IBM turning Power into Open Source?

  • Thread starter Thread starter Black Jack
  • Start date Start date
|> What exactly is this article saying? That IBM is now encouraging other
|> manufacturers to clone and extend its Power chips?
|>
|> http://story.news.yahoo.com/news?tmpl=story&cid=620&e=2&u=/nf/20040401/bs_nf/23584

No, but IBM have done that since the start of PowerPC, as did
SGI with MIPS and Sun with SPARC (the last even more so). It
is an encouragement to design and build supporting chips and
systems using the PowerPC.

As far as I know, there has been no major change in policy since
the very first PowerPC designs, and this is merely yet another
twitch of the publicity machine.

Note that nowhere above did I say "for free" :-)


Regards,
Nick Maclaren.
 

My take on it? Mostly a lot of hot air. I call this one 99%
marketing hype and maybe 1% substance.

IBM does plan to make it a little easier to license PowerPC cores, but
it's not really going to be "open" at all, you'll still have to pay
IBM to license the cores and IBM still controls the instruction set
and most of the IP.

Their plan seems to be exactly what they were already doing with
Nintendo and their Gamecube processor (custom PowerPC core, based off
of IBM's 400 series chips but with some customer Nintendo stuff).
They might be making things a little bit easier and maybe even a bit
cheaper, but it's basically just business as usual.
 
My take on it? Mostly a lot of hot air. I call this one 99%
marketing hype and maybe 1% substance.

IBM does plan to make it a little easier to license PowerPC cores, but
it's not really going to be "open" at all, you'll still have to pay
IBM to license the cores and IBM still controls the instruction set
and most of the IP.

Their plan seems to be exactly what they were already doing with
Nintendo and their Gamecube processor (custom PowerPC core, based off
of IBM's 400 series chips but with some customer Nintendo stuff).
They might be making things a little bit easier and maybe even a bit
cheaper, but it's basically just business as usual.
There is of cause the case of PPC cores in FPGA (Xilinx Virtex IIRC).
In such a case 'adding' something is a lot easier in a HDL language.
Then one could think of the possibility such a change was 'ported back'
to silicon.
Having not used that PPC core I am only speculating though.
JP
 
Jan Panteltje said:
There is of cause the case of PPC cores in FPGA (Xilinx Virtex IIRC).
In such a case 'adding' something is a lot easier in a HDL language.
Then one could think of the possibility such a change was 'ported back'
to silicon.
Having not used that PPC core I am only speculating though.
JP

You all can check out http://www-1.ibm.com/technology/power/ where there is
some kind of "design kit" that is downloadable to play with. I also looked
at the licensing page, and they have a synthesizable 440 core that it says
can be taken anywhere. Also that they are porting the hard core to other
foundries.

So yes indeed it would seem from the web site that one could take the soft
core, modify it, and synthesize to the foundry process of your choice using
the library vendor of your choice. Synopsis and Cadence are mentioned.

It would be my belief that there probably is some exchange of money involved
in this licencing but you should check for yourself.

Looks like all sorts of merriment could ensue. And if you need help, don't
forget Engineering and Technology Services :-)

Del Cecchi (personal opinions only)
 
|> What exactly is this article saying? That IBM is now encouraging other
|> manufacturers to clone and extend its Power chips?
|>
|> http://story.news.yahoo.com/news?tmpl=story&cid=620&e=2&u=/nf/20040401/bs_nf/23584

No, but IBM have done that since the start of PowerPC, as did
SGI with MIPS and Sun with SPARC (the last even more so). It
is an encouragement to design and build supporting chips and
systems using the PowerPC.

Not sure where "even more so" comes from? There are many more MIPS
licensees, products, and shipments than for SPARC, though SPARC has a
larger installed base of workstations & servers.
Jon
__@/
 
My take on it? Mostly a lot of hot air. I call this one 99%
marketing hype and maybe 1% substance.

Could be. It's been done before.
IBM does plan to make it a little easier to license PowerPC cores, but
it's not really going to be "open" at all, you'll still have to pay
IBM to license the cores and IBM still controls the instruction set
and most of the IP.

