the coupe does 190 but it feels like 60 mph

include/vmctx.hpp

@@ -3,27 +3,28 @@
 #include <vmhandlers.hpp>
 #include <vmp2.hpp>
 
-namespace vm
-{
+namespace vm {
 /// <summary>
-    /// vm::ctx_t class is used to auto generate vm_entry, calc_jmp, and other per-vm entry information...
-    /// creating a vm::ctx_t object can make it easier to pass around information pertaining to a given vm entry...
+/// vm::ctx_t class is used to auto generate vm_entry, calc_jmp, and other
+/// per-vm entry information... creating a vm::ctx_t object can make it easier
+/// to pass around information pertaining to a given vm entry...
 /// </summary>
-    class ctx_t
-    {
+class ctx_t {
  public:
   /// <summary>
-        /// default constructor for vm::ctx_t... all information for a given vm entry must be provided...
+  /// default constructor for vm::ctx_t... all information for a given vm entry
+  /// must be provided...
   /// </summary>
-        /// <param name="module_base">the linear virtual address of the module base...</param>
-        /// <param name="image_base">image base from optional nt header... <a
+  /// <param name="module_base">the linear virtual address of the module
+  /// base...</param> <param name="image_base">image base from optional nt
+  /// header... <a
   /// href="https://docs.microsoft.com/en-us/windows/win32/api/winnt/ns-winnt-image_optional_header64">IMAGE_OPTIONAL_HEADER64</a>...</param>
   /// <param name="image_size">image size from optional nt header... <a
   /// href="https://docs.microsoft.com/en-us/windows/win32/api/winnt/ns-winnt-image_optional_header64">IMAGE_OPTIONAL_HEADER64</a>...</param>
-        /// <param name="vm_entry_rva">relative virtual address from the module base address to the first push prior to
-        /// a vm entry...</param>
-        explicit ctx_t( std::uintptr_t module_base, std::uintptr_t image_base, std::uintptr_t image_size,
-                        std::uintptr_t vm_entry_rva );
+  /// <param name="vm_entry_rva">relative virtual address from the module base
+  /// address to the first push prior to a vm entry...</param>
+  explicit ctx_t(std::uintptr_t module_base, std::uintptr_t image_base,
+                 std::uintptr_t image_size, std::uintptr_t vm_entry_rva);
 
   /// <summary>
   /// init all per-vm entry data such as vm_entry, calc_jmp, and vm handlers...

include/vmutils.hpp

@@ -3,6 +3,7 @@
 #include <Zydis/Zydis.h>
 #include <xmmintrin.h>
 
+#include <memory>
 #include <optional>
 #include <vector>
 
@@ -76,6 +77,23 @@ using zydis_routine_t = std::vector<zydis_instr_t>;
 /// utils used by the other cpp files... misc things that get used a lot...
 /// </summary>
 namespace vm::util {
+
+inline std::shared_ptr<ZydisDecoder> g_decoder = nullptr;
+inline std::shared_ptr<ZydisFormatter> g_formatter = nullptr;
+
+inline void init() {
+  if (!vm::util::g_decoder && !vm::util::g_formatter) {
+    vm::util::g_decoder = std::make_shared<ZydisDecoder>();
+    vm::util::g_formatter = std::make_shared<ZydisFormatter>();
+
+    ZydisDecoderInit(vm::util::g_decoder.get(), ZYDIS_MACHINE_MODE_LONG_64,
+                     ZYDIS_ADDRESS_WIDTH_64);
+
+    ZydisFormatterInit(vm::util::g_formatter.get(),
+                       ZYDIS_FORMATTER_STYLE_INTEL);
+  }
+}
+
 /// <summary>
 /// utils pertaining to native registers...
 /// </summary>

src/vmctx.cpp

@@ -1,27 +1,29 @@
+#include <Zydis/Zydis.h>
+
 #include <vmprofiler.hpp>
 
-namespace vm
-{
-    ctx_t::ctx_t( std::uintptr_t module_base, std::uintptr_t image_base, std::uintptr_t image_size,
-                  std::uintptr_t vm_entry_rva )
-        : module_base( module_base ), image_base( image_base ), image_size( image_size ), vm_entry_rva( vm_entry_rva )
-    {
-    }
+namespace vm {
+ctx_t::ctx_t(std::uintptr_t module_base, std::uintptr_t image_base,
+             std::uintptr_t image_size, std::uintptr_t vm_entry_rva)
+    : module_base(module_base),
+      image_base(image_base),
+      image_size(image_size),
+      vm_entry_rva(vm_entry_rva) {}
 
