Merge branch 'main' into users/minglotus-6/spr/summary1

Author: mingmingl-llvm

.github/CODEOWNERS

@@ -115,6 +115,10 @@ clang/test/AST/Interp/ @tbaederr
 /mlir/**/LLVMIR/**/BasicPtxBuilderInterface* @grypp
 /mlir/**/NVVM*/ @grypp
 
+# MLIR Python Bindings
+/mlir/test/python/ @makslevental @stellaraccident
+/mlir/python/ @makslevental @stellaraccident
+
 # BOLT
 /bolt/ @aaupov @maksfb @rafaelauler @ayermolo @dcci
 

.github/workflows/pr-code-format.yml

@@ -1,4 +1,8 @@
 name: "Check code formatting"
+
+permissions:
+  contents: read
+
 on:
   pull_request:
     branches:

bolt/docs/BAT.md

@@ -81,9 +81,10 @@ Hot indices are delta encoded, implicitly starting at zero.
 | `FuncHash` | 8b | Function hash for input function | Hot |
 | `NumBlocks` | ULEB128 | Number of basic blocks in the original function | Hot |
 | `NumSecEntryPoints` | ULEB128 | Number of secondary entry points in the original function | Hot |
+| `ColdInputSkew` | ULEB128 | Skew to apply to all input offsets | Cold |
 | `NumEntries` | ULEB128 | Number of address translation entries for a function | Both |
-| `EqualElems` | ULEB128 | Number of equal offsets in the beginning of a function | Hot |
-| `BranchEntries` | Bitmask, `alignTo(EqualElems, 8)` bits | If `EqualElems` is non-zero, bitmask denoting entries with `BRANCHENTRY` bit | Hot |
+| `EqualElems` | ULEB128 | Number of equal offsets in the beginning of a function | Both |
+| `BranchEntries` | Bitmask, `alignTo(EqualElems, 8)` bits | If `EqualElems` is non-zero, bitmask denoting entries with `BRANCHENTRY` bit | Both |
 
 Function header is followed by *Address Translation Table* with `NumEntries`
 total entries, and *Secondary Entry Points* table with `NumSecEntryPoints`
@@ -99,8 +100,8 @@ entry is encoded. Input offsets implicitly start at zero.
 | `BBHash` | Optional, 8b | Basic block hash in input binary | BB |
 | `BBIdx`  | Optional, Delta, ULEB128 | Basic block index in input binary | BB |
 
-For hot fragments, the table omits the first `EqualElems` input offsets
-where the input offset equals output offset.
+The table omits the first `EqualElems` input offsets where the input offset
+equals output offset.
 
 `BRANCHENTRY` bit denotes whether a given offset pair is a control flow source
 (branch or call instruction). If not set, it signifies a control flow target

bolt/include/bolt/Core/BinaryData.h

@@ -107,7 +107,6 @@ class BinaryData {
   std::vector<MCSymbol *> &getSymbols() { return Symbols; }
 
   bool hasName(StringRef Name) const;
-  bool hasNameRegex(StringRef Name) const;
   bool nameStartsWith(StringRef Prefix) const;
 
   bool hasSymbol(const MCSymbol *Symbol) const {

bolt/include/bolt/Profile/BoltAddressTranslation.h

@@ -149,9 +149,9 @@ class BoltAddressTranslation {
   /// entries in function address translation map.
   APInt calculateBranchEntriesBitMask(MapTy &Map, size_t EqualElems);
 
-  /// Calculate the number of equal offsets (output = input) in the beginning
-  /// of the function.
-  size_t getNumEqualOffsets(const MapTy &Map) const;
+  /// Calculate the number of equal offsets (output = input - skew) in the
+  /// beginning of the function.
+  size_t getNumEqualOffsets(const MapTy &Map, uint32_t Skew) const;
 
   std::map<uint64_t, MapTy> Maps;
 

bolt/lib/Core/BinaryContext.cpp

@@ -555,6 +555,9 @@ bool BinaryContext::analyzeJumpTable(const uint64_t Address,
                                      const uint64_t NextJTAddress,
                                      JumpTable::AddressesType *EntriesAsAddress,
                                      bool *HasEntryInFragment) const {
+  // Target address of __builtin_unreachable.
+  const uint64_t UnreachableAddress = BF.getAddress() + BF.getSize();
+
   // Is one of the targets __builtin_unreachable?
   bool HasUnreachable = false;
 
@@ -564,9 +567,15 @@ bool BinaryContext::analyzeJumpTable(const uint64_t Address,
   // Number of targets other than __builtin_unreachable.
   uint64_t NumRealEntries = 0;
 