You think IBM is giving hardware away? Come on Tony! They're
looking for *business* opportunities. It's really that simple.
Their plan seems to be exactly what they were already doing with
Nintendo and their Gamecube processor (custom PowerPC core, based off
of IBM's 400 series chips but with some customer Nintendo stuff).

Wrong. The Nintendo Gekko is a 750(FX, IIRC - can't keep the
chips/marketing straight). It is most certainly not derived from
the PPC4xx series.
They might be making things a little bit easier and maybe even a bit
cheaper, but it's basically just business as usual.

Business is business. What I get out of this is the tools will
become more freely available. IBM released its compilers for
PPC970 (for Apple) some time back, I see this as more of the
same. Think of it this way; some one has figured out that if
you give the tools away, someone might just build a house with
your lumber.
 
There is of cause the case of PPC cores in FPGA (Xilinx Virtex IIRC).
In such a case 'adding' something is a lot easier in a HDL language.

Sure, as long as you haven't a care about performance. ...or
cost.
Then one could think of the possibility such a change was 'ported
back' to silicon.

Hard PPC4xx's are already available in the Xilinx Virtex-3 parts
(and several PPCs in the larger parts). Certainly if you want to
do an ASIC the PPC4xx core has been available for some years.
Having not used that PPC core I am only speculating though.

*that*???

There are many more than one.
 
Not sure where "even more so" comes from? There are many more MIPS
licensees, products, and shipments than for SPARC, though SPARC has a
larger installed base of workstations & servers.

Sun made the SPARC architecture explicitly free to all users; IBM
and SGI did not do quite that with PowerPC and MIPS, and it was and
is legally unclear whether you can build clones (at that level) with
no form of licence. I believe that the answer is "yes", but only in
most jurisdictions.

The Intel x86 scenario was and is very different, and vastly more
confused.


Regards,
Nick Maclaren.
 
Sure, as long as you haven't a care about performance. ...or
cost.


Hard PPC4xx's are already available in the Xilinx Virtex-3 parts
(and several PPCs in the larger parts). Certainly if you want to
do an ASIC the PPC4xx core has been available for some years.


*that*???

There are many more than one.
I went to that IBM site (other reply in thread), and found
PPC_QRG_2-22-04.pdf.
Now that is nice stuff, look at the low power it consumes.
I really did not know that.
 
So yes indeed it would seem from the web site that one could take the soft

What the toolkit seems to offer is a simulation model. No RTL.

modify it, and synthesize to the foundry process of your choice using
the library vendor of your choice. Synopsis and Cadence are mentioned.

It would be my belief that there probably is some exchange of money involved
in this licencing but you should check for yourself.

Of course there is. What makes this any more "Open" than ARM/ARC/Tensilica?

The very fact that you have to license it, means that it isn't "Open".

Cheers,
JonB
 
Jon said:
The very fact that you have to license it, means that it isn't "Open".

Ah, so all that 'Open Source' software under the GPL and similar
licenses isn't "open". I can see some merit in that argument, but
I suspect most Linux folk might disagree.

mfc
 
In comp.arch Mike Cowlishaw said:
Ah, so all that 'Open Source' software under the GPL and similar
licenses isn't "open". I can see some merit in that argument, but
I suspect most Linux folk might disagree.

I don't think so - with linux, you get the licence with the materials
and there is no need to talk to Linus top use it (same with any other
gpl software and its 'owner'). This is not so with Power and IBM.

The page doesn't even say 'free of cost and no patent crosslicencing
deals imposed' or similar. It does not really appear different from
SCSL that Sun released some of its past processors under some years
ago.
 
The very fact that you have to license it, means that it isn't "Open".

You pretty much have to license everything. Without a license
requirement, it's very difficult to disclaim liability. Even if you put in
liability disclaimers, you can't stop others from removing the disclaimers
and redistributing.

It's one thing to guarantee that what you developed worked as it is.
It's quite another thing to be sure it won't have bugs (or things others can
claim are bugs) that will show up when it's modified. You can then be sued.
The only way you can say you took reasonable steps to be sure anyone who
received the code/design/whatever was warned about the need to test its
suitability for a particular purpose is to put that in a license.

