Reland "[build, vm] Access TLS with less code."

- Avoid TLS initialization checks by using inline initialization.
 - Avoid global offset table indirection by reducing -fPIC to -fPIE.

out/ReleaseXARM64/exe.stripped/dart_precompiled_runtime
11137992 -> 11274776 (-1.21%)

We still need -fPIC in some places because we build a few shared libraries for FFI, so copy some of Fuchsia's GN setup to use -fPIE or -fPIC as appropriate.

Account for older gcc that does not default to -fpie.

TEST=ci
Bug: https://github.com/dart-lang/sdk/issues/51602
Change-Id: I85072153cb1aef9047c1adbf36c7496fbeb11e10
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/286221
Reviewed-by: Daco Harkes <dacoharkes@google.com>
Commit-Queue: Ryan Macnak <rmacnak@google.com>

build/config/BUILD.gn

@@ -18,3 +18,10 @@ config("product") {
     "PRODUCT",
   ]
 }
+
+config("shared_library_config") {
+  if (current_os == "android" || current_os == "fuchsia" ||
+      current_os == "linux" || current_os == "mac") {
+    cflags = [ "-fPIC" ]
+  }
+}

build/config/BUILDCONFIG.gn

@@ -342,7 +342,8 @@ set_defaults("static_library") {
 }
 
 # Shared library defaults (also for components in component mode).
-_shared_library_configs = _native_compiler_configs
+_shared_library_configs =
+    _native_compiler_configs + [ "//build/config:shared_library_config" ]
 if (is_win) {
   _shared_library_configs += _windows_linker_configs
 } else if (is_mac) {

build/config/compiler/BUILD.gn

@@ -249,15 +249,9 @@ config("compiler") {
     ]
   }
 
-  # We need -fPIC:
-  # 1. On ARM for tcmalloc.
-  # 2. On Android.
-  # 3. When using the sanitizers.
-  # Otherwise there is a performance hit, in particular on ia32.
-  if (is_android || is_asan || is_lsan || is_msan || is_tsan || is_ubsan ||
-      (is_linux && current_cpu != "x86") || is_fuchsia) {
-    cflags += [ "-fPIC" ]
-    ldflags += [ "-fPIC" ]
+  if (is_android || is_linux || is_mac || is_fuchsia) {
+    cflags += [ "-fPIE" ]
+    ldflags += [ "-fPIE" ]
   }
 
   # Linux-specific compiler flags setup.

runtime/bin/BUILD.gn

@@ -1036,9 +1036,6 @@ shared_library("entrypoints_verification_test") {
     # The only effect of DART_SHARED_LIB is to export the Dart API.
     "DART_SHARED_LIB",
   ]
-  if (is_linux || is_android) {
-    cflags = [ "-fPIC" ]
-  }
   if (is_win) {
     # TODO(dartbug.com/40579): This wrongly links in dart.exe on precompiled.
     libs = [ "dart.lib" ]
@@ -1055,9 +1052,6 @@ shared_library("ffi_test_dynamic_library") {
     # The only effect of DART_SHARED_LIB is to export the Dart API.
     "DART_SHARED_LIB",
   ]
-  if (is_linux || is_android) {
-    cflags = [ "-fPIC" ]
-  }
   if (is_win) {
     # TODO(dartbug.com/40579): This wrongly links in dart.exe on precompiled.
     libs = [ "dart.lib" ]
@@ -1091,9 +1085,6 @@ shared_library("ffi_test_functions") {
     # The only effect of DART_SHARED_LIB is to export the Dart API.
     "DART_SHARED_LIB",
   ]
-  if (is_linux || is_android) {
-    cflags = [ "-fPIC" ]
-  }
   if (is_win) {
     # TODO(dartbug.com/40579): This wrongly links in dart.exe on precompiled.
     libs = [ "dart.lib" ]

runtime/vm/compiler/assembler/disassembler_test.cc

@@ -21,7 +21,7 @@ ISOLATE_UNIT_TEST_CASE(Disassembler) {
   assembler.Stop("testing disassembler");
 
