Merge branch 'main' into main

Author: vedantjk

bolt/test/link_fdata.py

@@ -32,7 +32,7 @@
 
 # Regexes to extract FDATA lines from input and parse FDATA and pre-aggregated
 # profile data
-prefix_pat = re.compile(f"^# {args.prefix}: (.*)")
+prefix_pat = re.compile(f"^(#|//) {args.prefix}: (.*)")
 
 # FDATA records:
 # <is symbol?> <closest elf symbol or DSO name> <relative FROM address>
@@ -61,7 +61,7 @@
         prefix_match = prefix_pat.match(line)
         if not prefix_match:
             continue
-        profile_line = prefix_match.group(1)
+        profile_line = prefix_match.group(2)
         fdata_match = fdata_pat.match(profile_line)
         preagg_match = preagg_pat.match(profile_line)
         nolbr_match = nolbr_pat.match(profile_line)

bolt/test/runtime/copy_file.py

@@ -0,0 +1,15 @@
+import sys
+import shutil
+
+with open(sys.argv[1] + ".output") as log_file:
+    lines = log_file.readlines()
+    for line in lines:
+        if line.startswith(sys.argv[2]):
+            pid = line.split(" ")[1].strip()
+            shutil.copy(
+                sys.argv[1] + "." + pid + ".fdata",
+                sys.argv[1] + "." + sys.argv[3] + ".fdata",
+            )
+            sys.exit(0)
+
+sys.exit(1)

bolt/test/runtime/instrumentation-indirect-2.c

@@ -50,7 +50,7 @@ int main() {
   return 0;
 }
 /*
-REQUIRES: system-linux,shell,fuser
+REQUIRES: system-linux,fuser
 
 RUN: %clang %cflags %s -o %t.exe -Wl,-q -pie -fpie
 
@@ -61,10 +61,14 @@ RUN: --instrumentation-wait-forks
 
 # Instrumented program needs to finish returning zero
 # Both output and profile must contain all 16 functions
-RUN: %t.instrumented_conservative > %t.output
-# Wait for profile and output to be fully written
-RUN: bash %S/wait_file.sh %t.output
-RUN: bash %S/wait_file.sh %t.fdata
+# We need to use bash to invoke this as otherwise we hang inside a
+# popen.communicate call in lit's internal shell. Eventually we should not
+# need this.
+# TODO(boomanaiden154): Remove once
+# https://github.com/llvm/llvm-project/issues/156484 is fixed.
+RUN: bash -c "%t.instrumented_conservative; wait" > %t.output
+# We can just read because we ensure the profile will be fully written by
+# calling wait inside the bash invocation.
 RUN: cat %t.output | FileCheck %s --check-prefix=CHECK-OUTPUT
 RUN: cat %t.fdata | FileCheck %s --check-prefix=CHECK-COMMON-PROF
 
@@ -112,14 +116,8 @@ RUN: bash %S/wait_file.sh %t.output
 # Make sure all functions were called
 RUN: cat %t.output | FileCheck %s --check-prefix=CHECK-OUTPUT
 
-RUN: child_pid=$(cat %t.output | grep funcA | awk '{print $2;}')
-RUN: par_pid=$(cat %t.output | grep funcB | awk '{print $2;}')
-
-RUN: bash %S/wait_file.sh %t.$child_pid.fdata
-RUN: bash %S/wait_file.sh %t.$par_pid.fdata
-
-RUN: mv %t.$child_pid.fdata %t.child.fdata
-RUN: mv %t.$par_pid.fdata %t.parent.fdata
+RUN: %python %S/copy_file.py %t funcA child
+RUN: %python %S/copy_file.py %t funcB parent
 
 # Instrumented binary must produce two profiles with only local calls
 # recorded. Functions called only in child should not appear in parent's

bolt/test/timers.c

@@ -1,22 +1,21 @@
-/* This test checks timers for metadata manager phases.
-# RUN: %clang %cflags %s -o %t.exe
-# RUN: link_fdata %s %t.exe %t.fdata
-# RUN: llvm-bolt %t.exe -o %t.null --data %t.fdata -w %t.yaml --time-rewrite \
-# RUN:   2>&1 | FileCheck %s
-# RUN: link_fdata %s %t.exe %t.preagg PREAGG
-# RUN: perf2bolt %t.exe -o %t.null -p %t.preagg --pa --time-rewrite \
-# RUN:   2>&1 | FileCheck %s --check-prefix=CHECK-P2B
+// This test checks timers for metadata manager phases.
+// RUN: %clang %cflags %s -o %t.exe
+// RUN: link_fdata %s %t.exe %t.fdata
+// RUN: llvm-bolt %t.exe -o %t.null --data %t.fdata -w %t.yaml --time-rewrite \
+// RUN:   2>&1 | FileCheck %s
+// RUN: link_fdata %s %t.exe %t.preagg PREAGG
+// RUN: perf2bolt %t.exe -o %t.null -p %t.preagg --pa --time-rewrite \
+// RUN:   2>&1 | FileCheck %s --check-prefix=CHECK-P2B
 
