[ImplicitNullChecks] Work with implicit defs.

[sanjoy]

Summary

This change generalizes the implicit null checks pass to work with
instructions that don't have any explicit register defs. This lets us
use X86's cmp against memory as faulting load instructions.

[JosephTremoulet]

LGTM

lib/CodeGen/ImplicitNullChecks.cpp

bool ImplicitNullChecks::analyzeBlockForNullChecks(
   };
 
   for (auto MII = NotNullSucc->begin(), MIE = NotNullSucc->end(); MII != MIE;
        ++MII) {
     MachineInstr *MI = &*MII;
     unsigned BaseReg, Offset;
     if (TII->getMemOpBaseRegImmOfs(MI, BaseReg, Offset, TRI))
       if (MI->mayLoad() && !MI->isPredicable() && BaseReg == PointerReg &&
-          Offset < PageSize && MI->getDesc().getNumDefs() == 1 &&
+          Offset < PageSize && MI->getDesc().getNumDefs() <= 1 &&
           IsSafeToHoist(MI)) {
         NullCheckList.emplace_back(MI, MBP.ConditionDef, &MBB, NotNullSucc,
                                    NullSucc);
         return true;
       }
 
     // MI did not match our criteria for conversion to a trapping load.  Check
     // if we can continue looking.

 
 /// Wrap a machine load instruction, LoadMI, into a FAULTING_LOAD_OP machine
 /// instruction.  The FAULTING_LOAD_OP instruction does the same load as LoadMI
 /// (defining the same register), and branches to HandlerLabel if the load
 /// faults.  The FAULTING_LOAD_OP instruction is inserted at the end of MBB.
 MachineInstr *ImplicitNullChecks::insertFaultingLoad(MachineInstr *LoadMI,
                                                      MachineBasicBlock *MBB,
                                                      MCSymbol *HandlerLabel) {
+  const unsigned NoRegister = 0; // Guaranteed to be the NoRegister value for
+                                 // all targets.
+
   DebugLoc DL;
   unsigned NumDefs = LoadMI->getDesc().getNumDefs();
-  assert(NumDefs == 1 && "other cases unhandled!");
-  (void)NumDefs;
+  assert(NumDefs <= 1 && "other cases unhandled!");
 
-  unsigned DefReg = LoadMI->defs().begin()->getReg();
+  unsigned DefReg = NoRegister;
+  if (NumDefs != 0) {
+    DefReg = LoadMI->defs().begin()->getReg();
     assert(std::distance(LoadMI->defs().begin(), LoadMI->defs().end()) == 1 &&
            "expected exactly one def!");
+  }
 
   auto MIB = BuildMI(MBB, DL, TII->get(TargetOpcode::FAULTING_LOAD_OP), DefReg)
                  .addSym(HandlerLabel)
                  .addImm(LoadMI->getOpcode());
 
   for (auto &MO : LoadMI->uses())
     MIB.addOperand(MO);
 

lib/Target/X86/X86MCInstLower.cpp

void X86AsmPrinter::LowerFAULTING_LOAD_OP(const MachineInstr &MI,
   MCSymbol *HandlerLabel = MI.getOperand(1).getMCSymbol();
   unsigned LoadOpcode = MI.getOperand(2).getImm();
   unsigned LoadOperandsBeginIdx = 3;
 
   FM.recordFaultingOp(FaultMaps::FaultingLoad, HandlerLabel);
 
   MCInst LoadMI;
   LoadMI.setOpcode(LoadOpcode);
+
+  if (LoadDefRegister != X86::NoRegister)
     LoadMI.addOperand(MCOperand::createReg(LoadDefRegister));
+
   for (auto I = MI.operands_begin() + LoadOperandsBeginIdx,
             E = MI.operands_end();
        I != E; ++I)
     if (auto MaybeOperand = MCIL.LowerMachineOperand(&MI, *I))
       LoadMI.addOperand(MaybeOperand.getValue());
 
   OutStreamer->EmitInstruction(LoadMI, getSubtargetInfo());
 }

test/CodeGen/X86/implicit-null-check.ll

 
  not_null:
   %t0 = load i32, i32* %y
   %t1 = load i32, i32* %x
   store i32 %t0, i32* %z
   ret i32 %t1
 }
 
