Hello,
Some promising news,
After compiling Liquidsoap with Ocaml 4.01.0+dev10-2012-10-16 from SVN
trunk, as well as some optimizations (setting LIQ nice level to -20,
replace noise() with blank() for our backup stream, removing all amplify()
operators), Liquidsoap no longer stutters during playback. At idle it takes
15%, and when decoding a 128Kbps mp3, takes about 30%.
Funny thing is when checking the svn history of my download, I don't see
commit 13125 as referenced here
http://caml.inria.fr/mantis/view.php?id=5798
, so I will revisit that tomorrow and probably recompile everything again :/
Is there any difference between running mksafe on a source vs simply
setting the output sinks as fallible=true? That's what Im thinking of
trying next, but maybe it's silly?
Right now I can still induce output stuttering by loading a ajax heavy page
(starts nginx, php, postgresql in the background), so there's still some
work to do, but it appears that a no FP floating point branch is not
necessary (thanks for offering samuel :)
M
On Wed, Feb 13, 2013 at 4:36 PM, Romain Beauxis <[email protected]>wrote:
> Hi,
>
> 2013/2/13 Martin Konečný <[email protected]>:
> > Samuel,
> >
> > I'll report back with OCAML 4.00.1 results. However if that does not
> work,
> > it would be amazing to have this nofpu Liquidsoap.
> >
> > Just out of curiosity I know floating point is used to represent a wide
> > range of values with a relative error constant, whereas fixed point is
> used
> > to represent a much smaller range of values but with higher precision.
> How
> > would this affect Liquidsoap - is there a reason Liquidsoap was
> originally
> > written in floating point, when these instructions are so much more
> > expensive? Is it easier to work with Floating P in the audio domain, or
> is
> > it some other reason?
>
> Good questions.. My 2 cents:
>
> Fixed-point arithmetic is a pain in the butt and, from the programmer
> point of view, it's always easier to let the compiler/CPU deal with
> those details. Beside, there is no standard representation for
> fixed-point (or fractional) numbers in C and OCaml, which means we'd
> have to re-implement our own arithmetic operations in both sides,
> which would again be a pain.
>
> Romain
>
>
> ------------------------------------------------------------------------------
> Free Next-Gen Firewall Hardware Offer
> Buy your Sophos next-gen firewall before the end March 2013
> and get the hardware for free! Learn more.
> http://p.sf.net/sfu/sophos-d2d-feb
> _______________________________________________
> Savonet-users mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/savonet-users
>
--
Open source radio in the cloud. Get yours now! ---> http://airtime.pro
Martin Konecny
Software Developer, Sourcefabric
[email protected]
720 Bathurst St. Suite 203
M5S 2R4, Toronto, ON, Canada
+1 (416) 892-8420 (Cell)
Skype: martin.konecny15
http://www.sourcefabric.org
http://www.twitter.com/Sourcefabric
------------------------------------------------------------------------------
Free Next-Gen Firewall Hardware Offer
Buy your Sophos next-gen firewall before the end March 2013
and get the hardware for free! Learn more.
http://p.sf.net/sfu/sophos-d2d-feb
_______________________________________________
Savonet-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/savonet-users
