Parallelcoding.com » C#

Intel Threading Challenge #1

Robert Green — Fri, 10 Jul 2009 15:16:57 +0000

While I never participated in the Intel Threading Challenge, I still find the problems really intriguing. Why? Because they are problems designed to test threaded development which is not only cool but is also going to play a large part in the future of computing. I would call all this problem for the most part concurrent, not parallel. Why? Read this and you’ll understand completely. Now, on to the challenge!

Problem # 1 states:

Problem description: Given a set of unsorted items with keys that can be considered as a binary representation of an integer, the bits within the key can be used to sort the set of items. This method of sorting is known as Radix Sort.

Write a program that includes a threaded version of a Radix Sort algorithm that sorts the keys read from an input file, then output the sorted keys to another file. The input and output file names shall be the first and second arguments on the command line of the application execution.

The first line of the input text file is the total number of keys (N) to be sorted; this is followed by N keys, one per line, in the file. A key will be a seven-character string made up of printable characters not including the space character (ASCII 0×20). The number of keys within the file is less than 2^31 – 1. Sorted output must be stored in a text file, one key per line.

Timing: If you put timing code into your application to time the sorting process and report the elapsed time, this time will be used for scoring. If no timing code is added, the entire execution time (including time for input and output) will be used for scoring.

Example Input file:
8
H@skell
surVEYs
sysTEMS
HASKELL
Surveys
1234567
SURveys
systEMS

Example Output file:
1234567
H@skell
HASKELL
SURveys
Surveys
surVEYs
sysTEMS
systEMS

My solution (both serial and parallel):

using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
using System.Collections;
using System.Diagnostics;
using System.Threading;
using System.IO;
 
namespace RadixSort {
 
    class Program {
        static void Main(string[] args) {
            StreamReader sr;
            int length;
            TimeSpan serial, parallel;
            sr = File.OpenText(@"C:\Documents and Settings\rgreen\Desktop\Threading\Threading\rsTestK100.dat");
 
            length = Convert.ToInt32(sr.ReadLine().Trim());
            string[] values = new string[length];
            string[] newValues = new string[length];
            for(int x = 0; x < length; x++) {
                values[x] = sr.ReadLine().Trim();
            }
 
            Stopwatch sw = new Stopwatch();
 
            //
            // Serial
            //
            sw.Start();
            RadixSort(values).CopyTo(newValues, 0);
            sw.Stop();
            serial = sw.Elapsed;
 
            //
            // Parallel
            //
            sw.Reset();
            sw.Start();
            ParallelRadixSort(values).CopyTo(newValues, 0);
            sw.Stop();
            parallel = sw.Elapsed;
 
 
            Console.WriteLine("Serial Time: " + serial);
            Console.WriteLine("Parallel Time: " + parallel);
            Console.ReadLine();
        }
 
        public static string[] ParallelRadixSort(string[] array) {
            int length = array.Length;
 
            Parallel.For(0, array[0].Length - 1, delegate(int curRadix) {
                int index = array[0].Length - 1 - curRadix;
                array = new MergeSort(array, array[0].Length - 1 - index).Results;
            });
 
            return array;
        }
 
        public static string[] RadixSort(string[] array) {
 
            int length = array.Length;
 
            for(int curRadix = array[0].Length - 1; curRadix >= 0; curRadix--) {
                array = new MergeSort(array, array[0].Length - 1 - curRadix).Results;
            }
 
            return array;
        }
    }
}

And there you have it!

Project Euler #6

Robert Green — Wed, 01 Jul 2009 13:52:40 +0000

Project Euler problem #6 is

Find the difference between the sum of the squares of the first one hundred natural numbers and the square of the sum.

When I first began looking at this problem I wanted to experiment a bit with the Parallel Extensions for .NET, so I started where any parallel algorithm begins: with a sequential algorithm. The algorithm here is rather trivial: Loop over the natural numbers from 1 to 100. In order to sum the squares I will sum the square of each number. In order to square the sum I will sum the numbers and then square them. The difference is the answer. So, first, I came up with two functions: SumOfSquares and SquareOfSums.

