OSX/Intel/Cheetah?

[propstuff]

Well, I was JUST about to buy a new G5 and I thought I'd wait till WWDC was over..................
Sheesh! Now I'm not sure what to do....

Martin, can you give us any idea, at this early stage, what it all means to you and Cheetah?

cheers,
N.

[peer]

I think you can buy that G5. The new OS X will be compatable with G5 as well as Intel. I know that the current OS X apps will run in Leopard (OS 10.5) through the help of "Rosetta" A conversion program. Maybe we hear more with time.

Regards,
Peter

ps. I just read that it is very simple to convert cocoa and carbon apps to the new fat binary standard. It's a matter of days with the new developers tools (xcode 2.1). But Martin probably know these things better. I'm only an architect and no programmer :D

[propstuff]

Hi Peer,
Yes, I've read forums taking in turns to argue loudly that it is, or it is not a simple matter to port. :-/

Every layer of "conversion" just means slower and slower operation, and when we're talking about hi res rendering of 3D...................

I't's very early, but I was hoping Martin might be able to give an "informed" opinion of the consequences of the coding rewrite.

cheers,
N.

[peer]

Your right! There are different stories. I guess will just have to wait and see. At least the processing power will grow with the introduction of Intel. So that's good for 3Ding.

Regards,
Peter

[Martin]

Hi,
I haven't tried to create a universal binary yet but I don't expect that it would be to difficult. There are only some file loaders which have to be fixed manually because they care about big- or little- endians. But that are at most a hundert lines of code. The rest should compile without any modifications. So there should be a Intel version of Cheetah3D available on time. I'm actually considering if I should rent such a transition kit from Apple.

I don't expect that the transition will have any effect on application speed as long as the app was recompiled for Intel. I think it would even increase because the compilers are much better optimized of x86 CPUs.

In general I welcome the change to Intel because all the discusions about which is faster PPC or x86 will be over and the focus will return to the OS. The soul of the Mac was always the OS. And the OS will be as great on Intel as it is on the PPC.

By,
Martin

[azteck2]

One of the most well written and correct responses I heard since Jobs's speech, good job Martin!

[Startyger]

can someone explain little endian to me? by the way... i think the x86 move should be great for Cheetah3D since so many 3Dapps are already there... this should just ake things easier.... did Altivec help with rendering at all? i was just wondering....

[RandyWang]

Startyger, if you know a few basics about data and how it's stored, then this (http://en.wikipedia.org/wiki/Little_endian) should explain the difference between Big- and Little-Endian quite nicely.

i think the x86 move should be great for Cheetah3D since so many 3Dapps are already there...

Not quite - I think this would only really increase the competition for Martin, and he's only one man after all. On the plus side, Cheetah has the unique position of being the most useable and powerful sub-$100 3D renderer/modeller/animator on the Mac, and the move to Intel won't be changing that.

did Altivec help with rendering at all?

While I, personally, don't know the answer to this I can imagine that having a machine with Altivec support would marginally increase the performance of Cheetah indirectly, by allowing other processes to do their jobs without using as much of the CPU's time.

The same applies for having a powerful graphics card: Cheetah's renderer only uses the CPU, but a more powerful graphics card (specifically, one with Quartz Extreme support) would allow the screen to redraw with less strain on the system, giving more time to Cheetah.

[Martin]

Hi,
making Cheetah3D run on x86 does not mean that it will run on Windows so that I don't have to care about these competitors.

The endianness has only very little effect on Cheetah3D. I only have to fix some file loaders like .3ds or .lwo. That shouldn't make to much work.

Cheetah3D doesn't use Altivec yet. I played around with it some time ago. But the speed increase wasn't to big. Altivec is fine if you can process big chunks of data. But it your code makes many jumps it's quite difficult to itch out a speed increase. It's even possible that it becomes slower.

By,
Martin

[RandyWang]

Cheetah3D doesn't use Altivec yet.

I always assumed that 3D rendering was like video encoding, in that you could use vector processing effectively. I guess I was wrong. =)

[Martin]

Altivec is fine for accelerating a rasterizer renderer (OpenGL) where you have to transform big arrays of vertices for example. But even that job is done by the graphics card nowadays.

Accelerating a raytracer is much more difficult. But it is possible of course. There are some quite promising papers out there. But it is a very tricky job to itch out the maximum speed.

By,
Martin

[jr57k]

Well I just bought one of the new dual Core 2.0Ghz iMacs. What ever happened to the Intel version (universal binary). I could use it next week.

[Martin]

Hi,
I've also ordered a Intel iMac. I should get it quite soon since I was already notified that it shipped.
I will try me best to create as fast as possible a universal binary version of Cheetah3D.

By,
Martin

[Martin]

Hi jr57k,
if you are interested to test the Intel Beta please send me an email.

By,
Martin