DS
 
Sander said:
I don't think so - with linux, you get the licence with the materials
and there is no need to talk to Linus top use it (same with any other
gpl software and its 'owner'). This is not so with Power and IBM.

Not sure I follow; there's no need for an open source project
to send out its license with the materials (that's a GPL quirk,
which mostly seems to keep lawyers employed :-)).
The page doesn't even say 'free of cost..'
Nor does the typical open source license (though GPL
armwaves about patents, allowing patents which are free
to all users, but not necessarily without cost).

mfc
 
In comp.sys.ibm.pc.hardware.chips Mike Cowlishaw said:
Not sure I follow; there's no need for an open source project
to send out its license with the materials (that's a GPL quirk,

Actually, there _is_ a need for some sort of licencing
statement otherwise it isn't Open Source nor Free Software.
Copyrighted materials cannot be used without some sort of
permission or licence. Usually, you will see something like
"Licence granted under GPL 2.0" near the copyright notice.
Including a copy of the GPL is recommended but not required.
which mostly seems to keep lawyers employed :-)).

Hardly. AFAIK, the GPL has not been litigated, and doesn't
occupy much lawyers time. Particularly this -- including by
reference is hardly controversial.

Sander's point remains valid -- the GPL is a unilaterial
licence in that unknown persons can use the software without
needing separate permission, notice or other from the author
so long as they abide by the GPL terms. They aren't required
to follow the GPL, but then nothing else gives them any right
to use the software.

I don't think this is true of Microsoft's latest licences which
require you submit certain information (perhaps automated)
before the product is fully licenced and functional.

-- Robert
 
In comp.arch Mike Cowlishaw said:
Not sure I follow; there's no need for an open source project
to send out its license with the materials (that's a GPL quirk,
which mostly seems to keep lawyers employed :-)).

If there is no licence that spells out the terms under which you can
distribute it, then you *cannot* distribute it, and thats about
as far from Open Source as you can get. The licence need not be
distributed alongside it but there is need for a statement saying what
teh licence is. Without such you simply cannot copy copyrighted material.

It does not get any simpler than that.

For a licence to be a open source licence it may in fcat not require any
payment for licencening or other limitations on distribution of the source
or resulting 'binaries'. That IBM may have patents that you may need in order
to legaly implement it anywhere where those patents are valid is completely
besides the point - not only are there likely to be aspects that can be
studied and reused without any reference to IBM patents, a licence that
does not give you a basic set of freedoms simply isn't 'Open'.

Again, go see the cpu-s Sun is licencning under SCSL (which is not a open
licence either).
 
Robert said:
Sander's point remains valid -- the GPL is a unilaterial
licence in that unknown persons can use the software without
needing separate permission, notice or other from the author
so long as they abide by the GPL terms. They aren't required
to follow the GPL, but then nothing else gives them any right
to use the software.

Close. Note that copyright is about *copying*, not *use*. The GPL
allows you to distribute (modified) copies of the Programming Work.

Just as you can read a copyright book legally obtained from a library,
you can legally use a copyright program - you just cannot legally (re-)
distribute it without an additional license.
 
In comp.sys.ibm.pc.hardware.chips Toon Moene said:
Close. Note that copyright is about *copying*, not *use*. The GPL
allows you to distribute (modified) copies of the Programming Work.

Just as you can read a copyright book legally obtained from a library,
you can legally use a copyright program - you just cannot legally (re-)
distribute it without an additional license.

Again close :) When you "use" a program, you actually make many
copies of it -- at least in RAM and usually on the Hard Disk.
Additional copies in caches, buffers and swap probably count
as transient.

The difference isn't entirely trivial, because many people
can use the single copy that's in RAM. Multiple windows.
Modern OSes will automatically share codepages.

I believe there's legislation in the EU to specifically permit
copying software as necessary for use, but don't believe there's
anything beyond caselaw in the US.

-- Robert
 
Back
Top