[ARM][LowOverheadLoops] Fix generated code for "revert".

[efriedma]

Summary

Two issues:

1. t2CMPri shouldn't use CPSR if it isn't predicated. This doesn't really have any visible effect at the moment, but it might matter in the future.
2. The t2CMPri generated for t2WhileLoopStart might need to use a register that isn't LR.

My team found this because we have a patch to track register liveness late in the pass pipeline. I'll look into upstreaming it to help catch issues like this earlier.

[SjoerdMeijer]

Thanks for fixing.

My team found this because we have a patch to track register liveness late in the pass pipeline. I'll look into upstreaming it to help catch issues like this earlier.

that sounds very interesting/useful.

[efriedma]

Posted the liveness patch: https://reviews.llvm.org/D66319

llvm/trunk/lib/Target/ARM/ARMLowOverheadLoops.cpp

 // FIXME: Need to check that we're not trashing the CPSR when generating the
 // cmp. We could also try to generate a cbz if the value in LR is also in
 // another low register.
 void ARMLowOverheadLoops::RevertWhile(MachineInstr *MI) const {
   LLVM_DEBUG(dbgs() << "ARM Loops: Reverting to cmp: " << *MI);
   MachineBasicBlock *MBB = MI->getParent();
   MachineInstrBuilder MIB = BuildMI(*MBB, MI, MI->getDebugLoc(),
                                     TII->get(ARM::t2CMPri));
-  MIB.addReg(ARM::LR);
+  MIB.add(MI->getOperand(0));
   MIB.addImm(0);
   MIB.addImm(ARMCC::AL);
-  MIB.addReg(ARM::CPSR);
+  MIB.addReg(ARM::NoRegister);
 
   // TODO: Try to use tBcc instead
   MIB = BuildMI(*MBB, MI, MI->getDebugLoc(), TII->get(ARM::t2Bcc));
   MIB.add(MI->getOperand(1));   // branch target
   MIB.addImm(ARMCC::EQ);        // condition code
   MIB.addReg(ARM::CPSR);
   MI->eraseFromParent();
 }
void ARMLowOverheadLoops::RevertLoopEnd(MachineInstr *MI) const {
 
   // Create cmp
   MachineBasicBlock *MBB = MI->getParent();
   MachineInstrBuilder MIB = BuildMI(*MBB, MI, MI->getDebugLoc(),
                                     TII->get(ARM::t2CMPri));
   MIB.addReg(ARM::LR);
   MIB.addImm(0);
   MIB.addImm(ARMCC::AL);
-  MIB.addReg(ARM::CPSR);
+  MIB.addReg(ARM::NoRegister);
 
   // TODO Try to use tBcc instead.
   // Create bne
   MIB = BuildMI(*MBB, MI, MI->getDebugLoc(), TII->get(ARM::t2Bcc));
   MIB.add(MI->getOperand(1));   // branch target
   MIB.addImm(ARMCC::NE);        // condition code
   MIB.addReg(ARM::CPSR);
   MI->eraseFromParent();

llvm/trunk/test/CodeGen/Thumb2/LowOverheadLoops/end-positive-offset.mir

 # RUN: llc -mtriple=thumbv8.1m.main -run-pass=arm-low-overhead-loops %s -verify-machineinstrs -o - | FileCheck %s
 
 # CHECK-NOT: DoLoopStart
 # CHECK-NOT: DLS
 # CHECK: bb.1.for.body:
-# CHECK:   t2CMPri $lr, 0, 14, $cpsr, implicit-def $cpsr
+# CHECK:   t2CMPri $lr, 0, 14, $noreg, implicit-def $cpsr
 # CHECK:   t2Bcc %bb.3, 1, $cpsr
 # CHECK:   tB %bb.2, 14, $noreg
 # CHECK: bb.2.for.cond.cleanup:
 # CHECK: bb.3.for.header:
 
 --- |  
   define void @size_limit(i32* nocapture %a, i32* nocapture readonly %b, i32* nocapture readonly %c, i32 %N) {
   entry:

llvm/trunk/test/CodeGen/Thumb2/LowOverheadLoops/revert-non-loop.mir

 # RUN: llc -mtriple=thumbv8.1m.main -run-pass=arm-low-overhead-loops %s --verify-machineinstrs -o - | FileCheck %s
 
 # CHECK: name: non_loop
 # CHECK: bb.0.entry:
 # CHECK:   tBcc %bb.2, 3
 # CHECK: bb.1.not.preheader:
-# CHECK:   t2CMPri $lr, 0, 14
+# CHECK:   t2CMPri renamable $lr, 0, 14
 # CHECK:   t2Bcc %bb.4, 0
 # CHECK:   tB %bb.2
 # CHECK: bb.3.while.body:
 # CHECK:   t2CMPri $lr, 0, 14
 # CHECK:   t2Bcc %bb.3, 1
 # CHECK:   tB %bb.4
 # CHECK: bb.4.while.end:
 

llvm/trunk/test/CodeGen/Thumb2/LowOverheadLoops/revert-while.mir

 # RUN: llc -mtriple=thumbv8.1m.main -mattr=+lob -run-pass=arm-low-overhead-loops --verify-machineinstrs %s -o - | FileCheck %s
 # CHECK:      body:
 # CHECK:      bb.0.entry:
-# CHECK:        t2CMPri $lr, 0, 14
+# CHECK:        t2CMPri $r3, 0, 14
 # CHECK-NEXT:   t2Bcc %bb.3, 0, $cpsr
 # CHECK-NEXT:   tB %bb.1
 # CHECK:      bb.1.do.body.preheader:
 # CHECK:        $lr = tMOVr killed $r3
 # CHECK:      bb.2.do.body:
 # CHECK:        $lr = t2SUBri killed renamable $lr, 1, 14
-# CHECK-NEXT:   t2CMPri $lr, 0, 14, $cpsr
+# CHECK-NEXT:   t2CMPri $lr, 0, 14, $noreg, implicit-def $cpsr
 # CHECK-NEXT:   t2Bcc %bb.2, 1, $cpsr
 # CHECK-NEXT:   tB %bb.3, 14
 
 --- |
   define void @ne_trip_count(i1 zeroext %t1, i32* nocapture %a, i32* nocapture readonly %b, i32 %N) #0 {
   entry:
     %0 = call i1 @llvm.test.set.loop.iterations.i32(i32 %N)
     br i1 %0, label %do.body.preheader, label %if.end