xref: /freebsd/contrib/llvm-project/llvm/lib/Target/AMDGPU/AMDGPUPrintfRuntimeBinding.cpp (revision 65f8467e3351c38a9d57b538b85cf6c5fab5818e)
1 //=== AMDGPUPrintfRuntimeBinding.cpp - OpenCL printf implementation -------===//
2 //
3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4 // See https://llvm.org/LICENSE.txt for license information.
5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6 //
7 //===----------------------------------------------------------------------===//
8 // \file
9 //
10 // The pass bind printfs to a kernel arg pointer that will be bound to a buffer
11 // later by the runtime.
12 //
13 // This pass traverses the functions in the module and converts
14 // each call to printf to a sequence of operations that
15 // store the following into the printf buffer:
16 // - format string (passed as a module's metadata unique ID)
17 // - bitwise copies of printf arguments
18 // The backend passes will need to store metadata in the kernel
19 //===----------------------------------------------------------------------===//
20 
21 #include "AMDGPU.h"
22 #include "llvm/ADT/Triple.h"
23 #include "llvm/Analysis/InstructionSimplify.h"
24 #include "llvm/Analysis/TargetLibraryInfo.h"
25 #include "llvm/Analysis/ValueTracking.h"
26 #include "llvm/IR/DiagnosticInfo.h"
27 #include "llvm/IR/Dominators.h"
28 #include "llvm/IR/IRBuilder.h"
29 #include "llvm/IR/Instructions.h"
30 #include "llvm/InitializePasses.h"
31 #include "llvm/Support/DataExtractor.h"
32 #include "llvm/Transforms/Utils/BasicBlockUtils.h"
33 
34 using namespace llvm;
35 
36 #define DEBUG_TYPE "printfToRuntime"
37 #define DWORD_ALIGN 4
38 
39 namespace {
40 class AMDGPUPrintfRuntimeBinding final : public ModulePass {
41 
42 public:
43   static char ID;
44 
45   explicit AMDGPUPrintfRuntimeBinding();
46 
47 private:
48   bool runOnModule(Module &M) override;
49 
50   void getAnalysisUsage(AnalysisUsage &AU) const override {
51     AU.addRequired<TargetLibraryInfoWrapperPass>();
52     AU.addRequired<DominatorTreeWrapperPass>();
53   }
54 };
55 
56 class AMDGPUPrintfRuntimeBindingImpl {
57 public:
58   AMDGPUPrintfRuntimeBindingImpl(
59       function_ref<const DominatorTree &(Function &)> GetDT,
60       function_ref<const TargetLibraryInfo &(Function &)> GetTLI)
61       : GetDT(GetDT), GetTLI(GetTLI) {}
62   bool run(Module &M);
63 
64 private:
65   void getConversionSpecifiers(SmallVectorImpl<char> &OpConvSpecifiers,
66                                StringRef fmt, size_t num_ops) const;
67 
68   bool lowerPrintfForGpu(Module &M);
69 
70   Value *simplify(Instruction *I, const TargetLibraryInfo *TLI,
71                   const DominatorTree *DT) {
72     return simplifyInstruction(I, {*TD, TLI, DT});
73   }
74 
75   const DataLayout *TD;
76   function_ref<const DominatorTree &(Function &)> GetDT;
77   function_ref<const TargetLibraryInfo &(Function &)> GetTLI;
78   SmallVector<CallInst *, 32> Printfs;
79 };
80 } // namespace
81 
82 char AMDGPUPrintfRuntimeBinding::ID = 0;
83 
84 INITIALIZE_PASS_BEGIN(AMDGPUPrintfRuntimeBinding,
85                       "amdgpu-printf-runtime-binding", "AMDGPU Printf lowering",
86                       false, false)
87 INITIALIZE_PASS_DEPENDENCY(TargetLibraryInfoWrapperPass)
88 INITIALIZE_PASS_DEPENDENCY(DominatorTreeWrapperPass)
89 INITIALIZE_PASS_END(AMDGPUPrintfRuntimeBinding, "amdgpu-printf-runtime-binding",
90                     "AMDGPU Printf lowering", false, false)
91 
92 char &llvm::AMDGPUPrintfRuntimeBindingID = AMDGPUPrintfRuntimeBinding::ID;
93 
94 namespace llvm {
95 ModulePass *createAMDGPUPrintfRuntimeBinding() {
96   return new AMDGPUPrintfRuntimeBinding();
97 }
98 } // namespace llvm
99 
100 AMDGPUPrintfRuntimeBinding::AMDGPUPrintfRuntimeBinding() : ModulePass(ID) {
101   initializeAMDGPUPrintfRuntimeBindingPass(*PassRegistry::getPassRegistry());
102 }
103 
104 void AMDGPUPrintfRuntimeBindingImpl::getConversionSpecifiers(
105     SmallVectorImpl<char> &OpConvSpecifiers, StringRef Fmt,
106     size_t NumOps) const {
107   // not all format characters are collected.
