Re: [9fans] Plan9 on the Cell...

Re: [9fans] Plan9 on the Cell...

[jmk]

On Sat May 28 14:34:58 EDT 2005, plalonde@telus.net wrote:
> Really, for the PPC64/Cell compiler I don't care about the name...
> But then we also need a compiler for the SPUs; that will be much more  
> annoying to deal with :-)
> 
> Paul

You are right not to concern yourself with the name of
the compilers. Such things will be decided in the usual
way by a shadowy conclave of Plan 9 principals and a
successful deliberation announced to the hoi polloi by
the issuance of white steam from the cooling towers
behind Bldg. 3 at Murray Hill, NJ.

Cardinal Biggles

Re: [9fans] Plan9 on the Cell...

[Eric Van Hensbergen]

On 5/24/05, Eric Van Hensbergen <ericvh@gmail.com> wrote:
> 
> In the meantime, porting to the G5 (or any other readily available
> ppc64 platform) would probably be your best bet.  First step is the
> compiler (while you may be able to use the 32-bit power compiler,
> you'll really want 64-bit if you are looking to squeeze performance).
> Next is working with jmk (who is adding 64-bit support to the kernels
> for opteron) to do the same thing for PowerPC.
> 

As a step in providing resources towards this direction, I've put a
dual-2GHz G5 on my home network next to my grid systems.  If folks
want to take a crack at getting kencc to run under ppc64 Linux (or to
test output from cross-compilation) (it should also be possible to
test ppc32 output as well).  I will be trying to get the system
running the new version of rhype which is capable of running bare on
G5s, which will allow folks to try out cross-compiled kernels under
hypervisor on PPC64/PPC32.

I'll also be hosting a IBM Full System Simulator installation there
once its released - this will allow running kernels in an emulation
environment which should make bring up easier.

If folks are going to be actively pursuing this path and want access
to ppc64/ppc32 hardware, let me know and I'll set up accounts.

       -eric

Re: [9fans] Plan9 on the Cell...

[Martin C. Atkins]

On Sun, 12 Jun 2005 20:39:51 -0600 (MDT) "Ronald G. Minnich" <rminnich@lanl.gov> wrote:
> doubtful. It's a virtual machine environment, but not for different 
> architectures (yet). If you want emulation, you'll need to run it under 
> linux is my bet.

Or port qemu/etc to run natively (ie., as an OS) under the PS3's
virtual machine environment! :-) (Much sillier things have been done...)

Martin
-- 
Martin C. Atkins			martin_ml@parvat.com
Parvat Infotech Private Limited		http://www.parvat.com{/,/martin}

Re: [9fans] Plan9 on the Cell...

[Ronald G. Minnich]

On Mon, 13 Jun 2005 arisawa@ar.aichi-u.ac.jp wrote:

> Doesn't this mean PS3 will have virtual PC environment as a bonus?

doubtful. It's a virtual machine environment, but not for different 
architectures (yet). If you want emulation, you'll need to run it under 
linux is my bet.

ron

Re: [9fans] Plan9 on the Cell...

[George Gensure]

>hint at what this environment may be like, check out the IBM research
>hypervisor rHype.
>
>           -eric
>  
>
Along this line, the Revolution is supposed to have an IBM core called 
'Broadway'... anyone know anything about the architecture?

-George

Re: [9fans] Plan9 on the Cell...

[Eric Van Hensbergen]

On 6/12/05, arisawa@ar.aichi-u.ac.jp <arisawa@ar.aichi-u.ac.jp> wrote:
> 
> Doesn't this mean PS3 will have virtual PC environment as a bonus?
> 

It means PS3 will have a Hypervisor environment.  The Cell Hypervisor
architecture will likely be more akin to Xen versus vmware.  For a
hint at what this environment may be like, check out the IBM research
hypervisor rHype.

           -eric

Re: [9fans] Plan9 on the Cell...

