[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Byte-misaligned memcpy problems



Joel Sherrill wrote:
> 
> > What do you think about moving the source, Joel?  If we just put memcpy
> > in the RTEMS CPU libraries can we be sure we get it and not the one from
> > the newlib library?  Enquiring minds want to know.....
> 
> When I put on my "right thing to do" hat, it feels wrong.  I don't like
> the idea of having C library routines in the exec/score/cpu.  For one, it
> would make it a part of the "minimum executive set".
> 
> I could possibly be talked into putting it in c/src/lib/libc.
> 

I would be happy to see this happen. I have a block copy routine for
the 68302 which uses the IDMA controller. Having this sort of code
supported in newlib is not on but having RTEMS support it is. Moving
memory via software on the 68000 is slow unlike the 68360 which
supports loop mode, or 030 and higher which have cache.

This then raises the issue of configuration. Should we allow
configuration ?

The IDMA might be used in the design for the a special purpose so the
newlib routine would need to be used.

-- 
 Chris Johns  Networks, Plessey Asia Pacfic Pty. Ltd.
  mailto:ccj at acm.org    mailto:cjohns at plessey.com.au