Active Topics

 



Notices


Reply
Thread Tools
penguinbait's Avatar
Posts: 3,096 | Thanked: 1,525 times | Joined on Jan 2006 @ Michigan, USA
#11
Fullscreen is nice, and the 400x240 is pretty responsive, now how do you turn on the pixledoubling??

That would be awesome, fullscreen and fast (however pixelated)
 
pipeline's Avatar
Posts: 693 | Thanked: 502 times | Joined on Jul 2007
#12
Hmm looks do-able :
http://maemo.org/community/wiki/gamedevelopment/

I suppose i might try that this weekend as a mod to vid_sdl.c
 
penguinbait's Avatar
Posts: 3,096 | Thanked: 1,525 times | Joined on Jan 2006 @ Michigan, USA
#13
Is there a way to start the game with ctf?

Is there a reference for the command line options available for sdlquake? I can't seem to find one, although I have not tried the source code yet?
 
Posts: 2,102 | Thanked: 1,309 times | Joined on Sep 2006
#14
With the talk about whether the 770 has hardware floating point, I just wanted to check that you did compile to use hardware vfp (floating point)?

I.e. using CFLAGS=-mcpu=arm1136j-s -mfpu=vfp -mfloat-abi=softfp (you can remove the -mcpu flag if you need to as code generated with this flag can't be emulated in scratchbox so may break your build if intermediate binaries are generated and run).

The N800 toolchain naturally produces softfloat binaries, and you'll need to compile libm to use floating point too if you want to achieve the full effect as the standard N800 version is softfloat . If you already know all this then my apologies.

Cheers,


Simon
 
pipeline's Avatar
Posts: 693 | Thanked: 502 times | Joined on Jul 2007
#15
No i did not do anything special so it may be using softfloat. That means it might be playable on the 770 especially if pixel doubling can be made to work.

I'll see if i can compile as you describe to produce real vfp for 800.

Why would nokia release a math library for the n800 which isnt compiled to take advantage of the cpu fp... is that alot of asm code?

Last edited by pipeline; 2007-09-28 at 10:24.
 
pipeline's Avatar
Posts: 693 | Thanked: 502 times | Joined on Jul 2007
#16
WOW... thanks lardman! I just compiled using those flags and that seems to make a huge difference. I havent done anything with libm.

I can run at largest resolution and it is very playable. I'll keep two versions for download in case 770 users care to try (non-fp?) version.

The fp version for 800 users can be downloaded here :
http://wardenclyffetower.com/MaemoFiles/sdlquakefp.zip
 
pipeline's Avatar
Posts: 693 | Thanked: 502 times | Joined on Jul 2007
#17
penguinbait... quake1 didnt not have ctf by default... it was a mod called threewave.

I found threewave but im not sure how well mods will work on the arm... since i am running off internal memory i am using shareware version. If you've got the full version (both PAK files ~50megs), then you can try downloading the file here :
http://www.fileshack.com/file.x/124/...r+version+4.21

extract the 3wave421x directory underneath your quake dir and add -game 3wave421x to the command line. I have no idea if it will work though. Perhaps i will get mmc functionality working tonight/tomorrow where i can test full version and this too.
 
Posts: 2,102 | Thanked: 1,309 times | Joined on Sep 2006
#18
I think the reason they stayed with softfloat was partly compatibility and partly that the vfp (hardfloat) binaries were larger. Something funny like that. Iirc it was mentioned on the maemo-devel list a long while back when we were doing some benchmarking of soft-float vs vfp and looking at how fast vfp vorbis was vs fixed point tremor.


Simon
 
Posts: 2,102 | Thanked: 1,309 times | Joined on Sep 2006
#19
Did you compile your own vfp libm (assuming it requires libm that is)?

If not, there's a version here: http://people.bath.ac.uk/enpsgp/benchmarks/

I suggest you try linking against the static version (rather than having to alter LD_LIBRARY_PATH so you find the shared lib - don't replace your existing libm with this one!) and see if that adds any extra speed. If Quake uses few or no mathematical functions (exp(), trig stuff, etc.) or uses its own implementations you might find that it doesn't do anything much at all.


Simon
 
pipeline's Avatar
Posts: 693 | Thanked: 502 times | Joined on Jul 2007
#20
No i did not compile libm.. perhaps improvement came from mfloat-abi softfp lib? Perhaps the last build i did is still all software fp just more efficient soft fp?

I will try linking static vpf libm later... would that be libm.a?

Thanks
 
Reply


 
Forum Jump


All times are GMT. The time now is 07:57.