llvm 17 fixes

.github/workflows/ci.yml

@@ -3,11 +3,11 @@ name: VCPKG Continuous Integration
 on:
   # Run this workflow once every 6 hours against the master branch
   schedule:
-   - cron: "0 */6 * * *"
+    - cron: "0 */6 * * *"
 
   push:
     branches:
-      - 'master'
+      - "master"
 
     tags:
       - "*"
@@ -22,13 +22,11 @@ jobs:
       fail-fast: false
       matrix:
         image:
-          - { name: 'ubuntu', tag: '22.04' }
-        llvm: ['15', '16']
+          - { name: "ubuntu", tag: "22.04" }
+        llvm: ["17"]
         compiler:
-          - { CC: 'clang', CXX: 'clang++' }
-          - { CC: 'gcc', CXX: 'g++' }
-        exclude:
-          - { llvm: '15', compiler: { CC: 'gcc', CXX: 'g++' } }
+          - { CC: "clang", CXX: "clang++" }
+          - { CC: "gcc", CXX: "g++" }
 
     env:
       CC: ${{ matrix.compiler.CC }}
@@ -55,6 +53,7 @@ jobs:
         run: |
           git config --global --add safe.directory "$GITHUB_WORKSPACE"
           ./scripts/build.sh --llvm-version ${{ matrix.llvm }}
+          rm -r remill-build/_CPack_Packages
       - name: Build with build-presets script
         shell: bash
         run: |
@@ -110,10 +109,10 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        os: ['macos-12']
-        llvm: ['15', '16']
+        os: ["macos-13"]
+        llvm: ["17"]
 
-    runs-on: macos-12
+    runs-on: macos-13
 
     steps:
       - uses: actions/checkout@v2
@@ -137,7 +136,7 @@ jobs:
         shell: bash
         run: |
           brew install ninja
-          export CMAKE_TOOLCHAIN_FILE=$(pwd)/../lifting-bits-downloads/vcpkg_${{ matrix.os}}_llvm-${{ matrix.llvm }}_xcode-14.2_amd64/scripts/buildsystems/vcpkg.cmake
+          export CMAKE_TOOLCHAIN_FILE=$(pwd)/../lifting-bits-downloads/vcpkg_${{ matrix.os}}_llvm-${{ matrix.llvm }}_xcode-15.0_amd64/scripts/buildsystems/vcpkg.cmake
           export INSTALL_DIR=$(pwd)/remill-preset-install
           ./scripts/build-preset.sh release
       - name: Run tests
@@ -167,12 +166,9 @@ jobs:
           name: ${{ matrix.os }}_llvm${{ matrix.llvm }}_tgz_package
           path: ${{ steps.package_names.outputs.TGZ_PACKAGE_PATH }}
 
-
-
-
   release_packages:
     # Do not run the release procedure if any of the builds has failed
-    needs: [ build_linux, build_mac ]
+    needs: [build_linux, build_mac]
     runs-on: ubuntu-22.04
     if: github.event_name == 'push' && startsWith(github.event.ref, 'refs/tags')
 
@@ -238,20 +234,19 @@ jobs:
           asset_name: remill_macos-12_packages.zip
           asset_content_type: application/gzip
 
