Re: CVS commit: src/sys/arch/arm/arm32

[Steve Woodford <scw%netbsd.org@localhost>]

On Thursday 16 October 2003 9:02 am, Richard Earnshaw wrote:

> Isn't ok just to check for it unconditionally?

The problem is that if the strh/strd check fails, it falls through to 
the original strsh/strsb check, which incorrectly matches ldrd. My 
arm-arm doesn't document strs[hb], so I can't determine if that check 
can be narrowed down to avoid matching ldrd. Any ideas?

> This whole section really needs rewriting for modern ARM processors
> anyway, it doesn't handle Thumb and it wouldn't stand a chance with
> Jazelle.

Heh, and I haven't yet got my head around all the *existing* ARM 
architectures. ;-)

Cheers, Steve