Merge branch 'main' into wcscat-implmentation

Author: sribee8

.github/workflows/containers/github-action-ci-windows/Dockerfile

@@ -85,7 +85,7 @@ RUN powershell -Command \
 RUN git config --system core.longpaths true & \
     git config --global core.autocrlf false
 
-ARG RUNNER_VERSION=2.324.0
+ARG RUNNER_VERSION=2.325.0
 ENV RUNNER_VERSION=$RUNNER_VERSION
 
 RUN powershell -Command \

.github/workflows/containers/github-action-ci/Dockerfile

@@ -86,7 +86,7 @@ WORKDIR /home/gha
 
 FROM ci-container as ci-container-agent
 
-ENV GITHUB_RUNNER_VERSION=2.324.0
+ENV GITHUB_RUNNER_VERSION=2.325.0
 
 RUN mkdir actions-runner && \
     cd actions-runner && \

bolt/include/bolt/Core/BinaryFunction.h

@@ -2348,6 +2348,7 @@ class BinaryFunction {
       releaseCFG();
       CurrentState = State::Emitted;
     }
+    clearList(Relocations);
   }
 
   /// Process LSDA information for the function.

bolt/lib/Core/BinaryContext.cpp

@@ -1032,10 +1032,8 @@ void BinaryContext::adjustCodePadding() {
 
     if (!hasValidCodePadding(BF)) {
       if (HasRelocations) {
-        if (opts::Verbosity >= 1) {
-          this->outs() << "BOLT-INFO: function " << BF
-                       << " has invalid padding. Ignoring the function.\n";
-        }
+        this->errs() << "BOLT-WARNING: function " << BF
+                     << " has invalid padding. Ignoring the function\n";
         BF.setIgnored();
       } else {
         BF.setMaxSize(BF.getSize());

bolt/lib/Core/BinaryFunction.cpp

@@ -1526,8 +1526,6 @@ Error BinaryFunction::disassemble() {
   if (uint64_t Offset = getFirstInstructionOffset())
     Labels[Offset] = BC.Ctx->createNamedTempSymbol();
 
-  clearList(Relocations);
-
   if (!IsSimple) {
     clearList(Instructions);
     return createNonFatalBOLTError("");
@@ -3244,16 +3242,26 @@ void BinaryFunction::setTrapOnEntry() {
 }
 
 void BinaryFunction::setIgnored() {
+  IsIgnored = true;
+
   if (opts::processAllFunctions()) {
     // We can accept ignored functions before they've been disassembled.
-    // In that case, they would still get disassembled and emited, but not
+    // In that case, they would still get disassembled and emitted, but not
     // optimized.
-    assert(CurrentState == State::Empty &&
-           "cannot ignore non-empty functions in current mode");
-    IsIgnored = true;
+    if (CurrentState != State::Empty) {
+      BC.errs() << "BOLT-ERROR: cannot ignore non-empty function " << *this
+                << " in current mode\n";
+      exit(1);
+    }
     return;
   }
 
+  IsSimple = false;
+  LLVM_DEBUG(dbgs() << "Ignoring " << getPrintName() << '\n');
+
+  if (CurrentState == State::Empty)
+    return;
+
   clearDisasmState();
 
   // Clear CFG state too.
@@ -3273,9 +3281,11 @@ void BinaryFunction::setIgnored() {
 
   CurrentState = State::Empty;
 
-  IsIgnored = true;
-  IsSimple = false;
-  LLVM_DEBUG(dbgs() << "Ignoring " << getPrintName() << '\n');
+  // Fix external references in the original function body.
+  if (BC.HasRelocations) {
+    LLVM_DEBUG(dbgs() << "Scanning refs in " << *this << '\n');
+    scanExternalRefs();
+  }
 }
 
 void BinaryFunction::duplicateConstantIslands() {
@@ -3764,7 +3774,6 @@ void BinaryFunction::postProcessBranches() {
 
 MCSymbol *BinaryFunction::addEntryPointAtOffset(uint64_t Offset) {
   assert(Offset && "cannot add primary entry point");
-  assert(CurrentState == State::Empty || CurrentState == State::Disassembled);
 
   const uint64_t EntryPointAddress = getAddress() + Offset;
   MCSymbol *LocalSymbol = getOrCreateLocalLabel(EntryPointAddress);
@@ -3773,6 +3782,8 @@ MCSymbol *BinaryFunction::addEntryPointAtOffset(uint64_t Offset) {
   if (EntrySymbol)
     return EntrySymbol;
 