-    bool ctx_t::init()
-    {
-        if ( !vm::util::flatten( vm_entry, vm_entry_rva + module_base ) )
-            return false;
+bool ctx_t::init() {
+  vm::util::init();
+  if (!vm::util::flatten(vm_entry, vm_entry_rva + module_base)) return false;
 
   vm::util::deobfuscate(vm_entry);
-        if ( !vm::calc_jmp::get( vm_entry, calc_jmp ) )
-            return false;
+  if (!vm::calc_jmp::get(vm_entry, calc_jmp)) return false;
 
   if (auto vm_handler_table = vm::handler::table::get(vm_entry);
-             !vm::handler::get_all( module_base, image_base, vm_entry, vm_handler_table, vm_handlers ) )
+      !vm::handler::get_all(module_base, image_base, vm_entry, vm_handler_table,
+                            vm_handlers))
     return false;
 
-        if ( auto advancement = vm::calc_jmp::get_advancement( calc_jmp ); advancement.has_value() )
+  if (auto advancement = vm::calc_jmp::get_advancement(calc_jmp);
+      advancement.has_value())
     exec_type = advancement.value();
   else
     return false;

src/vmutils.cpp

@@ -60,21 +60,16 @@ std::optional<zydis_routine_t::iterator> get_fetch_operand(
 
 void print(const zydis_decoded_instr_t &instr) {
   char buffer[256];
-  ZydisFormatter formatter;
-  ZydisFormatterInit(&formatter, ZYDIS_FORMATTER_STYLE_INTEL);
-  ZydisFormatterFormatInstruction(&formatter, &instr, buffer, sizeof(buffer),
-                                  0u);
+  ZydisFormatterFormatInstruction(vm::util::g_formatter.get(), &instr, buffer,
+                                  sizeof(buffer), 0u);
   std::puts(buffer);
 }
 
 void print(zydis_routine_t &routine) {
   char buffer[256];
-  ZydisFormatter formatter;
-  ZydisFormatterInit(&formatter, ZYDIS_FORMATTER_STYLE_INTEL);
-
   for (auto [instr, raw, addr] : routine) {
-    ZydisFormatterFormatInstruction(&formatter, &instr, buffer, sizeof(buffer),
-                                    addr);
+    ZydisFormatterFormatInstruction(vm::util::g_formatter.get(), &instr, buffer,
+                                    sizeof(buffer), addr);
     std::printf("> %p %s\n", addr, buffer);
   }
 }
@@ -87,15 +82,12 @@ bool is_jmp(const zydis_decoded_instr_t &instr) {
 bool flatten(zydis_routine_t &routine, std::uintptr_t routine_addr,
              bool keep_jmps, std::uint32_t max_instrs,
              std::uintptr_t module_base) {
-  ZydisDecoder decoder;
   zydis_decoded_instr_t instr;
-
-  ZydisDecoderInit(&decoder, ZYDIS_MACHINE_MODE_LONG_64,
-                   ZYDIS_ADDRESS_WIDTH_64);
-
   std::uint32_t instr_cnt = 0u;
+
   while (ZYAN_SUCCESS(ZydisDecoderDecodeBuffer(
-      &decoder, reinterpret_cast<void *>(routine_addr), 0x1000, &instr))) {
+      vm::util::g_decoder.get(), reinterpret_cast<void *>(routine_addr), 0x1000,
+      &instr))) {
     if (++instr_cnt > max_instrs) return false;
     // detect if we have already been at this instruction... if so that means
     // there is a loop and we are going to just return...
@@ -116,9 +108,9 @@ bool flatten(zydis_routine_t &routine, std::uintptr_t routine_addr,
       }
 
       if (keep_jmps) routine.push_back({instr, raw_instr, routine_addr});
-
       ZydisCalcAbsoluteAddress(&instr, &instr.operands[0], routine_addr,
                                &routine_addr);
+
     } else if (instr.mnemonic == ZYDIS_MNEMONIC_RET) {
       routine.push_back({instr, raw_instr, routine_addr});
       return true;