OS X Intel/PPC Universal Binary - Not Yet

Post #1by **dirkpitt** » 11.01.2006, 17:36

I've just begun porting Celestia to run as a full-speed Universal Binary on the new Intel Macs. Sorry about the delay, but 1.4.0 and a few other commitments slowed things down.

Right now I'm getting the hang of setting up architecture-dependent build settings. What remains is the issue of obtaining universal versions of the dependent libraries (png, lua) as well as fixing some byte-ordering issues.

BlindedByTheLight · Post #2by **BlindedByTheLight** » 13.01.2006, 06:48

Thanks Dirkpitt... you rock!

Post #3by **dirkpitt** » 13.01.2006, 15:48

Just a quick update, I successfully compiled and tested a Universal "fat binary" of Celestia today on a Developer Transition Kit (an early Intel Mac) borrowed from Apple. Compiling a Universal binary is possible on any Tiger 10.4.x config, but the result will be something that will run even on 10.2 Jaguar on a PPC chip, all the way through 10.4+ on Intel. Exciting stuff.

I'll try to get the necessary changes to the source code committed to CVS asap, but in the mean time I'd be very interested to know if there are any OS X Celestia users who've recently acquired an Intel iMac and would like to test.

BlindedByTheLight · Post #4by **BlindedByTheLight** » 14.01.2006, 04:27

Waiting on the MacBook...

Yoo hoo!!!

marcof · Post #5by **marcof** » 16.01.2006, 14:14

I'm expecting my iMac Core Duo 20" this week..

SeaDour · Post #6by **SeaDour** » 16.01.2006, 20:25

Just got my Intel iMac this morning; I would be more than happy to test your build.

Post #7by **dirkpitt** » 17.01.2006, 17:39

SeaDour, marcof: I private-messaged you a url to a freshly-built Universal binary.
Thanks for offering to test and please let me know if you have any problems!

marcof · Post #8by **marcof** » 23.01.2006, 09:19

I'm downloading at the moment.. arrival of iMac took a little longer than expected (and I'm still waiting for my 2 Gig RAM to arrive!)

marcof · Post #9by **marcof** » 23.01.2006, 09:36

after 10 minutes with it, nothing special to mention, except it is pretty speedy. I don't have any real comparison, coming from a 867MHz G4 .. :)
It's running the tour system script in the background now. Controls seem to work fine, I don't have any addons installed though, but the app seems to behave pretty well. Full-screen mode switch is a bit slow, but that might be due to the lack of memory...

Post #10by **dirkpitt** » 23.01.2006, 16:06

That's great to know

I wonder what sort of framerates we're seeing near planets (the ` key toggles fps display)?
Also if you don't mind, I'm curious to know whether ring shadows are displaying properly.

marcof · Post #11by **marcof** » 23.01.2006, 20:19

FPS is around 60
I don't see any sign of shadows.. unfortunately.

Post #12by **dirkpitt** » 24.01.2006, 02:13

marcof wrote:I don't see any sign of shadows.. unfortunately.

Have you turned on Ring Shadows in the Preferences panel? Also make sure your Render Path is set to OpenGL 2.0.

marcof · Post #13by **marcof** » 24.01.2006, 10:16

that did the trick(OpenGL 2).. works!

watersb · Post #14by **watersb** » 24.01.2006, 22:53

I received the iMac today, it's sitting here next to the DTK machine.

I'd love to test the binary!

I could also test some CFLAGS settings, if necessary.

I pulled current from CVS, built an Intel binary. It Just Works! Except I cannot see the stars... none are rendered.

I had to install a CelestiaResources folder from the 1.4.0 release. I copied that to my ~/Library/Application Support folder and it works great. Nice new splash screen, it reports version 1.4.1 and it is an Intel binary.

Post #15by **dirkpitt** » 25.01.2006, 16:40

watersb wrote:I pulled current from CVS, built an Intel binary. It Just Works! Except I cannot see the stars... none are rendered.

