Microsoft MVP성태의 닷넷 이야기
.NET Framework: 517. calli IL 호출이 DllImport 호출보다 빠를까요? [링크 복사], [링크+제목 복사]
조회: 15153
글쓴 사람
정성태 (techsharer at outlook.com)
홈페이지
첨부 파일
(연관된 글이 2개 있습니다.)

calli IL 호출이 DllImport 호출보다 빠를까요?

네이티브 함수를 닷넷에서 직접 호출할 수 있는 IL 코드가 바로 calli인데요.

OpCodes.Calli Field
; https://learn.microsoft.com/en-us/dotnet/api/system.reflection.emit.opcodes.calli

예전에 OpenCover 코드 커버리지 도구를 살펴보면서 calli에 대한 동작 방식을 설명한 적도 있습니다. ^^

OpenCover 코드 커버리지 도구의 동작방식을 통해 살펴보는 Calli IL 코드 사용법
; https://www.sysnet.pe.kr/2/0/2882

근데... 이것이 일반적인 delegate 호출보다 성능이 더 좋다고 합니다.

Why is Calli Faster Than a Delegate Call?
; http://stackoverflow.com/questions/5893024/why-is-calli-faster-than-a-delegate-call

덧글에 보면 "The EmitCalli is faster because it is a raw byte code call"라고 하는데, 그럼 혹시 "Managed to Unmanaged"간의 전환이 좀 더 가벼울 수도 있지 않을까요?




그래서, 가정을 해봤습니다. .NET에서 native 코드 호출 방법이 대표적으로 DllImport가 있는데요. 그것과 비교해 보면 어떨까요? 예제를 간단히 하기 위해 GetCurrentThreadId Win32 API를 대상으로 다음의 3가지 방법에 대해 각각 성능 테스트를 해봤습니다.

  1. C++ DLL 프로젝트에서 GetCurrentThreadId Win32 API의 주소를 반환해 calli로 호출하는 예제
  2. BCL에서 제공되는 AppDomain.GetCurrentThreadId를 호출하는 예제
  3. DllImport로 직접 kernel32.dll의 GetCurrentThreadId를 호출하는 예제

소스 코드는 대충 다음과 같습니다.

// ============== C++ DLL ==============

WIN32PROJECT1_API __int64 __stdcall fnWin32Project1()
{
    return (__int64)&::GetCurrentThreadId;
}


// ============== C# DLL ==============

using System;
using System.Reflection.Emit;
using System.Runtime.InteropServices;

namespace ClassLibrary1
{
    [System.Security.SecuritySafeCritical]
    public class Class1
    {
        [DllImport("Win32Project1.dll", EntryPoint = "fnWin32Project1")]
        static extern long GetThisThreadId32();

        static GetThisThreadIdDelegate _GetThisThreadIdMethod = null;
        delegate int GetThisThreadIdDelegate();

        static Class1()
        {
            long result = 0;

            if (_GetThisThreadIdMethod == null)
            {
                if (IntPtr.Size == 4)
                {
                    result = GetThisThreadId32();
                }

                var type = typeof(Class1);
                DynamicMethod dynamicMethod = new DynamicMethod("", typeof(int), Type.EmptyTypes, type, true);

                var iLGenerator = dynamicMethod.GetILGenerator();

                if (IntPtr.Size == 4)
                {
                    iLGenerator.Emit(OpCodes.Ldc_I4, (int)result);
                }
                else
                {
                    iLGenerator.Emit(OpCodes.Ldc_I8, result);
                }

                iLGenerator.EmitCalli(OpCodes.Calli, CallingConvention.StdCall, typeof(int), Type.EmptyTypes);
                iLGenerator.Emit(OpCodes.Ret);

                GetThisThreadIdDelegate tempDelegate = dynamicMethod.CreateDelegate(typeof(GetThisThreadIdDelegate)) as GetThisThreadIdDelegate;
                _GetThisThreadIdMethod = tempDelegate;
            }
        }