[arisawa]

> On Fri, 10 Jun 2005, Paul Lalonde wrote:
>
>> And it gets even more interesting: Kutaragi-san seems to think we'll 
>> be able
>> to run (and write) new OSes to run on the PS3:
>> http://www.gamespot.com/news/2005/06/09/news_6127219.html
>>
>> Plan9 on a cheap and very effective piece of consumer electronics 
>> looks very
>> possible this morning :-)
>
> I heard a bit more today. Also, read the article carefully. Other OSes 
> are
> in the context of the virtual machine. PS/3 OSes run in a VM 
> environment.
>
> In other words, you can port an os but it will probably be running on 
> the
> Sony hypervisor.
>
> So, get ready -- we'll probably need more Plan 9 interfaces to virtual
> machine environments.

in http://www.gamespot.com/news/2005/06/09/news_6127219.html

> In order to run the OSes, we need an HDD. So in order to declare that 
> the PS3 is a computer, I think we'll have [the PS3's HDD] preinstalled 
> with Linux as a bonus.

Doesn't this mean PS3 will have virtual PC environment as a bonus?

Kenji Arisawa

Re: [9fans] Plan9 on the Cell...

[C H Forsyth]

>>you get, among other things, an 
>>emulated NE 2000 (That's my reading!). 

such magic, Goodgulf!

Re: [9fans] Plan9 on the Cell...

[Ronald G. Minnich]

On Fri, 10 Jun 2005, Tim Newsham wrote:

> What about standard VM perks like being able to share device drivers
> from a popular operating system?  Does Vanderbilt support that?

my reading (incomplete) of VT is that you get, among other things, an 
emulated NE 2000 (That's my reading!). 

So, I don't know what to say ...

You really do want virtual devices like the xen front end/back end 
drivers, since you can do better with a shared queue than with emulating 
some wacky legacy device (I think).

So, plan 9 will run unmodified under VT, but we do want to continue to 
have a plan 9 that is built for xen, long term.

Tim, I'll contact you w.r.t. Xen 3.0 and we can figure out what to do. 
I've started but 3.0 is still moving target mode -- first thing is to get 
it on my desktop and hope it will work. You can't have 2.0 and 3.0 on the 
same machine, so I am not willing to muck up my laptop just yet with 3.0 
given how well 2.0 is working for me. I did a demo today with a "cluster" 
on my laptop -- the compute nodes boot in 1 second on Xen.

ron

Re: [9fans] Plan9 on the Cell...

[Tim Newsham]

> So, anybody want to help with Plan 9 port to Xen 3.0?

I'm in.

> Other good news: current claim is that on VT technology Pentiums you can
> run an unmodified OS!

What about standard VM perks like being able to share device drivers
from a popular operating system?  Does Vanderbilt support that?

> ron

Tim Newsham
http://www.lava.net/~newsham/

Re: [9fans] Plan9 on the Cell...

[Ronald G. Minnich]

On Fri, 10 Jun 2005, Paul Lalonde wrote:

> And it gets even more interesting: Kutaragi-san seems to think we'll be able
> to run (and write) new OSes to run on the PS3:
> http://www.gamespot.com/news/2005/06/09/news_6127219.html
> 
> Plan9 on a cheap and very effective piece of consumer electronics looks very
> possible this morning :-)

I heard a bit more today. Also, read the article carefully. Other OSes are 
in the context of the virtual machine. PS/3 OSes run in a VM environment. 

In other words, you can port an os but it will probably be running on the 
Sony hypervisor. 

So, get ready -- we'll probably need more Plan 9 interfaces to virtual 
machine environments. 

So, anybody want to help with Plan 9 port to Xen 3.0? 

:-)

Other good news: current claim is that on VT technology Pentiums you can 
run an unmodified OS!

ron

Re: [9fans] Plan9 on the Cell...

[Bruce Ellis]

once the ps2 doc was out the inferno port took a day

Re: [9fans] Plan9 on the Cell...