-  auto addEntryAddress = [&](uint64_t EntryAddress) {
-    if (EntriesAsAddress)
-      EntriesAsAddress->emplace_back(EntryAddress);
+  // Size of the jump table without trailing __builtin_unreachable entries.
+  size_t TrimmedSize = 0;
+
+  auto addEntryAddress = [&](uint64_t EntryAddress, bool Unreachable = false) {
+    if (!EntriesAsAddress)
+      return;
+    EntriesAsAddress->emplace_back(EntryAddress);
+    if (!Unreachable)
+      TrimmedSize = EntriesAsAddress->size();
   };
 
   ErrorOr<const BinarySection &> Section = getSectionForAddress(Address);
@@ -618,8 +627,8 @@ bool BinaryContext::analyzeJumpTable(const uint64_t Address,
             : *getPointerAtAddress(EntryAddress);
 
     // __builtin_unreachable() case.
-    if (Value == BF.getAddress() + BF.getSize()) {
-      addEntryAddress(Value);
+    if (Value == UnreachableAddress) {
+      addEntryAddress(Value, /*Unreachable*/ true);
       HasUnreachable = true;
       LLVM_DEBUG(dbgs() << formatv("OK: {0:x} __builtin_unreachable\n", Value));
       continue;
@@ -673,6 +682,13 @@ bool BinaryContext::analyzeJumpTable(const uint64_t Address,
     addEntryAddress(Value);
   }
 
+  // Trim direct/normal jump table to exclude trailing unreachable entries that
+  // can collide with a function address.
+  if (Type == JumpTable::JTT_NORMAL && EntriesAsAddress &&
+      TrimmedSize != EntriesAsAddress->size() &&
+      getBinaryFunctionAtAddress(UnreachableAddress))
+    EntriesAsAddress->resize(TrimmedSize);
+
   // It's a jump table if the number of real entries is more than 1, or there's
   // one real entry and one or more special targets. If there are only multiple
   // special targets, then it's not a jump table.
@@ -1864,7 +1880,7 @@ MarkerSymType BinaryContext::getMarkerType(const SymbolRef &Symbol) const {
   // For aarch64 and riscv, the ABI defines mapping symbols so we identify data
   // in the code section (see IHI0056B). $x identifies a symbol starting code or
   // the end of a data chunk inside code, $d identifies start of data.
-  if ((!isAArch64() && !isRISCV()) || ELFSymbolRef(Symbol).getSize())
+  if (isX86() || ELFSymbolRef(Symbol).getSize())
     return MarkerSymType::NONE;
 
   Expected<StringRef> NameOrError = Symbol.getName();

bolt/lib/Core/BinaryData.cpp

@@ -55,14 +55,6 @@ bool BinaryData::hasName(StringRef Name) const {
   return false;
 }
 
-bool BinaryData::hasNameRegex(StringRef NameRegex) const {
-  Regex MatchName(NameRegex);
-  for (const MCSymbol *Symbol : Symbols)
-    if (MatchName.match(Symbol->getName()))
-      return true;
-  return false;
-}
-
 bool BinaryData::nameStartsWith(StringRef Prefix) const {
   for (const MCSymbol *Symbol : Symbols)
     if (Symbol->getName().starts_with(Prefix))

bolt/lib/Core/BinaryEmitter.cpp

@@ -512,7 +512,7 @@ void BinaryEmitter::emitFunctionBody(BinaryFunction &BF, FunctionFragment &FF,
 
       // Emit sized NOPs via MCAsmBackend::writeNopData() interface on x86.
       // This is a workaround for invalid NOPs handling by asm/disasm layer.
-      if (BC.MIB->isNoop(Instr) && BC.isX86()) {
+      if (BC.isX86() && BC.MIB->isNoop(Instr)) {
         if (std::optional<uint32_t> Size = BC.MIB->getSize(Instr)) {
           SmallString<15> Code;
           raw_svector_ostream VecOS(Code);

bolt/lib/Core/Relocation.cpp

@@ -1064,21 +1064,19 @@ MCBinaryExpr::Opcode Relocation::getComposeOpcodeFor(uint64_t Type) {
   }
 }
 