-
   Docker_Linux:
     runs-on: ubuntu-latest
     strategy:
       matrix:
-        llvm: ["15", "16"]
+        llvm: ["17"]
         ubuntu: ["22.04"]
     steps:
-    - uses: actions/checkout@v2
-    - name: Build LLVM ${{ matrix.llvm }} on ${{ matrix.ubuntu }}
-      run: |
-        docker build . -t ghcr.io/lifting-bits/remill/remill-llvm${{ matrix.llvm }}-ubuntu${{ matrix.ubuntu }}-amd64:latest -f Dockerfile --build-arg UBUNTU_VERSION=${{ matrix.ubuntu }} --build-arg ARCH=amd64 --build-arg LLVM_VERSION=${{ matrix.llvm }}
-    - name: Test Docker image
-      run: |
-        docker run --rm ghcr.io/lifting-bits/remill/remill-llvm${{ matrix.llvm }}-ubuntu${{ matrix.ubuntu }}-amd64:latest --arch amd64 --ir_out /dev/stdout --bytes c704ba01000000
-        docker run --rm ghcr.io/lifting-bits/remill/remill-llvm${{ matrix.llvm }}-ubuntu${{ matrix.ubuntu }}-amd64:latest --arch aarch64 --ir_out /dev/stdout --address 0x400544 --bytes FD7BBFA90000009000601891FD030091B7FFFF97E0031F2AFD7BC1A8C0035FD6
-        docker run --rm ghcr.io/lifting-bits/remill/remill-llvm${{ matrix.llvm }}-ubuntu${{ matrix.ubuntu }}-amd64:latest --arch aarch32 -ir_out /dev/stderr --bytes 0cd04de208008de504108de500208de508309de504009de500109de5903122e0c20fa0e110109fe5001091e5002081e5040081e50cd08de21eff2fe14000000000000000
+      - uses: actions/checkout@v2
+      - name: Build LLVM ${{ matrix.llvm }} on ${{ matrix.ubuntu }}
+        run: |
+          docker build . -t ghcr.io/lifting-bits/remill/remill-llvm${{ matrix.llvm }}-ubuntu${{ matrix.ubuntu }}-amd64:latest -f Dockerfile --build-arg UBUNTU_VERSION=${{ matrix.ubuntu }} --build-arg ARCH=amd64 --build-arg LLVM_VERSION=${{ matrix.llvm }}
+      - name: Test Docker image
+        run: |
+          docker run --rm ghcr.io/lifting-bits/remill/remill-llvm${{ matrix.llvm }}-ubuntu${{ matrix.ubuntu }}-amd64:latest --arch amd64 --ir_out /dev/stdout --bytes c704ba01000000
+          docker run --rm ghcr.io/lifting-bits/remill/remill-llvm${{ matrix.llvm }}-ubuntu${{ matrix.ubuntu }}-amd64:latest --arch aarch64 --ir_out /dev/stdout --address 0x400544 --bytes FD7BBFA90000009000601891FD030091B7FFFF97E0031F2AFD7BC1A8C0035FD6
+          docker run --rm ghcr.io/lifting-bits/remill/remill-llvm${{ matrix.llvm }}-ubuntu${{ matrix.ubuntu }}-amd64:latest --arch aarch32 -ir_out /dev/stderr --bytes 0cd04de208008de504108de500208de508309de504009de500109de5903122e0c20fa0e110109fe5001091e5002081e5040081e50cd08de21eff2fe14000000000000000

CMakeLists.txt

@@ -56,6 +56,8 @@ message(STATUS "LLVM Libraries: ${llvm_libs}")
 message(STATUS "Found LLVM ${LLVM_PACKAGE_VERSION}")
 message(STATUS "Using LLVMConfig.cmake in: ${LLVM_DIR}")
 
+include_directories(SYSTEM ${LLVM_INCLUDE_DIRS})
+
 string(REPLACE "." ";" LLVM_VERSION_LIST ${LLVM_PACKAGE_VERSION})
 list(GET LLVM_VERSION_LIST 0 LLVM_MAJOR_VERSION)
 list(GET LLVM_VERSION_LIST 1 LLVM_MINOR_VERSION)

Dockerfile

@@ -1,5 +1,5 @@
 # Choose your LLVM version
-ARG LLVM_VERSION=16
+ARG LLVM_VERSION=17
 ARG ARCH=amd64
 ARG UBUNTU_VERSION=22.04
 ARG DISTRO_BASE=ubuntu${UBUNTU_VERSION}

bin/differential_tester_x86/LiftAndCompare.cpp

@@ -16,6 +16,7 @@
 
 #include <glog/logging.h>
 #include <gtest/gtest.h>
+#include <llvm/ADT/StringExtras.h>
 #include <llvm/ExecutionEngine/ExecutionEngine.h>
 #include <llvm/ExecutionEngine/GenericValue.h>
 #include <llvm/ExecutionEngine/Interpreter.h>

include/remill/Arch/Arch.h

@@ -25,7 +25,7 @@
 #pragma clang diagnostic ignored "-Wswitch-enum"
 
