CODEWITHSUNDEEP
c#performance-testing
Thanos Darkadakis
Thanos Darkadakis

Reputation: 1729

C# performance of basic operations

I have found several articles (including some questions in SA), trying to find the cost of basic operations.

However, i tried to make my own little program, so that i could test it myself. When trying to test addition and subtraction i encountered something, that i present you in simple code

       int a,i,j;
        DateTime d1,d2;

        a = 0;
        d1= DateTime.Now;
        for (i = 1; i <= 10000; i++)
            for (j = 1; j <= 10000; j++)
            a = a + 1;
        d2=DateTime.Now;
        Console.WriteLine("a=a+1\t1E8\t"+(d2-d1));

        a = 0;
        d1 = DateTime.Now;
        for (i = 1; i <= 10000; i++)
            for (j = 1; j <= 10000; j++)
                a = a + 1;
        d2 = DateTime.Now;
        Console.WriteLine("a=a+1\t1E8\t" + (d2 - d1));

        a = 0;
        d1 = DateTime.Now;
        for (i = 1; i <= 10000; i++)
            for (j = 1; j <= 10000; j++)
                a = a + 1;
        d2 = DateTime.Now;
        Console.WriteLine("a=a+1\t1E8\t" + (d2 - d1));

        a = 0;
        d1 = DateTime.Now;
        for (i = 1; i <= 10000; i++)
            for (j = 1; j <= 10000; j++)
                a = a + 1;
        d2 = DateTime.Now;
        Console.WriteLine("a=a+1\t1E8\t" + (d2 - d1));
        Console.ReadKey();

First i declare some variables and then i perform the same task 4 times. every single time i do the exact same thing. Always , the result is the following:

a=a+1   1E8     00:00:00.2300230
a=a+1   1E8     00:00:00.2180218
a=a+1   1E8     00:00:00.2130213
a=a+1   1E8     00:00:00.2150215

or

a=a+1   1E8     00:00:00.2340000
a=a+1   1E8     00:00:00.2170000
a=a+1   1E8     00:00:00.2130000
a=a+1   1E8     00:00:00.2100000

meaning that the first time, it ALWAYS takes longer to compute (about 5% longer). All the other times, the time is about the same.

Then i just add the line a=0; after the first a=0;. The result is as follows:

a=a+1   1E8     00:00:00.2210000
a=a+1   1E8     00:00:00.2170000
a=a+1   1E8     00:00:00.2200000
a=a+1   1E8     00:00:00.2170000

or even

a=a+1   1E8     00:00:00.2160000
a=a+1   1E8     00:00:00.2160000
a=a+1   1E8     00:00:00.2200000
a=a+1   1E8     00:00:00.2230000

So 3 questions:

  1. Why is the first loop take longer than the rest in the first example?
  2. Why all the loops take about same time in the second example?
  3. Is there any better way to compute time interval than Datetime

Personal POV: Something needs to be initialized when the program starts running. By adding an extra command (a=0), i give the program the time to make its initializations.

Flaw in POV: When i change the number of loops from 10^8 to 10^9, the difference between time1 and time2 should again be ~0.015s(if my POV was valid). but now it is again ~0.15s (meaning again ~5%). If i add the second a=0, then again time1~time2~time3~time4

Upvotes: 2

Views: 156

Answers (1)

milan m
milan m

Reputation: 2254

Answer to your question number 3 -

Is there any better way to compute time interval than Datetime?

Use System.Diagnostics.Stopwatch for measuring time difference as it gives very high accuracy.

Upvotes: 3

Related Questions