   // Only verify that the disassembler does not crash.
-  AssemblerTest test("Disassembler", &assembler);
+  AssemblerTest test("Disassembler", &assembler, Thread::Current()->zone());
   test.Assemble();
   Disassembler::Disassemble(test.entry(), test.entry() + assembler.CodeSize());
 }

runtime/vm/os_thread.cc

@@ -20,11 +20,6 @@ OSThread* OSThread::thread_list_head_ = NULL;
 Mutex* OSThread::thread_list_lock_ = NULL;
 bool OSThread::creation_enabled_ = false;
 
-thread_local ThreadState* OSThread::current_vm_thread_ = NULL;
-#if defined(DEBUG)
-thread_local bool ThreadInterruptScope::in_thread_interrupt_scope_ = false;
-#endif
-
 #if defined(SUPPORT_TIMELINE)
 inline void UpdateTimelineTrackMetadata(const OSThread& thread) {
   RecorderSynchronizationLockScope ls;

runtime/vm/os_thread.h

@@ -306,7 +306,9 @@ class OSThread : public BaseThread {
   static OSThread* thread_list_head_;
   static bool creation_enabled_;
 
-  static thread_local ThreadState* current_vm_thread_;
+  // Inline initialization is important for avoiding unnecessary TLS
+  // initialization checks at each use.
+  static inline thread_local ThreadState* current_vm_thread_ = nullptr;
 
   friend class IsolateGroup;  // to access set_thread(Thread*).
   friend class OSThreadIterator;
@@ -412,7 +414,7 @@ class ThreadInterruptScope : public ValueObject {
 
  private:
   ThreadState* saved_current_vm_thread_;
-  static thread_local bool in_thread_interrupt_scope_;
+  static inline thread_local bool in_thread_interrupt_scope_ = false;
 #endif  // DEBUG
 };
 

runtime/vm/unit_test.h

@@ -107,7 +107,7 @@
       if (setjmp(*jump.Set()) == 0) {                                          \
         compiler::ObjectPoolBuilder object_pool_builder;                       \
         compiler::Assembler assembler(&object_pool_builder, far_branch_level); \
-        AssemblerTest test("" #name, &assembler);                              \
+        AssemblerTest test("" #name, &assembler, thread->zone());              \
         AssemblerTestGenerate##name(test.assembler());                         \
         test.Assemble();                                                       \
         AssemblerTestRun##name(&test);                                         \
@@ -504,11 +504,11 @@ struct is_double<double> {
 
 class AssemblerTest {
  public:
-  AssemblerTest(const char* name, compiler::Assembler* assembler)
+  AssemblerTest(const char* name, compiler::Assembler* assembler, Zone* zone)
       : name_(name),
         assembler_(assembler),
-        code_(Code::ZoneHandle()),
-        disassembly_(Thread::Current()->zone()->Alloc<char>(DISASSEMBLY_SIZE)) {
+        code_(Code::ZoneHandle(zone)),
+        disassembly_(zone->Alloc<char>(DISASSEMBLY_SIZE)) {
     ASSERT(name != NULL);
     ASSERT(assembler != NULL);
   }

tests/standalone/pie_test.dart

@@ -0,0 +1,28 @@
+// Copyright (c) 2023, the Dart project authors.  Please see the AUTHORS file
+// for details. All rights reserved. Use of this source code is governed by a
+// BSD-style license that can be found in the LICENSE file.
+
+import "dart:io";
+
+// Mmmm... pie!
+main() {
+  if (!Platform.isLinux) return; // readelf is a Linux tool.
+  // Modern Mac and Android binaries are always PIE.
+  // Fuchsia binaries are always PIE.
+
+  var result = Process.runSync("readelf", ["-h", Platform.resolvedExecutable]);
+  print("stdout:");
+  print(result.stdout);
+  print("stderr:");
+  print(result.stderr);
+
+  if (result.exitCode != 0) {
+    throw "readelf failed";
+  }
+
+  // A position-dependent executable outputs "EXEC (Executable file)".
+  if (!result.stdout.contains("DYN (Position-Independent Executable file)") &&
+      !result.stdout.contains("DYN (Shared object file)")) {
+    throw "Standalone VM should be a position-independent executable";
+  }
+}