108   // At this time the format characters of interest
109   // are %p and %s, which use to know if we
110   // are either storing a literal string or a
111   // pointer to the printf buffer.
112   static const char ConvSpecifiers[] = "cdieEfgGaosuxXp";
113   size_t CurFmtSpecifierIdx = 0;
114   size_t PrevFmtSpecifierIdx = 0;
115 
116   while ((CurFmtSpecifierIdx = Fmt.find_first_of(
117               ConvSpecifiers, CurFmtSpecifierIdx)) != StringRef::npos) {
118     bool ArgDump = false;
119     StringRef CurFmt = Fmt.substr(PrevFmtSpecifierIdx,
120                                   CurFmtSpecifierIdx - PrevFmtSpecifierIdx);
121     size_t pTag = CurFmt.find_last_of("%");
122     if (pTag != StringRef::npos) {
123       ArgDump = true;
124       while (pTag && CurFmt[--pTag] == '%') {
125         ArgDump = !ArgDump;
126       }
127     }
128 
129     if (ArgDump)
130       OpConvSpecifiers.push_back(Fmt[CurFmtSpecifierIdx]);
131 
132     PrevFmtSpecifierIdx = ++CurFmtSpecifierIdx;
133   }
134 }
135 
136 static bool shouldPrintAsStr(char Specifier, Type *OpType) {
137   return Specifier == 's' && isa<PointerType>(OpType);
138 }
139 
140 constexpr StringLiteral NonLiteralStr("???");
141 static_assert(NonLiteralStr.size() == 3);
142 
143 static StringRef getAsConstantStr(Value *V) {
144   StringRef S;
145   if (!getConstantStringInfo(V, S))
146     S = NonLiteralStr;
147 
148   return S;
149 }
150 
151 static void diagnoseInvalidFormatString(const CallBase *CI) {
152   DiagnosticInfoUnsupported UnsupportedFormatStr(
153       *CI->getParent()->getParent(),
154       "printf format string must be a trivially resolved constant string "
155       "global variable",
156       CI->getDebugLoc());
157   CI->getContext().diagnose(UnsupportedFormatStr);
158 }
159 
160 bool AMDGPUPrintfRuntimeBindingImpl::lowerPrintfForGpu(Module &M) {
161   LLVMContext &Ctx = M.getContext();
162   IRBuilder<> Builder(Ctx);
163   Type *I32Ty = Type::getInt32Ty(Ctx);
164 
165   // Instead of creating global variables, the printf format strings are
166   // extracted and passed as metadata. This avoids polluting llvm's symbol
167   // tables in this module. Metadata is going to be extracted by the backend
168   // passes and inserted into the OpenCL binary as appropriate.
