Merged main:9c5003cc0c70 into amd-gfx:606d2d42f73e

Local branch amd-gfx 606d2d4 Merged main:81e3e7e5d455 into amd-gfx:207601c46e4e
Remote branch main 9c5003c [RISCV] Implement RISCVInstrInfo::getMemOperandsWithOffsetWidth (llvm#73681)

clang/include/clang/Basic/AttrDocs.td

@@ -7073,7 +7073,7 @@ Parameters annotated with `in` or with no annotation are passed by value from
 the caller to the callee.
 
 Parameters annotated with `out` are written to the argument after the callee
-returns (Note: arguments values passed into `out` parameters _are not_ copied
+returns (Note: arguments values passed into `out` parameters *are not* copied
 into the callee).
 
 Parameters annotated with `inout` are copied into the callee via a temporary,

clang/lib/CodeGen/CGCoroutine.cpp

@@ -129,14 +129,48 @@ static SmallString<32> buildSuspendPrefixStr(CGCoroData &Coro, AwaitKind Kind) {
   return Prefix;
 }
 
-static bool memberCallExpressionCanThrow(const Expr *E) {
-  if (const auto *CE = dyn_cast<CXXMemberCallExpr>(E))
-    if (const auto *Proto =
-            CE->getMethodDecl()->getType()->getAs<FunctionProtoType>())
-      if (isNoexceptExceptionSpec(Proto->getExceptionSpecType()) &&
-          Proto->canThrow() == CT_Cannot)
-        return false;
-  return true;
+// Check if function can throw based on prototype noexcept, also works for
+// destructors which are implicitly noexcept but can be marked noexcept(false).
+static bool FunctionCanThrow(const FunctionDecl *D) {
+  const auto *Proto = D->getType()->getAs<FunctionProtoType>();
+  if (!Proto) {
+    // Function proto is not found, we conservatively assume throwing.
+    return true;
+  }
+  return !isNoexceptExceptionSpec(Proto->getExceptionSpecType()) ||
+         Proto->canThrow() != CT_Cannot;
+}
+
+static bool ResumeStmtCanThrow(const Stmt *S) {
+  if (const auto *CE = dyn_cast<CallExpr>(S)) {
+    const auto *Callee = CE->getDirectCallee();
+    if (!Callee)
+      // We don't have direct callee. Conservatively assume throwing.
+      return true;
+
+    if (FunctionCanThrow(Callee))
+      return true;
+
+    // Fall through to visit the children.
+  }
+
+  if (const auto *TE = dyn_cast<CXXBindTemporaryExpr>(S)) {
+    // Special handling of CXXBindTemporaryExpr here as calling of Dtor of the
+    // temporary is not part of `children()` as covered in the fall through.
+    // We need to mark entire statement as throwing if the destructor of the
+    // temporary throws.
+    const auto *Dtor = TE->getTemporary()->getDestructor();
+    if (FunctionCanThrow(Dtor))
+      return true;
+
+    // Fall through to visit the children.
+  }
+
+  for (const auto *child : S->children())
+    if (ResumeStmtCanThrow(child))
+      return true;
+
+  return false;
 }
 
 // Emit suspend expression which roughly looks like:
@@ -233,7 +267,7 @@ static LValueOrRValue emitSuspendExpression(CodeGenFunction &CGF, CGCoroData &Co
   // is marked as 'noexcept', we avoid generating this additional IR.
   CXXTryStmt *TryStmt = nullptr;
   if (Coro.ExceptionHandler && Kind == AwaitKind::Init &&
-      memberCallExpressionCanThrow(S.getResumeExpr())) {
+      ResumeStmtCanThrow(S.getResumeExpr())) {
     Coro.ResumeEHVar =
         CGF.CreateTempAlloca(Builder.getInt1Ty(), Prefix + Twine("resume.eh"));
     Builder.CreateFlagStore(true, Coro.ResumeEHVar);

clang/lib/Serialization/ASTReaderDecl.cpp

@@ -2032,7 +2032,7 @@ void ASTDeclReader::ReadCXXDefinitionData(
         break;
       case LCK_ByCopy:
       case LCK_ByRef:
-        auto *Var = readDeclAs<VarDecl>();
+        auto *Var = readDeclAs<ValueDecl>();
         SourceLocation EllipsisLoc = readSourceLocation();
         new (ToCapture) Capture(Loc, IsImplicit, Kind, Var, EllipsisLoc);
         ToCapture++;

clang/test/CodeGenCoroutines/coro-init-await-nontrivial-return.cpp

@@ -7,6 +7,11 @@ struct NontrivialType {
   ~NontrivialType() {}
 };
 
