On Sat, Jan 06, 2007 at 04:06:57PM +0100, Frank Wille wrote:
> Ignatios Souvatzis wrote:
> 
> On 06.01.07 10:54:52 you wrote:
> 
> >> Maybe implement a stub near the exception address space, which does
> >> something like:
> >> 
> >> s_trap:
> >>    bl  s_trap2
> >>    .long   s_trap_kernel   # R_PPC_ADDR32 reloc
> >> s_trap2:
> >>    mflr    r11
> >>    lwz r11,0(r11)
> >>    mtlr    r11
> >>    blr
> > 
> > Uh, no. This creates stalls waiting for the memory  read.
> 
> As the "bla" restrictions force the stub-code being in slow Chip-RAM there
> will be many stalls anyway. ;)
> 
> Mapping a piece of of Fast-RAM to 0xffff0000 and using a high-vector
> configuration (as under AmigaOS+PowerUp/WarpOS) would be much nicer, but I'm
> still not sure how to do that...

It is mapped - in hardware! Just don't change anything about that!

	-is