Microsoft MVP성태의 닷넷 이야기
글쓴 사람
정성태 (techsharer at outlook.com)
홈페이지
첨부 파일
(연관된 글이 1개 있습니다.)

C# - 재귀호출을 스택 자료구조와 반복문을 이용해 대체하는 방법을 Paralle.For와 함께?

아래의 글에 대해,

C# - 재귀호출을 스택 자료구조와 반복문을 이용해 대체하는 방법 
; https://www.sysnet.pe.kr/2/0/1599

Parallel.For를 이용한 재귀호출에도 사용이 가능하냐는 질문이 있었습니다. 예제로 보내온 소스코드는 다음과 같은데요.

using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
using System.Threading.Tasks;

namespace MonteCarlo
{
    class Program
    {
        static void Main(string[] args)
        {
            double dblSt = 0, dblEn = 0;

            Console.Write("Input the Start and End Value for Integration: ");
            string strBuff = Console.ReadLine();

            string[] strParams = strBuff.Split(' ');
            dblSt = double.Parse(strParams[0]);
            dblEn = double.Parse(strParams[1]);

            double dblInt = MonteCarloIntegral((x) =>
            {
                return (x + Math.Sin(x));
            }, dblSt, dblEn);
            Console.WriteLine("The Integration Result = {0}", dblInt);

            Console.ReadKey();
        }

        static double MonteCarloIntegral(Func<double, double> Func, double Init, double Fin)
        {
            double dblVolume = Fin - Init, dblIntg = 0;

            if (dblVolume <= 0.05) // Monte Carlo Integration: I = integral f(x) dx = (1/n) * Sum(i = 1 to n) f(x[i])
            {
                Random rndRand = new Random(100);
                long lngTrial = 10000 * ((long) ((Func(Fin) - Func(Init)) / dblVolume) + 1);

                for (long cnt = 0; cnt < lngTrial; cnt++) // Using I = integral f(x) dx = integral f(x(t)) * dx(t)/dt dt
                {
                    double dblRand = rndRand.NextDouble();
                    double dblReal = Math.Sqrt((dblRand * dblVolume * (Fin + Init)) + Math.Pow(Init, 2)); 
                    dblIntg += (Math.Pow(Fin, 2) - Math.Pow(Init, 2)) * Func(dblReal) / (2 * dblReal);
                }

                dblIntg /= lngTrial;
            }
            else
            {
                double[] dblPartSums = new double[8]; 
                double dblPartVol = dblVolume / dblPartSums.Length; 

                Parallel.For(0, dblPartSums.Length, (cnt) =>
                {
                    double dblStart = Init + (cnt * dblPartVol), dblEnd = Init + ((cnt + 1) * dblPartVol); 
                    dblPartSums[cnt] = MonteCarloIntegral(Func, dblStart, dblEnd);
                });

                dblIntg += dblPartSums.Sum();
            }

            return dblIntg;
        }
    }
}

정말 가능한지... 한번 해볼까요? ^^

우선, 예제 코드를 쉽게 이해할 수 있도록 병렬처리를 원래의 for 루프로 돌려놓겠습니다.

double[] dblPartSums = new double[8]; 
double dblPartVol = dblVolume / dblPartSums.Length; 

for (cnt i = 0; cnt < dblPartSums.Length; cnt ++)
{
    double dblStart = Init + (cnt * dblPartVol), dblEnd = Init + ((cnt + 1) * dblPartVol); 
    dblPartSums[cnt] = MonteCarloIntegral(Func, dblStart, dblEnd);
}

dblIntg += dblPartSums.Sum();

그럼, 이제 그다지 어렵지 않게 Stack + 반복문 유형으로 바꿀 수 있습니다.

public struct SnapShotStruct
{
    public double Init;
    public double Fin;

    public double dblVolume;
    public double dblIntg;

    public int stage;
}

static double MonteCarloIntegralLoop(Func<double, double> Func, double Init, double Fin)
{
    double retVal = 0.0;

    Stack<SnapShotStruct> snapshotStack = new Stack<SnapShotStruct>();

