> 1. 400ms buffers are hated by users, as was noted previously. They are
> passable for some applications, but way too laggy for others.
>
> 2. Even if you are will to have a 400-500 ms buffer, if you can prove that you
> will always meet that deadline, then it's a realtime system.
>
> 3. If you can show logically that you will nearly always meet the deadline,
> it's a soft realtime system. That's what we're after here. From a design
> standpoint, there are elegant soft realtime systems, and there are sucky
> ones.
>
> 4. So how do you propose to "program timings" so that it's really hard to miss
> those deadlines?
Having a backup of 400-500ms gives you an average hang-over of 200-250ms
that are hardly noticeable by a human in this topic. The issue is not if
you always meet the deadline, the issue is what amount of load will make
you miss it. If I have to hire five ppl clicking and dragging on my desktop
to make my player to skip, and it skips, I don't care if it missed the
deadline. This because my desktop will hardly see that load. But if you
have a 50-100ms backup, things turns out to be a little bit different.
If you pretend to run a `make -jUNREAL` and still have the audio not
skipping it is simply wrong. Running a `make -j20` in my machine shows an
average of 24 TASK_RUNNING tasks, that even if they're granted with a mere
40ms timeslice, it'll take a full second before an expired task will see
the light again. BTW, under such load RealPlayer skips badly, but I don't
really care since it never did while I was doing all the normal (and many
extra) stuff I'm doing on my desktop.
- Davide
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/