ARM: allow copying of CPSR when all else fails.

As with x86 and AArch64, certain situations can arise where we need to spill
CPSR in the middle of a calculation. These should be avoided where possible
(MRS/MSR is rather expensive), which ARM is actually better at than the other
two since it tries to Glue defs to uses, but as a last ditch effort, copying is
better than crashing.

rdar://problem/18011155

llvm-svn: 218789

llvm/lib/Target/ARM/ARMBaseInstrInfo.cpp

@@ -698,6 +698,49 @@ unsigned ARMBaseInstrInfo::getInstBundleLength(const MachineInstr *MI) const {
   return Size;
 }
 
+void ARMBaseInstrInfo::copyFromCPSR(MachineBasicBlock &MBB,
+                                    MachineBasicBlock::iterator I,
+                                    unsigned DestReg, bool KillSrc,
+                                    const ARMSubtarget &Subtarget) const {
+  unsigned Opc = Subtarget.isThumb()
+                     ? (Subtarget.isMClass() ? ARM::t2MRS_M : ARM::t2MRS_AR)
+                     : ARM::MRS;
+
+  MachineInstrBuilder MIB =
+      BuildMI(MBB, I, I->getDebugLoc(), get(Opc), DestReg);
+
+  // There is only 1 A/R class MRS instruction, and it always refers to
+  // APSR. However, there are lots of other possibilities on M-class cores.
+  if (Subtarget.isMClass())
+    MIB.addImm(0x800);
+
+  AddDefaultPred(MIB);
+
+  MIB.addReg(ARM::CPSR, RegState::Implicit | getKillRegState(KillSrc));
+}
+
+void ARMBaseInstrInfo::copyToCPSR(MachineBasicBlock &MBB,
+                                  MachineBasicBlock::iterator I,
+                                  unsigned SrcReg, bool KillSrc,
+                                  const ARMSubtarget &Subtarget) const {
+  unsigned Opc = Subtarget.isThumb()
+                     ? (Subtarget.isMClass() ? ARM::t2MSR_M : ARM::t2MSR_AR)
+                     : ARM::MSR;
+
+  MachineInstrBuilder MIB = BuildMI(MBB, I, I->getDebugLoc(), get(Opc));
+
+  if (Subtarget.isMClass())
+    MIB.addImm(0x800);
+  else
+    MIB.addImm(8);
+
+  MIB.addReg(SrcReg, getKillRegState(KillSrc));
+
+  AddDefaultPred(MIB);
+
+  MIB.addReg(ARM::CPSR, RegState::Implicit | RegState::Define);
+}
+
 void ARMBaseInstrInfo::copyPhysReg(MachineBasicBlock &MBB,
                                    MachineBasicBlock::iterator I, DebugLoc DL,
                                    unsigned DestReg, unsigned SrcReg,
@@ -785,6 +828,12 @@ void ARMBaseInstrInfo::copyPhysReg(MachineBasicBlock &MBB,
     Opc = ARM::VMOVS;
     BeginIdx = ARM::ssub_0;
     SubRegs = 2;
+  } else if (SrcReg == ARM::CPSR) {
+    copyFromCPSR(MBB, I, DestReg, KillSrc, Subtarget);
+    return;
+  } else if (DestReg == ARM::CPSR) {
+    copyToCPSR(MBB, I, SrcReg, KillSrc, Subtarget);
+    return;
   }
 
   assert(Opc && "Impossible reg-to-reg copy");

llvm/lib/Target/ARM/ARMBaseInstrInfo.h

@@ -156,6 +156,13 @@ public:
   unsigned isStoreToStackSlotPostFE(const MachineInstr *MI,
                                     int &FrameIndex) const override;
 
+  void copyToCPSR(MachineBasicBlock &MBB, MachineBasicBlock::iterator I,
+                  unsigned SrcReg, bool KillSrc,
+                  const ARMSubtarget &Subtarget) const;
+  void copyFromCPSR(MachineBasicBlock &MBB, MachineBasicBlock::iterator I,
+                    unsigned DestReg, bool KillSrc,
+                    const ARMSubtarget &Subtarget) const;
+
   void copyPhysReg(MachineBasicBlock &MBB, MachineBasicBlock::iterator I,
                    DebugLoc DL, unsigned DestReg, unsigned SrcReg,
                    bool KillSrc) const override;

llvm/lib/Target/ARM/ARMBaseRegisterInfo.cpp

@@ -182,7 +182,7 @@ ARMBaseRegisterInfo::getPointerRegClass(const MachineFunction &MF, unsigned Kind
 const TargetRegisterClass *
 ARMBaseRegisterInfo::getCrossCopyRegClass(const TargetRegisterClass *RC) const {
   if (RC == &ARM::CCRRegClass)
-    return nullptr;  // Can't copy CCR registers.
+    return &ARM::rGPRRegClass;  // Can't copy CCR registers.
   return RC;
 }
 

llvm/test/CodeGen/ARM/copy-cpsr.ll

@@ -0,0 +1,41 @@
+; RUN: llc -mtriple=armv7s-apple-ios7.0 -show-mc-encoding %s -o - | FileCheck %s --check-prefix=CHECK-ARM
+; RUN: llc -mtriple=thumbv7s-apple-ios7.0 -show-mc-encoding %s -o - | FileCheck %s --check-prefix=CHECK-THUMB
+; RUN: llc -mtriple=thumbv7m-none-eabi -show-mc-encoding %s -o - | FileCheck %s --check-prefix=CHECK-THUMB
+
+; In the ARM backend, most compares are glued to their uses so CPSR can't
+; escape. However, for long ADCS chains (and last ditch fallback) the dependency
+; is carried in the DAG because duplicating them can be more expensive than
+; copying CPSR.
+
+; Crafting a test for this was a little tricky, in case it breaks here are some
+; notes on what I was tring to achieve:
+;   + We want 2 long ADCS chains
+;   + We want them to split after an initial common prefix (so that a single
+;     CPSR is used twice).
+;   + We want both chains to write CPSR post-split (so that the copy can't be
+;     elided).
+;   + We want the chains to be long enough that duplicating them is expensive.
+
+define void @test_copy_cpsr(i128 %lhs, i128 %rhs, i128* %addr) {
+; CHECK-ARM: test_copy_cpsr:
+; CHECK-THUMB: test_copy_cpsr:
+
+; CHECK-ARM: mrs [[TMP:r[0-9]+]], apsr @ encoding: [0x00,0x{{[0-9a-f]}}0,0x0f,0xe1]
+; CHECK-ARM: msr APSR_nzcvq, [[TMP]] @ encoding: [0x0{{[0-9a-f]}},0xf0,0x28,0xe1]
+
+  ; In Thumb mode v7M and v7AR have different MRS/MSR instructions that happen
+  ; to overlap for the apsr case, so it's definitely worth checking both.
+; CHECK-THUMB: mrs [[TMP:r[0-9]+]], apsr @ encoding: [0xef,0xf3,0x00,0x8{{[0-9a-f]}}]
+; CHECK-THUMB: msr {{APSR|apsr}}_nzcvq, [[TMP]] @ encoding: [0x8{{[0-9a-f]}},0xf3,0x00,0x88]
+
+  %sum = add i128 %lhs, %rhs
+  store volatile i128 %sum, i128* %addr
+
+  %rhs2.tmp1 = trunc i128 %rhs to i64
+  %rhs2 = zext i64 %rhs2.tmp1 to i128
+
+  %sum2 = add i128 %lhs, %rhs2
+  store volatile i128 %sum2, i128* %addr
+
+  ret void
+}