EX-NAUGHTY DOG WORKING ON PS3 3D ENGINE

[Panajev2001a]

Gregg made tUME, Gregg is a programmer divinity...

And what he made at WoW was really something: he is the right man for 3D Engine and tools development.

I think SCE got the right man

 

[Deadmeat]

...

Edited by moderator

Do not post in such a derogatory manner again. If you do so you will be reccomended for a ban. Talking about a man committing suicide may be a joke to you, but it is not to others. This post is unexcusable, this is your only warning.

- Gee, I can't even joke here??? But the message remains the same. It is an impossible task to get PSX3 running even at 50% of capacity, and this poor coder is going to be facing the greatest obstacle in his entire programming career.

CELL was designed with no consideration for programmability, and its ill-effect is beginning to be felt in the developer community.

 

[oi]

If my memory doesn't fail me I believe it was Naughty Dog who gave the "It's only hard if you suck." comment about programming for the PS2 after Oddworld Inhabitants changed platform from PS2 to Xbox, because of how "hard" it was. And after seeing them deliver JD and JD2 I have no doubts about the competence of anyone from the ND crew. Definatly the right man for making a cool engine

 

[maskrider]

(I deleted my last worthless post)

Hopefully SCE will get it right this time with the right tools to kick start the development of the system.

If every game on PS3 will utilise the machine like J&D or even J&D2 level on PS2, graphics whores should be very happy.

 

[Paul]

It is an impossible task to get PSX3 running even at 50% of capacity, and this poor coder is going to be facing the greatest obstacle in his entire programming career.

Because final Playstation 3 developers kits are infact out in the open right?

CELL was designed with no consideration for programmability

and its ill-effect is beginning to be felt in the developer community.

Because developers are developing for Broadband Engine right now correct?

Your posts are just sad, it's not about just hating Sony anymore, you take it to the next level. It's called obsession and it's a severe disorder, perhaps you should get it checked out sometime.

 

[ChryZ]

Thanx Deadmeat ... and back to square one. Let's lock up "Console Talk", the madness is breaking loose again

 

[Vince]

Thanx Deadmeat ... and back to square one. Let's lock up "Console Talk", the madness is breaking loose again

Hey now! Just drop it bud, it's over and the mods have done their thing - let’s respect it and do our part which is to make like it never existed. I'm sure the powers that be are aware of what's going down.

Anyways, I'm curious about this comment in particular:

It is an impossible task to get PSX3 running even at 50% of capacity, and this poor coder is going to be facing the greatest obstacle in his entire programming career.

I'm curious as to why this is so. If we can first make the distinction that the Broadband Engine (as we assume it to be) has it's roots in the PS2's design, it will share with it many of the same traits as the Emotion Engine; be these contextually posititive or negative.

So, if we look at the EE and it's problem with underutilization explicitly (as DMGA has done) is it in correct to state that the most pronounced design deficit on the EE centers around VU0, perhaps closely followed/intertwined by some deficits in the MIPS core. If this is so, which I'm opening up to criticism and comment, then it would appear that the concept of a APU goes along way to resolving this efficiency problem. If the patent is to be believed (again, a known assumption) then it would appear that by making the APUs symmetric in conjunction with increasing the number of registers and amount of local cache, as well as their interconnectivity, would lend itself to comparing a singular APU more to VU1 - which AFAIK is getting pretty damn good efficiency out of programmers. The PA results would seem to bear this out. And while we're basically clueless to the microarchitectual details of the PE, be it Power/MIPS or ARM based, wouldn't it be true that since the APU's are now stand-alone processors capable of "feeding themselves" (elegantly said, I know. But to the point... ) as opposed to the PS2, the role of the superscalar component is dramatically decreased?

If anything, I would hazard that while the BE is hardly the 'bestest' thing since sliced bread and the glow-in-the-dark condom, it's design would appear to incorporate many architectural features which seem to be included due to lessons learned with the Emotion Engine and PS2. So, feel free to support or rip me a new one... heh.