+  assert(CurrentState == State::Empty || CurrentState == State::Disassembled);
+
   if (BinaryData *EntryBD = BC.getBinaryDataAtAddress(EntryPointAddress)) {
     EntrySymbol = EntryBD->getSymbol();
   } else {

bolt/lib/Rewrite/RewriteInstance.cpp

@@ -3446,6 +3446,7 @@ void RewriteInstance::disassembleFunctions() {
         BC->outs() << "BOLT-INFO: could not disassemble function " << Function
                    << ". Will ignore.\n";
       // Forcefully ignore the function.
+      Function.scanExternalRefs();
       Function.setIgnored();
     });
 

bolt/test/AArch64/patch-ignored.s

@@ -0,0 +1,33 @@
+## Check that llvm-bolt patches functions that are getting ignored after their
+## CFG was constructed.
+
+# RUN: %clang %cflags %s -o %t.exe -Wl,-q
+# RUN: llvm-bolt %t.exe -o %t.bolt --force-patch 2>&1 | FileCheck %s
+# RUN: llvm-objdump -d %t.bolt | FileCheck %s --check-prefix=CHECK-OBJDUMP
+
+  .text
+
+## The function is too small to be patched and BOLT is forced to ignore it under
+## --force-patch. Check that the reference to _start is updated.
+# CHECK: BOLT-WARNING: failed to patch entries in unpatchable
+	.globl unpatchable
+  .type unpatchable, %function
+unpatchable:
+  .cfi_startproc
+# CHECK-OBJDUMP:      <unpatchable>:
+# CHECK-OBJDUMP-NEXT: bl {{.*}} <_start>
+  bl _start
+  ret
+  .cfi_endproc
+  .size unpatchable, .-unpatchable
+
+  .globl _start
+  .type _start, %function
+_start:
+  .cfi_startproc
+  cmp  x0, 1
+  b.eq  .L0
+.L0:
+  ret  x30
+  .cfi_endproc
+  .size _start, .-_start

bolt/test/X86/patch-ignored.s

@@ -0,0 +1,44 @@
+## Check that llvm-bolt patches functions that are getting ignored after their
+## CFG was constructed.
+
+# RUN: %clang %cflags %s -o %t.exe -Wl,-q
+# RUN: llvm-bolt %t.exe -o %t.bolt 2>&1 | FileCheck %s
+# RUN: llvm-objdump -d %t.bolt | FileCheck %s --check-prefix=CHECK-OBJDUMP
+
+  .text
+
+## After the CFG is built, the following function will be marked as ignored
+## due to the presence of the internal call.
+# CHECK:      BOLT-WARNING: will skip the following function
+# CHECK-NEXT: internal_call
+	.globl internal_call
+  .type internal_call, %function
+internal_call:
+  .cfi_startproc
+# CHECK-OBJDUMP:      <internal_call>:
+  call .L1
+  jmp .L2
+# CHECK-OBJDUMP:      jmp
+.L1:
+  jmp _start
+# CHECK-OBJDUMP:      jmp
+# CHECK-OBJDUMP-SAME: <_start>
+  ret
+.L2:
+  jmp _start
+# CHECK-OBJDUMP:      jmp
+# CHECK-OBJDUMP-SAME: <_start>
+  .cfi_endproc
+  .size internal_call, .-internal_call
+
+  .globl _start
+  .type _start, %function
+_start:
+  .cfi_startproc
+  cmpq  %rdi, 1
+  jne  .L0
+  movq %rdi, %rax
+.L0:
+  ret
+  .cfi_endproc
+  .size _start, .-_start

clang/include/clang/CIR/Dialect/IR/CIROps.td

@@ -1464,9 +1464,13 @@ def ShiftOp : CIR_Op<"shift", [Pure]> {
     ```
   }];
 
-  let results = (outs CIR_AnyIntOrVecOfInt:$result);
-  let arguments = (ins CIR_AnyIntOrVecOfInt:$value, CIR_AnyIntOrVecOfInt:$amount,
-                       UnitAttr:$isShiftleft);
+  let arguments = (ins
+    CIR_AnyIntOrVecOfIntType:$value,
+    CIR_AnyIntOrVecOfIntType:$amount,
+    UnitAttr:$isShiftleft
+  );
+
+  let results = (outs CIR_AnyIntOrVecOfIntType:$result);
 
   let assemblyFormat = [{
     `(`
@@ -2050,7 +2054,7 @@ def VecCreateOp : CIR_Op<"vec.create", [Pure]> {
     in the vector type.
   }];
 
-  let arguments = (ins Variadic<CIR_AnyType>:$elements);
+  let arguments = (ins Variadic<CIR_VectorElementType>:$elements);
   let results = (outs CIR_VectorType:$result);
 
   let assemblyFormat = [{
@@ -2085,7 +2089,7 @@ def VecInsertOp : CIR_Op<"vec.insert", [Pure,
 
   let arguments = (ins
     CIR_VectorType:$vec,
-    AnyType:$value,
+    CIR_VectorElementType:$value,
     CIR_AnyFundamentalIntType:$index
   );
 
