RE: [Bug-gnubg] Building gnubg on windows - A newbies voyage

[macherius]

Well, I did a lot of benchmarking as well. First of all, the download
version was built without big optimizations and for a target of general 386.
I tried 3 compilers

Visual C 7.1
Intel C 8.1
gcc 3.4.1 (cygwin, only useable for no-gui)

So the regular release compiler, mingw, is missing in my arsenal. With the
above 3 I compiled with agressive P4 optimization and got the following
results in the no-gui version:

Release version as currently bundled: 16000 eval/s
MSVC: about 18000 eval/s
gcc: 21000 eval/s
intel C: 29500 eval/s

The situation is totally different for the gui version, it is like the evals
were not the same. They are, if you look in the code, but the displayed
figures differ significantly. Regardless of the compiler and settings used
they are around 20000 eval/s, with the existing release build slightly ahead
of the pack. No idea why. Either MingW gcc really rules, the gui measures
not only the eval/s, or I overlooked something very important. But remember,
I'm new to gnubg compiling. Maybe some more experienced compilator tries
mingw and reports figures here. I'd be very interested for 386 and P4
results.

I still wonder about the huge difference between command line and gui
display of eval/s.

Ingo Macherius


> -----Original Message-----
> From: address@hidden [mailto:address@hidden 
> Sent: Monday, January 31, 2005 6:01 PM
> To: Jon Kinsey; macherius
> Cc: address@hidden
> Subject: RE: [Bug-gnubg] Building gnubg on windows - A newbies voyage
> 
> 
> 
> 
> 
> > -----Original Message-----
> > From: address@hidden
> > [mailto:address@hidden 
> > On Behalf Of Jon Kinsey
> > Sent: 31 January 2005 14:21
> 
> > There were some problems when gtk 2.2 was tried, I think the
> > main issue was that it was quite a bit slower in some areas.
> > 
> > Jon
> > 
> 
> I ran a benchmark and got 16500 evaluations per second with 
> standard gnubg, but only 10500 with the 2.6 build.
> 
> -- Ian
>