[ORC][ORC_RT] Integrate ORC platforms with LLJIT and lli

This change enables integrating orc::LLJIT with the ORCv2
platforms (MachOPlatform and ELFNixPlatform) and the compiler-rt orc
runtime. Changes include:

- Adding SPS wrapper functions for the orc runtime's dlfcn emulation
  functions, allowing initialization and deinitialization to be invoked
  by LLJIT.

- Changing the LLJIT code generation default to add UseInitArray so
  that .init_array constructors are generated for ELF platforms.

- Integrating the ORCv2 Platforms into lli, and adding a
  PlatformSupport implementation to the LLJIT instance used by lli which
  implements initialization and deinitialization by calling the new
  wrapper functions in the runtime.

Reviewed By: lhames

Differential Revision: https://reviews.llvm.org/D126492

compiler-rt/lib/orc/CMakeLists.txt

@@ -9,6 +9,7 @@ set(ORC_SOURCES
   macho_platform.cpp
   elfnix_platform.cpp
   run_program_wrapper.cpp
+  dlfcn_wrapper.cpp
   )
 
 # Implementation files for all ORC architectures.

compiler-rt/lib/orc/dlfcn_wrapper.cpp

@@ -0,0 +1,52 @@
+//===- dlfcn_wrapper.cpp --------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+//
+// This file is a part of the ORC runtime support library.
+//
+//===----------------------------------------------------------------------===//
+
+#include "adt.h"
+#include "common.h"
+#include "wrapper_function_utils.h"
+
+#include <vector>
+
+using namespace __orc_rt;
+
+extern "C" const char *__orc_rt_jit_dlerror();
+extern "C" void *__orc_rt_jit_dlopen(const char *path, int mode);
+extern "C" int __orc_rt_jit_dlclose(void *dso_handle);
+
+ORC_RT_INTERFACE __orc_rt_CWrapperFunctionResult
+__orc_rt_jit_dlerror_wrapper(const char *ArgData, size_t ArgSize) {
+  return WrapperFunction<SPSString()>::handle(
+             ArgData, ArgSize,
+             []() { return std::string(__orc_rt_jit_dlerror()); })
+      .release();
+}
+
+ORC_RT_INTERFACE __orc_rt_CWrapperFunctionResult
+__orc_rt_jit_dlopen_wrapper(const char *ArgData, size_t ArgSize) {
+  return WrapperFunction<SPSExecutorAddr(SPSString, int32_t)>::handle(
+             ArgData, ArgSize,
+             [](const std::string &Path, int32_t mode) {
+               return ExecutorAddr::fromPtr(
+                   __orc_rt_jit_dlopen(Path.c_str(), mode));
+             })
+      .release();
+}
+
+ORC_RT_INTERFACE __orc_rt_CWrapperFunctionResult
+__orc_rt_jit_dlclose_wrapper(const char *ArgData, size_t ArgSize) {
+  return WrapperFunction<int32_t(SPSExecutorAddr)>::handle(
+             ArgData, ArgSize,
+             [](ExecutorAddr &DSOHandle) {
+               return __orc_rt_jit_dlclose(DSOHandle.toPtr<void *>());
+             })
+      .release();
+}

compiler-rt/test/orc/TestCases/Darwin/x86-64/lljit-ehframe.cpp

@@ -0,0 +1,15 @@
+// RUN: %clangxx -fPIC -emit-llvm -c -o %t %s
+// RUN: %lli_orc_jitlink -relocation-model=pic %t | FileCheck %s
+
+// CHECK: catch
+
+#include <stdio.h>
+
+int main(int argc, char *argv[]) {
+  try {
+    throw 0;
+  } catch (int X) {
+    puts("catch");
+  }
+  return 0;
+}

compiler-rt/test/orc/TestCases/Darwin/x86-64/lljit-initialize-deinitialize.ll

@@ -0,0 +1,32 @@
+; RUN: %lli_orc_jitlink -relocation-model=pic %s | FileCheck %s
+
+; CHECK: constructor
+; CHECK-NEXT: main
+; CHECK-NEXT: destructor
+
+@__dso_handle = external hidden global i8
+@.str = private unnamed_addr constant [5 x i8] c"main\00", align 1
+@.str.1 = private unnamed_addr constant [12 x i8] c"constructor\00", align 1
+@.str.2 = private unnamed_addr constant [11 x i8] c"destructor\00", align 1
+@llvm.global_ctors = appending global [1 x { i32, void ()*, i8* }] [{ i32, void ()*, i8* } { i32 65535, void ()* @constructor, i8* null }]
+
+define dso_local void @destructor(i8* %0) {
+  %2 = tail call i32 @puts(i8* nonnull dereferenceable(1) getelementptr inbounds ([11 x i8], [11 x i8]* @.str.2, i64 0, i64 0))
+  ret void
+}
+
+declare i32 @__cxa_atexit(void (i8*)*, i8*, i8*)
+
+; Function Attrs: nofree norecurse nounwind uwtable
+define dso_local i32 @main(i32 %0, i8** nocapture readnone %1) local_unnamed_addr #2 {
+  %3 = tail call i32 @puts(i8* nonnull dereferenceable(1) getelementptr inbounds ([5 x i8], [5 x i8]* @.str, i64 0, i64 0))
+  ret i32 0
+}
+
+declare i32 @puts(i8* nocapture readonly)
+
+define internal void @constructor() {
+  %1 = tail call i32 @puts(i8* nonnull dereferenceable(1) getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #5
+  %2 = tail call i32 @__cxa_atexit(void (i8*)* @destructor, i8* null, i8* nonnull @__dso_handle) #5
+  ret void
+}