@@ -2118,7 +2122,7 @@ def VecExtractOp : CIR_Op<"vec.extract", [Pure,
   }];
 
   let arguments = (ins CIR_VectorType:$vec, CIR_AnyFundamentalIntType:$index);
-  let results = (outs CIR_AnyType:$result);
+  let results = (outs CIR_VectorElementType:$result);
 
   let assemblyFormat = [{
     $vec `[` $index `:` type($index) `]` attr-dict `:` qualified(type($vec))
@@ -2180,7 +2184,7 @@ def VecShuffleDynamicOp : CIR_Op<"vec.shuffle.dynamic",
     ```
   }];
 
-  let arguments = (ins CIR_VectorType:$vec, IntegerVector:$indices);
+  let arguments = (ins CIR_VectorType:$vec, CIR_VectorOfIntType:$indices);
   let results = (outs CIR_VectorType:$result);
   let assemblyFormat = [{
     $vec `:` qualified(type($vec)) `,` $indices `:` qualified(type($indices))

clang/include/clang/CIR/Dialect/IR/CIRTypeConstraints.td

@@ -31,6 +31,18 @@ class CIR_ConfinedType<Type type, list<Pred> preds, string summary = "">
     : Type<And<[type.predicate, CIR_CastedSelfsToType<type.cppType, preds>]>,
          summary, type.cppType>;
 
+// Generates a type summary.
+// - For a single type: returns its summary.
+// - For multiple types: returns `any of <comma-separated summaries>`.
+class CIR_TypeSummaries<list<Type> types> {
+    assert !not(!empty(types)), "expects non-empty list of types";
+
+    list<string> summaries = !foreach(type, types, type.summary);
+    string joined = !interleave(summaries, ", ");
+
+    string value = !if(!eq(!size(types), 1), joined, "any of " # joined);
+}
+
 //===----------------------------------------------------------------------===//
 // Bool Type predicates
 //===----------------------------------------------------------------------===//
@@ -184,6 +196,24 @@ def CIR_PtrToVoidPtrType
 // Vector Type predicates
 //===----------------------------------------------------------------------===//
 
+def CIR_AnyVectorType : CIR_TypeBase<"::cir::VectorType", "vector type">;
+
+def CIR_VectorElementType : AnyTypeOf<[CIR_AnyIntOrFloatType, CIR_AnyPtrType],
+    "any cir integer, floating point or pointer type"
+> {
+    let cppFunctionName = "isValidVectorTypeElementType";
+}
+
+class CIR_ElementTypePred<Pred pred> : SubstLeaves<"$_self",
+    "::mlir::cast<::cir::VectorType>($_self).getElementType()", pred>;
+
+class CIR_VectorTypeOf<list<Type> types, string summary = "">
+    : CIR_ConfinedType<CIR_AnyVectorType,
+        [Or<!foreach(type, types, CIR_ElementTypePred<type.predicate>)>],
+        !if(!empty(summary),
+            "vector of " # CIR_TypeSummaries<types>.value,
+            summary)>;
+
 // Vector of integral type
 def IntegerVector : Type<
     And<[
@@ -196,8 +226,36 @@ def IntegerVector : Type<
     ]>, "!cir.vector of !cir.int"> {
 }
 
-// Any Integer or Vector of Integer Constraints
-def CIR_AnyIntOrVecOfInt: AnyTypeOf<[CIR_AnyIntType, IntegerVector]>;
+// Vector of type constraints
+def CIR_VectorOfIntType : CIR_VectorTypeOf<[CIR_AnyIntType]>;
+def CIR_VectorOfUIntType : CIR_VectorTypeOf<[CIR_AnyUIntType]>;
+def CIR_VectorOfSIntType : CIR_VectorTypeOf<[CIR_AnySIntType]>;
+def CIR_VectorOfFloatType : CIR_VectorTypeOf<[CIR_AnyFloatType]>;
+
+// Vector or Scalar type constraints
+def CIR_AnyIntOrVecOfIntType
+    : AnyTypeOf<[CIR_AnyIntType, CIR_VectorOfIntType],
+        "integer or vector of integer type"> {
+    let cppFunctionName = "isIntOrVectorOfIntType";
+}
+
+def CIR_AnySIntOrVecOfSIntType
+    : AnyTypeOf<[CIR_AnySIntType, CIR_VectorOfSIntType],
+        "signed integer or vector of signed integer type"> {
+    let cppFunctionName = "isSIntOrVectorOfSIntType";
+}
+
+def CIR_AnyUIntOrVecOfUIntType
+    : AnyTypeOf<[CIR_AnyUIntType, CIR_VectorOfUIntType],
+        "unsigned integer or vector of unsigned integer type"> {
+    let cppFunctionName = "isUIntOrVectorOfUIntType";
+}
+
+def CIR_AnyFloatOrVecOfFloatType
+    : AnyTypeOf<[CIR_AnyFloatType, CIR_VectorOfFloatType],
+        "floating point or vector of floating point type"> {
+    let cppFunctionName = "isFPOrVectorOfFPType";
+}
 
 //===----------------------------------------------------------------------===//
 // Scalar Type predicates