169   NamedMDNode *metaD = M.getOrInsertNamedMetadata("llvm.printf.fmts");
170   unsigned UniqID = metaD->getNumOperands();
171 
172   for (auto *CI : Printfs) {
173     unsigned NumOps = CI->arg_size();
174 
175     SmallString<16> OpConvSpecifiers;
176     Value *Op = CI->getArgOperand(0);
177 
178     if (auto LI = dyn_cast<LoadInst>(Op)) {
179       Op = LI->getPointerOperand();
180       for (auto *Use : Op->users()) {
181         if (auto SI = dyn_cast<StoreInst>(Use)) {
182           Op = SI->getValueOperand();
183           break;
184         }
185       }
186     }
187 
188     if (auto I = dyn_cast<Instruction>(Op)) {
189       Value *Op_simplified =
190           simplify(I, &GetTLI(*I->getFunction()), &GetDT(*I->getFunction()));
191       if (Op_simplified)
192         Op = Op_simplified;
193     }
194 
195     StringRef FormatStr;
196     if (!getConstantStringInfo(Op, FormatStr)) {
197       Value *Stripped = Op->stripPointerCasts();
198       if (!isa<UndefValue>(Stripped) && !isa<ConstantPointerNull>(Stripped))
199         diagnoseInvalidFormatString(CI);
200       continue;
201     }
202 
203     // We need this call to ascertain that we are printing a string or a
204     // pointer. It takes out the specifiers and fills up the first arg.
205     getConversionSpecifiers(OpConvSpecifiers, FormatStr, NumOps - 1);
206 
207     // Add metadata for the string
208     std::string AStreamHolder;
209     raw_string_ostream Sizes(AStreamHolder);
210     int Sum = DWORD_ALIGN;
211     Sizes << CI->arg_size() - 1;
212     Sizes << ':';
213     for (unsigned ArgCount = 1;
214          ArgCount < CI->arg_size() && ArgCount <= OpConvSpecifiers.size();
215          ArgCount++) {
216       Value *Arg = CI->getArgOperand(ArgCount);
217       Type *ArgType = Arg->getType();
218       unsigned ArgSize = TD->getTypeAllocSize(ArgType);
219       //
220       // ArgSize by design should be a multiple of DWORD_ALIGN,
221       // expand the arguments that do not follow this rule.
222       //
223       if (ArgSize % DWORD_ALIGN != 0) {
224         Type *ResType = Type::getInt32Ty(Ctx);
225         if (auto *VecType = dyn_cast<VectorType>(ArgType))
226           ResType = VectorType::get(ResType, VecType->getElementCount());
227         Builder.SetInsertPoint(CI);
228         Builder.SetCurrentDebugLocation(CI->getDebugLoc());
229 
230         if (ArgType->isFloatingPointTy()) {
231           Arg = Builder.CreateBitCast(
232               Arg,
233               IntegerType::getIntNTy(Ctx, ArgType->getPrimitiveSizeInBits()));
234         }
235 
236         if (OpConvSpecifiers[ArgCount - 1] == 'x' ||
237             OpConvSpecifiers[ArgCount - 1] == 'X' ||
238             OpConvSpecifiers[ArgCount - 1] == 'u' ||
239             OpConvSpecifiers[ArgCount - 1] == 'o')
240           Arg = Builder.CreateZExt(Arg, ResType);
241         else
242           Arg = Builder.CreateSExt(Arg, ResType);
243         ArgType = Arg->getType();
244         ArgSize = TD->getTypeAllocSize(ArgType);
245         CI->setOperand(ArgCount, Arg);
246       }
