我的计算机具有:24核,32 GB RAM
我已经了解了一段时间,并了解到在使用Parallel.ForEach
执行繁重的任务时可以利用计算机中的更多/所有内核
但是我从不完全了解如何准确地做到这一点。我将借此机会给出一个实时示例,该示例使用一个包含以下代码的单循环。
现在执行代码需要57秒。
此循环中进行了超过300万次循环/计算。
仅给出此循环是不可能的,因为我没有在此处粘贴的代码,因此无法从我在此处粘贴的代码进行测试。
有没有解决方案,如何使用与我的计算机一样多的内核来执行此代码? (我的计算机中有24个内核)
代码返回这两个列表:
List<String> minusLIST
List<String> plusLIST
public void calculationFunction(Dictionary<String, List<String>> tr1, Dictionary<String, List<ArrayList>> v1, Dictionary<String, bool> v5,
Dictionary<String, int> infoIndex, out List<String> minusLIST, out List<String> plusLIST)
{
System.Threading.Thread.CurrentThread.CurrentCulture = new System.Globalization.CultureInfo("en-US");
double num1 = 0; double num2 = 0; double num3 = 0; double number1 = 0; double number2 = 0; double number3 = 0; double thenum1 = 0; double thenum2 = 0; double thenum3 = 0;
double NUM1 = 0; double NUM2 = 0; double NUM3 = 0; double NUMBER1 = 0; double NUMBER2 = 0; double NUMBER3 = 0; String str = ""; String num11 = ""; String num22 = ""; String num33 = "";
String number11 = ""; String number22 = ""; String number33 = ""; double calc1 = 0;
minusLIST = new List<String>(); plusLIST = new List<String>();
foreach (KeyValuePair<String, List<String>> entry in tr1)
{
List<String> allparams = new List<String>(entry.Value);
if (allparams.Count == 10)
{
if (v5.ContainsKey(allparams[7]) && v5.ContainsKey(allparams[8]) && v5.ContainsKey(allparams[9]))
{
if (v5[allparams[7]] && v5[allparams[8]] && v5[allparams[9]])
{
//Calculate both calc scenarios!
num1 = (double)v1[allparams[0]][infoIndex[allparams[7]]][0]; number1 = (double)v1[allparams[0]][infoIndex[allparams[7]]][1]; thenum1 = (double)v1[allparams[0]][infoIndex[allparams[7]]][2];
num2 = (double)v1[allparams[1]][infoIndex[allparams[8]]][0]; number2 = (double)v1[allparams[1]][infoIndex[allparams[8]]][1]; thenum2 = (double)v1[allparams[1]][infoIndex[allparams[8]]][2];
num3 = (double)v1[allparams[2]][infoIndex[allparams[9]]][0]; number3 = (double)v1[allparams[2]][infoIndex[allparams[9]]][1]; thenum3 = (double)v1[allparams[2]][infoIndex[allparams[9]]][2];
NUM1 = num1; NUM2 = num2; NUM3 = num3; NUMBER1 = number1; NUMBER2 = number2; NUMBER3 = number3;
if (num1 <= 0 || number1 <= 0) { NUM1 = thenum1; NUMBER1 = thenum1; }
if (num2 <= 0 || number2 <= 0) { NUM2 = thenum2; NUMBER2 = thenum2; }
if (num3 <= 0 || number3 <= 0) { NUM3 = thenum3; NUMBER3 = thenum3; }
if (NUM1 > 0 && NUM2 > 0 && NUM3 > 0 && NUMBER1 > 0 && NUMBER2 > 0 && NUMBER3 > 0)
{
str = ""; num11 = ""; num22 = ""; num33 = ""; number11 = ""; number22 = ""; number33 = "";
if (num1 > 0 && num2 > 0 && num3 > 0 && number1 > 0 && number2 > 0 && number3 > 0) { } else { str = string.Format("{0:F10}", thenum1) + " / " + string.Format("{0:F10}", thenum2) + " / " + string.Format("{0:F10}", thenum3); }
