[lkml]   [2010]   [Jan]   [20]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectRe: [x86] Unify semaphore_32.S and rwlock_64.S

On Wed, 20 Jan 2010, H. Peter Anvin wrote:
> The motivation for rwsem_count_t seemed to be making it easier to switch
> over. I leave it up to Linus to motivate the typedef...

Indeed. I wanted to separate the issue of cleaning up the mask values from
cleaning up the code. That required that we'd have a separate type that
can change in _one_ place with the bias values.

It was also an unholy type mess before, so the rwsem_count_t also was
about making that unholy mess clearer. That's especially true since the
type will also have to percolate down to the slow cases before you can
actually fix the BIAS values.

As to keeping the 64-bit and 32-bit slow case wrappers separate: this is
very much one case where we _have_ to. Trying to pretend that 32-bit code
and 64-bit code is similar in this case is just lying. The calling
conventions are totally different, there is nothing similar there at all
(different argument registers, different clobbered registers, different
_everything_ that matters for the wrapper).


 \ /
  Last update: 2010-01-21 00:49    [W:0.082 / U:0.452 seconds]
©2003-2017 Jasper Spaans. hosted at Digital OceanAdvertise on this site