247       if (OpConvSpecifiers[ArgCount - 1] == 'f') {
248         ConstantFP *FpCons = dyn_cast<ConstantFP>(Arg);
249         if (FpCons)
250           ArgSize = 4;
251         else {
252           FPExtInst *FpExt = dyn_cast<FPExtInst>(Arg);
253           if (FpExt && FpExt->getType()->isDoubleTy() &&
254               FpExt->getOperand(0)->getType()->isFloatTy())
255             ArgSize = 4;
256         }
257       }
258       if (shouldPrintAsStr(OpConvSpecifiers[ArgCount - 1], ArgType))
259         ArgSize = alignTo(getAsConstantStr(Arg).size() + 1, 4);
260 
261       LLVM_DEBUG(dbgs() << "Printf ArgSize (in buffer) = " << ArgSize
262                         << " for type: " << *ArgType << '\n');
263       Sizes << ArgSize << ':';
264       Sum += ArgSize;
265     }
266     LLVM_DEBUG(dbgs() << "Printf format string in source = " << FormatStr
267                       << '\n');
268     for (char C : FormatStr) {
269       // Rest of the C escape sequences (e.g. \') are handled correctly
270       // by the MDParser
271       switch (C) {
272       case '\a':
273         Sizes << "\\a";
274         break;
275       case '\b':
276         Sizes << "\\b";
277         break;
278       case '\f':
279         Sizes << "\\f";
280         break;
281       case '\n':
282         Sizes << "\\n";
283         break;
284       case '\r':
285         Sizes << "\\r";
286         break;
287       case '\v':
288         Sizes << "\\v";
289         break;
290       case ':':
291         // ':' cannot be scanned by Flex, as it is defined as a delimiter
292         // Replace it with it's octal representation \72
293         Sizes << "\\72";
294         break;
295       default:
296         Sizes << C;
297         break;
298       }
299     }
300 
301     // Insert the printf_alloc call
302     Builder.SetInsertPoint(CI);
303     Builder.SetCurrentDebugLocation(CI->getDebugLoc());
304 
305     AttributeList Attr = AttributeList::get(Ctx, AttributeList::FunctionIndex,
306                                             Attribute::NoUnwind);
307 
308     Type *SizetTy = Type::getInt32Ty(Ctx);
309 
310     Type *Tys_alloc[1] = {SizetTy};
311     Type *I8Ty = Type::getInt8Ty(Ctx);
312     Type *I8Ptr = PointerType::get(I8Ty, 1);
313     FunctionType *FTy_alloc = FunctionType::get(I8Ptr, Tys_alloc, false);
314     FunctionCallee PrintfAllocFn =
315         M.getOrInsertFunction(StringRef("__printf_alloc"), FTy_alloc, Attr);
316 
317     LLVM_DEBUG(dbgs() << "Printf metadata = " << Sizes.str() << '\n');
318     std::string fmtstr = itostr(++UniqID) + ":" + Sizes.str();
319     MDString *fmtStrArray = MDString::get(Ctx, fmtstr);
320 
321     MDNode *myMD = MDNode::get(Ctx, fmtStrArray);
322     metaD->addOperand(myMD);
323     Value *sumC = ConstantInt::get(SizetTy, Sum, false);
324     SmallVector<Value *, 1> alloc_args;
325     alloc_args.push_back(sumC);
326     CallInst *pcall =
327         CallInst::Create(PrintfAllocFn, alloc_args, "printf_alloc_fn", CI);
328 
329     //
330     // Insert code to split basicblock with a
331     // piece of hammock code.
