Creating High Performance Radar Applications with the Cell Broadband Engine

Cool ! Can't watch now. Will it be archived ? Can you summarize ?
Porting radar tracking system was my first job (long long ago).

 

R:FoM is being...'evaluated' as a soldier training programme...

 

Any word on if an Axis of Evil<TM> nation has purchased untold amounts of PS3s to use as missile guidance system?

Cool thread. Thanks!
Even though it's a military application it's still fascinating. Mopre adoptation of Cell means more coders having familiarity with the chip = btter for devs and in the end us gamers!
Peace.

 

A couple of our research guy bought a PS3 to learn Cell programming. It cost too much and takes a while to get a Mercury or IBM Blade.

 

You should check the F@H world map every day to watch if Kim Jong-il is running it in North Korea...

For another Cell adoption news item,
http://www.hpcwire.com/hpc/1322708.html

 

Well it's no good for jeneral processings. Game code, you know. Maybe if radar programming was 87% streaming it would be a different story but the fact of the matter is that all that stuff is done on gpus now.

I kid.

Sorry, I just got thru poking around the locked propaganda thread and I couldn't believe what I was reading. Great job in there, XBD and others, trying to explain things. Even if I had the ability to explain that stuff as well as you did, I would have gone crazy doing so much typing.

On topic, I'm pleased to see the take-up of cell in various industries. It gives me a glimmer of hope one day that consumer computing will be accomplished by more lightweight, elegant silicon than it is today.

Here's to a cheap, powerful, flexible and (imo) pleasingly nuanced new architecture. May it be the first of many!

 

Thanks for the synopsis!
i missed the event

 

Cell supports your war of^H^Hon Terror

So the Cell is potentially good at making advanced weapons.

I guess if the console "war" ever manifests itself as a real war, we know who will have the upper hand

 

How is this console related?

 

I was curious about this. I suspected that a Cell fabbed at 90nm for a commercial product wouldn't be up to task for applications that need hardened electronics.

Perhaps if a run of chips at 90nm were given process tweaks like thickened gate oxides, and they significantly lowered clocks, there would be wider applicability. It may be that Cell has too much hardware, and they could dump half the SPEs.

At those low geometries, though, it may necessitate a variant of Cell with much better RAS and error correction. The PS3 doesn't put too much of a premium on data and system integrity.

A hardened Cell wouldn't be as impressive at peak numbers compared to the commercial variants, but it could still be very impressive.


The other thing I noticed is that the PPE again is turning up as a bottleneck. The transpose operations and the redistributing of data to the SPEs could be improved with a better PPE.

What this means for future Cells that will further increase the ratio of SPEs to PPEs is unclear.

 

Somewhat loosely! It all ties in to Cell's performance, particularly it's advantages in certain algorithms, and all evidence of this helps get an idea of what applications the chip can be put to - on and out of games.

Basically, seeing Cell perform well at these tasks makes people go 'woooo' at the prospect of a fast chip driving their games

 

Wasn't that bottleneck already overcome in comparable projects by simply shifting some of the restribution logic to SPE? For instance, a PPE sends data to two SPES and those in turn delegate some work to three more SPEs each (or two).

Wasn't that what they did in HAVOK among others?