if (num1 <= 0) { num11 = "0"; num1 = thenum1; } else { num11 = string.Format("{0:F10}", num1); }
if (num2 <= 0) { num22 = "0"; num2 = thenum2; } else { num22 = string.Format("{0:F10}", num2); }
if (num3 <= 0) { num33 = "0"; num3 = thenum3; } else { num33 = string.Format("{0:F10}", num3); }
if (number1 <= 0) { number11 = "0"; number1 = thenum1; } else { number11 = string.Format("{0:F10}", number1); }
if (number2 <= 0) { number22 = "0"; number2 = thenum2; } else { number22 = string.Format("{0:F10}", number2); }
if (number3 <= 0) { number33 = "0"; number3 = thenum3; } else { number33 = string.Format("{0:F10}", number3); }
//Calculate
if (allparams[6] == "0")
{
calc1 = ((num1 * number2 * number3) - 45) / 10;
}
else
{
calc1 = (((num1 * number2) / number3) + 45) / 10;
}
//String
str = calc1 + "," + allparams[0] + " - " + allparams[1] + " - " + allparams[2] + "," +
allparams[3] + " - " + allparams[4] + " - " + allparams[5] + "," +
num11 + " / " + num22 + " / " + num33 + "," +
number11 + " / " + number22 + " / " + number33 + "," +
str + "," +
calc1 + "%";
if (calc1 > 0) { plusLIST.Add(str); }
else
{
minusLIST.Add(str);
}
}
}
}
}
}
}
答案 0 :(得分:0)
我确实将Parallel.ForEach循环放入代码中。我还注意到我“必须??”放置一个锁对象,因为与其他线程共享变量时,以下列表似乎完全搞砸了。
if (calc1 > 0) { plusLIST2.Add(str); }
else
{
minusLIST2.Add(str);
}
Parallel.ForEach似乎不是线程安全的?
运行此代码后,花了62秒-vs- 57秒(单循环)。
使用24个内核是否不可能更快地做到这一点? 我知道锁对象也会阻塞代码,这是代码自由运行的“瓶颈”?
public void calculationFunction(Dictionary<String, List<String>> tr1, Dictionary<String, List<ArrayList>> v1, Dictionary<String, bool> v5,
Dictionary<String, int> infoIndex, out List<String> minusLIST, out List<String> plusLIST)
{
System.Threading.Thread.CurrentThread.CurrentCulture = new System.Globalization.CultureInfo("en-US");
double num1 = 0; double num2 = 0; double num3 = 0; double number1 = 0; double number2 = 0; double number3 = 0; double thenum1 = 0; double thenum2 = 0; double thenum3 = 0;
double NUM1 = 0; double NUM2 = 0; double NUM3 = 0; double NUMBER1 = 0; double NUMBER2 = 0; double NUMBER3 = 0; String str = ""; String num11 = ""; String num22 = ""; String num33 = "";
String number11 = ""; String number22 = ""; String number33 = ""; double calc1 = 0;
minusLIST = new List<String>(); plusLIST = new List<String>();
List<String> minusLIST2 = new List<String>(); List<String> plusLIST2 = new List<String>(); object lockobj = new object();
Parallel.ForEach(tr1, entry =>
{
List<String> allparams = new List<String>(entry.Value);
if (allparams.Count == 10)
{
if (v5.ContainsKey(allparams[7]) && v5.ContainsKey(allparams[8]) && v5.ContainsKey(allparams[9]))
{
if (v5[allparams[7]] && v5[allparams[8]] && v5[allparams[9]])
{
//Calculate both calc scenarios!
