[gccsdk] Dynamic libraries vs pthreads

Ben Avison bavison at riscosopen.org
Mon Aug 25 15:58:32 PDT 2008

On Mon, 25 Aug 2008 20:53:14 +0100, Lee Noar <leenoar at sky.com> wrote:
> Eventually, I traced this to the STMFD instruction in pthread_yield:
>    STMDB	a1, {r4-r14}^

There is a bug with STM^ on the early SA110s - you must have a rev K one
(it's fixed in rev S and T, the other two common ones in Risc PC processor
cards). Rev K SA110s are ancient history as far as ARM is concerned, I doubt
they'll have had many customers other than Acorn.

If you look at the kernel sources, you'll see exactly the same workaround in
a few places.


More information about the gcc mailing list