[PATCH] Atomic swap and test-and-set

[Jung-uk Kim]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 2013-07-31 16:01:48 -0400, Jung-uk Kim wrote:
> On 2013-07-31 13:32:47 -0400, Konstantin Belousov wrote:
>> On Tue, Jul 30, 2013 at 03:56:36PM -0400, Jung-uk Kim wrote:
>>> -----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
>>> 
>>> The attached patch implements atomic_swap_<type>() and 
>>> atomic_testandset_<type>() for x86.  <type> is int or long
>>> ATM.
>>> 
>>> The semantics of each operation (in atomic(9) style):
>>> 
>>> atomic_swap(p, v) r = *p; *p = v; return (r);
>>> 
>>> atomic_testandset(p, s) v = (<type>)1 << s % (sizeof(<type>) * 
>>> NBBY); r = (*p & v) != 0; *p |= v; return (r);
>>> 
>>> Note atomic_readandclear_<type>() is now obsolete and
>>> implemented as a macro around atomic_swap_<type>() where the
>>> new value v is 0.
>>> 
>>> Please let me know if you have any objection.
>> I think that the atomic.diff is fine.
> 
>> You did not documented swap and testandset, which would be good
>> to have in the manpage.
> 
> Sure.
> 
>> I think that constraints for all added inlines contradict the 
>> explicit requirement in the gcc documentation, which states that 
>> rmw operands should be specified using the 'same location'
>> output operand for the input operand (hopefully this can be
>> parsed).  It just happens so that for "m" compiler cannot do
>> anything else then passing the address.
> 
>> The problem is not specific to the new functions, it seems to be 
>> present in other functions as well, e.g. cmpset.  This probably
>> can be postponed.
> 
> What a coincidence!  I was cleaning up the constraints before I
> read your e-mail.  Also, I am implementing 64-bit versions for i386
> and moving atomic_cmpset_64 from pmap.h to atomic.h.
> 
> I'll post new patches when I am done.

I just created a branch and committed the patches because the final
version grew too big.

http://svnweb.freebsd.org/base/projects/atomic64

Please let me know what you think.  Especially, I need more eyes to
review the last two changes:

http://svnweb.freebsd.org/changeset/base/253876
http://svnweb.freebsd.org/changeset/base/253877

Feel free to implement the new atomic ops for other architectures,
BTW. :-)

Thanks!

Jung-uk Kim

* Note: drm2 patches for AMD GPUs were updated as well:

http://people.freebsd.org/~jkim/drm2_atomic.diff
http://people.freebsd.org/~jkim/drm2_radeon.diff

The second patch is a new addition, which makes it build on i386.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.20 (FreeBSD)

iQEcBAEBAgAGBQJR+v/fAAoJECXpabHZMqHOL3kH/1Ak7vGgolriV3XP3vontBjf
6Wj0MkuzA3ytF6uhdoV5XQaDT6De2hTOOJfSDzLvaUiRI7gwlGW+WlGgE03Raly6
bsqcEUID9JCSUDpLc+MaxVBnbSZM6bkenTaSgDzMVCqtHt28BLsFrEHP+yModqaY
f/FrVR8RRJc7kjpKu0kpbSwTAUBVR73oJeWP4z58dQvN4EjRd3VbrZDrP42ppzjq
CEKjaub4pd7AEOPH18aB4j/WNp1uY04ybsbFubsOdfjOJbSz7oDRhC+rM8rKEuYL
XBkNFE5xRL14CxPAluEkgrSNd4RQYTlPPQbup998tgN5gOu9M2wVYPtYlUXQCYk=
=CzrX
-----END PGP SIGNATURE-----