The following code gave me the results below:
isinst Total: 00:34:42.5468750
castclass Total: 00:35:05.6875000
Total Difference: 00:00:23.1406250
It appears that "as" and then testing for null is a little bit quicker, but
the results could be an anomaly and may not be reproducible on someone
else's machine. There are a million other factors that could lead one to
readily dismiss my results as inconclusive (as Ian has pointed out in the
past). For now though, I think I'm going to stick with "as" and test for
null.
I've included the code that I used. Please feel free to help me improve it.
using System;
using System.Collections.Generic;
using System.Text;
using System.Runtime.CompilerServices;
#endregion
namespace ConsoleApplication1
{
class Program
{
[MethodImpl(MethodImplOptions.NoInlining)]
static void Main(string[] args)
{
Collection<TimeSpan> isinstTestResults = new
Collection<TimeSpan>();
Collection<TimeSpan> castclassTestResults = new
Collection<TimeSpan>();
long length = 100000000;
TimeSpan timeSpan = new TimeSpan(0);
DateTime startDateTime = DateTime.Now;
for (int counter = 0; counter < 100; counter++)
{
startDateTime = DateTime.Now;
for (long i = 0; i < length; i++)
{
Base b = new Derived();
Derived d = b as Derived;
if (b != null)
{
Console.Write("");
}
else
{
throw new InvalidCastException();
}
}
timeSpan = DateTime.Now - startDateTime;
isinstTestResults.Add(timeSpan);
Console.WriteLine("isinst test {0} --> {1}", counter,
timeSpan);
startDateTime = DateTime.Now;
for (long i = 0; i < length; i++)
{
Base b = new Derived();
Derived d = (Derived)b;
Console.Write("");
}
timeSpan = DateTime.Now - startDateTime;
castclassTestResults.Add(timeSpan);
Console.WriteLine("castclass test {0} --> {1}", counter,
timeSpan);
}
TimeSpan isinstTotal = new TimeSpan(0);
TimeSpan castclassTotal = new TimeSpan(0);
for (int i = 0; i < isinstTestResults.Count; i++)
{
isinstTotal += isinstTestResults[i];
castclassTotal += castclassTestResults[i];
}
Console.WriteLine("isinst Total: {0}", isinstTotal);
Console.WriteLine("castclass Total: {0}", castclassTotal);
if (isinstTotal < castclassTotal)
{
Console.WriteLine("Total Difference: {0}", castclassTotal -
isinstTotal);
}
else
{
Console.WriteLine("Total Difference: {0}", isinstTotal -
castclassTotal);
}
}
}
public class Base
{
}
public class Derived : Base
{
}
}
Thanks,
Fred
> -----Original Message-----
> From: Unmoderated discussion of advanced .NET topics. [mailto:ADVANCED-
> [EMAIL PROTECTED] On Behalf Of Ted Neward
> Sent: Sunday, February 20, 2005 3:29 AM
> To: [email protected]
> Subject: Re: [ADVANCED-DOTNET] IL internals (castclass and isinst)
>
> Have you tried running a benchmark? Try each 100k times and see what the
> timing differences are?
>
> From looking at the Rotor code, it looks like they're both basically
> implemented in the same terms; that is, they both end up consulting the
> object's MethodTable to see if there's correlation and/or goodness.
> Assuming the CLR itself isn't too terribly different from Rotor on this
> score, you can see the gory details in sscli\clr\src\vm\jitinterface.cpp;
> do a search for "IsInst" and "ChkClass", as there are a couple of helpers
> that get invoked during the process.
>
> Ted Neward
> Co-Author, SSCLI Essentials
> http://www.neward.net/ted
>
> ---------- Original Message ----------------------------------
> From: Palmer Fred <[EMAIL PROTECTED]>
> Reply-To: "Unmoderated discussion of advanced .NET topics."
> ===================================
> This list is hosted by DevelopMentorR http://www.develop.com
>
> View archives and manage your subscription(s) at
> http://discuss.develop.com
===================================
This list is hosted by DevelopMentor� http://www.develop.com
View archives and manage your subscription(s) at http://discuss.develop.com
