Add Zydis x86/x86-64 disassembler support

.gitmodules

@@ -49,3 +49,7 @@
 [submodule "External/range-v3"]
 	path = External/range-v3
 	url = https://github.com/ericniebler/range-v3.git
+[submodule "External/zydis"]
+	shallow = true
+	path = External/zydis
+	url = https://github.com/zyantific/zydis.git

CMakeLists.txt

@@ -28,6 +28,7 @@ option(ENABLE_LIBCXX "Use LLVM's libc++ instead of the GNU libstdc++" FALSE)
 option(ENABLE_CCACHE "Enable ccache for build caching" TRUE)
 option(ENABLE_VIXL_SIMULATOR "Use the VIXL simulator for emulation (only useful for CI testing)" FALSE)
 option(ENABLE_VIXL_DISASSEMBLER "Enable debug disassembler output with VIXL" FALSE)
+option(ENABLE_ZYDIS "Enable x86/x86-64 guest disassembler output with Zydis" FALSE)
 option(USE_LEGACY_BINFMTMISC "Use legacy method of setting up binfmt_misc" FALSE)
 option(ENABLE_FEXCORE_PROFILER "Enable FEXCore's timeline profiling capabilities" FALSE)
 set(FEXCORE_PROFILER_BACKEND "gpuvis" CACHE STRING "Set which backend to use for FEXCore's profiler")
@@ -344,11 +345,27 @@ if (BUILD_TESTING OR ENABLE_VIXL_DISASSEMBLER OR ENABLE_VIXL_SIMULATOR)
   include_directories(SYSTEM External/vixl/src/)
 endif()
 
+find_package(PkgConfig REQUIRED)
+
+if (ENABLE_ZYDIS)
+  pkg_search_module(Zydis QUIET IMPORTED_TARGET zydis)
+  pkg_search_module(Zycore QUIET IMPORTED_TARGET zycore)
+  if (TARGET PkgConfig::Zydis AND TARGET PkgConfig::Zycore AND NOT CMAKE_CROSSCOMPILING)
+    add_library(Zydis::Zydis ALIAS PkgConfig::Zydis)
+    add_library(Zycore::Zycore ALIAS PkgConfig::Zycore)
+    message(STATUS "Using system Zydis")
+  else()
+    message(STATUS "Using bundled Zydis")
+    set(ZYDIS_BUILD_TOOLS OFF CACHE BOOL "" FORCE)
+    set(ZYDIS_BUILD_EXAMPLES OFF CACHE BOOL "" FORCE)
+    add_subdirectory(External/zydis/)
+  endif()
+endif()
+
 if (ENABLE_FEXCORE_PROFILER AND FEXCORE_PROFILER_BACKEND STREQUAL "TRACY")
   add_subdirectory(External/tracy)
 endif()
 
-find_package(PkgConfig REQUIRED)
 find_package(Python 3.9 REQUIRED COMPONENTS Interpreter)
 
 set(BUILD_SHARED_LIBS OFF)

FEXCore/Source/CMakeLists.txt

@@ -94,6 +94,10 @@ if (ENABLE_VIXL_DISASSEMBLER)
   list(APPEND DEFINES -DVIXL_DISASSEMBLER=1)
 endif()
 
+if (ENABLE_ZYDIS)
+  list(APPEND DEFINES -DZYDIS_DISASSEMBLER=1)
+endif()
+
 if (ARCHITECTURE_arm64 AND HAS_CLANG_PRESERVE_ALL)
   list(APPEND DEFINES "-DFEXCORE_PRESERVE_ALL_ATTR=__attribute__((preserve_all));-DFEXCORE_HAS_PRESERVE_ALL_ATTR=1")
 else()
@@ -106,6 +110,10 @@ if (ENABLE_VIXL_DISASSEMBLER OR ENABLE_VIXL_SIMULATOR)
   list(APPEND LIBS vixl)
 endif()
 
+if (ENABLE_ZYDIS)
+  list(APPEND LIBS Zydis::Zydis)
+endif()
+
 if (NOT MINGW)
   list(APPEND LIBS dl)
 else()

