[Radioactiveman]

Compiling works fine with ArchLinux too. I also enabled SMP-support, why is it not enabled by default?

[mitsubishi]

'Cause ioquake3 development has determined that only some Mac OS drivers have proper multithreading that won't crash it dismally. They also believe that the advantage might be marginal, non existent or negative since it was originally developed for very old systems. It's open though to use if you compile it and you're lucky with your setup and doesn't crash it or works.

--

It's also hardcoded to be removed at compile time if it's not on a Mac so one has to remove an '#ifndef' somewhere (the var may be in but do nothing if that isn't done).

This post has been edited by mitsubishi: 29 May 2010 - 01:59 PM

[ghostrider]

Radioactiveman, on 29 May 2010 - 01:45 PM, said:

Compiling works fine with ArchLinux too. I also enabled SMP-support, why is it not enabled by default?

Ok, it compiles now after setting USE_LOCAL_HEADERS to false. So thanks for the help everyone! Is BUILD_CLIENT_SMP the flag to set for SMP or do I need to set something else?

[mitsubishi]

They have hardcoded it off even if SMP is #defined [that probably avoids the 'SMP was removed at compile time' but it doesn't let you know it still does nothing].

One has to also avoid an

#ifndef MACOS_X
return qfalse; /* better safe than sorry for now. */
#endif

in sdl_glimp.c

Further reasoning from the file:
"I have no idea if this will even work...most platforms don't offer thread-safe OpenGL libraries, and it looks like the original Linux code counted on each thread claiming the GL context with glXMakeCurrent(), which you can't currently do in SDL. We'll just have to hope for the best."

--

On Win64 here it crapped out dismally. I'm interested in seeing if anyone is successful.

This post has been edited by mitsubishi: 29 May 2010 - 06:50 PM

[newDude]

So i very likely have this wrong, but i removed those 3 lines of code in code/sdl/sdl_glimp.c and set BUILD_CLIENT_SMP.
Recompiled without issue and the binary runs - whether it's actually got SMP support, well that's another story, im not sure :(

[mitsubishi]

r_smp 1

[newDude]

Yeah, no love, run it and get a black screen :(

[mitsubishi]

not. surprised.

[Radioactiveman]

I didn't edit sdl_glimp.c and I don't have a notification about missing SMP-support.
But another question: I noticed longer loadingtimes than with the 'normal' binary.
If a new map starts I join after the first round. :/ What can I do to fix this?

[mitsubishi]

then it does nothing unless you're on max os.

if it's slow and it's x64 it's normal because vm_86_64 is relatively unoptimized.