+define i32 @imp_null_check_via_mem_comparision(i32* %x, i32 %val) {
+; CHECK-LABEL: _imp_null_check_via_mem_comparision
+; CHECK: Ltmp9:
+; CHECK: cmpl   %esi, 4(%rdi)
+; CHECK: jge    LBB4_2
+; CHECK: movl   $100, %eax
+; CHECK: retq
+; CHECK: Ltmp8:
+; CHECK: movl   $42, %eax
+; CHECK: retq
+; CHECK: LBB4_2:
+; CHECK: movl   $200, %eax
+; CHECK: retq
+
+ entry:
+  %c = icmp eq i32* %x, null
+  br i1 %c, label %is_null, label %not_null, !make.implicit !0
+
+ is_null:
+  ret i32 42
+
+ not_null:
+  %x.loc = getelementptr i32, i32* %x, i32 1
+  %t = load i32, i32* %x.loc
+  %m = icmp slt i32 %t, %val
+  br i1 %m, label %ret_100, label %ret_200
+
+ ret_100:
+  ret i32 100
+
+ ret_200:
+  ret i32 200
+}
+
 !0 = !{}
 
 ; CHECK-LABEL: __LLVM_FaultMaps:
 
 ; Version:
 ; CHECK-NEXT: .byte 1
 
 ; Reserved x2
 ; CHECK-NEXT: .byte 0
 ; CHECK-NEXT: .short 0
 
 ; # functions:
-; CHECK-NEXT: .long 4
+; CHECK-NEXT: .long 5
 
 ; FunctionAddr:
 ; CHECK-NEXT: .quad _imp_null_check_add_result
 ; NumFaultingPCs
 ; CHECK-NEXT: .long 1
 ; Reserved:
 ; CHECK-NEXT: .long 0
 ; Fault[0].Type:

 ; CHECK-NEXT: .long 0
 ; Fault[0].Type:
 ; CHECK-NEXT: .long 1
 ; Fault[0].FaultOffset:
 ; CHECK-NEXT: .long Ltmp1-_imp_null_check_load
 ; Fault[0].HandlerOffset:
 ; CHECK-NEXT: .long Ltmp0-_imp_null_check_load
 
+; FunctionAddr:
+; CHECK-NEXT: .quad     _imp_null_check_via_mem_comparision
+; NumFaultingPCs
+; CHECK-NEXT: .long   1
+; Reserved:
+; CHECK-NEXT: .long   0
+; Fault[0].Type:
+; CHECK-NEXT: .long   1
+; Fault[0].FaultOffset:
+; CHECK-NEXT: .long   Ltmp9-_imp_null_check_via_mem_comparision
+; Fault[0].HandlerOffset:
+; CHECK-NEXT: .long   Ltmp8-_imp_null_check_via_mem_comparision
+
 ; OBJDUMP: FaultMap table:
 ; OBJDUMP-NEXT: Version: 0x1
-; OBJDUMP-NEXT: NumFunctions: 4
+; OBJDUMP-NEXT: NumFunctions: 5
 ; OBJDUMP-NEXT: FunctionAddress: 0x000000, NumFaultingPCs: 1
 ; OBJDUMP-NEXT: Fault kind: FaultingLoad, faulting PC offset: 0, handling PC offset: 5
 ; OBJDUMP-NEXT: FunctionAddress: 0x000000, NumFaultingPCs: 1
 ; OBJDUMP-NEXT: Fault kind: FaultingLoad, faulting PC offset: 0, handling PC offset: 7
 ; OBJDUMP-NEXT: FunctionAddress: 0x000000, NumFaultingPCs: 1
 ; OBJDUMP-NEXT: Fault kind: FaultingLoad, faulting PC offset: 0, handling PC offset: 7
 ; OBJDUMP-NEXT: FunctionAddress: 0x000000, NumFaultingPCs: 1
 ; OBJDUMP-NEXT: Fault kind: FaultingLoad, faulting PC offset: 0, handling PC offset: 3