-# CHECK-DAG: update metadata post-emit
-# CHECK-DAG: process section metadata
-# CHECK-DAG: process metadata pre-CFG
-# CHECK-DAG: process metadata post-CFG
-# CHECK-DAG: finalize metadata pre-emit
+// CHECK-DAG: update metadata post-emit
+// CHECK-DAG: process section metadata
+// CHECK-DAG: process metadata pre-CFG
+// CHECK-DAG: process metadata post-CFG
+// CHECK-DAG: finalize metadata pre-emit
 
-# CHECK-P2B-DAG: process section metadata
-# CHECK-P2B-DAG: process metadata pre-CFG
+// CHECK-P2B-DAG: process section metadata
+// CHECK-P2B-DAG: process metadata pre-CFG
 
-# FDATA: 0 [unknown] 0 1 main 0 1 0
-# PREAGG: B X:0 #main# 1 0
-*/
+// FDATA: 0 [unknown] 0 1 main 0 1 0
+// PREAGG: B X:0 #main# 1 0
 int main() { return 0; }

clang-tools-extra/clangd/XRefs.cpp

@@ -1106,11 +1106,11 @@ class FindControlFlow : public RecursiveASTVisitor<FindControlFlow> {
     return true;
   }
   bool VisitBreakStmt(BreakStmt *B) {
-    found(Break, B->getBreakLoc());
+    found(Break, B->getKwLoc());
     return true;
   }
   bool VisitContinueStmt(ContinueStmt *C) {
-    found(Continue, C->getContinueLoc());
+    found(Continue, C->getKwLoc());
     return true;
   }
   bool VisitSwitchCase(SwitchCase *C) {

clang/docs/ReleaseNotes.rst

@@ -134,6 +134,7 @@ C Language Changes
 
 C2y Feature Support
 ^^^^^^^^^^^^^^^^^^^
+- Clang now supports `N3355 <https://www.open-std.org/jtc1/sc22/wg14/www/docs/n3355.htm>`_ Named Loops.
 
 C23 Feature Support
 ^^^^^^^^^^^^^^^^^^^

clang/include/clang/AST/JSONNodeDumper.h

@@ -333,6 +333,7 @@ class JSONNodeDumper
   void VisitStringLiteral(const StringLiteral *SL);
   void VisitCXXBoolLiteralExpr(const CXXBoolLiteralExpr *BLE);
 
+  void VisitLoopControlStmt(const LoopControlStmt *LS);
   void VisitIfStmt(const IfStmt *IS);
   void VisitSwitchStmt(const SwitchStmt *SS);
   void VisitCaseStmt(const CaseStmt *CS);

clang/include/clang/AST/Stmt.h

@@ -277,24 +277,14 @@ class alignas(void *) Stmt {
     SourceLocation GotoLoc;
   };
 
-  class ContinueStmtBitfields {
-    friend class ContinueStmt;
+  class LoopControlStmtBitfields {
+    friend class LoopControlStmt;
 
     LLVM_PREFERRED_TYPE(StmtBitfields)
     unsigned : NumStmtBits;
 
-    /// The location of the "continue".
-    SourceLocation ContinueLoc;
-  };
-
-  class BreakStmtBitfields {
-    friend class BreakStmt;
-
-    LLVM_PREFERRED_TYPE(StmtBitfields)
-    unsigned : NumStmtBits;
-
-    /// The location of the "break".
-    SourceLocation BreakLoc;
+    /// The location of the "continue"/"break".
+    SourceLocation KwLoc;
   };
 
   class ReturnStmtBitfields {
@@ -1325,8 +1315,7 @@ class alignas(void *) Stmt {
     DoStmtBitfields DoStmtBits;
     ForStmtBitfields ForStmtBits;
     GotoStmtBitfields GotoStmtBits;
-    ContinueStmtBitfields ContinueStmtBits;
-    BreakStmtBitfields BreakStmtBits;
+    LoopControlStmtBitfields LoopControlStmtBits;
     ReturnStmtBitfields ReturnStmtBits;
     SwitchCaseBitfields SwitchCaseBits;
 
