[ARM] Lower BR_CC for f16

[SjoerdMeijer]

Summary

This case wasn't handled yet.

[samparker]

LGTM.

lib/Target/ARM/ARMISelLowering.cpp

ARMTargetLowering::ARMTargetLowering(const TargetMachine &TM,
   setOperationAction(ISD::SELECT_CC, MVT::f64, Custom);
 
   // Thumb-1 cannot currently select ARMISD::SUBE.
   if (!Subtarget->isThumb1Only())
     setOperationAction(ISD::SETCCE, MVT::i32, Custom);
 
   setOperationAction(ISD::BRCOND,    MVT::Other, Custom);
   setOperationAction(ISD::BR_CC,     MVT::i32,   Custom);
+  setOperationAction(ISD::BR_CC,     MVT::f16,   Custom);
   setOperationAction(ISD::BR_CC,     MVT::f32,   Custom);
   setOperationAction(ISD::BR_CC,     MVT::f64,   Custom);
   setOperationAction(ISD::BR_JT,     MVT::Other, Custom);
 
   // We don't support sin/cos/fmod/copysign/pow
   setOperationAction(ISD::FSIN,      MVT::f64, Expand);
   setOperationAction(ISD::FSIN,      MVT::f32, Expand);
   setOperationAction(ISD::FCOS,      MVT::f32, Expand);
SDValue ARMTargetLowering::LowerBR_CC(SDValue Op, SelectionDAG &DAG) const {
   if (LHS.getValueType() == MVT::i32) {
     SDValue ARMcc;
     SDValue Cmp = getARMCmp(LHS, RHS, CC, ARMcc, DAG, dl);
     SDValue CCR = DAG.getRegister(ARM::CPSR, MVT::i32);
     return DAG.getNode(ARMISD::BRCOND, dl, MVT::Other,
                        Chain, Dest, ARMcc, CCR, Cmp);
   }
 
-  assert(LHS.getValueType() == MVT::f32 || LHS.getValueType() == MVT::f64);
-
   if (getTargetMachine().Options.UnsafeFPMath &&
       (CC == ISD::SETEQ || CC == ISD::SETOEQ ||
        CC == ISD::SETNE || CC == ISD::SETUNE)) {
     if (SDValue Result = OptimizeVFPBrcond(Op, DAG))
       return Result;
   }
 
   ARMCC::CondCodes CondCode, CondCode2;

test/CodeGen/ARM/fp16-instructions.ll

 }
 
 ; Check VCMPZH
 define zeroext i1 @VCMP2(float %F.coerce) {
 entry:
   %0 = bitcast float %F.coerce to i32
   %tmp.0.extract.trunc = trunc i32 %0 to i16
   %1 = bitcast i16 %tmp.0.extract.trunc to half
-  %cmp = fcmp une half %1, 0.000000e+00 
+  %cmp = fcmp une half %1, 0.000000e+00
   ret i1 %cmp
 
 ; CHECK-LABEL:             VCMP2:
 
 ; CHECK-SOFT:              bl __aeabi_fcmpeq
 ; CHECK-SOFTFP-FP16:       vcmp.f32        s0, #0
 ; CHECK-SOFTFP-FULLFP16:   vcmp.f16        s0, #0
 ; CHECK-HARDFP-FULLFP16:   vcmp.f16        s0, #0

 ; CHECK-LABEL:  VCMPE2:
 
 ; CHECK-SOFT:              bl  __aeabi_fcmplt
 ; CHECK-SOFTFP-FP16:       vcmpe.f32 s{{.}}, s{{.}}
 ; CHECK-SOFTFP-FULLFP16:   vcmpe.f16 s{{.}}, s{{.}}
 ; CHECK-HARDFP-FULLFP16:   vcmpe.f16 s{{.}}, s{{.}}
 }
 
+; Test lowering of BR_CC
+define hidden i32 @VCMPBRCC() {
+entry:
+  %f = alloca half, align 2
+  br label %for.cond
+
+for.cond:
+  %0 = load half, half* %f, align 2
+  %cmp = fcmp nnan ninf nsz ole half %0, 0xH6800
+  br i1 %cmp, label %for.body, label %for.end
+
+for.body:
+  ret i32 1
+
+for.end:
+  ret i32 0
+
+; CHECK-LABEL:            VCMPBRCC:
+
+; CHECK-SOFT:             bl  __aeabi_fcmple
+; CHECK-SOFT:             cmp r0, #0
+
+; CHECK-SOFTFP-FP16:      vcvtb.f32.f16 [[S2:s[0-9]]], [[S2]]
+; CHECK-SOFTFP-FP16:      vcmpe.f32 [[S2]], s0
+; CHECK-SOFTFP-FP16:      vmrs  APSR_nzcv, fpscr
+
+; CHECK-SOFTFP-FULLFP16:  vcmpe.f16 s{{.}}, s{{.}}
+; CHECK-SOFTFP-FULLFP16:  vmrs  APSR_nzcv, fpscr
+}
+
 ; 5. VCVT (between floating-point and fixed-point)
 ; Only assembly/disassembly support
 
 ; 6. VCVT (between floating-point and integer, both directions)
 define i32 @fptosi(i32 %A.coerce) {
 entry:
   %tmp.0.extract.trunc = trunc i32 %A.coerce to i16
   %0 = bitcast i16 %tmp.0.extract.trunc to half