    SnapShotStruct currentSnapshot = new SnapShotStruct();
    currentSnapshot.Init = Init;
    currentSnapshot.Fin = Fin;

    currentSnapshot.dblVolume = 0;
    currentSnapshot.dblIntg = 0;

    currentSnapshot.stage = 0;

    snapshotStack.Push(currentSnapshot);

    while (snapshotStack.Count != 0)
    {
        currentSnapshot = snapshotStack.Pop();

        switch (currentSnapshot.stage)
        {
            case 0:
                currentSnapshot.dblVolume = currentSnapshot.Fin - currentSnapshot.Init;

                if (currentSnapshot.dblVolume <= 0.05)
                {
                    Random rndRand = new Random(100);
                    long lngTrial = 10000 * ((long)((Func(currentSnapshot.Fin) - Func(currentSnapshot.Init)) / currentSnapshot.dblVolume) + 1);

                    for (long cnt = 0; cnt < lngTrial; cnt++)
                    {
                        double dblRand = rndRand.NextDouble();
                        double dblReal = Math.Sqrt((dblRand * currentSnapshot.dblVolume * (currentSnapshot.Fin + currentSnapshot.Init)) + Math.Pow(currentSnapshot.Init, 2));
                        currentSnapshot.dblIntg += (Math.Pow(currentSnapshot.Fin, 2) - Math.Pow(currentSnapshot.Init, 2)) * Func(dblReal) / (2 * dblReal);
                    }

                    currentSnapshot.dblIntg /= lngTrial;

                    retVal += currentSnapshot.dblIntg;
                    continue;
                }

                currentSnapshot.stage = 1;
                snapshotStack.Push(currentSnapshot);

                int length = 8;
                double dblPartVol = currentSnapshot.dblVolume / length;

                for (int i = 0; i < length; i++)
                {
                    double dblStrt = currentSnapshot.Init + (i * dblPartVol);
                    double dblEnd = currentSnapshot.Init + ((i + 1) * dblPartVol);

                    SnapShotStruct newSnapShot = new SnapShotStruct();
                    newSnapShot.Init = dblStrt;
                    newSnapShot.Fin = dblEnd;
                    newSnapShot.stage = 0;

                    snapshotStack.Push(newSnapShot);
                }
                break;

            case 1:
                retVal = currentSnapshot.dblIntg + retVal;
                continue;
        }
    }

    return retVal;
}

Stack + 반복문으로 구현된 이 코드를 이제 원래의 Parallel.For로 돌려보면 어떻게 될까요?

int length = 8;
double dblPartVol = currentSnapshot.dblVolume / length;

Parallel.For(0, length, (idx) =>
{
    double dblStrt = currentSnapshot.Init + (idx * dblPartVol);
    double dblEnd = currentSnapshot.Init + ((idx + 1) * dblPartVol);

    SnapShotStruct newSnapShot = new SnapShotStruct();
    newSnapShot.Init = dblStrt;
    newSnapShot.Fin = dblEnd;
    newSnapShot.stage = 0;

    snapshotStack.Push(newSnapShot);
});

별 쓸모가 없죠? ^^ 물론 이렇게 바꾼다고 해서 해답이 달라지는 것은 아닙니다. 하지만, 재귀호출을 Stack + 반복문으로 고치는 것 자체가 함수 호출을 리스트 형식으로 직렬화하는 것과 유사하기 때문에 Parallel.For를 호출하는 것도 역시 목록에 포함시켜 다음에 실행하도록 스케쥴링하는 것에 불과하므로 그다지 효용성이 없습니다.

만약 위의 코드를 굳이 병렬화 되도록 고쳐야 한다면 Parallel.For 내의 코드를 별도의 함수 처리로 바꿔야 하고 이는 다시 메서드의 재귀호출로 이어지기 때문에 결국 바꾸기 전과 후의 차이가 없게 됩니다.

그나저나, 이번 질문을 통해서 새롭게 알게 되는군요. 재귀호출이 loop 내에서 불려지는 경우, 재귀호출 방법으로는 병렬화가 가능하지만 Stack + 반복문으로 처리하는 경우 병렬화 처리가 무의미하다는 것!

(위의 처리 코드는 첨부 파일에 넣어두었으니 참고하세요.)