+struct NontrivialTypeWithThrowingDtor {
+  ~NontrivialTypeWithThrowingDtor() noexcept(false) {}
+};
+
+namespace can_throw {
 struct Task {
     struct promise_type;
     using handle_type = std::coroutine_handle<promise_type>;
@@ -38,9 +43,66 @@ Task coro_create() {
     co_return;
 }
 
-// CHECK-LABEL: define{{.*}} ptr @_Z11coro_createv(
+// CHECK-LABEL: define{{.*}} ptr @_ZN9can_throw11coro_createEv(
 // CHECK: init.ready:
 // CHECK-NEXT: store i1 true, ptr {{.*}}
-// CHECK-NEXT: call void @_ZN4Task23initial_suspend_awaiter12await_resumeEv(
+// CHECK-NEXT: call void @_ZN9can_throw4Task23initial_suspend_awaiter12await_resumeEv(
 // CHECK-NEXT: call void @_ZN14NontrivialTypeD1Ev(
 // CHECK-NEXT: store i1 false, ptr {{.*}}
+}
+
+template <typename R>
+struct NoexceptResumeTask {
+    struct promise_type;
+    using handle_type = std::coroutine_handle<promise_type>;
+
+    struct initial_suspend_awaiter {
+        bool await_ready() {
+            return false;
+        }
+
+        void await_suspend(handle_type h) {}
+
+        R await_resume() noexcept { return {}; }
+    };
+
+    struct promise_type {
+        void return_void() {}
+        void unhandled_exception() {}
+        initial_suspend_awaiter initial_suspend() { return {}; }
+        std::suspend_never final_suspend() noexcept { return {}; }
+        NoexceptResumeTask get_return_object() {
+            return NoexceptResumeTask{handle_type::from_promise(*this)};
+        }
+    };
+
+    handle_type handler;
+};
+
+namespace no_throw {
+using InitNoThrowTask = NoexceptResumeTask<NontrivialType>;
+
+InitNoThrowTask coro_create() {
+    co_return;
+}
+
+// CHECK-LABEL: define{{.*}} ptr @_ZN8no_throw11coro_createEv(
+// CHECK: init.ready:
+// CHECK-NEXT: call void @_ZN18NoexceptResumeTaskI14NontrivialTypeE23initial_suspend_awaiter12await_resumeEv(
+// CHECK-NEXT: call void @_ZN14NontrivialTypeD1Ev(
+}
+
+namespace throwing_dtor {
+using InitTaskWithThrowingDtor = NoexceptResumeTask<NontrivialTypeWithThrowingDtor>;
+
+InitTaskWithThrowingDtor coro_create() {
+    co_return;
+}
+
+// CHECK-LABEL: define{{.*}} ptr @_ZN13throwing_dtor11coro_createEv(
+// CHECK: init.ready:
+// CHECK-NEXT: store i1 true, ptr {{.*}}
+// CHECK-NEXT: call void @_ZN18NoexceptResumeTaskI30NontrivialTypeWithThrowingDtorE23initial_suspend_awaiter12await_resumeEv(
+// CHECK-NEXT: call void @_ZN30NontrivialTypeWithThrowingDtorD1Ev(
+// CHECK-NEXT: store i1 false, ptr {{.*}}
+}

clang/test/Modules/pr72828.cppm

@@ -0,0 +1,24 @@
+// Test that we can handle capturing structured bindings.
+//
+// RUN: rm -fr %t
+// RUN: mkdir %t
+//
+// RUN: %clang_cc1 -std=c++23 -triple %itanium_abi_triple \
+// RUN:     %s -emit-module-interface -o %t/m.pcm
+// RUN: %clang_cc1 -std=c++23 -triple %itanium_abi_triple \
+// RUN:     -S -emit-llvm -disable-llvm-passes %t/m.pcm \
+// RUN:     -o - | FileCheck %s
+
+export module m;
+
+struct s {
+	int m;
+};
+
+void f() {
+	auto [x] = s();
+	[x] {};
+}
+
+// Check that we can generate the LLVM IR expectedly.
+// CHECK: define{{.*}}@_ZGIW1m

compiler-rt/include/sanitizer/memprof_interface.h

@@ -59,6 +59,12 @@ const char *SANITIZER_CDECL __memprof_default_options(void);
 /// \returns 0 on success.
 int SANITIZER_CDECL __memprof_profile_dump(void);
 