332     // basicblock splits after buffer overflow check
333     //
334     ConstantPointerNull *zeroIntPtr =
335         ConstantPointerNull::get(PointerType::get(I8Ty, 1));
336     auto *cmp = cast<ICmpInst>(Builder.CreateICmpNE(pcall, zeroIntPtr, ""));
337     if (!CI->use_empty()) {
338       Value *result =
339           Builder.CreateSExt(Builder.CreateNot(cmp), I32Ty, "printf_res");
340       CI->replaceAllUsesWith(result);
341     }
342     SplitBlock(CI->getParent(), cmp);
343     Instruction *Brnch =
344         SplitBlockAndInsertIfThen(cmp, cmp->getNextNode(), false);
345 
346     Builder.SetInsertPoint(Brnch);
347 
348     // store unique printf id in the buffer
349     //
350     GetElementPtrInst *BufferIdx = GetElementPtrInst::Create(
351         I8Ty, pcall, ConstantInt::get(Ctx, APInt(32, 0)), "PrintBuffID", Brnch);
352 
353     Type *idPointer = PointerType::get(I32Ty, AMDGPUAS::GLOBAL_ADDRESS);
354     Value *id_gep_cast =
355         new BitCastInst(BufferIdx, idPointer, "PrintBuffIdCast", Brnch);
356 
357     new StoreInst(ConstantInt::get(I32Ty, UniqID), id_gep_cast, Brnch);
358 
359     // 1st 4 bytes hold the printf_id
360     // the following GEP is the buffer pointer
361     BufferIdx = GetElementPtrInst::Create(I8Ty, pcall,
362                                           ConstantInt::get(Ctx, APInt(32, 4)),
363                                           "PrintBuffGep", Brnch);
364 
365     Type *Int32Ty = Type::getInt32Ty(Ctx);
366     for (unsigned ArgCount = 1;
367          ArgCount < CI->arg_size() && ArgCount <= OpConvSpecifiers.size();
368          ArgCount++) {
369       Value *Arg = CI->getArgOperand(ArgCount);
370       Type *ArgType = Arg->getType();
371       SmallVector<Value *, 32> WhatToStore;
372       if (ArgType->isFPOrFPVectorTy() && !isa<VectorType>(ArgType)) {
373         if (OpConvSpecifiers[ArgCount - 1] == 'f') {
374           if (auto *FpCons = dyn_cast<ConstantFP>(Arg)) {
375             APFloat Val(FpCons->getValueAPF());
376             bool Lost = false;
377             Val.convert(APFloat::IEEEsingle(), APFloat::rmNearestTiesToEven,
378                         &Lost);
379             Arg = ConstantFP::get(Ctx, Val);
380           } else if (auto *FpExt = dyn_cast<FPExtInst>(Arg)) {
381             if (FpExt->getType()->isDoubleTy() &&
382                 FpExt->getOperand(0)->getType()->isFloatTy()) {
383               Arg = FpExt->getOperand(0);
384             }
385           }
386         }
387         WhatToStore.push_back(Arg);
388       } else if (isa<PointerType>(ArgType)) {
389         if (shouldPrintAsStr(OpConvSpecifiers[ArgCount - 1], ArgType)) {
390           StringRef S = getAsConstantStr(Arg);
391           if (!S.empty()) {
392             const uint64_t ReadSize = 4;
393 
394             DataExtractor Extractor(S, /*IsLittleEndian=*/true, 8);
395             DataExtractor::Cursor Offset(0);
396             while (Offset && Offset.tell() < S.size()) {
397               uint64_t ReadNow = std::min(ReadSize, S.size() - Offset.tell());
398               uint64_t ReadBytes = 0;
399               switch (ReadNow) {
400               default: llvm_unreachable("min(4, X) > 4?");
401               case 1:
402                 ReadBytes = Extractor.getU8(Offset);
403                 break;
404               case 2:
405                 ReadBytes = Extractor.getU16(Offset);
406                 break;
407               case 3:
408                 ReadBytes = Extractor.getU24(Offset);
409                 break;
410               case 4:
411                 ReadBytes = Extractor.getU32(Offset);
412                 break;
413               }
414 
415               cantFail(Offset.takeError(),
416                        "failed to read bytes from constant array");
417 
418               APInt IntVal(8 * ReadSize, ReadBytes);
419 
420               // TODO: Should not bothering aligning up.
