Re: powerpc gcc strange assembler for bitfields

[Nathan Sidwell <nathan at codesourcery dot com>]

Florian Weimer wrote:
> * Nathan Sidwell:

> Ahem, you (or Pete) should use explicit assembler instructions if you
> want to make sure that you end up with certain instructions.
>
> Future compiler versions might optimize the explicit masking as well.
> (Even the volatile keyword would not prevent this.)

I disagree.  A volatile memory access must be visible outside the
abstract machine.  Where is the abstract machine?  Well, that's a little
vague and implementation defined.  But I would consider that any compiler
which did not perform word accesses on the following code to be broken.

void SetBit (volatile int *ptr, unsigned char mask)
{
   int tmp = *ptr;
   tmp |= mask;  // we know only the bottom 8 bits can be affected
   *ptr = tmp;
}

The conclusion of your train of thought would be that the compiler is
free to remove the following dereference, regardless of volatility
  (void)*ptr;

nathan
--
Nathan Sidwell    ::   http://www.codesourcery.com   ::     CodeSourcery LLC
nathan@codesourcery.com    ::     http://www.planetfall.pwp.blueyonder.co.uk