[Paul Lalonde]

And it gets even more interesting: Kutaragi-san seems to think we'll  
be able to run (and write) new OSes to run on the PS3:
http://www.gamespot.com/news/2005/06/09/news_6127219.html

Plan9 on a cheap and very effective piece of consumer electronics  
looks very possible this morning :-)

Paul

Re: [9fans] Plan9 on the Cell...

[Eric Van Hensbergen]

On 5/24/05, Paul Lalonde <plalonde@telus.net> wrote:
> IBM looks like they are going to open up the Cell architecture:

http://www.gamespot.com/news/2005/06/09/news_6127219.html

Okay, Sony has basically announced that they will be allowing "hosted"
operating systems on top of their "kernel" -- which to me sounds like
they are calling their hypervisor "the kernel".  Either they are going
to release documentation on their hypervisor interface, or it will be
apparent in their forthcoming Linux patches -- so I definitely
interested in working with folks to make the Plan 9 port happen.

In my view, we'd be best served by working on getting Plan 9 on a
commerically available ppc64 (or more cost-effectively a ppc64
simulator) preferably on top of a hypervisor architecture (rHype would
be the open-source candidate here).

More from me in a couple of weeks after IBM makes some external
announcements which I can't talk about yet.

        -eric

Re: [9fans] Plan9 on the Cell...

[David Leimbach]

On 6/3/05, Ronald G. Minnich <rminnich@lanl.gov> wrote:
> 
> 
> On Fri, 3 Jun 2005, Dave Eckhardt wrote:
> 
> > One thing to be aware of:  PearPC doesn't (at least as of early spring)
> > really implement OpenFirmware.  When (e.g.) MacOS tries to run a piece
> > of forth code, PearPC does switch(strlen(code)) to figure out which
> > hard-coded mock-up to invoke.
> 
> what a plus. Skip that OF stuff and run it in the OS. This is cool. I'll
> have to try to take a look at that -- is that support stuff available in
> source form?
> 
> Bios'es are for booting, not normal operation :-)
> 

This concept is far too simple to be popular.

> ron
>

Re: [9fans] Plan9 on the Cell...

[Brantley Coile]

> Bios'es are for booting, not normal operation :-)
> 
> ron

Unless you're running cp/m.

Re: [9fans] Plan9 on the Cell...

[Ronald G. Minnich]

On Fri, 3 Jun 2005, Dave Eckhardt wrote:

> One thing to be aware of:  PearPC doesn't (at least as of early spring)
> really implement OpenFirmware.  When (e.g.) MacOS tries to run a piece
> of forth code, PearPC does switch(strlen(code)) to figure out which
> hard-coded mock-up to invoke.

what a plus. Skip that OF stuff and run it in the OS. This is cool. I'll 
have to try to take a look at that -- is that support stuff available in 
source form?

Bios'es are for booting, not normal operation :-)

ron

Re: [9fans] Plan9 on the Cell...

[Dave Eckhardt]

> What would you need to do to get a working Plan 9 port up and
> running on pearpc?

One thing to be aware of:  PearPC doesn't (at least as of early
spring) really implement OpenFirmware.  When (e.g.) MacOS tries
to run a piece of forth code, PearPC does switch(strlen(code))
to figure out which hard-coded mock-up to invoke.

Dave Eckhardt

Re: [9fans] Plan9 on the Cell...

[Eric Van Hensbergen]

On 6/3/05, Noah Evans <noah.evans@gmail.com> wrote:
> What would you need to do to get a working Plan 9 port up and running
> on pearpc? Would the kernel's vesa drivers would work? Pearpc has
> etherlink support so you wouldn't need to write a new ethernet driver
> either(initially), right? What else would you need to do?
> 

Given that you have a network, the best thing would be to avoid video
and just set the sucker up as a CPU server then drawterm into it.

          -eric

Re: [9fans] Plan9 on the Cell...

[Noah Evans]

What would you need to do to get a working Plan 9 port up and running
on pearpc? Would the kernel's vesa drivers would work? Pearpc has
etherlink support so you wouldn't need to write a new ethernet driver
either(initially), right? What else would you need to do?

