[NVPTX] do not rely on cached subtarget info.

[tra]

Summary

If a module has function references, but no functions
themselves, we may end up never calling runOnMachineFunction
and therefore would never initialize nvptxSubtarget field
which would eventually cause a crash.

Instead of relying on nvptxSubtarget being initialized by
one of the methods, retrieve subtarget info directly.

@echristo : Added you as a reviewer as you were refactoring this code some time back (rL229785)

[echristo]

LGTM. Thanks!

llvm/trunk/lib/Target/NVPTX/NVPTXAsmPrinter.h

private:
   const MachineRegisterInfo *MRI;
   // The contents are specific for each
   // MachineFunction. But the size of the
   // array is not.
   typedef DenseMap<unsigned, unsigned> VRegMap;
   typedef DenseMap<const TargetRegisterClass *, VRegMap> VRegRCMap;
   VRegRCMap VRegMapping;
 
-  // Cache the subtarget here.
-  const NVPTXSubtarget *nvptxSubtarget;
-
   // List of variables demoted to a function scope.
   std::map<const Function *, std::vector<const GlobalVariable *>> localDecls;
 
   void emitPTXGlobalVariable(const GlobalVariable *GVar, raw_ostream &O);
   void emitPTXAddressSpace(unsigned int AddressSpace, raw_ostream &O) const;
   std::string getPTXFundamentalTypeStr(Type *Ty, bool = true) const;
   void printScalarConstant(const Constant *CPV, raw_ostream &O);
   void printFPConstant(const ConstantFP *Fp, raw_ostream &O);

llvm/trunk/lib/Target/NVPTX/NVPTXAsmPrinter.cpp

void NVPTXAsmPrinter::lowerToMCInst(const MachineInstr *MI, MCInst &OutMI) {
   // Special: Do not mangle symbol operand of CALL_PROTOTYPE
   if (MI->getOpcode() == NVPTX::CALL_PROTOTYPE) {
     const MachineOperand &MO = MI->getOperand(0);
     OutMI.addOperand(GetSymbolRef(
       OutContext.getOrCreateSymbol(Twine(MO.getSymbolName()))));
     return;
   }
 
+  const NVPTXSubtarget &STI = MI->getMF()->getSubtarget<NVPTXSubtarget>();
   for (unsigned i = 0, e = MI->getNumOperands(); i != e; ++i) {
     const MachineOperand &MO = MI->getOperand(i);
 
     MCOperand MCOp;
-    if (!nvptxSubtarget->hasImageHandles()) {
+    if (!STI.hasImageHandles()) {
       if (lowerImageHandleOperand(MI, i, MCOp)) {
         OutMI.addOperand(MCOp);
         continue;
       }
     }
 
     if (lowerOperand(MO, MCOp))
       OutMI.addOperand(MCOp);
MCOperand NVPTXAsmPrinter::GetSymbolRef(const MCSymbol *Symbol) {
   const MCExpr *Expr;
   Expr = MCSymbolRefExpr::create(Symbol, MCSymbolRefExpr::VK_None,
                                  OutContext);
   return MCOperand::createExpr(Expr);
 }
 
 void NVPTXAsmPrinter::printReturnValStr(const Function *F, raw_ostream &O) {
   const DataLayout &DL = getDataLayout();
-  const TargetLowering *TLI = nvptxSubtarget->getTargetLowering();
+  const NVPTXSubtarget &STI = TM.getSubtarget<NVPTXSubtarget>(*F);
+  const TargetLowering *TLI = STI.getTargetLowering();
 
   Type *Ty = F->getReturnType();
 
-  bool isABI = (nvptxSubtarget->getSmVersion() >= 20);
+  bool isABI = (STI.getSmVersion() >= 20);
 
   if (Ty->getTypeID() == Type::VoidTyID)
     return;
 
   O << " (";
 
   if (isABI) {
     if (Ty->isFloatingPointTy() || (Ty->isIntegerTy() && !Ty->isIntegerTy(128))) {
void NVPTXAsmPrinter::EmitFunctionEntryLabel() {
 
   VRegMapping.clear();
   // Emit open brace for function body.
   OutStreamer->EmitRawText(StringRef("{\n"));
   setAndEmitFunctionVirtualRegisters(*MF);
 }
 
 bool NVPTXAsmPrinter::runOnMachineFunction(MachineFunction &F) {
-  nvptxSubtarget = &F.getSubtarget<NVPTXSubtarget>();
   bool Result = AsmPrinter::runOnMachineFunction(F);
   // Emit closing brace for the body of function F.
   // The closing brace must be emitted here because we need to emit additional
   // debug labels/data after the last basic block.
   // We need to emit the closing brace here because we don't have function that
   // finished emission of the function body.
   OutStreamer->EmitRawText(StringRef("}\n"));
   return Result;

