[ValueTracking] Add CharWidth argument to getConstantStringInfo (NFC)

The method assumes that host chars and target chars have the same width.
Add a CharWidth argument so that it can bail out if the requested char
width differs from the host char width.

Alternatively, the check could be done at call sites, but this is more
error-prone.

In the future, this method will be replaced with a different one that
allows host/target chars to have different widths. The prototype will
be the same except that StringRef is replaced with something that is
byte width agnostic. Adding CharWidth argument now reduces the future
diff.

Author: s-barannikov

clang/lib/CodeGen/TargetBuiltins/AMDGPU.cpp

@@ -306,7 +306,7 @@ void CodeGenFunction::ProcessOrderScopeAMDGCN(Value *Order, Value *Scope,
 
   // Some of the atomic builtins take the scope as a string name.
   StringRef scp;
-  if (llvm::getConstantStringInfo(Scope, scp)) {
+  if (llvm::getConstantStringInfo(Scope, scp, /*CharWidth=*/8)) {
     if (getTarget().getTriple().isSPIRV())
       scp = mapScopeToSPIRV(scp);
     SSID = getLLVMContext().getOrInsertSyncScopeID(scp);
@@ -369,7 +369,7 @@ void CodeGenFunction::AddAMDGPUFenceAddressSpaceMMRA(llvm::Instruction *Inst,
   for (unsigned K = 2; K < E->getNumArgs(); ++K) {
     llvm::Value *V = EmitScalarExpr(E->getArg(K));
     StringRef AS;
-    if (llvm::getConstantStringInfo(V, AS)) {
+    if (llvm::getConstantStringInfo(V, AS, /*CharWidth=*/8)) {
       MMRAs.push_back({Tag, AS});
       // TODO: Delete the resulting unused constant?
       continue;

llvm/include/llvm/Analysis/ValueTracking.h

@@ -399,7 +399,7 @@ LLVM_ABI bool getConstantDataArrayInfo(const Value *V,
 /// trailing null characters as well as any other characters that come after
 /// it.
 LLVM_ABI bool getConstantStringInfo(const Value *V, StringRef &Str,
-                                    bool TrimAtNul = true);
+                                    unsigned CharWidth, bool TrimAtNul = true);
 
 /// If we can compute the length of the string pointed to by the specified
 /// pointer, return 'len+1'.  If we can't, return 0.

llvm/lib/Analysis/ValueTracking.cpp

@@ -6628,9 +6628,12 @@ bool llvm::getConstantDataArrayInfo(const Value *V,
 /// return true.  When TrimAtNul is set, Str will contain only the bytes up
 /// to but not including the first nul.  Return false on failure.
 bool llvm::getConstantStringInfo(const Value *V, StringRef &Str,
-                                 bool TrimAtNul) {
+                                 unsigned CharWidth, bool TrimAtNul) {
+  if (CharWidth != CHAR_BIT)
+    return false;
+
   ConstantDataArraySlice Slice;
-  if (!getConstantDataArrayInfo(V, Slice, 8))
+  if (!getConstantDataArrayInfo(V, Slice, CharWidth))
     return false;
 
   if (Slice.Array == nullptr) {

llvm/lib/Target/AMDGPU/AMDGPUPrintfRuntimeBinding.cpp

@@ -121,7 +121,7 @@ static_assert(NonLiteralStr.size() == 3);
 
 static StringRef getAsConstantStr(Value *V) {
   StringRef S;
-  if (!getConstantStringInfo(V, S))
+  if (!getConstantStringInfo(V, S, /*CharWidth=*/8))
     S = NonLiteralStr;
 
   return S;
@@ -154,7 +154,7 @@ bool AMDGPUPrintfRuntimeBindingImpl::lowerPrintfForGpu(Module &M) {
     Value *Op = CI->getArgOperand(0);
 
     StringRef FormatStr;
-    if (!getConstantStringInfo(Op, FormatStr)) {
+    if (!getConstantStringInfo(Op, FormatStr, /*CharWidth=*/8)) {
       Value *Stripped = Op->stripPointerCasts();
       if (!isa<UndefValue>(Stripped) && !isa<ConstantPointerNull>(Stripped))
         diagnoseInvalidFormatString(CI);

llvm/lib/Target/SPIRV/SPIRVAsmPrinter.cpp

@@ -707,7 +707,7 @@ void SPIRVAsmPrinter::outputAnnotations(const Module &M) {
 
       StringRef AnnotationString;
       [[maybe_unused]] bool Success =
-          getConstantStringInfo(GV, AnnotationString);
+          getConstantStringInfo(GV, AnnotationString, /*CharWidth=*/8);
       assert(Success && "Failed to get annotation string");
       MCInst Inst;
       Inst.setOpcode(SPIRV::OpDecorate);