FEXCore/Source/Interface/Config/Config.json.in

@@ -334,13 +334,21 @@
           "STATS": "stats"
         },
         "Desc": [
-          "Allows controlling of the vixl disassembler.",
+          "Allows controlling of the vixl disassembler for generated ARM code.",
           "\toff: No disassembly will be output",
           "\tdispatcher: Will enable disassembly of the JIT dispatcher loop",
           "\tblocks: Will enable disassembly of the translated instruction code blocks",
           "\tstats: Will print stats when disassembling the code"
         ]
       },
+      "X86Disassemble": {
+        "Type": "bool",
+        "Default": "false",
+        "Desc": [
+          "Enables x86/x86-64 guest disassembly output for compiled blocks.",
+          "Requires FEX to be built with -DENABLE_ZYDIS=TRUE"
+        ]
+      },
       "ForceSVEWidth": {
         "Type": "uint32",
         "Default": "0",

FEXCore/Source/Interface/Core/Core.cpp

@@ -9,6 +9,9 @@ desc: Glues Frontend, OpDispatcher and IR Opts & Compilation, LookupCache, Dispa
 */
 
 #include <cstdint>
+#ifdef ZYDIS_DISASSEMBLER
+#include <Zydis/Zydis.h>
+#endif
 #include "Interface/Core/ArchHelpers/Arm64Emitter.h"
 #include "Interface/Core/LookupCache.h"
 #include "Interface/Core/CPUBackend.h"
@@ -539,9 +542,24 @@ ContextImpl::GenerateIR(FEXCore::Core::InternalThreadState* Thread, uint64_t Gue
 
     const auto GPRSize = Thread->OpDispatcher->GetGPROpSize();
 
+#ifdef ZYDIS_DISASSEMBLER
+    const auto ZydisMachineMode = Config.Is64BitMode ? ZYDIS_MACHINE_MODE_LONG_64 : ZYDIS_MACHINE_MODE_LEGACY_32;
+    if (FEXCore::Config::Get_X86DISASSEMBLE()) {
+      const uint64_t DecodedMin = Thread->FrontendDecoder->DecodedMinAddress;
+      const uint64_t DecodedMax = Thread->FrontendDecoder->DecodedMaxAddress;
+      LogMan::Msg::IFmt("Guest x86 Begin (RIP={:#x}, {:#x}-{:#x})", GuestRIP, DecodedMin, DecodedMax);
+    }
+#endif
+
     for (size_t j = 0; j < CodeBlocks->size(); ++j) {
       const FEXCore::Frontend::Decoder::DecodedBlocks& Block = CodeBlocks->at(j);
 
+#ifdef ZYDIS_DISASSEMBLER
+      if (FEXCore::Config::Get_X86DISASSEMBLE() && CodeBlocks->size() > 1) {
+        LogMan::Msg::IFmt("  Block {} Entry={:#x} NumInsts={}", j, Block.Entry, Block.NumInstructions);
+      }
+#endif
+
       bool BlockInForceTSOValidRange = false;
       auto InstForceTSOIt = ForceTSOInstructions.end();
       if (ForceTSOValidRanges.Contains({Block.Entry, Block.Entry + Block.Size})) {
@@ -573,6 +591,19 @@ ContextImpl::GenerateIR(FEXCore::Core::InternalThreadState* Thread, uint64_t Gue
 
         TableInfo = Block.DecodedInstructions[i].TableInfo;
         DecodedInfo = &Block.DecodedInstructions[i];
+
+#ifdef ZYDIS_DISASSEMBLER
+        if (FEXCore::Config::Get_X86DISASSEMBLE()) {
+          const uint8_t* InstBytes = reinterpret_cast<const uint8_t*>(InstAddress);
+          ZydisDisassembledInstruction ZydisInst;
+          if (ZYAN_SUCCESS(ZydisDisassembleIntel(ZydisMachineMode, InstAddress, InstBytes, DecodedInfo->InstSize, &ZydisInst))) {
+            LogMan::Msg::IFmt("    {:#x}: {}", InstAddress, ZydisInst.text);
+          } else {
+            LogMan::Msg::IFmt("    {:#x}: (decode failed, {} bytes)", InstAddress, DecodedInfo->InstSize);
+          }
+        }
+#endif
+
         bool IsLocked = DecodedInfo->Flags & FEXCore::X86Tables::DecodeFlags::FLAG_LOCK;
 
         // Do a partial register cache flush before every instruction. This
@@ -693,6 +724,12 @@ ContextImpl::GenerateIR(FEXCore::Core::InternalThreadState* Thread, uint64_t Gue
       }
     }
 
+#ifdef ZYDIS_DISASSEMBLER
+    if (FEXCore::Config::Get_X86DISASSEMBLE()) {
+      LogMan::Msg::IFmt("Guest x86 End");
+    }
+#endif
+
     Thread->OpDispatcher->Finalize();
 
     Thread->FrontendDecoder->DelayedDisownBuffer();