Re: [9fans] Plan9 on the Cell...

[Eric Van Hensbergen]

On 5/29/05, Charles Forsyth <forsyth@terzarima.net> wrote:
> >>What would a g5 port entail?
> 
> a fair bit of kernel work.
> 

It could be slightly less.  There are efforts underway to get either
IBM's rhyper or Xen-ppc to boot on the G5.  This would simplify the
driver support, but we'd still need to do a lot of work to support
hypervisor based mmu.

           -eric

Re: [9fans] Plan9 on the Cell...

[Charles Forsyth]

>>What would a g5 port entail? 

a fair bit of kernel work.

Re: [9fans] Plan9 on the Cell...

[Charles Forsyth]

i suspect it can probably be done as an option to the existing compiler and linker.
amd64 needed a new one because the floating-point is completely different,
with proper registers this time, and code generation strategy is also a little different
for integers.  by contrast, ppc64 is just the same as the 32-bit mode,
with some extra instructions and somewhat different code required to load values.

Re: [9fans] Plan9 on the Cell...

[Noah Evans]

What would a g5 port entail? 

On 5/29/05, Ronald G. Minnich <rminnich@lanl.gov> wrote:
> Actually, I learned a good bit more about the whole Cell situation this
> morning, from the chipset level on up to the 'when will systems get out'
> level. The situation is, to say the least, murky, and will be for a while.
> 
> However: If we were to put effort into a G5 port, which is cheap to do
> (buy a mac!), we'd be ready to rock and roll on the Cell when systems
> become more generally available.
> 
> So, Plan 9 is getting fixed for 64-bit mode, we have a K8 compiler now as
> a template. What do we name the compiler?
> 
> Cell C (or cc)? Just joking!
> 
> G5 C or gc? I'll stop now.
> 
> ron
> 
>

Re: [9fans] Plan9 on the Cell...

[Paul Lalonde]

Really, for the PPC64/Cell compiler I don't care about the name...
But then we also need a compiler for the SPUs; that will be much more  
annoying to deal with :-)

Paul

On 28-May-05, at 10:56 AM, Ronald G. Minnich wrote:

> Actually, I learned a good bit more about the whole Cell situation  
> this
> morning, from the chipset level on up to the 'when will systems get  
> out'
> level. The situation is, to say the least, murky, and will be for a  
> while.
>
> However: If we were to put effort into a G5 port, which is cheap to do
> (buy a mac!), we'd be ready to rock and roll on the Cell when systems
> become more generally available.
>
> So, Plan 9 is getting fixed for 64-bit mode, we have a K8 compiler  
> now as
> a template. What do we name the compiler?
>
> Cell C (or cc)? Just joking!
>
> G5 C or gc? I'll stop now.
>
> ron
>
>

Re: [9fans] Plan9 on the Cell...

[jmk]

On Sat May 28 14:11:40 EDT 2005, knapjack@gmail.com wrote:
> On 5/28/05, Ronald G. Minnich <rminnich@lanl.gov> wrote:
> > So, Plan 9 is getting fixed for 64-bit mode, we have a K8 compiler now as
> > a template. What do we name the compiler?
> > 
> > Cell C (or cc)? Just joking!
> > 
> > G5 C or gc? I'll stop now.
> 
> What's the K8 compiler called? kc?  'Cause I want to put on my my my
> my my boogie shoes.
> 

6[acl]

it's already in some of the header files like mach.h.

> We'll be here all week.
> 
> -Jack

great, we'll stop by.

--jim

Re: [9fans] Plan9 on the Cell...

[Jack Johnson]

On 5/28/05, Ronald G. Minnich <rminnich@lanl.gov> wrote:
> So, Plan 9 is getting fixed for 64-bit mode, we have a K8 compiler now as
> a template. What do we name the compiler?
> 
> Cell C (or cc)? Just joking!
> 
> G5 C or gc? I'll stop now.