        [System.Security.SecuritySafeCritical]
        public static int GetThisThreadId()
        {
            try
            {
                return _GetThisThreadIdMethod();
            }
            catch
            {
            }

            return 0;
        }
    }
}

// ============== C# EXE ==============

using System;
using System.Diagnostics;
using System.Runtime.InteropServices;

namespace ConsoleApplication1
{
    class Program
    {
        [DllImport("kernel32.dll")]
        static extern int GetCurrentThreadId();

        static void Main(string[] args)
        {
            Action<int, string, Func<int>> action = (loopCount, title, work) =>
            {
                Stopwatch st = new Stopwatch();
                st.Start();

                Random rand = new Random(Environment.TickCount);

                for (int i = 0; i < loopCount; i++)
                {
                    work();
                }

                st.Stop();

                Console.WriteLine(title + " : " + st.ElapsedTicks);
            };

            action(1, "touch-JIT", ClassLibrary1.Class1.GetThisThreadId);
            action(1, "touch-JIT", AppDomain.GetCurrentThreadId); // 이 코드는 닷넷 프레임워크 환경에서 테스트한 것입니다. (참고: 닷넷 런타임에 따라 달라지는 AppDomain.GetCurrentThreadId의 반환값)
            action(1, "touch-JIT", GetCurrentThreadId);

            Console.WriteLine();

            action(10000, "Calli", ClassLibrary1.Class1.GetThisThreadId);
            action(10000, "BCL", AppDomain.GetCurrentThreadId);
            action(10000, "DllImport", GetCurrentThreadId);
        }
    }
}

결과는 Calli가 코드가 복잡한 거에 비하면 감동적인 수준은 아니지만 그래도 빨라지긴 했습니다.

// x86 Release 빌드로 테스트 (낮을수록 좋음)
Calli : 851
BCL : 2170
DllImport : 1260

Calli : 843
BCL : 1303
DllImport : 1577

Calli : 881
BCL : 1445
DllImport : 1247

(첨부한 파일은 위의 코드 테스트를 포함합니다.)



결론은, 성능을 조금이라도 개선할 필요가 있을 때만 calli 호출을 쓰시고 그 외에는 그냥 DllImport를 쓰시는 것이 좋습니다. 또는, 특수한 사례로 네이티브 코드를 제공할 때 symbol export가 안되도록 숨기고 싶은 경우에 한해 calli를 쓰는 것도 좋겠고. ^^




[이 글에 대해서 여러분들과 의견을 공유하고 싶습니다. 틀리거나 미흡한 부분 또는 의문 사항이 있으시면 언제든 댓글 남겨주십시오.]

[연관 글]






[최초 등록일: ]
[최종 수정일: 3/20/2023]

Creative Commons License
이 저작물은 크리에이티브 커먼즈 코리아 저작자표시-비영리-변경금지 2.0 대한민국 라이센스에 따라 이용하실 수 있습니다.
by SeongTae Jeong, mailto:techsharer at outlook.com

비밀번호

댓글 작성자
 



2020-02-27 10시25분
{
    // public delegate void VoidIntPtrDelegate(IntPtr objAddress);

    var type = typeof(Program);
    DynamicMethod dynamicMethod = new DynamicMethod("", typeof(void), new Type[] { typeof(IntPtr) }, type, true);

    var iLGenerator = dynamicMethod.GetILGenerator();

    iLGenerator.Emit(OpCodes.Ldarg_0);

    if (IntPtr.Size == 4)
    {
        iLGenerator.Emit(OpCodes.Ldc_I4, ...32비트 코드 주소...);
    }
    else
    {
        iLGenerator.Emit(OpCodes.Ldc_I8, ...645비트 코드 주소...);
    }

    iLGenerator.EmitCalli(OpCodes.Calli, CallingConvention.StdCall, typeof(void), new Type[] { typeof(IntPtr) });
    iLGenerator.Emit(OpCodes.Ret);

    VoidIntPtrDelegate tempDelegate = dynamicMethod.CreateDelegate(typeof(VoidIntPtrDelegate)) as VoidIntPtrDelegate;
}
정성태