+/// Closes the existing file descriptor, if it is valid and not stdout or
+/// stderr, and resets the internal state such that the profile filename is
+/// reopened on the next profile dump attempt. This can be used to enable
+/// multiple rounds of profiling on the same binary.
+void SANITIZER_CDECL __memprof_profile_reset(void);
+
 #ifdef __cplusplus
 } // extern "C"
 #endif

compiler-rt/lib/memprof/memprof_allocator.cpp

@@ -738,3 +738,13 @@ int __memprof_profile_dump() {
   // detected during the dumping process.
   return 0;
 }
+
+void __memprof_profile_reset() {
+  if (report_file.fd != kInvalidFd && report_file.fd != kStdoutFd &&
+      report_file.fd != kStderrFd) {
+    CloseFile(report_file.fd);
+    // Setting the file descriptor to kInvalidFd ensures that we will reopen the
+    // file when invoking Write again.
+    report_file.fd = kInvalidFd;
+  }
+}

compiler-rt/lib/memprof/memprof_interface_internal.h

@@ -49,6 +49,7 @@ extern uptr __memprof_shadow_memory_dynamic_address;
 SANITIZER_INTERFACE_ATTRIBUTE SANITIZER_WEAK_ATTRIBUTE extern char
     __memprof_profile_filename[1];
 SANITIZER_INTERFACE_ATTRIBUTE int __memprof_profile_dump();
+SANITIZER_INTERFACE_ATTRIBUTE void __memprof_profile_reset();
 
 SANITIZER_INTERFACE_ATTRIBUTE void __memprof_load(uptr p);
 SANITIZER_INTERFACE_ATTRIBUTE void __memprof_store(uptr p);

compiler-rt/test/memprof/TestCases/profile_reset.cpp

@@ -0,0 +1,39 @@
+// Test to ensure that multiple rounds of dumping, using the
+// __memprof_profile_reset interface to close the initial file
+// and cause the profile to be reopened, works as expected.
+
+// RUN: %clangxx_memprof  %s -o %t
+
+// RUN: rm -f %t.log.*
+// RUN: %env_memprof_opts=print_text=true:log_path=%t.log %run %t
+
+// Check both outputs, starting with the renamed initial dump, then remove it so
+// that the second glob matches a single file.
+// RUN: FileCheck %s < %t.log.*.sv
+// RUN: rm -f %t.log.*.sv
+// RUN: FileCheck %s < %t.log.*
+// CHECK: Memory allocation stack id
+
+#include <sanitizer/memprof_interface.h>
+#include <stdio.h>
+
+#include <stdlib.h>
+#include <string.h>
+#include <string>
+int main(int argc, char **argv) {
+  char *x = (char *)malloc(10);
+  memset(x, 0, 10);
+  free(x);
+  __memprof_profile_dump();
+  // Save the initial dump in a different file.
+  std::string origname = __sanitizer_get_report_path();
+  std::string svname = origname + ".sv";
+  rename(origname.c_str(), svname.c_str());
+  // This should cause the current file descriptor to be closed and the
+  // the internal state reset so that the profile filename is reopened
+  // on the next write.
+  __memprof_profile_reset();
+  // This will dump to origname again.
+  __memprof_profile_dump();
+  return 0;
+}

libc/cmake/modules/LLVMLibCFlagRules.cmake

@@ -132,8 +132,9 @@ endfunction(get_fq_dep_list_without_flag)
 # Special flags
 set(FMA_OPT_FLAG "FMA_OPT")
 set(ROUND_OPT_FLAG "ROUND_OPT")
-# SSE2 is the baseline for x86_64, so we add a negative flag to disable it if needed.
-set(DISABLE_SSE2_OPT_FLAG "DISABLE_SSE2_OPT")
+# This flag will define macros that gated away vectorization code such that
+# one can always test the fallback generic code path.
+set(PREFER_GENERIC_FLAG "PREFER_GENERIC")
 