My first attempt at these functions ended up something like this:

static double SumOfSquares(int min, int max) {
    double result = 0;
    for(int x = min; x <= max; x++) {
        result += Math.Pow(x, 2);
     }
     return result;
}
static double SquareOfSums(int min, int max) {
    double result = 0;
    for(int x = min; x <= max; x++) {
        result += x;
    }
    return Math.Pow(result, 2);
}

Both of those functions are very straight forward. My first thought after writing these was, “Hey, why not use some LINQ?” So I did. Here is how the functions change:

static double LinqSumOfSquares(int min, int max) {
    return Enumerable.Range(min, max).Select(d => Math.Pow(d, 2)).Sum();           
}
static double LinqSquareOfSums(int min, int max) {
    return Math.Pow(Enumerable.Range(min, max).Select(d => d).Sum(), 2);    
}

Wow! Talk about incredible, shrinking functions! I love finding a way to make code more succinct, readable, and elegant and these changes seem to have hit the nail on the head! Anyways, that is basically all the pieces for the sequential algorithm. All you have to do is call each of those functions and take the difference. Simple, huh? But the real question is how can we parallelize these bad boys? I have a few thoughts.

Call each function in parallel. In other words let the sumOfSquares and SquareOfSum function run at the same time in different threads.
Parallelize each function invidvidually. In other words leverage the loops inside of each function in order to parallelize them.

So let’s take a look at each method using the Parallel Extensions for .NET. The first method is calling each function in parallel. My first thought here was to use Parallel.Invoke in order to call each function at the same time. A little further research quickly revealed that Parallel.Invoke cannot return any values. My initial response to that: “Well that sucks.” Luckily there’s another class in the Parallel library called Futures. What’s a future? From DevX –

“In TPL terms, a Future is basically a task that returns a value. It’s like a deferred function. You start it running and then use its value later. If the Future hasn’t finished calculating its value by the time you need it, it makes you wait while it finishes.”

Sounds good to me. So how do we use futures? Like this:

static long Problem6Futures(int min, int max) {
 
    Future<double> fSumOfSquares = Future.Create(() => SumOfSquares(min, max));
    Future<double> fSquareOfSums = Future.Create(() => SquareOfSums(min, max));
    double result = fSquareOfSums.Value - fSumOfSquares.Value;
 
    return result;
}

Easy, huh? All you have to remember is that Futures are deferred functions that return values. The result of Future operation gets stored in an object of type Future and the value is stored in Object.Value. So how about the second method of parallelizing this algorithm? Well, it’s even easier because of PLINQ – or Parallel LINQ. Let’s see what it looks like.

static double ParallelSumOfSquares(int min, int max) {
    return Enumerable.Range(min, max).AsParallel().Select(d => Math.Pow(d, 2)).Sum();
}
static double ParallelSquareOfSums(int min, int max) {
    return Math.Pow(Enumerable.Range(min, max).AsParallel().Sum(), 2);
}

Hah! Even easier! All that I did was add .AsParrallel() to our data. That tells LINQ to do the processing in parallel!

So, there’s Project Euler Problem #6 for you. Was it a hard problem? Not really. Are you going to see major performance results through the parallelization of this algorithm? No. But you gotta’ start somewhere when you’re learning how to parallelize algorithms using a new library. Today I used Futures and PLINQ and that sound’s like a pretty solid start to me.

Project Euler #3

Robert Green — Tue, 30 Jun 2009 12:09:06 +0000

Project Euler #3 is :

Find the largest prime factor of a composite number.

