Board index » kylix » Linux ported to the Cell Processor

Linux ported to the Cell Processor


2005-06-29 08:50:53 PM
kylix2
linuxdevices.com/news/NS6219524044.html
Cell CPU (PowerPC) based systems could become popular, so I hope the
Kylix compiler developlers are aware of it.
siegs
 
 

Re:Linux ported to the Cell Processor

Quote
Cell CPU (PowerPC) based systems could become popular, so I hope the
Kylix compiler developlers are aware of it.
There is already pascal compiler for PPC:
www.freepascal.org/fpc.html
Why bothering on migrating Kylix x86 compiler to new platform if hasn't
been updated since 2002 on the same architecture?
--- posted by geoForum on delphi.newswhat.com
 

Re:Linux ported to the Cell Processor

SiegfriedN wrote:
Quote
linuxdevices.com/news/NS6219524044.html

Cell CPU (PowerPC) based systems could become popular, so I hope the
Kylix compiler developlers are aware of it.

siegs
AFAIK, Linux will only directly make use of one of the multiple PPCs on
the chip. The others might be used by drivers (e.g. screen). Thus the
issue boils down to "Delphi language on PPC-Linux". Here we have two
promising candidates: Free-Pascal (with Lararus IDE if desired) and
Chrome on Mono.
-Michael
 

{smallsort}

Re:Linux ported to the Cell Processor

Michael Schnell wrote:
Quote
Here we have two promising candidates: Free-Pascal (with Lararus IDE
if desired) and Chrome on Mono.
I count at least three promising candidates, because Delphi for .NET
can produce assemblies that work on Mono.
--
Dave Nottage [TeamB]
 

Re:Linux ported to the Cell Processor

On 2005-07-10, Michael Schnell < XXXX@XXXXX.COM >wrote:
Quote
SiegfriedN wrote:
>linuxdevices.com/news/NS6219524044.html
>
>Cell CPU (PowerPC) based systems could become popular, so I hope the
>Kylix compiler developlers are aware of it.
>
>siegs

AFAIK, Linux will only directly make use of one of the multiple PPCs on
the chip.
Why that? Linux is NUMA SMP and should be able to use multiple cores. Or do you
mean the vectorising units?
Quote
The others might be used by drivers (e.g. screen).
It seems you mean vector units. True, they need specialist support. Not unlike e.g.
MMX or SSE, only a magnitude more complicated.
Quote
Thus the issue boils down to "Delphi language on PPC-Linux". Here we have
two promising candidates: Free-Pascal (with Lararus IDE if desired)
First, probably the beast needs to be hacked/modded so that it will run PPC at all. This
is not always trivial, and increasingly difficult. (because it also opens the way to pirated
games)
Quote
and Chrome on Mono.
Is more a candidate for the "Delphi.NET language on PPC-linux" IMHO ;-)
 

Re:Linux ported to the Cell Processor

Quote
>AFAIK, Linux will only directly make use of one of the multiple PPCs on
>the chip.


Why that? Linux is NUMA SMP and should be able to use multiple cores. Or do you
mean the vectorising units?


There is only one "normal" PPC on a cell chip. The others are
"symbiotic" processors and _not_ able to do SMP. The testing target at
IBM is a box with two Cells. Here you do have SMP between the two main PPCs.
Quote
>The others might be used by drivers (e.g. screen).


It seems you mean vector units. True, they need specialist support. Not unlike e.g.
MMX or SSE, only a magnitude more complicated.
Linux support for the "symbiotic" processors is in the works with IBM.
The "symbiotic" processors are accessed via a special file system. A
Thread in the main PPC needs to start a program in a "symbiotic"
processor and the thread is halted until the external program is done.
Thus the processes appear in the normal Linux thread list by their
associated threads..
Quote


>Thus the issue boils down to "Delphi language on PPC-Linux". Here we have
>two promising candidates: Free-Pascal (with Lararus IDE if desired)


First, probably the beast needs to be hacked/modded so that it will run PPC at all.
The MMU ant the interrupt controllers are different from the standard
desktop PPCs. That work is already done and is said to have been easy.
Quote
This
is not always trivial, and increasingly difficult. (because it also opens the way to pirated
games)


The cell is not only for games (e.g. Sony PS 3), but IBM will release
workstations with that processor. AFAIK, the PS3 will have a slot for a
hard disk and there will be a hard disk with Linux installed available
officially by Sony.
-Michael
 

Re:Linux ported to the Cell Processor

Dave Nottage [TeamB] wrote:
Quote
Michael Schnell wrote:


>Here we have two promising candidates: Free-Pascal (with Lararus IDE
>if desired) and Chrome on Mono.


I count at least three promising candidates, because Delphi for .NET
can produce assemblies that work on Mono.

Until Borland officially supports the Mono platform/framwork it would be
a mistake to rely on the fact that some Delphi for .Net assemblies work
with Mono as an option to develope business tools. Perhaps in the future
when Mono is more mature and widespread in the business world - I am
sure Borland will provide support for Mono in that case.
OT
I consider Mono the 'true' '.Net' implementation i.e. OSS and supports
more OSs.
siegs
 

Re:Linux ported to the Cell Processor

siegfriedn wrote:
Quote
Until Borland officially supports the Mono platform/framwork it would
be a mistake to rely on the fact that some Delphi for .Net assemblies
work with Mono as an option to develope business tools.
Yes, that's a very valid point.
Quote
Perhaps in the future when Mono is more mature and widespread
in the business world - I am sure Borland will provide support
for Mono in that case.
I'm hoping so.
--
Dave Nottage [TeamB]