What's the K8 compiler called? kc?  'Cause I want to put on my my my
my my boogie shoes.

We'll be here all week.

-Jack

Re: [9fans] Plan9 on the Cell...

[Ronald G. Minnich]

Actually, I learned a good bit more about the whole Cell situation this
morning, from the chipset level on up to the 'when will systems get out'
level. The situation is, to say the least, murky, and will be for a while.

However: If we were to put effort into a G5 port, which is cheap to do
(buy a mac!), we'd be ready to rock and roll on the Cell when systems 
become more generally available. 

So, Plan 9 is getting fixed for 64-bit mode, we have a K8 compiler now as 
a template. What do we name the compiler?

Cell C (or cc)? Just joking!

G5 C or gc? I'll stop now.

ron

Re: [9fans] Plan9 on the Cell...

[Ronald G. Minnich]

On Sat, 28 May 2005, andrey mirtchovski wrote:

> ibm to offer cells to appear in blades. plan9 port even more necessary:

And, geez, the blades only cost $32K. And, I just found out, only 1000 
will be available this year. If one were to buy, say, 8 blades, for only 
$256K, one would have own 10% of total production. (No, I'm not the One).

Let's wait a bit :-)

ron

Re: [9fans] Plan9 on the Cell...

[Eric Van Hensbergen]

On 5/28/05, andrey mirtchovski <mirtchovski@gmail.com> wrote:
> ibm to offer cells to appear in blades. plan9 port even more necessary:
> 
> http://www.theregister.co.uk/2005/05/27/ibm_demos_cell/
> 

Don't count on these being generally available.  Rumor is that the
primary target is game development customers, so they'll be as
expensive as the "Cell Workstations" (which are pricey - not anywhere
close to commodity).

         -eric

Re: [9fans] Plan9 on the Cell...

[andrey mirtchovski]

ibm to offer cells to appear in blades. plan9 port even more necessary:

http://www.theregister.co.uk/2005/05/27/ibm_demos_cell/

Re: [9fans] Plan9 on the Cell...

[Paul Lalonde]

http://lkml.org/lkml/2005/5/13/218 has some interesting information:  
IBM's linux access to the SPE's communications mechanisms is through  
a simple virtual file system.  It looks pretty easy to port to our  
well-loved environment :-)

Paul

Re: [9fans] Plan9 on the Cell...

[C H Forsyth]

for the compiler suite, the 32-bit powerpcs are compatible
(ignoring the original 601 and not including Altivec)

kernels see bigger differences, where there are three or
more possibilities for memory management, and two different
schemes for trap handling, and even different instructions
for processor control, but generally the differences are
fairly small.   the surrounding IO structure is usually completely
different.

64-bit operation is a fairly small extension to the
32-bit set, as for several other RISCS, compared (say) to x86

Re: [9fans] Plan9 on the Cell...

[Ronald G. Minnich]

On Wed, 25 May 2005, C H Forsyth wrote:

> yes.

what CPUs? is 440 supported?

anybody know of a 440 port of Plan9? 

ron

Re: [9fans] Plan9 on the Cell...

[C H Forsyth]

> Is the 32 bit power compiler heavilly used by anyone?

yes.

Re: [9fans] Plan9 on the Cell...

[Bruce Ellis]

Sound like fun.  A bit of PS2 and a bit of Froggie.  I'll put it
on the list.  Unfortunately the list is long.

brucee

Re: [9fans] Plan9 on the Cell...

