[mythtv] [mythtv-commits] Ticket #6330: [RFC] New ThreadedFileWriter (lockless)
Mark Buechler
mark.buechler at gmail.com
Mon May 11 23:09:11 UTC 2009
Hi
On Mon, May 11, 2009 at 8:32 AM, MythTV <mythtv at cvs.mythtv.org> wrote:
> #6330: [RFC] New ThreadedFileWriter (lockless)
>
> --------------------------------------------------------------+-------------
> Reporter: Matthias "mortalmatt" Dahl <devel at mortal-soul.de |
> Owner: ijr
> Type: enhancement |
> Status: new
> Priority: minor |
> Milestone: 0.22
> Component: MythTV - General |
> Version: head
> Severity: medium |
> Resolution:
> Mlocked: 0 |
>
> --------------------------------------------------------------+-------------
>
> Comment(by Matthias "mortalmatt" Dahl <devel at mortal-soul.de>):
>
> Paul, the old code base is better tested on a wider range of platforms. I
> have only been able to test this very thoroughly on a x86_64 based machine
> which ironically uncovered a corruption bug with the old TFW.
>
> The lockless TFW relies on the atomic operations support of Qt which is
> not supported natively on all platforms, meaning that they are emulated
> with e.g. mutexes. Even though I doubt PA-RISC, SH or Blackfin matter for
> MythTV, so that shouldn't be a problem.
>
> On PowerPC, S390, MIPS and SH-4a, the atomic operations are supported
> natively but are not guaranteed to be wait free. I haven't been able to
> test on those systems, so I can't say how much that affects performance.
> Based on some research I did and some talking on the Qt irc channel, that
> won't be a problem too performance wise but I still would like to see some
> profiling data (old TFW vs. lockless TFW) for those systems, if they
> matter for MythTV... I guess PowerPC still does?
>
> That's all that jumps to my mind actually. Like I initially said, I trust
> it enough to use it myself. :-)
>
> Matthias, can you explain the x86_64 corruption you saw?
Thanks, Mark.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mythtv.org/pipermail/mythtv-dev/attachments/20090511/b037a799/attachment.htm>
More information about the mythtv-dev
mailing list