John Ballance writes:
> Phil suggested enabling block char devices.. this got over the problem
> with char/selection.c.. it is only used with char devices... but its
> still an incorrect piece of code.. FYI
Indeed. I'll look into it once I've finished characterising an apparant
as yet unknown SA110 rev. S hardware bug for Intel which has been causing
other people grief for the past month.
> I actually got as far as a zImage or a vmlinux yesterday eve.. BUT
> loader suggested it could not understand the zimage compression method,
> and vmlinux just froze in early boot. Does the loader have to be elf
> aware?..
No. The zImage file contains it's own decompression code, which should
be ok. I haven't checked .126 on the RiscPC, and I'm now on .129, but
I'm not in a position whereby I can compile it for the RiscPC (I'm
currently working on the Netwinder port, and to switch over will take
out several evenings).
Could you supply version numbers, and the exact error messages you see
please? I tend to be quite particular about both these points.
> now.. I still cannot get to a rpc hosted compile environment for 2.1.126..
> blow ups left/right/centre in trying to get phil's suggested things running
I can't help on that - I'm not even using Phil's tools yet on the RiscPC,
and my tools are hacked up GCC 2.7.2.2 and binutils 2.7 for ELF, both of
which I don't want to go to the hastle of releasing for many reasons, one
major one is that the files they produce are totally incompatable with the
ARM ELF standard. (they were purely to get the 2.1 kernels rolling).
> Is there any way to get a working system with current working compilers
> etc for 2.1.126 for riscpc on a disc?..e.g. a single tar-gz on a zip disc?
> this would really help. I would even consider the prospect of grabbing
> several tens of Mb via the net if this would simplify things...
I'm not sure...
> n.b. i note a considerable difference in speed between redhat 5.1 on i686 and
> what we have on the rpc.. wherein lies the main slowness.. is it an
> inappropriate cache on the SA, or something more mundane???
The RiscPC:
1. Does not support StrongARM's half-word operations, resulting in more
code required to do the same job.
2. The bus speed is slow.
3. The IDE interface is really slow (1.5MB/s) compared with a PC at 3-10MB/s.
4. The RiscPC display is not character based, so requires a lot of
processor power to manipulate it (especially when you have a VT100
scrolling region enabled).
Basically, IMHO the RiscPC is usable, but there are faster machines on the
market. It is actually interesting to note that the RiscPC is faster at
some things than PCI based machines at the moment, but I'm not going to go
into that here, except to quote a (formatted and spellchecked) mail I've just
read from the Corel mailing lists:
> One way to answer this question would be to give the equivilent in
> terms of a Pentium or PowerPC architecutre. Is the SA-110 275MHz
> processor equivilent to a penitum 275MHz in integer performance?
Pretty much. I some areas it's faster (when you can take advantage of
simultaneous shift & op for example) and in others it will be slower
(smaller simpler cache) but overall it always seemsa pretty good match.
Until you come to FP.... however, it looks as if the ARM10 architecture
finally solves that little problem albeit at some cost in power
consumption and die size. If Intel ever release it, the SA1500 also has
hardware support for FP by virtue of a on-chip coporocessor DSP unit.
It is claimed that a 300MHz version would offer a total of 1500mips!
Things like decoding two MPEG-2 streams in realtime are claimed.
> On Tue, 01 Dec, Russell King wrote:
> > It will *always* appear to those who have normal working hours
> > that any mail sent to me will take one day for a reply to reach
> > them.
>
> I wasn't worried about the *odd* day's delay.. it is just that its nice to
> actually get a response from someone wrt the arm port of linux.. so far I've
> wasted a great deal of time fighting the system to (so far unsuccessfully)
> get to the point where I can compile kernel sources, cross or otherwise, for
> current rpc kernels!
I'm sorry, but as I see it, you've only just come into the domain that I
know about, as of the weekend. Hence, as far as I'm concerned, I've only
delayed answering one of your mails.
_____
|_____| ------------------------------------------------- ---+---+-
| | Russell King [EMAIL PROTECTED] --- ---
| | | | http://www.arm.linux.org.uk/~rmk/armlinux.html / / |
| +-+-+ --- -+-
/ | THE developer of ARM Linux |+| /|\
/ | | | --- |
+-+-+ ------------------------------------------------- /\\\ |
unsubscribe: body of `unsubscribe linux-arm' to [EMAIL PROTECTED]
