:
float t1(int argc) {
return (argc == 1123) ? 1.234f : 2.38213f;
}
We would generate truly awful code on ARM (those with a weak stomach should look
away):
_t1:
movw r1, #1123
movs r2, #1
movs r3, #0
cmp r0, r1
mov.w r0, #0
it eq
moveq r0, r2
movs r1, #4
cmp r0, #0
it ne
movne r3, r1
adr r0, #LCPI1_0
ldr r0, [r0, r3]
bx lr
The problem was that legalization was creating a cascade of SELECT_CC nodes, for
for the comparison of "argc == 1123" which was fed into a SELECT node for the ?:
statement which was itself converted to a SELECT_CC node. This is because the
ARM back-end doesn't have custom lowering for SELECT nodes, so it used the
default "Expand".
I added a fairly simple "LowerSELECT" to the ARM back-end. It takes care of this
testcase, but can obviously be expanded to include more cases.
Now we generate this, which looks optimal to me:
_t1:
movw r1, #1123
movs r2, #0
cmp r0, r1
adr r0, #LCPI0_0
it eq
moveq r2, #4
ldr r0, [r0, r2]
bx lr
.align 2
LCPI0_0:
.long 1075344593 @ float 2.382130e+00
.long 1067316150 @ float 1.234000e+00
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@110799 91177308-0d34-0410-b5e6-96231b3b80d8
diff --git a/lib/Target/ARM/ARMISelLowering.cpp b/lib/Target/ARM/ARMISelLowering.cpp
index 56adcf2..e4b5562 100644
--- a/lib/Target/ARM/ARMISelLowering.cpp
+++ b/lib/Target/ARM/ARMISelLowering.cpp
@@ -483,9 +483,9 @@
setOperationAction(ISD::SETCC, MVT::i32, Expand);
setOperationAction(ISD::SETCC, MVT::f32, Expand);
setOperationAction(ISD::SETCC, MVT::f64, Expand);
- setOperationAction(ISD::SELECT, MVT::i32, Expand);
- setOperationAction(ISD::SELECT, MVT::f32, Expand);
- setOperationAction(ISD::SELECT, MVT::f64, Expand);
+ setOperationAction(ISD::SELECT, MVT::i32, Custom);
+ setOperationAction(ISD::SELECT, MVT::f32, Custom);
+ setOperationAction(ISD::SELECT, MVT::f64, Custom);
setOperationAction(ISD::SELECT_CC, MVT::i32, Custom);
setOperationAction(ISD::SELECT_CC, MVT::f32, Custom);
setOperationAction(ISD::SELECT_CC, MVT::f64, Custom);
@@ -2314,6 +2314,52 @@
return DAG.getNode(ARMISD::FMSTAT, dl, MVT::Flag, Cmp);
}
+SDValue ARMTargetLowering::LowerSELECT(SDValue Op, SelectionDAG &DAG) const {
+ SDValue Cond = Op.getOperand(0);
+ SDValue SelectTrue = Op.getOperand(1);
+ SDValue SelectFalse = Op.getOperand(2);
+ DebugLoc dl = Op.getDebugLoc();
+
+ // Convert:
+ //
+ // (select (cmov 1, 0, cond), t, f) -> (cmov t, f, cond)
+ // (select (cmov 0, 1, cond), t, f) -> (cmov f, t, cond)
+ //
+ if (Cond.getOpcode() == ARMISD::CMOV && Cond.hasOneUse()) {
+ const ConstantSDNode *CMOVTrue =
+ dyn_cast<ConstantSDNode>(Cond.getOperand(0));
+ const ConstantSDNode *CMOVFalse =
+ dyn_cast<ConstantSDNode>(Cond.getOperand(1));
+
+ if (CMOVTrue && CMOVFalse) {
+ unsigned CMOVTrueVal = CMOVTrue->getZExtValue();
+ unsigned CMOVFalseVal = CMOVFalse->getZExtValue();
+
+ SDValue True;
+ SDValue False;
+ if (CMOVTrueVal == 1 && CMOVFalseVal == 0) {
+ True = SelectTrue;
+ False = SelectFalse;
+ } else if (CMOVTrueVal == 0 && CMOVFalseVal == 1) {
+ True = SelectFalse;
+ False = SelectTrue;
+ }
+
+ if (True.getNode() && False.getNode()) {
+ EVT VT = Cond.getValueType();
+ SDValue ARMcc = Cond.getOperand(2);
+ SDValue CCR = Cond.getOperand(3);
+ SDValue Cmp = Cond.getOperand(4);
+ return DAG.getNode(ARMISD::CMOV, dl, VT, True, False, ARMcc, CCR, Cmp);
+ }
+ }
+ }
+
+ return DAG.getSelectCC(dl, Cond,
+ DAG.getConstant(0, Cond.getValueType()),
+ SelectTrue, SelectFalse, ISD::SETNE);
+}
+
SDValue ARMTargetLowering::LowerSELECT_CC(SDValue Op, SelectionDAG &DAG) const {
EVT VT = Op.getValueType();
SDValue LHS = Op.getOperand(0);
@@ -3687,6 +3733,7 @@
return Subtarget->isTargetDarwin() ? LowerGlobalAddressDarwin(Op, DAG) :
LowerGlobalAddressELF(Op, DAG);
case ISD::GlobalTLSAddress: return LowerGlobalTLSAddress(Op, DAG);
+ case ISD::SELECT: return LowerSELECT(Op, DAG);
case ISD::SELECT_CC: return LowerSELECT_CC(Op, DAG);
case ISD::BR_CC: return LowerBR_CC(Op, DAG);
case ISD::BR_JT: return LowerBR_JT(Op, DAG);