-#define ELF_RELOC(name, value) #name,
-
 void Relocation::print(raw_ostream &OS) const {
-  static const char *X86RelocNames[] = {
-#include "llvm/BinaryFormat/ELFRelocs/x86_64.def"
-  };
-  static const char *AArch64RelocNames[] = {
-#include "llvm/BinaryFormat/ELFRelocs/AArch64.def"
-  };
   switch (Arch) {
   default:
     OS << "RType:" << Twine::utohexstr(Type);
     break;
 
   case Triple::aarch64:
+    static const char *const AArch64RelocNames[] = {
+#define ELF_RELOC(name, value) #name,
+#include "llvm/BinaryFormat/ELFRelocs/AArch64.def"
+#undef ELF_RELOC
+    };
+    assert(Type < ArrayRef(AArch64RelocNames).size());
     OS << AArch64RelocNames[Type];
     break;
 
@@ -1088,16 +1086,22 @@ void Relocation::print(raw_ostream &OS) const {
     switch (Type) {
     default:
       llvm_unreachable("illegal RISC-V relocation");
-#undef ELF_RELOC
 #define ELF_RELOC(name, value)                                                 \
   case value:                                                                  \
     OS << #name;                                                               \
     break;
 #include "llvm/BinaryFormat/ELFRelocs/RISCV.def"
+#undef ELF_RELOC
     }
     break;
 
   case Triple::x86_64:
+    static const char *const X86RelocNames[] = {
+#define ELF_RELOC(name, value) #name,
+#include "llvm/BinaryFormat/ELFRelocs/x86_64.def"
+#undef ELF_RELOC
+    };
+    assert(Type < ArrayRef(X86RelocNames).size());
     OS << X86RelocNames[Type];
     break;
   }

bolt/lib/Profile/BoltAddressTranslation.cpp

@@ -153,12 +153,13 @@ APInt BoltAddressTranslation::calculateBranchEntriesBitMask(MapTy &Map,
   return BitMask;
 }
 
