merge amd-staging into amd-feature/wave-transform (#472)

Author: cdevadas

.github/instructions/lldb.instructions.md

@@ -0,0 +1,79 @@
+---
+applyTo: lldb/**/*
+---
+
+When reviewing code, focus on:
+
+## Language, Libraries & Standards
+
+- Target C++17 and avoid vendor-specific extensions.
+- For Python scripts, follow PEP 8.
+- Prefer standard library or LLVM support libraries instead of reinventing data structures.
+
+## Comments & Documentation
+
+- Each source file should include the standard LLVM file header.
+- Header files must have proper header guards.
+- Non-trivial classes and public methods should have Doxygen documentation.
+- Use `//` or `///` comments normally; avoid block comments unless necessary.
+- Non-trivial code should have comments explaining what it does and why. Avoid comments that explain how it does it at a micro level.
+
+## Language & Compiler Issues
+
+- Write portable code; wrap non-portable code in interfaces.
+- Do not use RTTI or exceptions.
+- Prefer C++-style casts over C-style casts.
+- Do not use static constructors.
+- Use `class` or `struct` consistently; `struct` only for all-public data.
+- When then same class is declared or defined multiple times, make sure it's consistently done using either `class` or `struct`.
+
+## Headers & Library Layering
+
+- Include order: module header → local/private headers → project headers → system headers.
+- Headers must compile standalone (include all dependencies).
+- Maintain proper library layering; avoid circular dependencies.
+- Include minimally; use forward declarations where possible.
+- Keep internal headers private to modules.
+- Use full namespace qualifiers for out-of-line definitions.
+
+## Control Flow & Structure
+
+- Prefer early exits over deep nesting.
+- Do not use `else` after `return`, `continue`, `break`, or `goto`.
+- Encapsulate loops that compute predicates into helper functions.
+
+## Naming
+
+- LLDB's code style differs from LLVM's coding style.
+- Variables are `snake_case`.
+- Functions and methods are `UpperCamelCase`.
+- Static, global and member variables have `s_`, `g_` and `m_` prefixes respectively.
+
+## General Guidelines
+
+- Use `assert` liberally; prefer `llvm_unreachable` for unreachable states.
+- Do not use `using namespace std;` in headers.
+- Provide a virtual method anchor for classes defined in headers.
+- Do not use default labels in fully covered switches over enumerations.
+- Use range-based for loops wherever possible.
+- Capture `end()` outside loops if not using range-based iteration.
+- Including `<iostream>` is forbidded. Use LLVM’s `raw_ostream` instead.
+- Don’t use `inline` when defining a function in a class definition.
+
+## Microscopic Details
+
+- Preserve existing style in modified code.
+- Prefer pre-increment (`++i`) when value is unused.
+- Use `private`, `protected`, or `public` keyword as appropriate to restrict class member visibility.
+- Omit braces for single-statement `if`, `else`, `while`, `for` unless needed.
+
+## Review Style
+
+- Be specific and actionable in feedback.
+- Explain the "why" behind recommendations.
+- Link back to the LLVM Coding Standards: https://llvm.org/docs/CodingStandards.html.
+- Ask clarifying questions when code intent is unclear.
+
+Ignore formatting and assume that's handled by external tools like `clang-format` and `black`.
+Remember that these standards are **guidelines**.
+Always prioritize consistency with the style that is already being used by the surrounding code.

.github/copilot-instructions.md renamed to .github/instructions/llvm.instructions.md

@@ -1,3 +1,7 @@
+---
+applyTo: llvm/**/*
+---
+
 When performing a code review, pay close attention to code modifying a function's
 control flow. Could the change result in the corruption of performance profile
 data? Could the change result in invalid debug information, in particular for

.github/workflows/release-binaries-all.yml

@@ -90,7 +90,6 @@ jobs:
         runs-on:
           - ubuntu-22.04
           - ubuntu-22.04-arm
-          - macos-13
           - macos-14
 
     uses: ./.github/workflows/release-binaries.yml

.github/workflows/release-binaries.yml

@@ -21,7 +21,6 @@ on:
         options:
           - ubuntu-22.04
           - ubuntu-22.04-arm
-          - macos-13
           - macos-14
 
