CELL Programming

[version]

http://www-128.ibm.com/developerworks/linux/library/pa-cell/index.html?

 

[version]

Don't post whole articles

 

[AlgebraicRing]

*Shakes Fists in Air*


Dammit, I want to know if and when there's going to be an emulator for the freaking processor. Looks like we're waiting on the GCC, GDB, and Linux ports to be complete. But they should have started (or one should start up) an opensource project to make a Cell emulator/simulator.

Also waiting for the technical specs, What are the assembly languages, what are the ISAs, how do you bootstrap the damn thing.

AR

I HATE WAITING!!!!

 

[passby]

Excellent posts. These are the kind of forum-posted articles B3D should have. Technical, unbiased, no PR FUD.

 

[ChryZ]

http://www.pro-linux.de/news/2005/8313.html

Linux Tag 2005 (german linux con)

-IBM dual cell blade prototype @ 2.4GHz
-300W PSU, estimated power consumption 100W per cell (guestimate)
-one cell was defective
-satellite demo was on display, but only as movie footage
-PPE was 99% compliant to Fedora PPC Linux, only a few adjustments were necessary, SPE implementation in progress
-the staff got tangled up in some (IMHO uneducated) speculations regarding cell for PS3

 

[Ragemare]

Also waiting for the technical specs, What are the assembly languages

Because the Cell's PPE uses the same instruction set as the PowerPC 970 CPU, it does not require changes to the compiler or the binutils. However, compiled code runs more efficiently on it when using a compiler that schedules instructions specifically for the CPU's pipeline structure. You can get a patch to GCC to add a pipeline definition for the PPE so you can create optimized code.

Since the SPU instruction set is not directly related to an existing CPU architecture, a new back end was written for both GCC and binutils. SPU code is compiled separately from the PPC code and gets loaded at run time.

Well that tells you which assembly language the PPE uses and somewhere else in the article it says that the assembly language for the SPUs mostly consists of arithmatic and branch instructions if thet tells you anything.

 

[Shifty Geezer]

All this talk of Linux on Cell, but who's going to be writing to the spufs for example? There's next to no Cell kit out there! Or is that only PS3 kits? How many devs have Cell gear to write to?

 

[Ragemare]

All this talk of Linux on Cell, but who's going to be writing to the spufs for example? There's next to no Cell kit out there! Or is that only PS3 kits? How many devs have Cell gear to write to?

Alot when IBM starts selling CELL workstations, servers and integrated systems. I'm sure there are many devs out there fairly interested in using CELL to power their applications despite not yet haveing the hardware.

 

[Carl B]

I agree, I think if nothing else Cell just has an immense novelty factor behind it that will keep it strong in the open-source community for quite awhile after this stuff gets released. And, if in that time the barrier to entry can be brought down for coding on Cell, then maybe it will have the chance to catch on in a wide range of fields as STI originally hoped.

 

[Titanio]

An interview with Arnd Bergmann:

http://www-128.ibm.com/developerworks/power/library/pa-expert4/

 

[AlgebraicRing]

An interview with Arnd Bergmann:

http://www-128.ibm.com/developerworks/power/library/pa-expert4/

That was what was deleted from the second post in this thread. What is interesting is that Arnd has no knowledge of any intention by IBM to sell cell workstations. And he's heading up the Linux port. Now either they are keeping Arnd in the dark, he has his head in the sand, or IBM really doesn't know what the hell to do with the processors apart from selling them to Sony. This actually makes me worried.

dW: But we will be having a Cell-based Blade, if nothing else?

Bergmann: Well, this is the one thing that I'm working on, and I don't know about the availability for customers at all.

What we are working on now is really a technology study. And people who are interested in that can ask the IBM Engineering and Technology Services [ET&S] about how to evaluate this technology for their purposes. But as far as I know, there will not be a Cell-based IBM server product.

There is a cell based prototype Blade board, but it doesn't sound like there is a business plan built around selling the thing. It sounds more like an experimental situation. Which the dev kits for the PS3 are probably on par. Makes the whole Cell operation sound pretty shaky. A big experiment at this point.

 

[one]

Makes the whole Cell operation sound pretty shaky. A big experiment at this point.

They won't have mission critical server products with Cell until his Linux port is fully developed and evaluated... the Cell workstation for Sony/IBM is OTOH a graphics workstation and is not a server.

 

[Shifty Geezer]

I thought I'd heard announcements of Cell being used...medical and military apps (imaging and image recognition). I think it comes down to custom products for the time being, and if the market grows, off the shelf solutions would follow. I wouldn't release a new server model, untried, that competes with my existing servers and would need a whole new level of support and development to maintain.

 

[Titanio]

IBM haven't got their own products planned, but they have a whole service setup to sell and incorporate it into other companys' products, IIRC. So they see it used beyond Sony, but not necessarily in their own systems, but others'.

 

[c0_re]

I mean no offence to anyone here, but by this time next year there will be quad core P4's(and hyperthreaded) and Athlons why would anyone buy a cell server blade when they can have a 32-way intel\Athlon cheaper more flexible and much more powerful overall.

It's obvious that except for some extremely specialized code the cell and the Xenos(or whatevr the 360CPU) is called just doesn't cut it for general computing purposes.

 

[Carl B]

The Cell has the potential to greatly outperform both the Intel and the AMD offerings in a number of areas - the question is software support. As far as scaling goes, just as future Intel and AMD chips will pack more in, Cell would be no less versatile if need be. Future revisions and process shrinks should allow for added functionality or cores or just allow pricing that makes the chip as a whole more competetive.

But again it'll really be an issue of software support.

 

[c0_re]

I agree in theory the potential is most certainly there I just see it being pushed to the wayside as many great products have been before because of market leader dominance and development woes. I hope your right, the CPU world is ready for something that’s REALLY new I just don’t see it happening.

 

[Shifty Geezer]

For home use, where Cell is 'weak' like wordprocessing it's still more than adequate for personal use. It'll be strong in media functions like photo retouching and movie editing. This, as I understand it, is where Sony and friends were approaching the design of Cell from. Rather than building up an existing technology to make it more powerful, where that original technology wasn't at all designed for modern uses, design something from scratch to better support modern uses.

Where Cell is lowsy for server situations (web servers, huge databases), for home use is it not ideal?

Question : What home user applications will Cell be weak at compared to Intel/AMD, and will it be so weak that people will notice? What areas will Cell be stronger at and will home users notice?

Assuming Cell does have very apparent advantages and non-apparent disadvantages there could certainly be a market for an alternative to MS if cheaper than Apple, IMO

 

[c0_re]

I'd like to see the Cell to be used for proffessional video editing equipment or used in consumer HDTV 1080p camcorders and TV's or somthing to that nature, would that be a possibuility?

I mean we all know pretty much nothing can come close to matching it's performance for encoding, right?

 

[hey69]

wouldnt Cell be very fast in 3dmax, maya and lightwave rendering?
i believe these are FPU hungry?