 }
 
 void NVPTXAsmPrinter::emitImplicitDef(const MachineInstr *MI) const {
   unsigned RegNo = MI->getOperand(0).getReg();
   if (TargetRegisterInfo::isVirtualRegister(RegNo)) {
     OutStreamer->AddComment(Twine("implicit-def: ") +
                             getVirtualRegisterName(RegNo));
   } else {
+    const NVPTXSubtarget &STI = MI->getMF()->getSubtarget<NVPTXSubtarget>();
     OutStreamer->AddComment(Twine("implicit-def: ") +
-                            nvptxSubtarget->getRegisterInfo()->getName(RegNo));
+                            STI.getRegisterInfo()->getName(RegNo));
   }
   OutStreamer->AddBlankLine();
 }
 
 void NVPTXAsmPrinter::emitKernelFunctionDirectives(const Function &F,
                                                    raw_ostream &O) const {
   // If the NVVM IR has some of reqntid* specified, then output
   // the reqntid directive, and set the unspecified ones to 1.
void NVPTXAsmPrinter::printParamName(Function::const_arg_iterator I,
                                      int paramIndex, raw_ostream &O) {
   getSymbol(I->getParent())->print(O, MAI);
   O << "_param_" << paramIndex;
 }
 
 void NVPTXAsmPrinter::emitFunctionParamList(const Function *F, raw_ostream &O) {
   const DataLayout &DL = getDataLayout();
   const AttributeList &PAL = F->getAttributes();
-  const TargetLowering *TLI = nvptxSubtarget->getTargetLowering();
+  const NVPTXSubtarget &STI = TM.getSubtarget<NVPTXSubtarget>(*F);
+  const TargetLowering *TLI = STI.getTargetLowering();
   Function::const_arg_iterator I, E;
   unsigned paramIndex = 0;
   bool first = true;
   bool isKernelFunc = isKernelFunction(*F);
-  bool isABI = (nvptxSubtarget->getSmVersion() >= 20);
+  bool isABI = (STI.getSmVersion() >= 20);
+  bool hasImageHandles = STI.hasImageHandles();
   MVT thePointerTy = TLI->getPointerTy(DL);
 
   if (F->arg_empty()) {
     O << "()\n";
     return;
   }
 
   O << "(\n";
 
   for (I = F->arg_begin(), E = F->arg_end(); I != E; ++I, paramIndex++) {
     Type *Ty = I->getType();
 
     if (!first)
       O << ",\n";
 
     first = false;
 
     // Handle image/sampler parameters
     if (isKernelFunction(*F)) {
       if (isSampler(*I) || isImage(*I)) {
         if (isImage(*I)) {
           std::string sname = I->getName();
           if (isImageWriteOnly(*I) || isImageReadWrite(*I)) {
-            if (nvptxSubtarget->hasImageHandles())
+            if (hasImageHandles)
               O << "\t.param .u64 .ptr .surfref ";
             else
               O << "\t.param .surfref ";
             CurrentFnSym->print(O, MAI);
             O << "_param_" << paramIndex;
           }
           else { // Default image is read_only
-            if (nvptxSubtarget->hasImageHandles())
+            if (hasImageHandles)
               O << "\t.param .u64 .ptr .texref ";
             else
               O << "\t.param .texref ";
             CurrentFnSym->print(O, MAI);
             O << "_param_" << paramIndex;
           }
         } else {
-          if (nvptxSubtarget->hasImageHandles())
+          if (hasImageHandles)
             O << "\t.param .u64 .ptr .samplerref ";
           else
             O << "\t.param .samplerref ";
           CurrentFnSym->print(O, MAI);
           O << "_param_" << paramIndex;
         }
         continue;
       }

llvm/trunk/test/CodeGen/NVPTX/nofunc.ll

+; RUN: llc < %s -march=nvptx -mcpu=sm_20 | FileCheck %s
+; RUN: llc < %s -march=nvptx64 -mcpu=sm_20 | FileCheck %s
+
+; Test that we don't crash if we're compiling a module with function references,
+; but without any functions in it.
+
+target datalayout = "e-i64:64-i128:128-v16:16-v32:32-n16:32:64"
+target triple = "nvptx64-nvidia-cuda"
+
+@Funcs = local_unnamed_addr addrspace(1) externally_initialized
+         global [1 x void (i8*)*] [void (i8*)* @func], align 8
+
+declare void @func(i8*)
+
+; CHECK: Funcs[1] = {func}