[9fans] ports ...

[9fans] ports ...

[ron minnich]

I noticed somebody asking about the MIPS and SPARC ports.

Hey, you want to do a port, do something interesting like the K8 ... AMD
is flooding the world with docs on all the internals, and they're being
friendly. It's really amazing how open they are being.

ron

Re: [9fans] ports ...

[Christopher Nielsen]

Peter Wemm of Yahoo! has been working on the FreeBSD
port of amd64. It'd be a fun project, if I could get
my hands on the hardware. :)

When, recently, was someone talking about MIPS and
SPARC ports?

Porting to a 64-bit architecture, while not pointless,
seems not as exciting, since Plan 9 cannot currently
use the the more interesting capabilities a 64-bit
processor provides. And from the sounds of it, it is
non-trivial to make Plan 9 64-bit. I remember something
being said about the reliance on pointers being 32-bits.
Please correct me if I am mistaken; this is completely
from memory.

On Wed, Jun 04, 2003 at 10:31:55PM -0600, ron minnich wrote:
>
> I noticed somebody asking about the MIPS and SPARC ports.
>
> Hey, you want to do a port, do something interesting like the K8 ... AMD
> is flooding the world with docs on all the internals, and they're being
> friendly. It's really amazing how open they are being.

--
Christopher Nielsen
"They who can give up essential liberty for temporary
safety, deserve neither liberty nor safety." --Benjamin Franklin

Re: [9fans] ports ...

[David Presotto]

It would be work to move to 64 bit pointers.  Not infinite
though.  The reliance is more on ulong and pointer being
the same size.  If they're both 64 bits, it'll be pretty
easy.  The truly hard part is changing the compiler and
debugging it afterwards.  If you have to simultaneously
rewrite a compiler and do a port, its much easier to have
multiple people.

Re: [9fans] ports ...

[jmk]

if we had a need and the hardware we'd port to the k8.

Re: [9fans] ports ...

[Jack Johnson]

On Wed, 4 Jun 2003, ron minnich wrote:
> I noticed somebody asking about the MIPS and SPARC ports.
>
> Hey, you want to do a port, do something interesting like the K8 ...

It's funny the dichotomy about the porting issue.  Some people
like/want/need to use new hardware and others like/want/need to
resuscitate older hardware.  Linux has this legacy hardware support for
which (I think) they do an increasingly poorer job of supporting (NetBSD
probably does a better job of continuing support).  So, for everybody,
recent hardware tends to get decent support for a while, then a kind of
"recovery support" where the popular items get fixed when they break under
version x, then you get the trailing stuff where popular hardware from
five generations ago is free or nearly free and there's a request or push
for new support for legacy hardware.

Plan 9 always intrigued me with its thin workstation concept, where the
workstation can be so-so to barely bearable but always supported by decent
(or exceptional) file/cpu servers.

So, I suppose I'm just rambling, but my vote is equally split between
SPARC and K8.

-Jack

Re: [9fans] ports ...

[Christopher Nielsen]

On Thu, Jun 05, 2003 at 09:18:21AM -0400, David Presotto wrote:
> It would be work to move to 64 bit pointers.  Not infinite
> though.  The reliance is more on ulong and pointer being
> the same size.  If they're both 64 bits, it'll be pretty
> easy.  The truly hard part is changing the compiler and
> debugging it afterwards.  If you have to simultaneously
> rewrite a compiler and do a port, its much easier to have
> multiple people.

Taking a break from coding something else, I did a tiny
bit of research.

All of amd64, alpha, and sparc64 have ulong and pointer
being 64-bit on all the platforms I surveyed. That would
make the part about reliance on ulong and pointer being
64-bit a non-issue.

Of course, there are the other hurdles... :)

--
Christopher Nielsen
"They who can give up essential liberty for temporary
safety, deserve neither liberty nor safety." --Benjamin Franklin

Re: [9fans] ports ...

[boyd, rounin]

> All of amd64, alpha, and sparc64 have ulong and pointer
> being 64-bit on all the platforms I surveyed.

yes, i remember yelling at CRL about making long being
64 bits on the alpha and int being 32 bits.  AS-irc int was
intended to map straight onto the machine's word size.

obviously, i'm talking about the compiler.

Re: [9fans] ports ...

[Douglas A. Gwyn]

"boyd, rounin" wrote:
> yes, i remember yelling at CRL about making long being
> 64 bits on the alpha and int being 32 bits.  AS-irc int was
> intended to map straight onto the machine's word size.

Unfortunately there are several, somewhat conflicting,
demands being made on the various integer types.  If you
have a specific width requirement you should use <stdint.h>
rather than assuming things about the standard types.

Re: [9fans] ports ...

[boyd, rounin]

> Unfortunately there are several, somewhat conflicting,
> demands being made on the various integer types.  If you
> have a specific width requirement you should use <stdint.h>
> rather than assuming things about the standard types.

yes, very true.  no way am i going to fight a broken compiler.

how did it used to go?  "everything's a vax" ...

Re: [9fans] ports ...

[Joel Salomon]

I believe it was "All the world's a VAX", but of course now all the
world's a PC ;-)

--Joel