My first attempt at this whipped up some typical code that simply brute forced my way to the solution. My code looked like this:

static void Problem3() {
    long n = 600851475143;
    int factor = 2;
    int lastFactor = 1;
 
 
    if(n % 2 == 0) {
        lastFactor = 2;
        n = n / 2;
        while(n % 2 == 0) {
            n = n / 2;
        }
     } else {
         lastFactor = 1;
     }
     factor = 3;
 
     double maxFactor = Math.Sqrt(n);
     while(n > 1 && factor <= maxFactor) {
          if(n % factor == 0) {
              n = n / factor;
              lastFactor = factor;
              while(n % factor == 0) {
                  n = n / factor;
              }
              maxFactor = Math.Sqrt(n);
          }
          factor += 2;
      }
 
      if(n == 1) {
          Console.Write(lastFactor.ToString() + " ");
      } else {
          Console.Write(n.ToString() + " ");
      }
}

Not very elegant, but it works. So I set out to find something a bit prettier and I came across I LINQ solution here. All I did was tack on the .AsParallel() in order to give it a little speed boost. The code looks like:

static void LinqProblem3() {
    long largeNumber = 600851475143;
    var allPrimeFactors = from p in Primes.PrimeFactors(largeNumber).AsParallel()
                                 orderby p descending
                                 select p;
 
    foreach(var f in allPrimeFactors){
        Console.WriteLine(f);
    }
}
 
public static class Primes {
 
    // Find all prime factors.
    public static IEnumerable<int> PrimeFactors(long number) {
        // Start by removing the lowest prime (2)
        return MorePrimeFactors(number, 2);
    }
 
    // This recursive method finds all prime factors.
    private static IEnumerable<int> MorePrimeFactors(long number, int factor) {
        // Find the next prime factor
	while(number % factor != 0)
	    factor++;
	// Return it.
	yield return factor;
 
	// look again...
	if(number > factor)
		// recursively look for this factor again, using Num/factor
		// as the new big number
		foreach(int factors in MorePrimeFactors(number / factor, factor))
			yield return factors;
	}
}

Left Joins in LINQ

Robert Green — Fri, 19 Jun 2009 14:15:10 +0000

For some reason it seems that I often need to perform a left join in LINQ. Every time I need to do this I find myself scouring the web one more time in order to remember how a left join works in LINQ. So how does it work? The best example that I’ve found is here. The example looks something like this:

var list = from r in dc.tblRooms
             join ui in dc.tblUserInfos
             on r.UserName equals ui.UserName into userrooms
             where r.CourseID == 1848
             from ur in userrooms.DefaultIfEmpty()
             select new{
                 FirstName = (ur.FirstName == null) ? "N/A" : ur.FirstName,
                 LastName = (ur.LastName == null) ? "N/A" : ur.LastName,
                 RoomName = r.Name
              };

I like this example a lot because it is very straight forward. Personally I would like to make the statement a bit more generic. In order to do that all we need to remember is that every Left Join has a left table A and a right table B. All the results from the A will be returned regardless of the join with B. In my example I will call table A the LEFT_TABLE and table B the RIGHT_TABLE.

var list =  from LT in LEFT_TABLE
	      join RT in RIGHT_TABLE
	      on LT.key equals RT.KEY into NEW_TABLE
              where <CONDITIONS>
              from NT in NEW_TABLE.DefaultIfEmpty()
              <SELECT_STATEMENT>;

The only problem that may occur with this left join occurs with the DefaultIfEmpty() operator. A better practice would be to pass in a default value so that we can know what to expect in return.

Project Euler #2

Robert Green — Thu, 28 May 2009 20:07:09 +0000

Problem #2 in the Project Euler series is:

Each new term in the Fibonacci sequence is generated by adding the previous two terms. By starting with 1 and 2, the first 10 terms will be:

1, 2, 3, 5, 8, 13, 21, 34, 55, 89, …

Find the sum of all the even-valued terms in the sequence which do not exceed four million.

