Page 1 of 1

Spear Of Destiny

Posted: Thu May 01, 2014 1:36 am
by purpleeagle01
Yesterday the game (Spear Of Destiny) on my Falcon with 14MB tested and the game is EXTREMELY slow unplayable can test with an accelerated Falcon anyone?

Re: Spear Of Destiny

Posted: Thu May 01, 2014 6:54 pm
by Marcer
Could you Please post a working package?

Skickat från min C6903 via Tapatalk

Re: Spear Of Destiny

Posted: Fri May 02, 2014 4:08 am
by purpleeagle01
The game is from your Game DVD 2.0
I booted my Falcon with Control and the game loaded after about 3-4 minutes and after the eternally long intro scenes was a typical ID Software menu
and the game is reminiscent of Wolfenstein and ca.all 30 'seconds you can make a step.
Cheers

Re: Spear Of Destiny

Posted: Fri May 02, 2014 8:50 pm
by evil
purpleeagle01 wrote:Yesterday the game (Spear Of Destiny) on my Falcon with 14MB tested and the game is EXTREMELY slow unplayable can test with an accelerated Falcon anyone?


Generally direct-ported PC games will run terrible on a Falcon without TT-RAM.
With a CT2, AB040 or CT60/3 things usually turn out quite good though.

That said, I haven't tested the above mentioned port.

Re: Spear Of Destiny

Posted: Sat May 03, 2014 9:44 am
by Eero Tamminen
If there are sources or debug symbols, and it runs on 14MB Falcon, it can be profiled with Hatari:
http://hg.tuxfamily.org/mercurialroot/h ... #Profiling

One problem with PC game ports compiled with GCC/MiNTlib is the Unix timing functions. When I profiled Quake I port running on TOS, half of the time went to timing functions used by the game.

I contributed year ago a MiNTlib fix to short-circuiting nanosleep() on TOS, but any calls to gettimeofday() or mktime() are still really bad performance-wise. On Linux gettimeofday() has fast-path to kernel because it's used so often, on TOS/MiNT there's nothing. Main reason for slowdown with MiNTlib is timezone stuff (accessing timezone database information on disk). If "TZ" environment variable is not set, timezone stuff is done on every call. The old MiNTlib version in Sparemint distro isn't quite as bad in this respect, but MiNTlib coming with the Atari gcc cross-compiler is.

One of the problems in Quake source was also it giving timezone argument to gettimeofday() call, that should always be set to NULL instead.

Best is if those calls are just replaced outright with something more suitable for TOS.

If you're interested on more information on this, see "MiNTlib timing related functions behavior/performance" on MiNT mailing list.