 # Skip FMA_OPT flag for targets that don't support fma.
 if(NOT((LIBC_TARGET_ARCHITECTURE_IS_X86 AND (LIBC_CPU_FEATURES MATCHES "FMA")) OR
@@ -145,8 +146,3 @@ endif()
 if(NOT(LIBC_TARGET_ARCHITECTURE_IS_X86 AND (LIBC_CPU_FEATURES MATCHES "SSE4_2")))
   set(SKIP_FLAG_EXPANSION_ROUND_OPT TRUE)
 endif()
-
-# Skip DISABLE_SSE2_OPT flag for targets that don't support SSE2.
-if(NOT(LIBC_TARGET_ARCHITECTURE_IS_X86 AND (LIBC_CPU_FEATURES MATCHES "SSE2")))
-  set(SKIP_FLAG_EXPANSION_DISABLE_SSE2_OPT TRUE)
-endif()

libc/cmake/modules/LLVMLibCObjectRules.cmake

@@ -18,12 +18,12 @@ function(_get_common_compile_options output_var flags)
     set(ADD_SSE4_2_FLAG TRUE)
   endif()
 
-  list(FIND flags ${DISABLE_SSE2_OPT_FLAG} no_sse2)
-  if(${no_sse2} LESS 0)
-    list(FIND flags "${DISABLE_SSE2_OPT_FLAG}__ONLY" no_sse2)
+  list(FIND flags ${PREFER_GENERIC_FLAG} prefer_generic)
+  if(${prefer_generic} LESS 0)
+    list(FIND flags "${PREFER_GENERIC_FLAG}__ONLY" prefer_generic)
   endif()
-  if((${no_sse2} GREATER -1) AND (LIBC_CPU_FEATURES MATCHES "SSE2"))
-    set(DISABLE_SSE2_FLAG TRUE)
+  if(${prefer_generic} GREATER -1)
+    set(ADD_PREFER_GENERIC_FLAG TRUE)
   endif()
 
   set(compile_options ${LIBC_COMPILE_OPTIONS_DEFAULT} ${ARGN})
@@ -66,17 +66,17 @@ function(_get_common_compile_options output_var flags)
     if(ADD_SSE4_2_FLAG)
       list(APPEND compile_options "-msse4.2")
     endif()
-    if(DISABLE_SSE2_FLAG)
-      list(APPEND compile_options "-mno-sse2")
+    if(ADD_PREFER_GENERIC_FLAG)
+      list(APPEND compile_options "-D__LIBC_PREFER_GENERIC")
     endif()
   elseif(MSVC)
     list(APPEND compile_options "/EHs-c-")
     list(APPEND compile_options "/GR-")
     if(ADD_FMA_FLAG)
       list(APPEND compile_options "/arch:AVX2")
     endif()
-    if(DISABLE_SSE2_FLAG)
-      list(APPEND compile_options "/arch:SSE")
+    if(ADD_PREFER_GENERIC_FLAG)
+      list(APPEND compile_options "/D__LIBC_PREFER_GENERIC")
     endif()
   endif()
   if (LIBC_TARGET_ARCHITECTURE_IS_GPU)

libc/config/linux/aarch64/entrypoints.txt

@@ -130,14 +130,6 @@ set(TARGET_LIBC_ENTRYPOINTS
     #libc.src.stdio.scanf
     #libc.src.stdio.fscanf
 
-    # search.h entrypoints
-    libc.src.search.hcreate
-    libc.src.search.hcreate_r
-    libc.src.search.hsearch
-    libc.src.search.hsearch_r
-    libc.src.search.hdestroy
-    libc.src.search.hdestroy_r
-
     # sys/mman.h entrypoints
     libc.src.sys.mman.madvise
     libc.src.sys.mman.mmap
@@ -459,6 +451,14 @@ if(LLVM_LIBC_FULL_BUILD)
     libc.src.signal.sigfillset
     libc.src.signal.signal
 
+    # search.h entrypoints
+    libc.src.search.hcreate
+    libc.src.search.hcreate_r
+    libc.src.search.hsearch
+    libc.src.search.hsearch_r
+    libc.src.search.hdestroy
+    libc.src.search.hdestroy_r
+
     # threads.h entrypoints
     libc.src.threads.call_once
     libc.src.threads.cnd_broadcast

libc/config/linux/arm/entrypoints.txt

@@ -88,15 +88,7 @@ set(TARGET_LIBC_ENTRYPOINTS
     libc.src.stdlib.strtoll
     libc.src.stdlib.strtoul
     libc.src.stdlib.strtoull
-
-    # search.h entrypoints
-    libc.src.search.hcreate
-    libc.src.search.hcreate_r
-    libc.src.search.hsearch
-    libc.src.search.hsearch_r
-    libc.src.search.hdestroy
-    libc.src.search.hdestroy_r
-
+
     # sys/mman.h entrypoints
     libc.src.sys.mman.mmap
     libc.src.sys.mman.munmap