   workflow_call:
@@ -130,13 +129,6 @@ jobs:
           target_cmake_flags="$target_cmake_flags -DBOOTSTRAP_BOOTSTRAP_COMPILER_RT_ENABLE_IOS=OFF"
           if [ "$RUNNER_ARCH" = "ARM64" ]; then
             arches=arm64
-          else
-            arches=x86_64
-            # Disable Flang builds on macOS x86_64.  The FortranLower library takes
-            # 2-3 hours to build on macOS, much slower than on Linux.
-            # The long build time causes the release build to time out on x86_64,
-            # so we need to disable flang there.
-            target_cmake_flags="$target_cmake_flags -DLLVM_RELEASE_ENABLE_PROJECTS='clang;lld;lldb;clang-tools-extra;polly;mlir'"
           fi
           target_cmake_flags="$target_cmake_flags -DBOOTSTRAP_BOOTSTRAP_DARWIN_osx_ARCHS=$arches -DBOOTSTRAP_BOOTSTRAP_DARWIN_osx_BUILTIN_ARCHS=$arches"
         fi
@@ -152,14 +144,6 @@ jobs:
             build_runs_on="depot-${{ inputs.runs-on }}-16"
             test_runs_on=$build_runs_on
             ;;
-          macos-13)
-            if [ "$GITHUB_EVENT_NAME" = "pull_request" ]; then
-              build_runs_on="${{ inputs.runs-on }}"
-            else
-              build_runs_on="macos-13-large"
-            fi
-            test_runs_on="${{ inputs.runs-on }}"
-            ;;
           macos-14)
             if [ "$GITHUB_EVENT_NAME" = "pull_request" ]; then
               build_runs_on="${{ inputs.runs-on }}"

bolt/include/bolt/Core/BinaryContext.h

@@ -781,11 +781,6 @@ class BinaryContext {
     uint64_t PseudoProbeLooseMatchedSampleCount{0};
     ///   the count of call matched samples
     uint64_t CallMatchedSampleCount{0};
-    ///   the number of stale functions that have matching number of blocks in
-    ///   the profile
-    uint64_t NumStaleFuncsWithEqualBlockCount{0};
-    ///   the number of blocks that have matching size but a differing hash
-    uint64_t NumStaleBlocksWithEqualIcount{0};
   } Stats;
 
   // Original binary execution count stats.
@@ -937,6 +932,16 @@ class BinaryContext {
   std::pair<const MCSymbol *, uint64_t>
   handleAddressRef(uint64_t Address, BinaryFunction &BF, bool IsPCRel);
 
+  /// When \p Address inside function \p BF is a target of a control transfer
+  /// instruction (branch) from another function, return a corresponding symbol
+  /// that should be used by the branch. For example, main or secondary entry
+  /// point.
+  ///
+  /// If \p Address is an invalid destination, such as a constant island, return
+  /// nullptr and mark \p BF as ignored, since we cannot properly handle a
+  /// branch to a constant island.
+  MCSymbol *handleExternalBranchTarget(uint64_t Address, BinaryFunction &BF);
+
   /// Analyze memory contents at the given \p Address and return the type of
   /// memory contents (such as a possible jump table).
   MemoryContentsType analyzeMemoryAt(uint64_t Address, BinaryFunction &BF);

bolt/include/bolt/Passes/MarkRAStates.h

@@ -13,11 +13,16 @@
 #define BOLT_PASSES_MARK_RA_STATES
 
 #include "bolt/Passes/BinaryPasses.h"
+#include <mutex>
 
 namespace llvm {
 namespace bolt {
 
 class MarkRAStates : public BinaryFunctionPass {
+  // setIgnored() is not thread-safe, but the pass is running on functions in
+  // parallel.
+  std::mutex IgnoreMutex;
+
 public:
   explicit MarkRAStates() : BinaryFunctionPass(false) {}
 

bolt/lib/Core/BinaryContext.cpp

@@ -518,6 +518,23 @@ BinaryContext::handleAddressRef(uint64_t Address, BinaryFunction &BF,
   return std::make_pair(TargetSymbol, 0);
 }
 
+MCSymbol *BinaryContext::handleExternalBranchTarget(uint64_t Address,
+                                                    BinaryFunction &BF) {
+  if (BF.isInConstantIsland(Address)) {
+    BF.setIgnored();
+    this->outs() << "BOLT-WARNING: ignoring entry point at address 0x"
+                 << Twine::utohexstr(Address)
+                 << " in constant island of function " << BF << '\n';
+    return nullptr;
+  }
+
+  const uint64_t Offset = Address - BF.getAddress();
+  assert(Offset < BF.getSize() &&
+         "Address should be inside the referenced function");
+
+  return Offset ? BF.addEntryPointAtOffset(Offset) : BF.getSymbol();
+}
+
 MemoryContentsType BinaryContext::analyzeMemoryAt(uint64_t Address,
                                                   BinaryFunction &BF) {
   if (!isX86())
@@ -1399,17 +1416,10 @@ void BinaryContext::processInterproceduralReferences() {
             << Function.getPrintName() << " and "
             << TargetFunction->getPrintName() << '\n';
       }
-      if (uint64_t Offset = Address - TargetFunction->getAddress()) {
-        if (!TargetFunction->isInConstantIsland(Address)) {
-          TargetFunction->addEntryPointAtOffset(Offset);
-        } else {
-          TargetFunction->setIgnored();
-          this->outs() << "BOLT-WARNING: Ignoring entry point at address 0x"
-                       << Twine::utohexstr(Address)
-                       << " in constant island of function " << *TargetFunction
-                       << '\n';
-        }
-      }
+
+      // Create an extra entry point if needed. Can also render the target
+      // function ignored if the reference is invalid.
+      handleExternalBranchTarget(Address, *TargetFunction);
 