[Paul Lalonde]

I can't comment too deeply without treading into NDA land, but from  
public sources it's safe to say that the difficulty in programming  
the SPEs isn't so much in code generation (although automatic  
parallelization will continue to be a grail) but in data movement.   
The SPEs have only explicit "cache" in the sense that the local  
memory is very fast, but doesn't share an address space with the PU.   
Instead, you DMA chunks back and forth as needed and rely on DMA  
bandwidth (and sufficiency of channels) to fill relatively large  
chunks of "cache" explicitly instead of relying on automated cache- 
line granularity.  That makes you think of the machine architecture  
pretty much any time you design an algorithm to run on the SPE -  
which I guess puts it in line with other parallelization methods :- 
(   Wherefrom, of course, comes the interesting systems work :-)

Paul

On 24-May-05, at 4:18 PM, Jack Johnson wrote:

> On 5/24/05, Paul Lalonde <plalonde@telus.net> wrote:
>
>> The SPEs, of course, are the interesting part from the systems point
>> of view.
>>
>
> I'm a layman, so speaking completely out my posterior here, but I read
> a paper somewhere that lead me to believe that some of the vectorizing
> techniques used on the Crays could be applied well to the cell
> processors.  True?
>
> Not that I recall seeing a Cray port on sources.... ;)
>
> -Jack
>

Re: [9fans] Plan9 on the Cell...

[Eric Van Hensbergen]

On 5/24/05, Jack Johnson <knapjack@gmail.com> wrote:
> On 5/24/05, Paul Lalonde <plalonde@telus.net> wrote:
> > The SPEs, of course, are the interesting part from the systems point
> > of view.
> 
> I'm a layman, so speaking completely out my posterior here, but I read
> a paper somewhere that lead me to believe that some of the vectorizing
> techniques used on the Crays could be applied well to the cell
> processors.  True?
> 
> Not that I recall seeing a Cray port on sources.... ;)
> 

The Cell architecture is quite different from a traditional vector
architecture.  "It's like nothing we've ever seen before"...sorta....

      -eric

Re: [9fans] Plan9 on the Cell...

[Eric Van Hensbergen]

On 5/24/05, Paul Lalonde <plalonde@telus.net> wrote:
> Yes, I've been reading the compiler code in anticipation.  It's been
> close to 15 years since I last did any real compiler (code
> generation) work; it will take me some effort to get back up to speed.
> I've been itching for an excuse to get a power mac, and this might be
> it.
>

I've offered access to my G5 to forsyth for some time now (of course,
I haven't gotten around to setting it up ;)  Now that I have 9grid.us
up, I'll turn to getting my G5 up-to-date and available for folks who
want to take a crack at a ppc64 compiler.
 
>
> Is the 32 bit power compiler heavilly used by anyone?
>

Not that I know of, but it was working fairly well for Motorolla
32-bit PowerPC chips, so it should be acceptable for IBM ppc32 - at
least as a starting point.
 
> The SPEs, of course, are the interesting part from the systems point
> of view.  It would be interesting to find a clean way of offering
> them up (along with the required PU code - that's the hard part) as a
> cpu-like computing resource.  The tricky part is that SPE code seems
> to like to set up pipelines using multiple SPEs, which makes
> allocation trickier.  Pre-emption looks expensive because of the
> local memories.
> 

Yes, there are many interesting issues.  It would be great if Brucee
could comment on the "Froggie" architecture (which was more of an
Inferno port, but regardless).  There are some rudimentary
similarities, at least from a high-level organizational view, and his
approach to handling the various "legs" under Inferno might be a
starting point for how to manage SPEs under Plan 9.

         -eric

Re: [9fans] Plan9 on the Cell...

[Jack Johnson]

On 5/24/05, Paul Lalonde <plalonde@telus.net> wrote:
> The SPEs, of course, are the interesting part from the systems point
> of view.

I'm a layman, so speaking completely out my posterior here, but I read
a paper somewhere that lead me to believe that some of the vectorizing
techniques used on the Crays could be applied well to the cell
processors.  True?

Not that I recall seeing a Cray port on sources.... ;)

-Jack

Re: [9fans] Plan9 on the Cell...

[Paul Lalonde]

Yes, I've been reading the compiler code in anticipation.  It's been  
close to 15 years since I last did any real compiler (code  
generation) work; it will take me some effort to get back up to speed.
I've been itching for an excuse to get a power mac, and this might be  
it.

