UE3 evaluation/ProDG/PhysX/Havok in PS3 SDK

Technology partnerships with software providers is becoming just as important a part of building a modern platform as partnerships with hardware providers.

 

So just to confirm, this isn't any closer deal than the MS development inclusion of Aegia or UE3 as development options? Devs are provided evaluation software for these libraries when they buy the PS3 kit and then choose to buy into these libraries or not.(?)

 

Bah - the console forum comes back when I wasn't looking. My post about this is on the developers forum...

 

So let me try this again after reading Deano's post. The 2 physics tools are 'sub-license', and come as part of the SDK - with all the appropriate integration and optimizations - if a studio license it? Or are portions already within the SDK even if a studio decide not to license them? Because the PR slides are a bit confusing, and Deano seems to imply that they are default?

I'm also lost about the Transmeta deal. I expected an IBM deal for CPU optimization instead! But since these are simple cores, it may be right up Transmeta's alley.

 

From what I've heard, Deano is correct.. both Havok and AGEIA's physics libraries are included as a part of the SDK - for 'free' if you like (ignoring the $$$ that the PS3 devkits will cost). Only UE3 is an evaluation/trial version.

I'm not so much confused about the Transmeta deal (for SPU work, although I'd hope some of their stuff could be applied to PU too) as I am about how the SN Systems compiler (SNC) work fits in with (or replaces) the existing GCC toolchain for PS3. Will the PS3 SDK ship with both? What about all the cool compiler work going on in GCC 4.x.. will PS3 developers have that?

Seeing SCEI take development tools issues seriously is good news though. I've been in a good mood since hearing about this stuf.

 

Well those features have been in commercial compilers for a long long long time. Really, it deserves that many 'long's. I'm sure as a long-time commercial-grade tool for console development, SN's compiler will live up to those rather common commercial standards.

Any dev here wants to contrdict this? Because if true, this is really ...WOW. SCE software plans are really way more competent than I expected. We had 2 threads per week wondering how to get Physics working on such an unconventional CPU. This seems to alleviate those concerns quite a bit - though probably not completely.

 

I think one of the outcome of this deal between SCEI, Havok, and Ageia is that it single-handedly virtually killed Meqon.

Another good news for George Broussard's team.

 

These should be nice to dev, they should be able to speed up the production, at least in the begining.

After all Sony is not so behind MS as we thought before, in terms of tools.

 

Well, I've been developing using commercial compilers for a long long long time too.. and the stuff in GCC 4.x still impresses me. It's radically different to GCC 3.4.x. But if you compare to VS.NET's compiler (2005 edition) and, to a lesser extent, Intel's then yes.. you're completely correct.

Of course, SN haven't been writing compilers for very long.. they've usually just modified GCC. Their PSP compiler, for example, is not yet mature.. so I wonder how long it will take them to get a *good* PU compiler up and running.

The next few months will be *very* interesting.. that's for sure!

 

My 2c on the SNsystems thing - I suspect it's primarily for their toolchain, debuggers and other tools they make (their DVD emulator for PS2 was a very nice thing for instance), as well as bringing tools to Windows platform.
The compilers as mentioned aren't that strong - they do the job, but nothing spectacular, so I don't think that's primary goal of this deal.
Sony also did a fair amount of internal work on compile tech so this could be part of consolidating things into a more focused compiler development.

 

if those sdk's are going to be properly loaded they are going to have to memory instance encoded, rather than object instance threaded; otherwise, the bandwidth use is going to a major nut buster.

 

http://www.winbeta.org/comments.php?id=3352&catid=1

http://www.tomshardware.com/hardnews/20050721_140938.html

 

Deanoc , what makes Havock better than Ageia ,in your eyes ?
Does it have to do with maturity ?

 

We've been using Havok for longer than Ageia have had a product, which is an important point (when you use middleware you don't want them disappearing, I was on the side-lines when Mathengine disappeared, it wasn't a pretty sight...)

Also we have a fairly close relationship with the Havok guys and we get excellent support of them. In many ways this is a vital consideration, it may be that Novodex has great support but as I've never used them so I don't know. Better the devil you know and all...

Havok has a fast robust physics system. Havok don't do fluids or particles system but what they do, is of a very high standard.

Both are multi-threaded aware (though how useful there support is under PS3 is doubtful, there multi-thread kernels seem to be designed for SMP systems). Havok has a continous collision system, which IMO is very important (I don't think Novodex's collision is continous but I'm not sure, it seems to have some anti-tunneling stuff).

Novodex is free for non commercial use, which is very cool (I'm thinking of using it for my personal projects... I currently use ODE at home).

Havok have now got some fairly good animation and IK code, I don't think Novodex has anything like that... While technically physics and anims aren't connected, in practise it makes life easier if they are. We aren't currently using Havok anims but we are looking closely to see if makes sense.

The biggest reason is simply because Havok is already intergrated into our code, so changing to a different (better or worse) physics engine would be a lot of work.

 

It's a shame that in future your support will be coming from SCEI then, isn't it..?

 

Tkz Deanoc.

I was asking ,because novodex seams a bit late on features ,particulary when you don't want a 100% fully physics based solving ,let's say to get SF stuff controled behavior (and then things get pretty uncrontrolable very fast).
Their strenght seems to be fast collision detection,and fluis is most probably a PPU only thing.
Destruction based on Vertex creation and material properties looks interesting ,but far from being available yet.

 

I'm sure they'd miss that Irish brogue at the other end of the phone

I think SCEI's support is just for Japan.

 

I'm not so sure. From what I remember of previous dealings with Havok, as a company they've been *very* careful of just how many people they licensed to - or allowed to evaluate - purely because they think, quite rightly, that the support developers get is just as (or more) important as the actual product. Put your support under too much stress (in terms of sheer numbers of developers), and the quality of support starts to fall. Actually, at a company I used to work for, we almost had to plead to get a 'slot' to evaluate.. that was pretty bizarre.

So anyway, back to my point (!). When SCEI include Havok with every PS3 SDK, there will be more developers that Havok can handle. And given that most of the cost of Havok is for support, I believe that all PS3 Havok support will be handled via SCEI. Not just for Japan.

 

I think you're probably right on this; the feel I get is that Sony probably got a significant discount on a per license basis from these guys in order to be able to bundle the full versions (with optomizations) within every PS3 SDK. And I think part of that agreement was probably that Sony also pay the cost of providing support. Whether that will mean funding on their part of an expanded and sub-contracted support team from these two companies or that they will train an in-house force, who knows - but it does seem Sony will bare the support costs, for better or for worse.

By the way I have to say I'm very impressed with the efforts Sony is making here. Post Playstation Meeting they've really exceeded my expectations as to how seriously they're taking erasing the errors of the past, and the Transmeta deal is an example of just how far they're taking it.