llvm/lib/Target/SPIRV/SPIRVPrepareFunctions.cpp

@@ -166,7 +166,7 @@ static std::string getAnnotation(Value *AnnoVal, Value *OptAnnoVal) {
   std::string Anno;
   if (auto *C = dyn_cast_or_null<Constant>(AnnoVal)) {
     StringRef Str;
-    if (getConstantStringInfo(C, Str))
+    if (getConstantStringInfo(C, Str, /*CharWidth=*/8))
       Anno = Str;
   }
   // handle optional annotation parameter in a way that Khronos Translator do

llvm/lib/Target/WebAssembly/WebAssemblyAsmPrinter.cpp

@@ -572,7 +572,7 @@ void WebAssemblyAsmPrinter::EmitFunctionAttributes(Module &M) {
     // The second field is a pointer to a global annotation string.
     auto *GV = cast<GlobalVariable>(CS->getOperand(1)->stripPointerCasts());
     StringRef AnnotationString;
-    getConstantStringInfo(GV, AnnotationString);
+    getConstantStringInfo(GV, AnnotationString, /*CharWidth=*/8);
     auto *Sym = static_cast<MCSymbolWasm *>(getSymbol(F));
     CustomSections[AnnotationString].push_back(Sym);
   }

llvm/lib/Transforms/AggressiveInstCombine/AggressiveInstCombine.cpp

@@ -1195,6 +1195,8 @@ class StrNCmpInliner {
 /// handled by the instcombine pass.
 ///
 bool StrNCmpInliner::optimizeStrNCmp() {
+  unsigned CharWidth = DL.getByteWidth();
+
   if (StrNCmpInlineThreshold < 2)
     return false;
 
@@ -1208,8 +1210,10 @@ bool StrNCmpInliner::optimizeStrNCmp() {
     return false;
 
   StringRef Str1, Str2;
-  bool HasStr1 = getConstantStringInfo(Str1P, Str1, /*TrimAtNul=*/false);
-  bool HasStr2 = getConstantStringInfo(Str2P, Str2, /*TrimAtNul=*/false);
+  bool HasStr1 =
+      getConstantStringInfo(Str1P, Str1, CharWidth, /*TrimAtNul=*/false);
+  bool HasStr2 =
+      getConstantStringInfo(Str2P, Str2, CharWidth, /*TrimAtNul=*/false);
   if (HasStr1 == HasStr2)
     return false;
 
@@ -1347,12 +1351,14 @@ void StrNCmpInliner::inlineCompare(Value *LHS, StringRef RHS, uint64_t N,
 /// Convert memchr with a small constant string into a switch
 static bool foldMemChr(CallInst *Call, DomTreeUpdater *DTU,
                        const DataLayout &DL) {
+  unsigned CharWidth = DL.getByteWidth();
+
   if (isa<Constant>(Call->getArgOperand(1)))
     return false;
 
   StringRef Str;
   Value *Base = Call->getArgOperand(0);
-  if (!getConstantStringInfo(Base, Str, /*TrimAtNul=*/false))
+  if (!getConstantStringInfo(Base, Str, CharWidth, /*TrimAtNul=*/false))
     return false;
 
   uint64_t N = Str.size();

llvm/lib/Transforms/Utils/AMDGPUEmitPrintf.cpp

@@ -250,7 +250,7 @@ static Value *callBufferedPrintfStart(
   for (size_t i = 1; i < Args.size(); i++) {
     if (SpecIsCString.test(i)) {
       StringRef ArgStr;
-      if (getConstantStringInfo(Args[i], ArgStr)) {
+      if (getConstantStringInfo(Args[i], ArgStr, /*CharWidth=*/8)) {
         auto alignedLen = alignTo(ArgStr.size() + 1, 8);
         StringContents.push_back(StringData(
             ArgStr,
@@ -432,7 +432,7 @@ Value *llvm::emitAMDGPUPrintfCall(IRBuilder<> &Builder, ArrayRef<Value *> Args,
   SparseBitVector<8> SpecIsCString;
   StringRef FmtStr;
 
-  if (getConstantStringInfo(Fmt, FmtStr))
+  if (getConstantStringInfo(Fmt, FmtStr, /*CharWidth=*/8))
     locateCStrings(SpecIsCString, FmtStr);
 
   if (IsBuffered) {