> On Jan. 6, 2011, 5:37 p.m., Aleric Inglewood wrote:
> > This is really not how you want to deal with this bug :/. It's a known
> > fact that audio mixers are very bad with low volumes. Setting a volume to 0
> > (or something really small) can put a very high load on the CPU for the
> > audio mixer, which causes severe problems. See
> > https://jira.secondlife.com/browse/VWR-14914
> >
> > So, on the contrary (as VWR-14914 fixed a horrible bug that made FPS drop
> > drastically): when the volume is set to 0 (or even close to zero) the audio
> > channel has to be muted and not mixed, ever. Assuming that VWR-14914 is in
> > Viewer 2, and wasn't broken in the meantime, a volume of 0 would cause the
> > channel to be muted, but setting it to 0.000001 will not cause it to be
> > muted, but result in a high CPU load.
> >
After discussion on IRC Jonathan explained to me what this is all about. The
patch is correct.
Nevertheless, I was confused about the fact that this value of 0.000001 is
never going to be USED.
Perhaps a different value and comment can avoid that in the future when other
coders look at it.
mInternalGain is only ever compared with, and the whole point of this patch is
to make sure
that the first comparison fails (I now understand). A common value used to make
sure that
a first comparison fails is a value outside the valid range of that variable.
The valid
range being [0, 1], I'd suggest using -1 and adding a comment in the lines of:
// Make sure that the first call to setMasterGain will cause
setInternalGain to be called.
mInternalGain = -1.f;
- Aleric
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
http://codereview.secondlife.com/r/72/#review128
-----------------------------------------------------------
On Jan. 6, 2011, 2:37 p.m., Jonathan Yap wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> http://codereview.secondlife.com/r/72/
> -----------------------------------------------------------
>
> (Updated Jan. 6, 2011, 2:37 p.m.)
>
>
> Review request for Viewer.
>
>
> Summary
> -------
>
> There is an edge case in setMasterGain during startup which prevents
> setInternalGain from being called if the master volume setting and
> mInternalGain both equal 0.
>
> Setting mInternalGain to a very low but non-zero value fixes this issue.
>
>
> This addresses bug STORM-830.
> http://jira.secondlife.com/browse/STORM-830
>
>
> Diffs
> -----
>
> indra/llaudio/llaudioengine.cpp 6d44f0d85a80
>
> Diff: http://codereview.secondlife.com/r/72/diff
>
>
> Testing
> -------
>
> In Preferences / Sound & Media tested:
> Buttons
> Ambient
> Sound Effects
> Stream Music
> Media
> Voice Chat
>
>
> Thanks,
>
> Jonathan
>
>
_______________________________________________
Policies and (un)subscribe information available here:
http://wiki.secondlife.com/wiki/OpenSource-Dev
Please read the policies before posting to keep unmoderated posting privileges
