Hardware Fp


Nmn

Still Fresh
Joined
Dec 27, 2005
Messages
80
Hey guys... I've been busy for a while. I'm the n00b who thought of using the NAND for ram... Well, I'm back. And I'm much less busy. I'm a bit better at programming, especially SDL. Recently, i have become interested in writing directly to the FB etc. But I noticed something, finally, that was posted at the GP2X.de archive when screwing around... GP2X11 !!!111 I know, its prolly gonna crawl on anything not designed for it... and its a low resolution(But then again, thats really what the scaler was designed for)... But its definitely a proven window solution for desktop and embedded platforms. Now, i really wanted to port Matchbox. Not just for me, but for the community. Besides, I'm sure everyone would enjoy running Dillo on your 2x!!

(SPEAKING OF WHICH..., MB hasn't been ported has it? How bout compiz metacity? twm? fvwm? fvwm95 :)? cdwm? kwin? explorer.exe? jk lol, closed source :) too big and full of bugs 'nyway. What about ZWM? AfterStep? FluxBox? IceWM? Sawfish? gnome? kde?... I wish?)

So, with my newly setup GP2X dev kit, i began devving. I started with LibMB. Created a dev file, and began.

NOTICE: The below italics type is slightly offtopic, But i REALLY don't want to explain later why I'm using Dev-Cpp besides liking it.

Let me note something: I'm using a special GP2X dev kit, one with a special build and setup of devcpp. This is because the rest of the devkits have problems. Mainly because I'm on vista, and some programs don't run(I installed it, and a bit later XP died. What a suprise...?)

Luckily, Dev-cpp is FLAWLESS. I was amazed, and very happy. Most devkits had trouble. One lacked a minsys, the other was terribly outdated and not enough libs, and the other had a bad minsys which i couldn't figure out how to get it to find out my prefixes and settings to compile with the GP2X (Among some problems with RXVT and the autotools I need very badly without an IDE.) This one has Dev-Cpp, and thats enough to port and build, and then some.


Now... this kit didn't include the GP2X11 devel files. So i downloaded them. Extracted them in my /arm-bla-bla/tools dir. A few other libraries (Namely freetype) were gotton from theoddbot's libs pack.

Now, for what you've all been waiting for: The problem.
I now have a bunch of hardware fp libraries yet my program uses software floating point. So, what should i do? Can i get the core libraries in HW-FP and make my proggies compile that way? (And how do i do that?) Should i possibly reinstall the kit and find some SW-FP versions? (Where would i find those?) OR: Can somebody help me with installing a minsys setup? (I SERIOUSLY am against this. RXVT+Windows=Linux annoyances)

Or worseyet, will i have to buy some more cds and burn SuSE Linux 10.1? Jk, i love linux :)


OT: Does anyone have the MK2? is it literally faster or just default over clocked?

To those who ported X to GP2X: I hope someday this becomes a reality.

Hope thats nuff info. Anyone think this is a good idea?
 
Matchbox has already been ported in (it's in)GPE which includes Dillo.

GP2X11 works fine, it is not slow at all.

The hw fp/sw fp is a problem, you need to use all one all the other. The X11 libs are compiled with soft-float though.
 
As Parkydr says, Matchbox is part of GPE, but if you could get it working stand-alone, that would be interesting (since we've all given up trying to compile all of the components :D). Other than GPE, nothing else has been ported that makes use of the server (apart from the standard X utilities which Parkydr compiled a while ago and I keep forgetting to upload :)).

EDIT: There are a few bits and bobs here: http://x11.gp2x.de/
 
Sweet. Thanks guys. I might migrate to linux again. After all, compiz beats areo any day, and wine will too eventually :) I can compile my own gcc and lib sets with HW floating points. Anyways, Thanks alot. This has truely made my day; This REALLY puts the X in GP2X :)

Oh... and can anyone tell me how to compile with hw-floating points in GCC? or better yet how to set this setting?
 
No hardware floating points on this thing, unfortunately. I seem to get a slight performace increase by compiling with -msoft-float, but it's still all done in software.
 
A kernel fp emualtor handels the fp if you compile your stuff with hardfloat, wich is way slower than letting gcc handle it, especialy with gcc 3.5 and higher.
 
Wow, really? i thought it had hardware floating points on the 2x... Thanks for letting me know, i have gcc 4.0.3.
 
Back
Top