 #include <llvm/ADT/SmallVector.h>
-#include <llvm/ADT/Triple.h>
+#include <llvm/TargetParser/Triple.h>
 #include <llvm/IR/DataLayout.h>
 #include <llvm/IR/IRBuilder.h>
 #include <remill/BC/InstructionLifter.h>

lib/Arch/AArch64/Arch.cpp

@@ -16,12 +16,12 @@
 
 #include <gflags/gflags.h>
 #include <glog/logging.h>
-#include <llvm/ADT/Triple.h>
 #include <llvm/IR/Attributes.h>
 #include <llvm/IR/DataLayout.h>
 #include <llvm/IR/Function.h>
 #include <llvm/IR/IRBuilder.h>
 #include <llvm/IR/Module.h>
+#include <llvm/TargetParser/Triple.h>
 
 #include <algorithm>
 #include <cctype>

lib/Arch/Arch.cpp

@@ -20,6 +20,7 @@
 #include <glog/logging.h>
 #include <llvm/ADT/APInt.h>
 #include <llvm/ADT/SmallVector.h>
+#include <llvm/IR/AttributeMask.h>
 #include <llvm/IR/BasicBlock.h>
 #include <llvm/IR/Function.h>
 #include <llvm/IR/IRBuilder.h>

lib/Arch/Name.cpp

@@ -16,7 +16,7 @@
 
 #include "remill/Arch/Name.h"
 
-#include <llvm/ADT/Triple.h>
+#include <llvm/TargetParser/Triple.h>
 
 namespace remill {
 

lib/Arch/Sleigh/AArch32Arch.cpp

@@ -17,12 +17,12 @@
 #include "AArch32Arch.h"
 
 #include <glog/logging.h>
-#include <llvm/ADT/Triple.h>
 #include <llvm/IR/Attributes.h>
 #include <llvm/IR/DataLayout.h>
 #include <llvm/IR/Function.h>
 #include <llvm/IR/IRBuilder.h>
 #include <llvm/IR/Module.h>
+#include <llvm/TargetParser/Triple.h>
 #include <remill/Arch/AArch32/ArchContext.h>
 
 #include <iomanip>

lib/Arch/Sleigh/ARMBase.cpp

@@ -1,10 +1,10 @@
 #include <glog/logging.h>
-#include <llvm/ADT/Triple.h>
 #include <llvm/IR/Attributes.h>
 #include <llvm/IR/DataLayout.h>
 #include <llvm/IR/Function.h>
 #include <llvm/IR/IRBuilder.h>
 #include <llvm/IR/Module.h>
+#include <llvm/TargetParser/Triple.h>
 #include <remill/Arch/AArch32/AArch32Base.h>
 #include <remill/Arch/Name.h>
 #include <remill/BC/ABI.h>

lib/Arch/Sleigh/X86Arch.cpp

@@ -15,12 +15,12 @@
  */
 
 #include <glog/logging.h>
-#include <llvm/ADT/Triple.h>
 #include <llvm/IR/Attributes.h>
 #include <llvm/IR/DataLayout.h>
 #include <llvm/IR/Function.h>
 #include <llvm/IR/IRBuilder.h>
 #include <llvm/IR/Module.h>
+#include <llvm/TargetParser/Triple.h>
 #include <remill/Arch/Context.h>
 #include <remill/Arch/Name.h>
 #include <remill/Arch/X86/X86Base.h>