421               if (ReadNow < ReadSize)
422                 IntVal = IntVal.zext(8 * ReadSize);
423 
424               Type *IntTy = Type::getIntNTy(Ctx, IntVal.getBitWidth());
425               WhatToStore.push_back(ConstantInt::get(IntTy, IntVal));
426             }
427           } else {
428             // Empty string, give a hint to RT it is no NULL
429             Value *ANumV = ConstantInt::get(Int32Ty, 0xFFFFFF00, false);
430             WhatToStore.push_back(ANumV);
431           }
432         } else {
433           WhatToStore.push_back(Arg);
434         }
435       } else {
436         WhatToStore.push_back(Arg);
437       }
438       for (unsigned I = 0, E = WhatToStore.size(); I != E; ++I) {
439         Value *TheBtCast = WhatToStore[I];
440         unsigned ArgSize = TD->getTypeAllocSize(TheBtCast->getType());
441         SmallVector<Value *, 1> BuffOffset;
442         BuffOffset.push_back(ConstantInt::get(I32Ty, ArgSize));
443 
444         Type *ArgPointer = PointerType::get(TheBtCast->getType(), 1);
445         Value *CastedGEP =
446             new BitCastInst(BufferIdx, ArgPointer, "PrintBuffPtrCast", Brnch);
447         StoreInst *StBuff = new StoreInst(TheBtCast, CastedGEP, Brnch);
448         LLVM_DEBUG(dbgs() << "inserting store to printf buffer:\n"
449                           << *StBuff << '\n');
450         (void)StBuff;
451         if (I + 1 == E && ArgCount + 1 == CI->arg_size())
452           break;
453         BufferIdx = GetElementPtrInst::Create(I8Ty, BufferIdx, BuffOffset,
454                                               "PrintBuffNextPtr", Brnch);
455         LLVM_DEBUG(dbgs() << "inserting gep to the printf buffer:\n"
456                           << *BufferIdx << '\n');
457       }
458     }
459   }
460 
461   // erase the printf calls
462   for (auto *CI : Printfs)
463     CI->eraseFromParent();
464 
465   Printfs.clear();
466   return true;
467 }
468 
469 bool AMDGPUPrintfRuntimeBindingImpl::run(Module &M) {
470   Triple TT(M.getTargetTriple());
471   if (TT.getArch() == Triple::r600)
472     return false;
473 
474   auto PrintfFunction = M.getFunction("printf");
475   if (!PrintfFunction || !PrintfFunction->isDeclaration())
476     return false;
477 
478   for (auto &U : PrintfFunction->uses()) {
479     if (auto *CI = dyn_cast<CallInst>(U.getUser())) {
480       if (CI->isCallee(&U))
481         Printfs.push_back(CI);
482     }
483   }
484 
485   if (Printfs.empty())
486     return false;
487 
488   TD = &M.getDataLayout();
489 
490   return lowerPrintfForGpu(M);
491 }
492 
493 bool AMDGPUPrintfRuntimeBinding::runOnModule(Module &M) {
494   auto GetDT = [this](Function &F) -> DominatorTree & {
495     return this->getAnalysis<DominatorTreeWrapperPass>(F).getDomTree();
496   };
497   auto GetTLI = [this](Function &F) -> TargetLibraryInfo & {
498     return this->getAnalysis<TargetLibraryInfoWrapperPass>().getTLI(F);
499   };
500 
501   return AMDGPUPrintfRuntimeBindingImpl(GetDT, GetTLI).run(M);
502 }
503 
504 PreservedAnalyses
505 AMDGPUPrintfRuntimeBindingPass::run(Module &M, ModuleAnalysisManager &AM) {
506   FunctionAnalysisManager &FAM =
507       AM.getResult<FunctionAnalysisManagerModuleProxy>(M).getManager();
508   auto GetDT = [&FAM](Function &F) -> DominatorTree & {
509     return FAM.getResult<DominatorTreeAnalysis>(F);
510   };
511   auto GetTLI = [&FAM](Function &F) -> TargetLibraryInfo & {
512     return FAM.getResult<TargetLibraryAnalysis>(F);
513   };
514   bool Changed = AMDGPUPrintfRuntimeBindingImpl(GetDT, GetTLI).run(M);
515   return Changed ? PreservedAnalyses::none() : PreservedAnalyses::all();
516 }
517