-size_t BoltAddressTranslation::getNumEqualOffsets(const MapTy &Map) const {
+size_t BoltAddressTranslation::getNumEqualOffsets(const MapTy &Map,
+                                                  uint32_t Skew) const {
   size_t EqualOffsets = 0;
   for (const std::pair<const uint32_t, uint32_t> &KeyVal : Map) {
     const uint32_t OutputOffset = KeyVal.first;
     const uint32_t InputOffset = KeyVal.second >> 1;
-    if (OutputOffset == InputOffset)
+    if (OutputOffset == InputOffset - Skew)
       ++EqualOffsets;
     else
       break;
@@ -196,12 +197,17 @@ void BoltAddressTranslation::writeMaps(std::map<uint64_t, MapTy> &Maps,
         SecondaryEntryPointsMap.count(Address)
             ? SecondaryEntryPointsMap[Address].size()
             : 0;
+    uint32_t Skew = 0;
     if (Cold) {
       auto HotEntryIt = Maps.find(ColdPartSource[Address]);
       assert(HotEntryIt != Maps.end());
       size_t HotIndex = std::distance(Maps.begin(), HotEntryIt);
       encodeULEB128(HotIndex - PrevIndex, OS);
       PrevIndex = HotIndex;
+      // Skew of all input offsets for cold fragments is simply the first input
+      // offset.
+      Skew = Map.begin()->second >> 1;
+      encodeULEB128(Skew, OS);
     } else {
       // Function hash
       size_t BFHash = getBFHash(HotInputAddress);
@@ -217,24 +223,21 @@ void BoltAddressTranslation::writeMaps(std::map<uint64_t, MapTy> &Maps,
                         << '\n');
     }
     encodeULEB128(NumEntries, OS);
-    // For hot fragments only: encode the number of equal offsets
-    // (output = input) in the beginning of the function. Only encode one offset
-    // in these cases.
-    const size_t EqualElems = Cold ? 0 : getNumEqualOffsets(Map);
-    if (!Cold) {
-      encodeULEB128(EqualElems, OS);
-      if (EqualElems) {
-        const size_t BranchEntriesBytes = alignTo(EqualElems, 8) / 8;
-        APInt BranchEntries = calculateBranchEntriesBitMask(Map, EqualElems);
-        OS.write(reinterpret_cast<const char *>(BranchEntries.getRawData()),
-                 BranchEntriesBytes);
-        LLVM_DEBUG({
-          dbgs() << "BranchEntries: ";
-          SmallString<8> BitMaskStr;
-          BranchEntries.toString(BitMaskStr, 2, false);
-          dbgs() << BitMaskStr << '\n';
-        });
-      }
+    // Encode the number of equal offsets (output = input - skew) in the
+    // beginning of the function. Only encode one offset in these cases.
+    const size_t EqualElems = getNumEqualOffsets(Map, Skew);
+    encodeULEB128(EqualElems, OS);
+    if (EqualElems) {
+      const size_t BranchEntriesBytes = alignTo(EqualElems, 8) / 8;
+      APInt BranchEntries = calculateBranchEntriesBitMask(Map, EqualElems);
+      OS.write(reinterpret_cast<const char *>(BranchEntries.getRawData()),
+               BranchEntriesBytes);
+      LLVM_DEBUG({
+        dbgs() << "BranchEntries: ";
+        SmallString<8> BitMaskStr;
+        BranchEntries.toString(BitMaskStr, 2, false);
+        dbgs() << BitMaskStr << '\n';
+      });
     }
     const BBHashMapTy &BBHashMap = getBBHashMap(HotInputAddress);
     size_t Index = 0;
@@ -315,10 +318,12 @@ void BoltAddressTranslation::parseMaps(std::vector<uint64_t> &HotFuncs,
     uint64_t HotAddress = Cold ? 0 : Address;
     PrevAddress = Address;
     uint32_t SecondaryEntryPoints = 0;
+    uint64_t ColdInputSkew = 0;
     if (Cold) {
       HotIndex += DE.getULEB128(&Offset, &Err);
       HotAddress = HotFuncs[HotIndex];
       ColdPartSource.emplace(Address, HotAddress);
+      ColdInputSkew = DE.getULEB128(&Offset, &Err);
     } else {
       HotFuncs.push_back(Address);
       // Function hash
@@ -339,28 +344,25 @@ void BoltAddressTranslation::parseMaps(std::vector<uint64_t> &HotFuncs,
                             getULEB128Size(SecondaryEntryPoints)));
     }
     const uint32_t NumEntries = DE.getULEB128(&Offset, &Err);
-    // Equal offsets, hot fragments only.
-    size_t EqualElems = 0;
+    // Equal offsets.
+    const size_t EqualElems = DE.getULEB128(&Offset, &Err);
     APInt BEBitMask;
-    if (!Cold) {
-      EqualElems = DE.getULEB128(&Offset, &Err);
-      LLVM_DEBUG(dbgs() << formatv("Equal offsets: {0}, {1} bytes\n",
-                                   EqualElems, getULEB128Size(EqualElems)));
-      if (EqualElems) {
-        const size_t BranchEntriesBytes = alignTo(EqualElems, 8) / 8;
-        BEBitMask = APInt(alignTo(EqualElems, 8), 0);
-        LoadIntFromMemory(
-            BEBitMask,
-            reinterpret_cast<const uint8_t *>(
-                DE.getBytes(&Offset, BranchEntriesBytes, &Err).data()),
-            BranchEntriesBytes);
-        LLVM_DEBUG({
-          dbgs() << "BEBitMask: ";
-          SmallString<8> BitMaskStr;
-          BEBitMask.toString(BitMaskStr, 2, false);
-          dbgs() << BitMaskStr << ", " << BranchEntriesBytes << " bytes\n";
-        });
-      }
+    LLVM_DEBUG(dbgs() << formatv("Equal offsets: {0}, {1} bytes\n", EqualElems,
+                                 getULEB128Size(EqualElems)));
+    if (EqualElems) {
+      const size_t BranchEntriesBytes = alignTo(EqualElems, 8) / 8;
+      BEBitMask = APInt(alignTo(EqualElems, 8), 0);
+      LoadIntFromMemory(
+          BEBitMask,
+          reinterpret_cast<const uint8_t *>(
+              DE.getBytes(&Offset, BranchEntriesBytes, &Err).data()),
+          BranchEntriesBytes);
+      LLVM_DEBUG({
+        dbgs() << "BEBitMask: ";
+        SmallString<8> BitMaskStr;
+        BEBitMask.toString(BitMaskStr, 2, false);
+        dbgs() << BitMaskStr << ", " << BranchEntriesBytes << " bytes\n";
+      });
     }
     MapTy Map;
 
@@ -375,7 +377,7 @@ void BoltAddressTranslation::parseMaps(std::vector<uint64_t> &HotFuncs,
       PrevAddress = OutputAddress;
       int64_t InputDelta = 0;
       if (J < EqualElems) {
-        InputOffset = (OutputOffset << 1) | BEBitMask[J];
+        InputOffset = ((OutputOffset + ColdInputSkew) << 1) | BEBitMask[J];
       } else {
         InputDelta = DE.getSLEB128(&Offset, &Err);
         InputOffset += InputDelta;