[Beowulf] Has anyone actually seen/used a cell system?

[Vincent Diepeveen]

Not wanting to sound too negative, but total nonsense concept.

First of all this 'sequoia' claims to be a new programming language. Meaning 
it'll take a year or 30
until some good compilers for it are there, provided someone is going to 
support it.

Which isn't going to happen.

The parallellization basically is based upon complex assumptions for 
programmers. So for programmers they don't actually make it easier than 
trivial parallellization is via C/C++ function calls.

The sequoia parallellization basically is simplistically over for loops that 
a programmer himself can
trivially parallellize too.

Further the optimization of sequoia simply doesn't happen. They assume 
"kernel libraries" solve the
problem. Interestingly it mentions explicitly:

"if kernel libraries could be obtained, such as FFTW and the intel MKL for 
PCs, or the IBM SPE matrix library for Cell, we call these libraries from 
Sequoia leaf tasks".

In short if some algorithm has not been implemented for sequoia, sequoia is 
unusable.
Others may do the work as usual to promote sequoia.

Not gonna happen.

If a processor has a lot of cores and isn't having a native support from 
intel nor ibm, then it is going to fail.

Lot of assumptions. In the end what you need is a 1000 compiler personnel 
who optimize a compiler,
especially when it has to be built from ground up.

Who is going to pay them for sequoia?

ANL (Another new language) is what everyone writing things on paper in an 
university likes to invent,
but that's not going to help us execute our programs easier let alone 
faster.

There is so many little tricks in implementation of a compiler that it is a 
sheer impossible task for a few people to create an efficient compiler. Why 
do all those paper guys not practical help to optimize GCC,
which definitely needs some help, especially a 64 bits port that works in 64 
bits windows at x86-64.

Right now there is only some possibilities to create 32 bits gcc exes for 
windows using for example
mingw with their msys.

I can garantuee you that those paper guys keep writing such papers and show 
up with ANL.

Vincent

----- Original Message ----- 
From: "Andrew Shewmaker" <agshew at gmail.com>
To: <J.A.Delcorso at larc.nasa.gov>
Cc: <beowulf at beowulf.org>
Sent: Sunday, October 01, 2006 7:05 AM
Subject: Re: [Beowulf] Has anyone actually seen/used a cell system?


> On 9/19/06, J.A.Delcorso at larc.nasa.gov <J.A.Delcorso at larc.nasa.gov> wrote:
>> If this has already been discussed on the list, a pointer
>> to the thread would be appreciated!
>>
>> Looking over the list I found a couple references to the
>> IBM cell processors being used in PS3.  Has anyone had a
>> chance to actually use/test one of these systems yet?
>> Considering that the PS3 is supposed to be competing with
>> the XBOX 360 and the Wii, I assume that it's out, or that
>> someone has actually used one of these systems.
>>
>> Can anyone point me to a url, or tell me what their
>> experience is with this technology?  Is it as fast as
>> it's purported to be?  Apparently RedHat is developing
>> EL 4.3 to run on the system?
>>
>> Any input is appreciated,
>
> People in Stanford's graphics lab are doing promising
> work making the Cell easier to program.
>
> They have been busy creating a programming language
> called Sequoia that focuses on memory hierarchies.  A
> programmer defines a task that recursively decomposes
> the input data until it reaches the bottom of the memory
> hierarchy, where it runs an optimized kernel.  The runtime
> environment uses a specification of the memory hierarchy
> for the generic PC cluster, Cell Blade Server, etc.  In the
> case of a cluster, the runtime is built on top of MPI.  In the
> case of a Cell Blade Server, the runtime manages task
> overlays and asynchronous DMA transfers of the data.
>
> They have a paper that explains it well and has some
> interesting benchmarks.
>
> http://sc06.supercomputing.org/schedule/pdf/pap225.pdf
>
> And the public release of their compiler and runtime
> system should be publically available soon.  The site
> does have some more documentation of Sequoia.
>
> http://www.stanford.edu/group/sequoia/cgi-bin/
>
> -- 
> Andrew Shewmaker
> _______________________________________________
> Beowulf mailing list, Beowulf at beowulf.org
> To change your subscription (digest mode or unsubscribe) visit 
> http://www.beowulf.org/mailman/listinfo/beowulf
>