@@ -2184,6 +2173,14 @@ class LabelStmt : public ValueStmt {
   SourceLocation getBeginLoc() const { return getIdentLoc(); }
   SourceLocation getEndLoc() const LLVM_READONLY { return SubStmt->getEndLoc();}
 
+  /// Look through nested labels and return the first non-label statement; e.g.
+  /// if this is 'a:' in 'a: b: c: for(;;)', this returns the for loop.
+  const Stmt *getInnermostLabeledStmt() const;
+  Stmt *getInnermostLabeledStmt() {
+    return const_cast<Stmt *>(
+        const_cast<const LabelStmt *>(this)->getInnermostLabeledStmt());
+  }
+
   child_range children() { return child_range(&SubStmt, &SubStmt + 1); }
 
   const_child_range children() const {
@@ -3056,26 +3053,53 @@ class IndirectGotoStmt : public Stmt {
   }
 };
 
-/// ContinueStmt - This represents a continue.
-class ContinueStmt : public Stmt {
-public:
-  ContinueStmt(SourceLocation CL) : Stmt(ContinueStmtClass) {
-    setContinueLoc(CL);
+/// Base class for BreakStmt and ContinueStmt.
+class LoopControlStmt : public Stmt {
+  /// If this is a named break/continue, the label whose statement we're
+  /// targeting, as well as the source location of the label after the
+  /// keyword; for example:
+  ///
+  ///   a: // <-- TargetLabel
+  ///   for (;;)
+  ///     break a; // <-- LabelLoc
+  ///
+  LabelDecl *TargetLabel = nullptr;
+  SourceLocation LabelLoc;
+
+protected:
+  LoopControlStmt(StmtClass Class, SourceLocation Loc, SourceLocation LabelLoc,
+                  LabelDecl *Target)
+      : Stmt(Class), TargetLabel(Target), LabelLoc(LabelLoc) {
+    setKwLoc(Loc);
   }
 
-  /// Build an empty continue statement.
-  explicit ContinueStmt(EmptyShell Empty) : Stmt(ContinueStmtClass, Empty) {}
+  LoopControlStmt(StmtClass Class, SourceLocation Loc)
+      : LoopControlStmt(Class, Loc, SourceLocation(), nullptr) {}
 
-  SourceLocation getContinueLoc() const { return ContinueStmtBits.ContinueLoc; }
-  void setContinueLoc(SourceLocation L) { ContinueStmtBits.ContinueLoc = L; }
+  LoopControlStmt(StmtClass Class, EmptyShell ES) : Stmt(Class, ES) {}
 
-  SourceLocation getBeginLoc() const { return getContinueLoc(); }
-  SourceLocation getEndLoc() const { return getContinueLoc(); }
+public:
+  SourceLocation getKwLoc() const { return LoopControlStmtBits.KwLoc; }
+  void setKwLoc(SourceLocation L) { LoopControlStmtBits.KwLoc = L; }
 
-  static bool classof(const Stmt *T) {
-    return T->getStmtClass() == ContinueStmtClass;
+  SourceLocation getBeginLoc() const { return getKwLoc(); }
+  SourceLocation getEndLoc() const {
+    return hasLabelTarget() ? getLabelLoc() : getKwLoc();
   }
 
+  bool hasLabelTarget() const { return TargetLabel != nullptr; }
+
+  SourceLocation getLabelLoc() const { return LabelLoc; }
+  void setLabelLoc(SourceLocation L) { LabelLoc = L; }
+
+  LabelDecl *getLabelDecl() { return TargetLabel; }
+  const LabelDecl *getLabelDecl() const { return TargetLabel; }
+  void setLabelDecl(LabelDecl *S) { TargetLabel = S; }
+
+  /// If this is a named break/continue, get the loop or switch statement
+  /// that this targets.
+  const Stmt *getNamedLoopOrSwitch() const;
+
   // Iterators
   child_range children() {
     return child_range(child_iterator(), child_iterator());
@@ -3084,35 +3108,42 @@ class ContinueStmt : public Stmt {
   const_child_range children() const {
     return const_child_range(const_child_iterator(), const_child_iterator());
   }
-};
 
-/// BreakStmt - This represents a break.
-class BreakStmt : public Stmt {
-public:
-  BreakStmt(SourceLocation BL) : Stmt(BreakStmtClass) {
-    setBreakLoc(BL);
+  static bool classof(const Stmt *T) {
+    StmtClass Class = T->getStmtClass();
+    return Class == ContinueStmtClass || Class == BreakStmtClass;
   }
+};
 