lib/Arch/X86/Arch.cpp

@@ -15,12 +15,12 @@
  */
 
 #include <glog/logging.h>
-#include <llvm/ADT/Triple.h>
 #include <llvm/IR/Attributes.h>
 #include <llvm/IR/DataLayout.h>
 #include <llvm/IR/Function.h>
 #include <llvm/IR/IRBuilder.h>
 #include <llvm/IR/Module.h>
+#include <llvm/TargetParser/Triple.h>
 #include <remill/Arch/ArchBase.h>  // For `Arch` and `ArchImpl`.
 #include <remill/Arch/X86/X86Base.h>
 

lib/BC/Optimizer.cpp

@@ -17,7 +17,8 @@
 #include "remill/BC/Optimizer.h"
 
 #include <glog/logging.h>
-#include <llvm/ADT/Triple.h>
+#include <llvm/Analysis/CGSCCPassManager.h>
+#include <llvm/Analysis/InlineCost.h>
 #include <llvm/Analysis/TargetLibraryInfo.h>
 #include <llvm/IR/Constants.h>
 #include <llvm/IR/DataLayout.h>
@@ -29,9 +30,15 @@
 #include <llvm/IR/MDBuilder.h>
 #include <llvm/IR/Metadata.h>
 #include <llvm/IR/Module.h>
+#include <llvm/IR/PassManager.h>
 #include <llvm/IR/Type.h>
+#include <llvm/Pass.h>
+#include <llvm/Passes/OptimizationLevel.h>
+#include <llvm/Passes/PassBuilder.h>
+#include <llvm/TargetParser/Triple.h>
 #include <llvm/Transforms/IPO.h>
-#include <llvm/Transforms/IPO/PassManagerBuilder.h>
+#include <llvm/Transforms/IPO/Inliner.h>
+#include <llvm/Transforms/IPO/ModuleInliner.h>
 #include <llvm/Transforms/Scalar.h>
 #include <llvm/Transforms/Utils/Cloning.h>
 #include <llvm/Transforms/Utils/Local.h>
@@ -46,81 +53,39 @@ namespace remill {
 void OptimizeModule(const remill::Arch *arch, llvm::Module *module,
                     std::function<llvm::Function *(void)> generator,
                     OptimizationGuide guide) {
-
-  llvm::legacy::FunctionPassManager func_manager(module);
-  llvm::legacy::PassManager module_manager;
-
-  auto TLI =
-      new llvm::TargetLibraryInfoImpl(llvm::Triple(module->getTargetTriple()));
-
-  TLI->disableAllFunctions();  // `-fno-builtin`.
-
-  llvm::PassManagerBuilder builder;
-  // TODO(alex): Some of the optimization passes that the builder adds still rely on typed pointers
-  // so we cannot use them. We should switch to using the new pass manager and choose which passes
-  // we want.
-  builder.OptLevel = 0;
-  builder.SizeLevel = 0;
-  builder.Inliner = llvm::createFunctionInliningPass(250);
-  builder.LibraryInfo = TLI;  // Deleted by `llvm::~PassManagerBuilder`.
-  builder.DisableUnrollLoops = false;  // Unroll loops!
-#if LLVM_VERSION_NUMBER < LLVM_VERSION(16, 0)
-  builder.RerollLoops = false;
-#endif
-  builder.SLPVectorize = guide.slp_vectorize;
-  builder.LoopVectorize = guide.loop_vectorize;
-  builder.VerifyInput = guide.verify_input;
-  builder.VerifyOutput = guide.verify_output;
-  builder.MergeFunctions = false;
-
-  builder.populateFunctionPassManager(func_manager);
-  builder.populateModulePassManager(module_manager);
-  func_manager.doInitialization();
-  llvm::Function *func = nullptr;
-  while (nullptr != (func = generator())) {
-    func_manager.run(*func);
-  }
-  func_manager.doFinalization();
-  module_manager.run(*module);
+  OptimizeBareModule(module, guide);
 }
 
