Re: ntdll: Use lockfree implementation for get_cached_fd. (try 2)

[Ken Thomases]

On May 31, 2014, at 12:05 PM, Sebastian Lackner wrote:
> Changes in try 2:
>
> * Always try to access the fd cache two times - this ensures that we
> don't release a file descriptor when its still in use by a second thread.

Heh.  Well, you can ignore that part of my reply to your previous try.  :)
> * Replace interlocked_xchg() with an assert(). The original code looks
> like its valid that the fd is nonzero at this point, but in fact this
> would cause releasing a file descriptor which is still in use.

I'll leave it to others to comment on that.
> Replacing with an assert() also saves a couple of CPU cycles.

I understand that this is an attempt to improve performance, but I don't think it's necessary to squeeze every last CPU cycle out of this code.

-Ken

ntdll: Use lockfree implementation for get_cached_fd. (try 2)

[Bruno Jesus]

Hi, all. I would just like to share my results:
http://influenza.blog.br/wine/testpatch.pdf

I tested in wine-git with and without the patch, plus a XP VM and
Windows 7 (no VM). At least in my computer it looks like wine will
beat Windows 7 with the patch while the current implementation is as
slow as XP running in the VM.

The test files:
http://influenza.blog.br/wine/clz.c
http://influenza.blog.br/wine/clz.exe

Best regards,
Bruno

On Sat, May 31, 2014 at 2:35 PM, Ken Thomases <ken(a)codeweavers.com> wrote:
> On May 31, 2014, at 12:05 PM, Sebastian Lackner wrote:
> > Changes in try 2:
> >
> > * Always try to access the fd cache two times - this ensures that we
> > don't release a file descriptor when its still in use by a second thread.
>
> Heh.  Well, you can ignore that part of my reply to your previous try.  :)
> > * Replace interlocked_xchg() with an assert(). The original code looks
> > like its valid that the fd is nonzero at this point, but in fact this
> > would cause releasing a file descriptor which is still in use.
>
> I'll leave it to others to comment on that.
> > Replacing with an assert() also saves a couple of CPU cycles.
>
> I understand that this is an attempt to improve performance, but I don't think it's necessary to squeeze every last CPU cycle out of this code.
>
> -Ken