-  /// Build an empty break statement.
-  explicit BreakStmt(EmptyShell Empty) : Stmt(BreakStmtClass, Empty) {}
-
-  SourceLocation getBreakLoc() const { return BreakStmtBits.BreakLoc; }
-  void setBreakLoc(SourceLocation L) { BreakStmtBits.BreakLoc = L; }
+/// ContinueStmt - This represents a continue.
+class ContinueStmt : public LoopControlStmt {
+public:
+  ContinueStmt(SourceLocation CL) : LoopControlStmt(ContinueStmtClass, CL) {}
+  ContinueStmt(SourceLocation CL, SourceLocation LabelLoc, LabelDecl *Target)
+      : LoopControlStmt(ContinueStmtClass, CL, LabelLoc, Target) {}
 
-  SourceLocation getBeginLoc() const { return getBreakLoc(); }
-  SourceLocation getEndLoc() const { return getBreakLoc(); }
+  /// Build an empty continue statement.
+  explicit ContinueStmt(EmptyShell Empty)
+      : LoopControlStmt(ContinueStmtClass, Empty) {}
 
   static bool classof(const Stmt *T) {
-    return T->getStmtClass() == BreakStmtClass;
+    return T->getStmtClass() == ContinueStmtClass;
   }
+};
 
-  // Iterators
-  child_range children() {
-    return child_range(child_iterator(), child_iterator());
-  }
+/// BreakStmt - This represents a break.
+class BreakStmt : public LoopControlStmt {
+public:
+  BreakStmt(SourceLocation BL) : LoopControlStmt(BreakStmtClass, BL) {}
+  BreakStmt(SourceLocation CL, SourceLocation LabelLoc, LabelDecl *Target)
+      : LoopControlStmt(BreakStmtClass, CL, LabelLoc, Target) {}
 
-  const_child_range children() const {
-    return const_child_range(const_child_iterator(), const_child_iterator());
+  /// Build an empty break statement.
+  explicit BreakStmt(EmptyShell Empty)
+      : LoopControlStmt(BreakStmtClass, Empty) {}
+
+  static bool classof(const Stmt *T) {
+    return T->getStmtClass() == BreakStmtClass;
   }
 };
 

clang/include/clang/AST/TextNodeDumper.h

@@ -255,6 +255,7 @@ class TextNodeDumper
   void VisitExpressionTemplateArgument(const TemplateArgument &TA);
   void VisitPackTemplateArgument(const TemplateArgument &TA);
 
+  void VisitLoopControlStmt(const LoopControlStmt *L);
   void VisitIfStmt(const IfStmt *Node);
   void VisitSwitchStmt(const SwitchStmt *Node);
   void VisitWhileStmt(const WhileStmt *Node);

clang/include/clang/Basic/BuiltinsAMDGPU.def

@@ -667,6 +667,9 @@ TARGET_BUILTIN(__builtin_amdgcn_global_load_monitor_b128, "V4iV4i*1Ii", "nc", "g
 TARGET_BUILTIN(__builtin_amdgcn_flat_load_monitor_b32, "ii*0Ii", "nc", "gfx1250-insts")
 TARGET_BUILTIN(__builtin_amdgcn_flat_load_monitor_b64, "V2iV2i*0Ii", "nc", "gfx1250-insts")
 TARGET_BUILTIN(__builtin_amdgcn_flat_load_monitor_b128, "V4iV4i*0Ii", "nc", "gfx1250-insts")
+TARGET_BUILTIN(__builtin_amdgcn_cluster_load_b32, "ii*1Iii", "nc", "gfx1250-insts,wavefrontsize32")
+TARGET_BUILTIN(__builtin_amdgcn_cluster_load_b64, "V2iV2i*1Iii", "nc", "gfx1250-insts,wavefrontsize32")
+TARGET_BUILTIN(__builtin_amdgcn_cluster_load_b128, "V4iV4i*1Iii", "nc", "gfx1250-insts,wavefrontsize32")
 TARGET_BUILTIN(__builtin_amdgcn_global_load_async_to_lds_b8, "vc*1c*3IiIi", "nc", "gfx1250-insts")
 TARGET_BUILTIN(__builtin_amdgcn_global_load_async_to_lds_b32, "vi*1i*3IiIi", "nc", "gfx1250-insts")
 TARGET_BUILTIN(__builtin_amdgcn_global_load_async_to_lds_b64, "vV2i*1V2i*3IiIi", "nc", "gfx1250-insts")