 // Optimize a normal module. This might not contain special Remill-specific
 // intrinsics functions like `__remill_jump`, etc.
 void OptimizeBareModule(llvm::Module *module, OptimizationGuide guide) {
-  llvm::legacy::FunctionPassManager func_manager(module);
-  llvm::legacy::PassManager module_manager;
 
-  auto TLI =
-      new llvm::TargetLibraryInfoImpl(llvm::Triple(module->getTargetTriple()));
+  llvm::ModuleAnalysisManager mam;
+  llvm::FunctionAnalysisManager fam;
+  llvm::LoopAnalysisManager lam;
+  llvm::CGSCCAnalysisManager cam;
+
+
+  llvm::PipelineTuningOptions opts;
+  opts.InlinerThreshold = 250;
+  llvm::PassBuilder pb(nullptr, opts);
+
+  pb.registerModuleAnalyses(mam);
+  pb.registerFunctionAnalyses(fam);
+  pb.registerLoopAnalyses(lam);
+  pb.registerCGSCCAnalyses(cam);
+  pb.crossRegisterProxies(lam, fam, cam, mam);
+
+  llvm::ModulePassManager mpm;
+  mpm.addPass(llvm::ModuleInlinerPass(llvm::getInlineParams(250)));
 
-  TLI->disableAllFunctions();  // `-fno-builtin`.
+  mpm.run(*module, mam);
 
-  llvm::PassManagerBuilder builder;
-  // TODO(alex): Some of the optimization passes that the builder adds still rely on typed pointers
-  // so we cannot use them. We should switch to using the new pass manager and choose which passes
-  // we want.
-  builder.OptLevel = 0;
-  builder.SizeLevel = 0;
-  builder.Inliner = llvm::createFunctionInliningPass(250);
-  builder.LibraryInfo = TLI;  // Deleted by `llvm::~PassManagerBuilder`.
-  builder.DisableUnrollLoops = false;  // Unroll loops!
-#if LLVM_VERSION_NUMBER < LLVM_VERSION(16, 0)
-  builder.RerollLoops = false;
-#endif
-  builder.SLPVectorize = guide.slp_vectorize;
-  builder.LoopVectorize = guide.loop_vectorize;
-  builder.VerifyInput = guide.verify_input;
-  builder.VerifyOutput = guide.verify_output;
-  builder.MergeFunctions = false;
 
-  builder.populateFunctionPassManager(func_manager);
-  builder.populateModulePassManager(module_manager);
-  func_manager.doInitialization();
-  for (auto &func : *module) {
-    func_manager.run(func);
-  }
-  func_manager.doFinalization();
-  module_manager.run(*module);
+  mam.clear();
+  fam.clear();
+  lam.clear();
+  cam.clear();
 }
 
 }  // namespace remill

lib/BC/Util.cpp

@@ -1205,17 +1205,6 @@ MoveConstantIntoModule(llvm::Constant *c, llvm::Module *dest_module,
         moved_c = ret;
         return ret;
       }
-      case llvm::Instruction::Select: {
-        auto ret = llvm::ConstantExpr::getSelect(
-            MoveConstantIntoModule(ce->getOperand(0), dest_module, value_map,
-                                   type_map),
-            MoveConstantIntoModule(ce->getOperand(1), dest_module, value_map,
-                                   type_map),
-            MoveConstantIntoModule(ce->getOperand(2), dest_module, value_map,
-                                   type_map));
-        moved_c = ret;
-        return ret;
-      }
       case llvm::Instruction::Shl: {
         const auto b = llvm::dyn_cast<llvm::ShlOperator>(ce);
         auto ret = llvm::ConstantExpr::getShl(

lib/OS/OS.cpp

@@ -18,7 +18,7 @@
 
 #include <gflags/gflags.h>
 #include <glog/logging.h>
-#include <llvm/ADT/Triple.h>
+#include <llvm/TargetParser/Triple.h>
 
 namespace remill {