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

[연관 글]






[최초 등록일: ]
[최종 수정일: 6/23/2021]

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
 



2014-01-20 04시07분
[자유주의] 오호~^^ 루프 내에서의 스택 + 반복문 처리는 병렬화가 무의미하군요. 저 코드를 비재귀적으로 바꿀 수 있을까 하고 고민을 했는데, 비로소 답을 알게 되었네요.^^ 정말로 감사합니다.^^
[guest]

[1]  2  3  4  5  6  7  8  9  10  11  12  13  14  15  ...
NoWriterDateCnt.TitleFile(s)
13945정성태6/7/2025447오류 유형: 960. 파이썬 + conda - mysqlclient 사용 시 "NameError: name '_mysql' is not defined" 에러
13944정성태6/7/2025472오류 유형: 959. The trust relationship between this workstation and the primary domain failed. - 네 번째 이야기
13943정성태6/6/2025691개발 환경 구성: 748. Windows + Foundry Local - 로컬에서 AI 모델 활용
13942정성태6/5/2025876오류 유형: 958. winget 설치 시 "0x80d02002 : unknown error"
13941정성태6/2/20251028닷넷: 2334. C# - cpuid 명령어를 이용한 CPU 제조사 문자열 가져오기파일 다운로드1
13940정성태6/1/20251416C/C++: 188. C++의 32비트 + Release 어셈블리 코드를 .NET으로 포팅할 때 주의할 점파일 다운로드1
13939정성태5/29/20251707오류 유형: 957. NVIDIA Triton Inference Server - version `GLIBCXX_3.4.32' not found (required by /opt/tritonserver/backends/python/triton_python_backend_stub)
13938정성태5/29/20251435개발 환경 구성: 747. 파이썬 - WSL/docker에 구성한 Triton 예제 개발 환경
13937정성태5/24/20251359개발 환경 구성: 746. Windows + WSL2 환경에서 (tensorflow 등의) NVIDIA GPU 인식
13936정성태5/23/20251188개발 환경 구성: 745. Linux / WSL 환경에 Miniconda 설치하기
13935정성태5/20/20251233파이썬 - pip 사용 시 "ImportError: cannot import name 'html5lib' from 'pip._vendor'" 오류
13934정성태5/20/20251710스크립트: 77. 파이썬 - 'urllib.request' 모듈의 명시적/암시적 로딩 차이
13933정성태5/19/20251290오류 유형: 956. Visual Studio 2022가 17.12 버전부터 업데이트 되지 않는다면?
13932정성태5/18/20251501스크립트: 76. 파이썬 - Version 문자열 다루기(semver 패키지)
13931정성태5/17/20251793스크립트: 75. 파이썬 - Cython 기본 예제 및 컴파일
13930정성태5/17/20251491개발 환경 구성: 744. 파이썬 - Windows embeddable package 환경에서 외부 패키지 사용하는 방법(ex: UFO² 환경 구성)
13929정성태5/16/20251516오류 유형: 955. 파이썬 - "Windows embeddable package" REPL 환경에서 "NameError: name 'exit' is not defined"
13928정성태5/15/20251558오류 유형: 954. UFO² - "'Invalid URL (POST /v1/chat/completions/chat/completions)'"
13927정성태5/15/20251545오류 유형: 953. OpenAI - The API request of HOST_AGENT failed: OpenAI API request exceeded rate limit: Error code: 429
13926정성태5/14/20251907개발 환경 구성: 743. LLM과 윈도우의 만남 - Desktop AgentOS UFO² 기본 환경 구성
13925정성태5/12/20252007닷넷: 2333. C# - (Console 유형의 프로젝트에서) Clipboard 연동파일 다운로드1
13924정성태5/8/20251758닷넷: 2332. C# - (JetBrains Omea Reader 대상으로) 런타임 시에 메서드 가로채기 [2]파일 다운로드1
13923정성태5/5/20251504스크립트: 74. 파이썬 - C# - Python.NET의 RunSimpleScript, Exec, Eval 차이점파일 다운로드1
13922정성태5/3/20251757스크립트: 73. 파이썬 - Windows embeddable package 버전에서 tkinter 환경 구성
13921정성태5/3/20252276오류 유형: 952. 듀얼 채널 메모리 정렬을 지키지 않은 컴퓨터의 Windows 비정상 종료 현상(Blue Screen) [2]
[1]  2  3  4  5  6  7  8  9  10  11  12  13  14  15  ...