num1 = (double)v1[allparams[0]][infoIndex[allparams[7]]][0]; number1 = (double)v1[allparams[0]][infoIndex[allparams[7]]][1]; thenum1 = (double)v1[allparams[0]][infoIndex[allparams[7]]][2];
num2 = (double)v1[allparams[1]][infoIndex[allparams[8]]][0]; number2 = (double)v1[allparams[1]][infoIndex[allparams[8]]][1]; thenum2 = (double)v1[allparams[1]][infoIndex[allparams[8]]][2];
num3 = (double)v1[allparams[2]][infoIndex[allparams[9]]][0]; number3 = (double)v1[allparams[2]][infoIndex[allparams[9]]][1]; thenum3 = (double)v1[allparams[2]][infoIndex[allparams[9]]][2];
NUM1 = num1; NUM2 = num2; NUM3 = num3; NUMBER1 = number1; NUMBER2 = number2; NUMBER3 = number3;
if (num1 <= 0 || number1 <= 0) { NUM1 = thenum1; NUMBER1 = thenum1; }
if (num2 <= 0 || number2 <= 0) { NUM2 = thenum2; NUMBER2 = thenum2; }
if (num3 <= 0 || number3 <= 0) { NUM3 = thenum3; NUMBER3 = thenum3; }
if (NUM1 > 0 && NUM2 > 0 && NUM3 > 0 && NUMBER1 > 0 && NUMBER2 > 0 && NUMBER3 > 0)
{
lock (lockobj)
{
str = ""; num11 = ""; num22 = ""; num33 = ""; number11 = ""; number22 = ""; number33 = "";
if (num1 > 0 && num2 > 0 && num3 > 0 && number1 > 0 && number2 > 0 && number3 > 0) { } else { str = string.Format("{0:F10}", thenum1) + " / " + string.Format("{0:F10}", thenum2) + " / " + string.Format("{0:F10}", thenum3); }
if (num1 <= 0) { num11 = "0"; num1 = thenum1; } else { num11 = string.Format("{0:F10}", num1); }
if (num2 <= 0) { num22 = "0"; num2 = thenum2; } else { num22 = string.Format("{0:F10}", num2); }
if (num3 <= 0) { num33 = "0"; num3 = thenum3; } else { num33 = string.Format("{0:F10}", num3); }
if (number1 <= 0) { number11 = "0"; number1 = thenum1; } else { number11 = string.Format("{0:F10}", number1); }
if (number2 <= 0) { number22 = "0"; number2 = thenum2; } else { number22 = string.Format("{0:F10}", number2); }
if (number3 <= 0) { number33 = "0"; number3 = thenum3; } else { number33 = string.Format("{0:F10}", number3); }
//Calculate
if (allparams[6] == "0")
{
calc1 = ((num1 * number2 * number3) - 45) / 10;
}
else
{
calc1 = (((num1 * number2) / number3) + 45) / 10;
}
//String
str = calc1 + "," + allparams[0] + " - " + allparams[1] + " - " + allparams[2] + "," +
allparams[3] + " - " + allparams[4] + " - " + allparams[5] + "," +
num11 + " / " + num22 + " / " + num33 + "," +
number11 + " / " + number22 + " / " + number33 + "," +
str + "," +
calc1 + "%";
if (calc1 > 0) { plusLIST2.Add(str); }
else
{
minusLIST2.Add(str);
}
}
}
}
}
}
});
plusLIST = new List<String>(plusLIST2);
minusLIST = new List<String>(minusLIST2);
}
答案 1 :(得分:0)
我成功地在“ Parallel.ForEach(tr1,entry =>”)循环中运行代码。 当我在循环内声明所有变量时。现在,变量没有在线程中混淆。另外,在.Add添加到2个列表中时,我还需要放置一个锁定对象。
现在花了12秒(Parallel.ForEach)-VS- 57秒(单循环) 因此,这提高了4.75倍。
但是我仍然有24个内核,所以我想知道为什么它没有快至少15到20倍?
我该怎么做才能使用尽可能多的/所有内核?
public void calculationFunction(Dictionary<String, List<String>> tr1, Dictionary<String, List<ArrayList>> v1, Dictionary<String, bool> v5,
Dictionary<String, int> infoIndex, out List<String> minusLIST, out List<String> plusLIST)
{
System.Threading.Thread.CurrentThread.CurrentCulture = new System.Globalization.CultureInfo("en-US");
minusLIST = new List<String>(); plusLIST = new List<String>();
List<String> minusLIST2 = new List<String>(); List<String> plusLIST2 = new List<String>(); object lockobj = new object();
Parallel.ForEach(tr1, entry =>
{
List<String> allparams = new List<String>(entry.Value);
if (allparams.Count == 10)
{
if (v5.ContainsKey(allparams[7]) && v5.ContainsKey(allparams[8]) && v5.ContainsKey(allparams[9]))
{
if (v5[allparams[7]] && v5[allparams[8]] && v5[allparams[9]])
{
double num1 = 0; double num2 = 0; double num3 = 0; double number1 = 0; double number2 = 0; double number3 = 0; double thenum1 = 0; double thenum2 = 0; double thenum3 = 0;
double NUM1 = 0; double NUM2 = 0; double NUM3 = 0; double NUMBER1 = 0; double NUMBER2 = 0; double NUMBER3 = 0; String str = ""; String num11 = ""; String num22 = ""; String num33 = "";
String number11 = ""; String number22 = ""; String number33 = ""; double calc1 = 0;
//Calculate both calc scenarios!