EDIT: I'm also not implying that we'll see 90%+ utilization rate on all software or any such thing one might somehow be led to believe - this is something I just wanted to make absolutly clear. Also, if anyone knows the Chicago cop who saw fit to give myself and around 10 other people tickets with codes for parking under a viaduct which doesn't exist... feel free to lead him to the head of the class, ass.

 

[V3]

If the patent is to be believed (again, a known assumption) then it would appear that by making the APUs symmetric in conjunction with increasing the number of registers and amount of local cache, as well as their interconnectivity, would lend itself to comparing a singular APU more to VU1 - which AFAIK is getting pretty damn good efficiency out of programmers. The PA results would seem to bear this out.

Will they expose the APUs the way they do with VU1 though ?

And while we're basically clueless to the microarchitectual details of the PE, be it Power/MIPS or ARM based, wouldn't it be true that since the APU's are now stand-alone processors capable of "feeding themselves" (elegantly said, I know. But to the point... ) as opposed to the PS2, the role of the superscalar component is dramatically decreased?

They still need to maintain control table of some sort for DMAC purposes. Besides that, they'll have to evaluate those apulets and load them to its proper place as well as sending interrupts to the APUs to get it starting, so

Even if APUs can keep themself busy, they will only be busy once the PU done its work. While the APUs is busy, the PU don't relax either.

So I don't think their role is dramatically decrease. If their role has dramatically decrease Sony would have stuck just one PU for the whole Broadband Engine.

 

[kaching]

I think you guys are making some headway with Deadmeat. The fact that he's allowing for the design of the PS3 to operate at as much as 49% seems like an upward revision of his earlier estimates.

I mean, that's pretty damn good for a design "with no consideration for programmability", isn't it?

 

[Panajev2001a]

If the patent is to be believed (again, a known assumption) then it would appear that by making the APUs symmetric in conjunction with increasing the number of registers and amount of local cache, as well as their interconnectivity, would lend itself to comparing a singular APU more to VU1 - which AFAIK is getting pretty damn good efficiency out of programmers. The PA results would seem to bear this out.

Will they expose the APUs the way they do with VU1 though ?

And while we're basically clueless to the microarchitectual details of the PE, be it Power/MIPS or ARM based, wouldn't it be true that since the APU's are now stand-alone processors capable of "feeding themselves" (elegantly said, I know. But to the point... ) as opposed to the PS2, the role of the superscalar component is dramatically decreased?

They still need to maintain control table of some sort for DMAC purposes. Besides that, they'll have to evaluate those apulets and load them to its proper place as well as sending interrupts to the APUs to get it starting, so

Even if APUs can keep themself busy, they will only be busy once the PU done its work. While the APUs is busy, the PU don't relax either.

So I don't think their role is dramatically decrease. If their role has dramatically decrease Sony would have stuck just one PU for the whole Broadband Engine.

Agreed, but compared to EE's VUs which could not DMA data themselves nor do any I/O at all to the APUs which can DMA data to and from the e-DRAM and do some kind of I/O with external devices we can see that there has been a change.

There is work to be done by the PU: decision of memory sandboxes access, I/O, message passing, OS execution, processes/threads scheduling, etc...

 

[Deadmeat]

...

I think you guys are making some headway with Deadmeat. The fact that he's allowing for the design of the PS3 to operate at as much as 49% seems like an upward revision of his earlier estimates.

I was simply making a case from PSX2 performance examples... of course the PSX3 will peak out at much less.

 

[kaching]

Ah, but you make no mention of PS2 performance in your previous post...you simply reference the CELL design in and of itself. How was anyone to know that your optimistic figures for PS3 were based on PS2 performance?

So how much less capacity should we expect PS3 code to run at, given the design makes NO consideration for programmability, as you say? 0%, maybe 5% tops?