[Asterisk-Dev] Sparc Ultra Support

[dking at pimpsoft.com]

I'm not sure if this announcement is needed here or not but I know a 
large amount of people emailed me about it when I originally brought 
this up, and I also know this list is the general thing that pops up 
when people search for information on the topic. And with the release 
of 1.0 more people are getting interested  in asterisk and the system 
on non PC arch systems, so here I go.

The current version of Asterisk DOES support compilation on 
ultrasparc cpu based systems thanks to work done on the asterisk code 
branch pre 1.0 by Myself. It does not however compile to 64bit by 
default due to limitations of the compiler used in many sun4u 
specific linux distribution ports.  Currently I have it set to 
compile for version 8 of the instruction set. Since version 9 is the 
64 bit version sun4u cpus are able to work with over time this will 
be changed but for now this is a good work around of the problem 
considering several things including the core asterisk application 
refused to compile on these systems before I did the work on it.

Also due to this zaptel does not currently compile for the correct 
kernel arch. The common linux kernel binaries available are built to 
make use of the 64 bit abilities of the version 9 cpu so modules 
built for a different arch (like the default of sparclite on most gcc 
installs in sparc distributions like debian-sparc) are trouble and 
refuse to load. This is one of the reasons many people who use sparc 
get the ‘invalid arch’ errors when they install the zaptel modules.

This could be fixed if the person was willing to make a custom kernel 
with little to no optimization and no benefits taken from the 
advanced hardware, But in my mind to do this would be to break a 
perfectly working system. You shouldn’t have to slow down your system 
or compile the kernel for a less functional hardware architecture 
subset to get asterisk and zaptel to work in my humble opinion. We 
also need a sunsparc 64 bit assembly base not unlike the k6 base that 
was causing me trouble in my work to port asterisk to sparc.

The reason better sparc support is needed is twofold.

One: Sparc based sun hardware has been used in the telecommunications 
industry for years now.  It’s always been a credit to sun that sparc 
based hardware has always been more reliable, functional, stable, and 
been able to provide more raw power under the hood then a similar 
speeded PC based cpu.
Two: Due to the reliability issue Many telecomm’s continue to use any 
sparc based computer systems they can get there hands on. I have seen 
perfectly working systems as old as ultra 5’s in the field.  By 
providing support for more sun based hardware not only would 
Asterisks usage base expand more, not only with hobbyists but telcos 
as well.

So I am making a request of not only the asterisk developer 
community, but the open source community, To do the following:

Mirror asterisk releases and cvs. We never know when a gpl backup 
will be needed.
Help with support of your favorite non i386 compatible hardware. 
Something as simple as adding the needed makefile info will help 
many.
Help keep the sparc based linux distributions out there up to date in 
terms of compilers and development tools.
Let me know if I’m wrong about something here so I can learn 
something and pass it on.

IT would be very nice to hear from others using sparc hardware. If we 
got together as a group I am sure more can be done. 

I’m also currently working on a project of my own that I hope will 
end up to be a viable tool to replace many of the repetitive commands 
used in open source software development build environments like 
asterisks.  I request that if your working on non i386 compatible 
hardware you please let me know via email offlist what cpu your 
using, and the version of gcc you have installed. Thanks.

 - D