compiler-rt/test/orc/TestCases/FreeBSD/x86-64/lljit-ehframe.cpp

@@ -0,0 +1,15 @@
+// RUN: %clangxx -fPIC -emit-llvm -c -o %t %s
+// RUN: %lli_orc_jitlink -relocation-model=pic %t | FileCheck %s
+
+// CHECK: catch
+
+#include <stdio.h>
+
+int main(int argc, char *argv[]) {
+  try {
+    throw 0;
+  } catch (int X) {
+    puts("catch");
+  }
+  return 0;
+}

compiler-rt/test/orc/TestCases/FreeBSD/x86-64/lljit-initialize-deinitialize.ll

@@ -0,0 +1,32 @@
+; RUN: %lli_orc_jitlink %s | FileCheck %s
+
+; CHECK: constructor
+; CHECK-NEXT: main
+; CHECK-NEXT: destructor
+
+@__dso_handle = external hidden global i8
+@.str = private unnamed_addr constant [5 x i8] c"main\00", align 1
+@.str.1 = private unnamed_addr constant [12 x i8] c"constructor\00", align 1
+@.str.2 = private unnamed_addr constant [11 x i8] c"destructor\00", align 1
+@llvm.global_ctors = appending global [1 x { i32, void ()*, i8* }] [{ i32, void ()*, i8* } { i32 65535, void ()* @constructor, i8* null }]
+
+define dso_local void @destructor(i8* %0) {
+  %2 = tail call i32 @puts(i8* nonnull dereferenceable(1) getelementptr inbounds ([11 x i8], [11 x i8]* @.str.2, i64 0, i64 0))
+  ret void
+}
+
+declare i32 @__cxa_atexit(void (i8*)*, i8*, i8*)
+
+; Function Attrs: nofree norecurse nounwind uwtable
+define dso_local i32 @main(i32 %0, i8** nocapture readnone %1) local_unnamed_addr #2 {
+  %3 = tail call i32 @puts(i8* nonnull dereferenceable(1) getelementptr inbounds ([5 x i8], [5 x i8]* @.str, i64 0, i64 0))
+  ret i32 0
+}
+
+declare i32 @puts(i8* nocapture readonly)
+
+define internal void @constructor() {
+  %1 = tail call i32 @puts(i8* nonnull dereferenceable(1) getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #5
+  %2 = tail call i32 @__cxa_atexit(void (i8*)* @destructor, i8* null, i8* nonnull @__dso_handle) #5
+  ret void
+}

compiler-rt/test/orc/TestCases/Linux/x86-64/lljit-ehframe.cpp

@@ -0,0 +1,15 @@
+// RUN: %clangxx -fPIC -emit-llvm -c -o %t %s
+// RUN: %lli_orc_jitlink -relocation-model=pic %t | FileCheck %s
+
+// CHECK: catch
+
+#include <stdio.h>
+
+int main(int argc, char *argv[]) {
+  try {
+    throw 0;
+  } catch (int X) {
+    puts("catch");
+  }
+  return 0;
+}

compiler-rt/test/orc/TestCases/Linux/x86-64/lljit-initialize-deinitialize.ll

@@ -0,0 +1,32 @@
+; RUN: %lli_orc_jitlink %s | FileCheck %s
+
+; CHECK: constructor
+; CHECK-NEXT: main
+; CHECK-NEXT: destructor
+
+@__dso_handle = external hidden global i8
+@.str = private unnamed_addr constant [5 x i8] c"main\00", align 1
+@.str.1 = private unnamed_addr constant [12 x i8] c"constructor\00", align 1
+@.str.2 = private unnamed_addr constant [11 x i8] c"destructor\00", align 1
+@llvm.global_ctors = appending global [1 x { i32, void ()*, i8* }] [{ i32, void ()*, i8* } { i32 65535, void ()* @constructor, i8* null }]
+
+define dso_local void @destructor(i8* %0) {
+  %2 = tail call i32 @puts(i8* nonnull dereferenceable(1) getelementptr inbounds ([11 x i8], [11 x i8]* @.str.2, i64 0, i64 0))
+  ret void
+}
+
+declare i32 @__cxa_atexit(void (i8*)*, i8*, i8*)
+
+; Function Attrs: nofree norecurse nounwind uwtable
+define dso_local i32 @main(i32 %0, i8** nocapture readnone %1) local_unnamed_addr #2 {
+  %3 = tail call i32 @puts(i8* nonnull dereferenceable(1) getelementptr inbounds ([5 x i8], [5 x i8]* @.str, i64 0, i64 0))
+  ret i32 0
+}
+
+declare i32 @puts(i8* nocapture readonly)
+
+define internal void @constructor() {
+  %1 = tail call i32 @puts(i8* nonnull dereferenceable(1) getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #5
+  %2 = tail call i32 @__cxa_atexit(void (i8*)* @destructor, i8* null, i8* nonnull @__dso_handle) #5
+  ret void
+}