I've seen the stars problem happen on the DTK I tested on too, but I assumed it would be fixed on release versions of OS X 10.4.4+ for Intel (i.e., on Intel iMacs and Mac Books).
Could you try out the binary I private-messaged you on an Intel iMac just to be sure?

If you're familiar with key value coding (KVC), the prerelease version of OS X 10.4.3 that runs on DTKs has a bug in that area, causing the faintestVisible preference to always be read as zero. This results in a generally messed up display, the side effect of which is intermittently visible stars.

neo albireo · Post #16by **neo albireo** » 27.01.2006, 13:04

dirkpitt, could you tell me the link to your binary? I'd like to test it on a new iMac ASAP...
Thx!

neo albireo · Post #17by **neo albireo** » 27.01.2006, 22:48

Ok, I tested the binary intensively on an Intel iMac and I'm totally amazed. Everything works very very well. Open GL 2.0 is great. I wonder though if the rings' light effects are accurate, I was surprised how dark they are when viewed against the sun.
Only one negative point, but I don't know if that's related to this build or to all builds: By default, there were no stars visible. When I adjust the magnitude and quit Celestia, it doesn't save the value, so the next time there won't be any stars shown again.
Ah and another negative point: The iMac isn't mine.

I used all of my previous used addons and many addons that I can't use on my own machine. It might be of interest to possible mac buyers, that with the 20'' intel core duo iMac, 2GB Ram, 256 MB ATI Radeon X1600 it is possible to view all the following addons at the same time in one screen (didn't split) with acceptable performance (sorry for the not so accurate details on those addons, it's just to give an idea):
- 2 mio stars database
- Jesters Earth Mark II complete with clouds, spec, normal maps (although when changing from one Level to another, there is a break of 2 seconds which might cause the viewer to unintentionally reach the ground, because the release of the "home" key is not reacted on properly)
- HeveliusConstellations_PNG_hires
- The big Satellite databases (Eccentric / Geostationary_full)
- Thorvalds' Highres Sun
- Detailled models of Hubble or ISS
- Big comets/asteroids databases with orbits with 300 points
However, when I tried all the above AND Selden's graticules AND the deepspace datasets 2df_300k-cmod and sdss-dr2-v2, the frame rate was unacceptable. Anyway I would never need to show all this at the same moment.

All in all:

Post #18by **dirkpitt** » 28.01.2006, 02:43

neo albireo wrote:Only one negative point, but I don't know if that's related to this build or to all builds: By default, there were no stars visible. When I adjust the magnitude and quit Celestia, it doesn't save the value, so the next time there won't be any stars shown again.

I'll make a few more test builds.

watersb · Post #19by **watersb** » 28.01.2006, 04:44

@dirkpitt:

Thanks for the pm to your build a few days ago; the stars were not visible with that build, either.

Actually it is weirder than that --- the stars can appear as very large, fuzzy spheres that are sometimes in the foreground. Or so it seems.

But it is dependent upon the slider setting for the "faintest stars to display" preference: for most settings of this, the stars are not visible.

I got roughly the same behavior from your build as from the default pull from CVS. Anyway I could not perceive a difference.

BTW: How to set "-mtune=pentium4" as an optimization option in XCode? Anyone?

neo albireo · Post #20by **neo albireo** » 28.01.2006, 09:35

watersb wrote:Actually it is weirder than that --- the stars can appear as very large, fuzzy spheres that are sometimes in the foreground. Or so it seems.

But it is dependent upon the slider setting for the "faintest stars to display" preference: for most settings of this, the stars are not visible.

I can confirm that, though I thought that had to do something with the galaxy spheres, so I set the galaxy magnitude to 0 and used the []-keys for the magnitude. After that I could not reproduce the spheres. I hadn't found out how to handle the slider, but with the keys the magnitude adjusting worked well.