Another New Watch Impress GDC/PS3 article

[London Geezer]

Anyone know how intensive SPH fluid simulation is? I did some googling, but only found links in the context of astrophysics rather than games

Oh believe you me, it's tough to do properly.
Fluid dynamics like this was something that was only done in CGI up until not too long ago. Personally i find it fascinating that we can already do this in realtime, i thought it would take a lot longer.

 

[Mmmkay]

They did show crossing of streams, but it was very brief, and hard to tell how or if they were interacting:

Yeah I remember that bit of the demo, but seeing how he's holding the cups on stage it doesn't look like they're pouring them into each other.

 

[London Geezer]

Yeah I remember that bit of the demo, but seeing how he's holding the cups on stage it doesn't look like they're pouring them into each other.

I'm almost 100% sure that in that eyetoy demo there was no Z axis movement. If you look at the video, the movement of the cups looks funny on the screen because they only move along the X and Y axis and not the Z axis. In that case, the streams should intersect as the cups have the same Z positions in the system.

Could be wrong but i'm quite sure i'm not...

 

[DUALDISASTER]

Quick question about the threading mentioned in the ppt and the OS, is the OS suppose to run on a thread of the SPE? If so, the OS wouldn't consume a whole SPE. That'd be great.

 

[Mmmkay]

Could be wrong but i'm quite sure i'm not...

Hmm you could indeed be right, but when he first scoops up the water with the left glass, he seems to move it to the lower left hand side of the pool while the right glass is hovering around the upper right. Though he doesn't make things easy because when they show him using the cups, he keeps them on the same plane most of the time. It doesn't matter either way though, since the streams didn't properly interact even if they were fixed on the z-axis or weren't given the opportunity to interact if they had free movement along the z-axis.

 

[London Geezer]

Hmm you could indeed be right, but when he first scoops up the water with the left glass, he seems to move it to the lower left hand side of the pool while the right glass is hovering around the upper right. Though he doesn't make things easy because when they show him using the cups, he keeps them on the same plane most of the time. It doesn't matter either way though, since the streams didn't properly interact even if they were fixed on the z-axis or weren't given the opportunity to interact if they had free movement along the z-axis.

I'm sure there is no movement along the Z axis, so the 2 cups were either locked at different points on it, which i think is the case, or the 2 streams just went through each other.
My opinion is that the 2 cups were just locked at 2 different point on the z axis therefore the streams just didn't touch each other.

 

[Mmmkay]

My opinion is that the 2 cups were just locked at 2 different point on the z axis therefore the streams just didn't touch each other.

Ah, well I can't agree with that because he poured water from one glass into the other a couple of times.

 

[London Geezer]

Ah, well I can't agree with that because he poured water from one glass into the other a couple of times.

Did he? Can't remember!
Then they're at the same points on Z.

 

[Shifty Geezer]

It looked to me like the water was precaptured, and the guy was just going through the motions. It didn't match exactly. Did anyone see an EyeToy camera on set? It'd have to be pretty close as it's a wideangle lens. Or they had a custom telephoto version. But personally I think that bit was a stage-fudge.

Which demo was that? The only one I'd heard of previously was the different simulations running on each SPE, in a cube..

That is the one. They only managed 8 pieces of cloth IIRC. Using all the SPEs. Not really useful for realtime game situations - one frigate at full sail and there'd be no resources for the rest of the game, let alone a whole armada!

 

[Titanio]

That is the one. They only managed 8 pieces of cloth IIRC. Using all the SPEs. Not really useful for realtime game situations - one frigate at full sail and there'd be no resources for the rest of the game, let alone a whole armada!

Well I guess that speaks to the relative sophistication of the simulations and/or 'cleverness' of their implementations I mean, the duckies demo was managing 3 pieces of cloth per boat with ripping and what looked like self-collision, with just a fraction of a SPE's time. My guess is the Alias implementation is more technically particular and 'correct', with denser meshes?

 

[Shifty Geezer]

Probably. Just we don't know, so can't glean any useful performance expectations from this, especially in game situations. It's that sort of info, on the complexity of the simulation, that we need as well as being told it's using <%50 of a SPE.

 

[Mmmkay]

It looked to me like the water was precaptured, and the guy was just going through the motions. It didn't match exactly. Did anyone see an EyeToy camera on set? It'd have to be pretty close as it's a wideangle lens. Or they had a custom telephoto version. But personally I think that bit was a stage-fudge.

Yeah I thought that too when I first saw it, which is why I was so impressed to see them discuss it in that presentation. It would be a bit err disingenuous to go to great lengths to explain how they implemented SPH (some 10 months after the demo) if they didn't really do it. They used two differently coloured plastic cups and it was pretty clear given how he was manipulating the cups (compared to what was happening on screen) that it wasn't staged. There was a white mound in the center of the podium where he was standing which would have been the camera.

 

[Titanio]

It looked to me like the water was precaptured, and the guy was just going through the motions. It didn't match exactly. Did anyone see an EyeToy camera on set? It'd have to be pretty close as it's a wideangle lens. Or they had a custom telephoto version. But personally I think that bit was a stage-fudge.

The eyetoy was there, you could see him adjust it just before he started. I would say there was a slight lag between his motion and what came up on the screen, rather than it not matching up. The lag could be explained by a variety of reasons (not least, perhaps, the projected image - synching audio with a home theatre projector can be tricky sometimes because of lag).

 

[Titanio]

BTW, it wasn't obvious from the slides, but this was apparently another demo they showed in the physics presentation:

pcostabel on GAF says it was one of the most impressive demos he saw at GDC, the cloth would wrinkle at a very fine level, no interpenetrations, not 'rubbery' etc.

 

[London Geezer]

I'd pay to see that moving...

 

[Shifty Geezer]

Cell is looking better and better for makeup and clothes design etc. Perhaps PS3 can be made to appeal to girlies as a result

 

[Shifty Geezer]

The eyetoy was there, you could see him adjust it just before he started. I would say there was a slight lag between his motion and what came up on the screen, rather than it not matching up.

Maybe. I don't remember the EyeToy adjusting. That doesn't make much sense to me though, as all the setup should have been done before the show starts. Bad stage crew if not.

Either way, I take the technology as real and indicative of the EyeToy interacting with fluid dynamics. Whether the show was live or a prerecorded live show makes no odds. Stagecraft is all about giving impressions. If you watch the original River Dance recording carefully you might catch glimpse of a musician playing the spoons to get the sound of Flatley's taps!