Is the 32 bit power compiler heavilly used by anyone?

The SPEs, of course, are the interesting part from the systems point  
of view.  It would be interesting to find a clean way of offering  
them up (along with the required PU code - that's the hard part) as a  
cpu-like computing resource.  The tricky part is that SPE code seems  
to like to set up pipelines using multiple SPEs, which makes  
allocation trickier.  Pre-emption looks expensive because of the  
local memories.

Paul

On 24-May-05, at 12:26 PM, Eric Van Hensbergen wrote:

> On 5/24/05, Paul Lalonde <plalonde@telus.net> wrote:
>
>> IBM looks like they are going to open up the Cell architecture:
>> http://www.eetimes.com/news/latest/showArticle.jhtml? 
>> articleID=163106213
>> Which, if someone can start shipping dev boards, would lead me to
>> attempt a port :-)
>> I've been lucky enough to be working on one of these puppies for a
>> bit (demo at http://ps3.ign.com/articles/614/614705p1.html) and the
>> thing is so sweet to squeeze performance out of that I'd love to be
>> running a CPU server native on it...
>>
>>
>
> I have no direct knowledge of these things, but development boards are
> going to be scarce for a while.  I've heard rumors that IBM will be
> releasing the Cell full-systems simulator as part of them opening up
> the architecture -- that will likely be your best candidate for a port
> until development systems become more widely available.
>
> In the meantime, porting to the G5 (or any other readily available
> ppc64 platform) would probably be your best bet.  First step is the
> compiler (while you may be able to use the 32-bit power compiler,
> you'll really want 64-bit if you are looking to squeeze performance).
> Next is working with jmk (who is adding 64-bit support to the kernels
> for opteron) to do the same thing for PowerPC.
>
> That pretty much sums up the hard parts (at least for an initial
> support).  How to deal with the SPE's is a really big question, but
> that can be dealt with after the core port is complete.
>
> As you can see, I haven't put much thought into this myself ;)
>
>     -eric
>

Re: [9fans] Plan9 on the Cell...

[Eric Van Hensbergen]

On 5/24/05, Paul Lalonde <plalonde@telus.net> wrote:
> IBM looks like they are going to open up the Cell architecture:
> http://www.eetimes.com/news/latest/showArticle.jhtml?articleID=163106213
> Which, if someone can start shipping dev boards, would lead me to
> attempt a port :-)
> I've been lucky enough to be working on one of these puppies for a
> bit (demo at http://ps3.ign.com/articles/614/614705p1.html) and the
> thing is so sweet to squeeze performance out of that I'd love to be
> running a CPU server native on it...
> 

I have no direct knowledge of these things, but development boards are
going to be scarce for a while.  I've heard rumors that IBM will be
releasing the Cell full-systems simulator as part of them opening up
the architecture -- that will likely be your best candidate for a port
until development systems become more widely available.

In the meantime, porting to the G5 (or any other readily available
ppc64 platform) would probably be your best bet.  First step is the
compiler (while you may be able to use the 32-bit power compiler,
you'll really want 64-bit if you are looking to squeeze performance). 
Next is working with jmk (who is adding 64-bit support to the kernels
for opteron) to do the same thing for PowerPC.

That pretty much sums up the hard parts (at least for an initial
support).  How to deal with the SPE's is a really big question, but
that can be dealt with after the core port is complete.

As you can see, I haven't put much thought into this myself ;)

    -eric

[9fans] Plan9 on the Cell...

[Paul Lalonde]

IBM looks like they are going to open up the Cell architecture:  
http://www.eetimes.com/news/latest/showArticle.jhtml?articleID=163106213
Which, if someone can start shipping dev boards, would lead me to  
attempt a port :-)
I've been lucky enough to be working on one of these puppies for a  
bit (demo at http://ps3.ign.com/articles/614/614705p1.html) and the  
thing is so sweet to squeeze performance out of that I'd love to be  
running a CPU server native on it...
Paul