num1 = (double)v1[allparams[0]][infoIndex[allparams[7]]][0]; number1 = (double)v1[allparams[0]][infoIndex[allparams[7]]][1]; thenum1 = (double)v1[allparams[0]][infoIndex[allparams[7]]][2];
num2 = (double)v1[allparams[1]][infoIndex[allparams[8]]][0]; number2 = (double)v1[allparams[1]][infoIndex[allparams[8]]][1]; thenum2 = (double)v1[allparams[1]][infoIndex[allparams[8]]][2];
num3 = (double)v1[allparams[2]][infoIndex[allparams[9]]][0]; number3 = (double)v1[allparams[2]][infoIndex[allparams[9]]][1]; thenum3 = (double)v1[allparams[2]][infoIndex[allparams[9]]][2];
NUM1 = num1; NUM2 = num2; NUM3 = num3; NUMBER1 = number1; NUMBER2 = number2; NUMBER3 = number3;
if (num1 <= 0 || number1 <= 0) { NUM1 = thenum1; NUMBER1 = thenum1; }
if (num2 <= 0 || number2 <= 0) { NUM2 = thenum2; NUMBER2 = thenum2; }
if (num3 <= 0 || number3 <= 0) { NUM3 = thenum3; NUMBER3 = thenum3; }
if (NUM1 > 0 && NUM2 > 0 && NUM3 > 0 && NUMBER1 > 0 && NUMBER2 > 0 && NUMBER3 > 0)
{
str = ""; num11 = ""; num22 = ""; num33 = ""; number11 = ""; number22 = ""; number33 = "";
if (num1 > 0 && num2 > 0 && num3 > 0 && number1 > 0 && number2 > 0 && number3 > 0) { } else { str = string.Format("{0:F10}", thenum1) + " / " + string.Format("{0:F10}", thenum2) + " / " + string.Format("{0:F10}", thenum3); }
if (num1 <= 0) { num11 = "0"; num1 = thenum1; } else { num11 = string.Format("{0:F10}", num1); }
if (num2 <= 0) { num22 = "0"; num2 = thenum2; } else { num22 = string.Format("{0:F10}", num2); }
if (num3 <= 0) { num33 = "0"; num3 = thenum3; } else { num33 = string.Format("{0:F10}", num3); }
if (number1 <= 0) { number11 = "0"; number1 = thenum1; } else { number11 = string.Format("{0:F10}", number1); }
if (number2 <= 0) { number22 = "0"; number2 = thenum2; } else { number22 = string.Format("{0:F10}", number2); }
if (number3 <= 0) { number33 = "0"; number3 = thenum3; } else { number33 = string.Format("{0:F10}", number3); }
//Calculate
if (allparams[6] == "0")
{
calc1 = ((num1 * number2 * number3) - 45) / 10;
}
else
{
calc1 = (((num1 * number2) / number3) + 45) / 10;
}
//String
str = calc1 + "," + allparams[0] + " - " + allparams[1] + " - " + allparams[2] + "," +
allparams[3] + " - " + allparams[4] + " - " + allparams[5] + "," +
num11 + " / " + num22 + " / " + num33 + "," +
number11 + " / " + number22 + " / " + number33 + "," +
str + "," +
calc1 + "%";
if (calc1 > 0)
{
lock (lockobj) { plusLIST2.Add(str); }
}
else
{
lock (lockobj) { minusLIST2.Add(str); }
}
}
}
}
}
});
plusLIST = new List<String>(plusLIST2);
minusLIST = new List<String>(minusLIST2);
}