1  2  3  4  5  [6]  7  8  9  10  11  12  13  14  15  ...
NoWriterDateCnt.TitleFile(s)
13476정성태12/8/20232211닷넷: 2181. C# - .NET 8 JsonStringEnumConverter의 AOT를 위한 개선파일 다운로드1
13475정성태12/7/20232281닷넷: 2180. .NET 8 - 함수 포인터에 대한 Reflection 정보 조회파일 다운로드1
13474정성태12/6/20232136개발 환경 구성: 690. 닷넷 코어/5+ 버전의 ilasm/ildasm 실행 파일 구하는 방법 - 두 번째 이야기
13473정성태12/5/20232330닷넷: 2179. C# - 값 형식(Blittable)을 메모리 복사를 이용해 바이트 배열로 직렬화/역직렬화파일 다운로드1
13472정성태12/4/20232156C/C++: 164. Visual C++ - InterlockedCompareExchange128 사용 방법
13471정성태12/4/20232185Copilot - To enable GitHub Copilot, authorize this extension using GitHub's device flow
13470정성태12/2/20232487닷넷: 2178. C# - .NET 8부터 COM Interop에 대한 자동 소스 코드 생성 도입파일 다운로드1
13469정성태12/1/20232199닷넷: 2177. C# - (Interop DLL 없이) CoClass를 이용한 COM 개체 생성 방법파일 다운로드1
13468정성태12/1/20232190닷넷: 2176. C# - .NET Core/5+부터 달라진 RCW(Runtime Callable Wrapper) 대응 방식파일 다운로드1
13467정성태11/30/20232212오류 유형: 882. C# - Unhandled exception. System.Runtime.InteropServices.COMException (0x800080A5)파일 다운로드1
13466정성태11/29/20232401닷넷: 2175. C# - DllImport 메서드의 AOT 지원을 위한 LibraryImport 옵션
13465정성태11/28/20232155개발 환경 구성: 689. MSBuild - CopyToOutputDirectory가 "dotnet publish" 시에는 적용되지 않는 문제파일 다운로드1
13464정성태11/28/20232277닷넷: 2174. C# - .NET 7부터 UnmanagedCallersOnly 함수 export 기능을 AOT 빌드에 통합파일 다운로드1
13463정성태11/27/20232216오류 유형: 881. Visual Studio - NU1605: Warning As Error: Detected package downgrade
13462정성태11/27/20232251오류 유형: 880. Visual Studio - error CS0246: The type or namespace name '...' could not be found
13461정성태11/26/20232272닷넷: 2173. .NET Core 3/5+ 기반의 COM Server를 registry 등록 없이 사용하는 방법파일 다운로드1
13460정성태11/26/20232225닷넷: 2172. .NET 6+ 기반의 COM Server 내에 Type Library를 내장하는 방법파일 다운로드1
13459정성태11/26/20232212닷넷: 2171. .NET Core 3/5+ 기반의 COM Server를 기존의 regasm처럼 등록하는 방법파일 다운로드1
13458정성태11/26/20232219닷넷: 2170. .NET Core/5+ 기반의 COM Server를 tlb 파일을 생성하는 방법(tlbexp)
13457정성태11/25/20232165VS.NET IDE: 187. Visual Studio - 16.9 버전부터 추가된 "Display inline type hints" 옵션
13456정성태11/25/20232463닷넷: 2169. C# - OpenAI를 사용해 PDF 데이터를 대상으로 OpenAI 챗봇 작성 [1]파일 다운로드1
13455정성태11/25/20232343닷넷: 2168. C# - Azure.AI.OpenAI 패키지로 OpenAI 사용파일 다운로드1
13454정성태11/23/20232686닷넷: 2167. C# - Qdrant Vector DB를 이용한 Embedding 벡터 값 보관/조회 (Azure OpenAI) [1]파일 다운로드1
13453정성태11/23/20232218오류 유형: 879. docker desktop 설치 시 "Invalid JSON string. (Exception from HRESULT: 0x83750007)"
13452정성태11/22/20232296닷넷: 2166. C# - Azure OpenAI API를 이용해 사용자가 제공하는 정보를 대상으로 검색하는 방법파일 다운로드1
13451정성태11/21/20232432닷넷: 2165. C# - Azure OpenAI API를 이용해 ChatGPT처럼 동작하는 콘솔 응용 프로그램 제작파일 다운로드1
1  2  3  4  5  [6]  7  8  9  10  11  12  13  14  15  ...