[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Debian testing/unstable on RiscPC ?

Am Mittwoch, 28. Januar 2004 16:18 schrieb Phil Blundell:
> It turns out that all versions of gcc since 3.2 have been generating
> ARMv4 code as default, which won't run on RiscPC or ARM7 machines.  This
> was due to a bug introduced upstream in late 2001.  As a result, any
> binary that has been generated since gcc-3.2 became the default compiler
> may potentially be tainted with v4 instructions.

So a lot of thing have to be recompiled...

> Since this problem has gone unnoticed for over a year, and RiscPC
> support has always been kind of sketchy in any case, I'm kind of
> inclined to just declare that Debian will no longer support ARMv3
> processors from Sarge onwards.  There is some precedent for this in
> other architectures: the SPARC port has dropped sun4c in Sarge, and the
> x86 port has likewise dropped i386.  Nobody has shown a great deal of
> interest in implementing debian-installer for RiscPC, which also
> suggests that there are not very many users on this platform any more.
> (It would still be perfectly possible for someone to maintain an
> unofficial port for v3 machines with its own binary archive, if there is
> a desire for such a thing.)

I personally would not like such a declaration. First, of course, because I 
only have this RiscPC to run Debian/ARM on. My Netwinder is needed as 
Internet gateway and so not to be used for experiments.

Furthermore, there is no real reason. The only difference between v3 and v4 is 
the presence of the long multiply and ldrh/strh instruction. I admit that 
there are applications that benefit greatly from these instructions, but 
overall the performance benefit will not be noticable. Especially, if one 
takes into account that, for typical Unix applications, the speed of ARM 
processors is limited by main memory speed, because there is no L2 cache. And 
I believe it should not be too difficult to fix that bug in gcc for a person 
who is involved in this! And it is possible, with proper configuration, to 
have several versions of the same shared library existing, optimized for 
different architectures, if desired (libmad would be a good example, where 
this is useful).

Do not forget: many people run Debian on rather outdated hardware. And the 
RiscPC is nearly 10 years younger that 386 and sun4c!

The fact that no one complained about this before (I could have done earlier, 
but had more important things to do in my limited time) does not necessarily 
give a hint about the amount of people using Debian on RiscPCs. They could 
just run stable, as probably most people do. And the thing with the Debian 
installer for RiscPCs: Yes, I also had the some problems with it, and fixed 
it by combining several versions.

Personally, I would not like to maintain an unofficial port for v3. So I will 
make an offer: If Debian will keep support for v3, I will care about 
debian-installer for the RiscPC. If someone can give me assistance from time 
to time, I think that is more economic for me.

Peter Teichmann

Reply to: