> I've finally had a chance to start looking at 1.3.62... and I've found
> that the assembler complains when I compile r3000.S in arch/mips/kernel:
> r3000.S:103: Error: lui expression not in range 0..65535
> This is due to the line:
> lui k1,%HI(TLBMAP)
> ...now this is exactly the same code as for the R4000... so I'm assuming
> it's too big an offset for an r3k. It's late at night, and I'm not up
> to rummmaging for the solution to this. Ralf? (I had cheated previously
> by doing a hack, now I want to fix it permanently.)
The lui instruction works exactly the same on all MIPS CPUs. Probably a
problem with your assembler? I've fixed this bug for my versions of
the GNU binutils 2.5.2 and 2.6; in vanilla Cygnus/FSF binutils %HI, %hi
and %lo are completly broken.
If you're using one of my assembler version this seems to be another bug.
In that case it would be great if you could reduce your code into a
minimum example that demostrates the bug.
> P.S. Anyone live near enough to Brussels, Geneve, or Neutraubling? Or,
> is anyone else attending the European High Performance Computing
> and Networking conference? I'll be in Brussels in mid-April for the
> conference, and then on to the other cities (maybe more), and
> wouldn't mind meeting up with some of our European contingent...
Paul, these locations are 200km to 700km or so aways from me and I think
for a meeting at Brussels would be ok. (Or will you come to the Linux
congress in Berlin ?-)