Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Symbolic: A doomed effort, or it's time to get my lead-lined jockstr

Author: Vincent Diepeveen

Date: 05:57:08 02/18/04

Go up one level in this thread


On February 17, 2004 at 23:45:55, Russell Reagan wrote:

>On February 17, 2004 at 19:44:35, Vincent Diepeveen wrote:
>
>>When not referring to API, you have no idea what you talk about. 10%. Make that
>>100% to start with.

Those open source things are horribly programmed, without a single exception.

>Here's what I did. I took several open source chess programs and compiled them
>under native C/C++, and then under managed C/C++, and the results were that the
>biggest difference was about 20% for TSCP. The rest were much closer, less than
>10% slower, and one was actually faster when compiled as managed C++. So what
>tests did you do?
>
>Since all of the .NET languages get compiled to native machine code, I see no
>reason why C# should be 100% slower just because it's C#. There are plenty of
>benchmarks that show C# to be fairly close to the speed of native C/C++, and the

Compare with a PGO'd C compile of a commercial program.

C# is intended especially for interface programming and there definitely is a
need for a language doing this. I see it as the microsoft proprietery
counterpart of JAVA.

A company doesn't choose for JAVA because it is fast. When we choose JAVA for a
small project that was because it was supposed to work multiplatform. It's easy
for Sun hardware to use (thanks to superior java compilers/interpreters there)
and the cpu load is not relevant; basically harddisk speeds and network
communication speeds and latencies are the only problem in these worlds.

When speaking about interface, there still is a lot of possibilities there for
new products/programming languages which can build you good environments. C# is
not an exception to that.

>gap should only narrow as newer compilers and newer versions of the languages
>arrive.
>
>There are plenty of benchmarks on the net that show C# to be within a few
>percent of natively compiled C++ already. There are a few areas that need some
>work, but it won't be long, and C# will be within a few percent of C++.

Again you do it. Comparing slow object allocating c++ code is never a good
compare to anything. It's a zillion times slower of course than non object
allocating C code.

Now compare with good c++ code, which we use for the diep interface by the way.

You aren't going to beat that *anyhow* with C# within 3 years of time.





This page took 0 seconds to execute

Last modified: Thu, 15 Apr 21 08:11:13 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.