Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: visual C.net

Author: martin fierz

Date: 16:58:40 03/15/02

Go up one level in this thread


On March 15, 2002 at 19:33:00, David Rasmussen wrote:

>On March 15, 2002 at 14:52:24, martin fierz wrote:
>
>>On March 15, 2002 at 10:24:34, Vincent Diepeveen wrote:
>>
>>>On March 15, 2002 at 00:54:23, martin fierz wrote:
>>>
>>>>aloha
>>>>
>>>>i think visual C.net is shipping now - has anybody already tested if the
>>>>compiler has improved over version 6.0?
>>>>
>>>>cheers
>>>>  martin
>>>
>>>yes visual c++ NET 7.0 is a joke for C anyway.
>>>
>>>In a database server to web conversion tool i compiled conversion
>>>from float to int already gets buggy when -O2 is used,
>>>with debug info it goes ok.
>>>
>>>This compiler is a joke everywhere. First of all looks of the interface,
>>>i do not like it. Secondly there are major bugs in creating projects,
>>>no way to create a project which you then can easily debug.
>>>
>>>Converting projects from msvc 6.0 goes ok and debugging them too.
>>>
>>>The real problem is the major bugs in the compiler. I don't know who wrote it,
>>>but it's a beginner who wrote it. Probably C/C++ compiler has no big
>>>priority anymore within visual c++ and they might have put their best
>>>man on the joke language C#, which is an obvious attempt to keep people
>>>away from getting cross platform in the future, after the successfull
>>>attempt of JAVA from SUN in this direction.
>>>
>>>When i compile diep with 7.0 i can't express very well what i feel when
>>>i see its output. Loads of eval code gets wrong compiled, loads of
>>>patterns give unexpected results back, even though it is very neatly
>>>coded code with pointers.
>>>
>>>Again the same problem arises which weirdly is in nearly all compilers
>>>that are tricked to do better on the specbenches: allocate some
>>>2 dimensional arrays from your own datatypes. then keep reusing 2
>>>variables to loop through all those arrays.
>>>
>>>It simply creates wrong code here. Amazingly it doesn't crash.
>>>
>>>Really pathetic is it is impossible to create a simple project which
>>>can get debugged without modifying loads of options.
>>>
>>>Designers of the user interface must be shipped back to school too. They
>>>are using Internet Explorer look and feel, something which i do not like.
>>>
>>>Now about speed of the compiler. It generates 2% slower code than
>>>visual c++ 6.0 sp4 with processor pack.
>>>
>>>When i realized this I have thrown the compiler out of my window of course,
>>>for a compiler which creates bugs in my program i have no respect, a
>>>compiler which generates slower code AND bugs, that's really too much
>>>for me.
>>>
>>>Best regards,
>>>Vincent
>>
>>
>>oh, that sounds really terrible... so i guess we'll have to wait for 7.1 or
>>something... i was also afraid that microsoft would work more on their .net
>>attempt to kill java and on C# than on the C compiler, but i would not have
>>imagined that it would actually get worse! but then, what do you expect from
>>microsoft :-)
>>
>>aloha
>>  martin
>
>Don't take Vincents word for it. My code is 100% ANSI/ISO-compliant C++, and it
>works fine in VC++ 7. And the code is about 30% faster than gcc (as usual). I am
>not saying this compiler is perfect. In fact I found a compiler bug in it. But I
>also found compiler bugs in gcc 2.96 and 3.0.x. For compliance, the gcc 3.0.x,
>and esp. 3.1 when it is released, are best. But many compliant programs compile
>wihtout problems on VC 7. Many many more than with VC 6. VC 6 isn't compliant
>with anything. It fails to compile even the simplest compliant C++ code. So if
>compliance and portability is important (and speed) try MSVC 7. MSVC 6 doesnt
>fill this bill at least.
>
>/David


for me, as for most here i guess, speed is what is important. if version 7
produces no faster code than version 6 i'm not interested.

aloha
  martin



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.