       continue;
     }

bolt/lib/Core/BinaryFunction.cpp

@@ -1697,11 +1697,12 @@ bool BinaryFunction::scanExternalRefs() {
       if (!TargetFunction || ignoreFunctionRef(*TargetFunction))
         continue;
 
-      const uint64_t FunctionOffset =
-          TargetAddress - TargetFunction->getAddress();
+      // Get a reference symbol for the function when address is a valid code
+      // reference.
       BranchTargetSymbol =
-          FunctionOffset ? TargetFunction->addEntryPointAtOffset(FunctionOffset)
-                         : TargetFunction->getSymbol();
+          BC.handleExternalBranchTarget(TargetAddress, *TargetFunction);
+      if (!BranchTargetSymbol)
+        continue;
     }
 
     // Can't find more references. Not creating relocations since we are not

bolt/lib/Passes/BinaryPasses.cpp

@@ -1508,12 +1508,6 @@ Error PrintProgramStats::runOnFunctions(BinaryContext &BC) {
   if (NumAllStaleFunctions) {
     const float PctStale =
         NumAllStaleFunctions / (float)NumAllProfiledFunctions * 100.0f;
-    const float PctStaleFuncsWithEqualBlockCount =
-        (float)BC.Stats.NumStaleFuncsWithEqualBlockCount /
-        NumAllStaleFunctions * 100.0f;
-    const float PctStaleBlocksWithEqualIcount =
-        (float)BC.Stats.NumStaleBlocksWithEqualIcount /
-        BC.Stats.NumStaleBlocks * 100.0f;
     auto printErrorOrWarning = [&]() {
       if (PctStale > opts::StaleThreshold)
         BC.errs() << "BOLT-ERROR: ";
@@ -1536,17 +1530,6 @@ Error PrintProgramStats::runOnFunctions(BinaryContext &BC) {
                 << "%) belong to functions with invalid"
                    " (possibly stale) profile.\n";
     }
-    BC.outs() << "BOLT-INFO: " << BC.Stats.NumStaleFuncsWithEqualBlockCount
-              << " stale function"
-              << (BC.Stats.NumStaleFuncsWithEqualBlockCount == 1 ? "" : "s")
-              << format(" (%.1f%% of all stale)",
-                        PctStaleFuncsWithEqualBlockCount)
-              << " have matching block count.\n";
-    BC.outs() << "BOLT-INFO: " << BC.Stats.NumStaleBlocksWithEqualIcount
-              << " stale block"
-              << (BC.Stats.NumStaleBlocksWithEqualIcount == 1 ? "" : "s")
-              << format(" (%.1f%% of all stale)", PctStaleBlocksWithEqualIcount)
-              << " have matching icount.\n";
     if (PctStale > opts::StaleThreshold) {
       return createFatalBOLTError(
           Twine("BOLT-ERROR: stale functions exceed specified threshold of ") +

bolt/lib/Passes/MarkRAStates.cpp

@@ -43,10 +43,11 @@ bool MarkRAStates::runOnFunction(BinaryFunction &BF) {
         // Not all functions have .cfi_negate_ra_state in them. But if one does,
         // we expect psign/pauth instructions to have the hasNegateRAState
         // annotation.
-        BF.setIgnored();
         BC.outs() << "BOLT-INFO: inconsistent RAStates in function "
                   << BF.getPrintName()
                   << ": ptr sign/auth inst without .cfi_negate_ra_state\n";
+        std::lock_guard<std::mutex> Lock(IgnoreMutex);
+        BF.setIgnored();
         return false;
       }
     }
@@ -67,6 +68,7 @@ bool MarkRAStates::runOnFunction(BinaryFunction &BF) {
           BC.outs() << "BOLT-INFO: inconsistent RAStates in function "
                     << BF.getPrintName()
                     << ": ptr signing inst encountered in Signed RA state\n";
+          std::lock_guard<std::mutex> Lock(IgnoreMutex);
           BF.setIgnored();
           return false;
         }
@@ -80,6 +82,7 @@ bool MarkRAStates::runOnFunction(BinaryFunction &BF) {
                     << BF.getPrintName()
                     << ": ptr authenticating inst encountered in Unsigned RA "
                        "state\n";
+          std::lock_guard<std::mutex> Lock(IgnoreMutex);
           BF.setIgnored();
           return false;
         }