Pthreads, part 2

Peter Naulls peter at
Tue Oct 22 02:53:09 PDT 2002

In message <3DB3F6B2.3060302 at>
          Alex Waugh <Alex.Waugh at> wrote:

> Peter Naulls wrote:

> > Can we avoid this?  Having different sets of code that works on
> > 26bit/32bit sounds all very well, but the upshot means having two
> > libraries.  Instead, can we write 26bit/32bit agnostic code, with the
> > eventual aim to be fully APCS-32 supporting ARM3 through to XScale.
> I changed this because John requested it, and it made it consistant with 
> the current implementation of longjmp. We could use SFM in all cases, 
> and insist that the user has a recent enough version of the FPEmulator.

I did mean the comment as a general one, not just specific to your FP
code of course.  But to address this point, insisting on a recent FPE
might be an acceptable thing to do.  The alternative is checking for
32-bit mode at runtime, which is a 2 instruction sequence.

 Peter Naulls - peter at
 Homepage                     -
 RISC OS C Programming        -
 Unix Programs on RISC OS     -

More information about the gcc mailing list