compiler-rt/test/orc/lit.cfg.py

@@ -13,6 +13,7 @@ def build_invocation(compile_flags):
 
 # Assume that llvm-jitlink is in the config.llvm_tools_dir.
 llvm_jitlink = os.path.join(config.llvm_tools_dir, 'llvm-jitlink')
+lli = os.path.join(config.llvm_tools_dir, 'lli')
 if config.host_os == 'Darwin':
   orc_rt_path = '%s/libclang_rt.orc_osx.a' % config.compiler_rt_libdir
 else:
@@ -30,9 +31,11 @@ def build_invocation(compile_flags):
      build_invocation(config.cxx_mode_flags + [config.target_cflags])))
 config.substitutions.append(
     ('%llvm_jitlink', (llvm_jitlink + ' -orc-runtime=' + orc_rt_path)))
+config.substitutions.append(
+    ('%lli_orc_jitlink', (lli + ' -jit-kind=orc -jit-linker=jitlink -orc-runtime=' + orc_rt_path)))
 
 # Default test suffixes.
-config.suffixes = ['.c', '.cpp', '.S']
+config.suffixes = ['.c', '.cpp', '.S', '.ll']
 
 # Exclude Inputs directories.
 config.excludes = ['Inputs']

llvm/include/llvm/ExecutionEngine/Orc/JITTargetMachineBuilder.h

@@ -125,7 +125,7 @@ class JITTargetMachineBuilder {
   /// Set TargetOptions.
   ///
   /// Note: This operation will overwrite any previously configured options,
-  /// including EmulatedTLS and ExplicitEmulatedTLS which
+  /// including EmulatedTLS, ExplicitEmulatedTLS, and UseInitArray which
   /// the JITTargetMachineBuilder sets by default. Clients are responsible
   /// for re-enabling these overwritten options.
   JITTargetMachineBuilder &setOptions(TargetOptions Options) {

llvm/lib/ExecutionEngine/Orc/ELFNixPlatform.cpp

@@ -263,6 +263,10 @@ ELFNixPlatform::standardRuntimeUtilityAliases() {
   static const std::pair<const char *, const char *>
       StandardRuntimeUtilityAliases[] = {
           {"__orc_rt_run_program", "__orc_rt_elfnix_run_program"},
+          {"__orc_rt_jit_dlerror", "__orc_rt_elfnix_jit_dlerror"},
+          {"__orc_rt_jit_dlopen", "__orc_rt_elfnix_jit_dlopen"},
+          {"__orc_rt_jit_dlclose", "__orc_rt_elfnix_jit_dlclose"},
+          {"__orc_rt_jit_dlsym", "__orc_rt_elfnix_jit_dlsym"},
           {"__orc_rt_log_error", "__orc_rt_log_error_to_stderr"}};
 
   return ArrayRef<std::pair<const char *, const char *>>(

llvm/lib/ExecutionEngine/Orc/JITTargetMachineBuilder.cpp

@@ -19,6 +19,7 @@ JITTargetMachineBuilder::JITTargetMachineBuilder(Triple TT)
     : TT(std::move(TT)) {
   Options.EmulatedTLS = true;
   Options.ExplicitEmulatedTLS = true;
+  Options.UseInitArray = true;
 }
 
 Expected<JITTargetMachineBuilder> JITTargetMachineBuilder::detectHost() {

llvm/lib/ExecutionEngine/Orc/MachOPlatform.cpp

@@ -302,6 +302,10 @@ MachOPlatform::standardRuntimeUtilityAliases() {
   static const std::pair<const char *, const char *>
       StandardRuntimeUtilityAliases[] = {
           {"___orc_rt_run_program", "___orc_rt_macho_run_program"},
+          {"___orc_rt_jit_dlerror", "___orc_rt_macho_jit_dlerror"},
+          {"___orc_rt_jit_dlopen", "___orc_rt_macho_jit_dlopen"},
+          {"___orc_rt_jit_dlclose", "___orc_rt_macho_jit_dlclose"},
+          {"___orc_rt_jit_dlsym", "___orc_rt_macho_jit_dlsym"},
           {"___orc_rt_log_error", "___orc_rt_log_error_to_stderr"}};
 
   return ArrayRef<std::pair<const char *, const char *>>(