My solution:

static void Problem2() {
    int term1 = 0;
    int term2 = 1;
    int term3 = 1;
    int max = 4000000;
    double sum = 0;
 
    while(term1 < max && term2 < max && term3 < max) {
        term1 = term2 + term3;
        term2 = term1 + term3;
        term3 = term2 + term1;
 
        if(term1 % 2 == 0) {
            sum += term1;
        }
        if(term2 % 2 == 0) {
            sum += term2;
         }
         if(term3 % 2 == 0) {
            sum += term3;
         }
 
    }
}

Capitalize A String in C#

Robert Green — Wed, 27 May 2009 17:34:34 +0000

Capitalizing a string is a rather trivial task in C#. There are 2 ways to approach single word capitalization where each method includes 3 steps. Method 1 uses only strings and string methods while method 2 treats the letter to be capitalized as a char.

Method 1:

Get the first character as a string ( stringToCapitalize.Substring(0,1) )
Transform the first character to uppercase ( stringToCapitalize.Substring(0,1).ToUpper() )
Append the rest of the string ( stringToCapitalize.Substring(0,1).ToUpper() + stringToCapitalize.Substring(1) )

public static string Capitalize(string toCapitalize) {
    try {
        if(toCapitalize.Length > 1) {
            toCapitalize = toCapitalize.Substring(0, 1).ToUpper() + toCapitalize.Substring(1);
        }
    } catch(Exception ex) {
        ExceptionHandling.ExceptionLogging(ex, "Error:Capitalize.");
    }
    return toCapitalize;
}

Method 2:

Get the first character as a char ( stringToCapitalize[0] )
Transform the first character to uppercase using the char.toUpper method ( char.toUpper(stringToCapitalize[0]) )
Append the rest of the string ( char.toUpper(stringToCapitalize.Substring[0]) + stringToCapitalize.Substring(1) )

public static string Capitalize(string toCapitalize) {
    try {
        if(toCapitalize.Length > 1) {
            toCapitalize = char.ToUpper(toCapitalize[0]) + toCapitalize.Substring(1);
        }
    } catch(Exception ex) {
        ExceptionHandling.ExceptionLogging(ex, "Error:Capitalize.");
    }
    return toCapitalize;
}

Using either of these methods you could create an extension method as well. Something like this:

public static class MyExtensions
{
    public static string Capitalize(this String toCapitalize)
    {
        if(toCapitalize.Length > 1) {
            toCapitalize = toCapitalize.Substring(0, 1).ToUpper() + toCapitalize.Substring(1);
        }
        return toCapitalize;
    }
}

Content-Disposition attachment vs inline

manatarms — Fri, 30 May 2008 16:46:03 +0000

Today I ran into an interesting issue. We have some legacy code in .NET 1.1 that exports an HTML table to Microsoft Excel. This export occurs by simply rendering the table via Response.Write and setting the header content-disposition to “attachment; filename=FileName.xls”. The original code looked something like this:

Response.Clear()
Response.AddHeader("content-disposition", "attachment;filename=SalesByProductReport.xls")
Response.Charset = "utf-8"
Response.Cache.SetCacheability(HttpCacheability.NoCache)
Response.ContentType = "application/vnd.ms-excel"
 
Dim stringWrite As IO.StringWriter = New System.IO.StringWriter
Dim htmlWrite As HtmlTextWriter = New HtmlTextWriter(stringWrite)
 
tblTable.RenderControl(htmlWrite)
 
Response.Write(stringWrite.ToString())
Response.Flush()
Response.End()

The problem that occurred was that any user using Internet Explorer (surprise, surprise!) would get a prompt to download the file but the file would not download! The file worked properly in all other browsers. The solution is to change

1	Response.AddHeader("content-disposition", "attachment;filename=SalesByProductReport.xls")

1	Response.AddHeader("content-disposition", "inline;filename=SalesByProductReport.xls")

Now, why exactly does this work? I’m not sure, so if you know please tell me.