Mail Archives: djgpp/2001/01/18/21:37:15
Yes, I am quite impressed with the DJGPP implementation of
gcc... unfortunately, gcc/g++
on my Mandrake Linux 7.2 made an executable which was about the same speed
as the Borland C++ 5.5 code. (Actually I'm still recovering from this
since Mandrake killed my C and D Win98 partitions (although I can access
them from /mnt/win_c and /mnt/win_d, strangely enough...))
I *would* use DJGPP for future development, but... does it support Windows
apps with OpenGL? Borland C++ 5.5 does, and... well, even if it is a bit
slower, might have to settle for second best in this case...
Btw, in response to an earlier post about alignment, I forced BCC5.5 to
align on 8-byte boundaries and still no faster...
Thanks for the thoughts
Mike
On Thu, 18 Jan 2001, Alexei A. Frounze wrote:
> Believe me GCC is quite good at optimization (well, when you helped it with
> nice algorithm and source).
> Around a year ago I had a bet in this NG that I can make a faster
> implementation of my texture mapper in C+ASM mix rather than nearly plain C.
> But in fact, eventually I got nearly the same thing when used 100% of GCC
> capabilities and no extra ASM at all (well, probably just 1 or 2 ASM
> instructions which is not a lot :). I was also wondered by the fact that GCC
> beats Watcom. I've thought that Watcom is one of best C compilers before I
> got GCC. But when I saw than my 3d-engine comiled by Watcom is about twice
> as slow as compiled by GCC...
>
> :)
>
> Happy coding with GCC!
> --
> Alexei A. Frounze
> alexfru [AT] chat [DOT] ru
> frounze [AT] ece [DOT] rochester [DOT] edu
> http://alexfru.chat.ru
> http://members.xoom.com/alexfru/
> http://welcome.to/pmode/
>
> "Mike Darrett" <ez073236 AT mailbox DOT ucdavis DOT edu> wrote in message
> news:Pine DOT GSO DOT 4 DOT 21 DOT 0101172014390 DOT 7634-100000 AT sandman DOT ucdavis DOT edu...
> > Hi guys,
> >
> > I was running some algorithm benchmarks on DJGPP vs Borland C++ 5.5, and
> > was shocked to see that DJGPP outperformed Borland C++ on some stack
> > tests. Using a linked list to simulate a stack, adding and removing 80,000
> > entries took 1.8 seconds on my AMDK6-2 350, but took 2.5 seconds on
> > Borland C++, compiled without the -tW option (since it is a console app).
> >
> > Any ideas? Is Borland C++ using thunking to access memory? Is DJGPP simply
> > more efficient? Would like to get any input before I try optimizing the
> > code any further.
> >
> > This was a homework assignment, but was meant only to test different
> > algorithms vs each other, and not vs other compilers. The homework
> > assignment can be seen
> > at: http://wwwcsif.cs.ucdavis.edu/~davis/110/prog1.html
> >
> > Thanks,
> >
> > Mike Darrett
> > mrdarrett AT ucdavis DOT edu
> > http://mdarrett.freeyellow.com
> >
> > Get a free Windows C++ compiler! With STL, OpenGL and DirectX support.
> > http://www.borland.com/bcppbuilder/freecompiler/
> >
> >
>
>
>
Mike Darrett
mrdarrett AT ucdavis DOT edu
http://mdarrett.freeyellow.com
Get a free Windows C++ compiler! With STL, OpenGL and DirectX support.
http://www.borland.com/bcppbuilder/freecompiler/
- Raw text -