[sldev] More crash rate statistics
stickman at gmail.com
Thu Aug 13 04:16:54 PDT 2009
> For example, a user might choose the following simplistic demand:
> * If the FPS is higher than 20, I don't care about FPS anymore,
> I just want the highest possible quality.
> * If the FPS drops below 5, I don't care about the quality anymore,
> do whatever is needed to give me at LEAST a FPS of 5.
Yay, fancy math! You even got some layman in there for the rest of us.
I remember seeing a train traveling earlier, and it disappeared into
the distance. I'm not sure where it will end up, but this is what I
saw, and where I saw it headed.
1) LL is worried about viewer performance involving very high
polycount items. In an effort improve the overall viewer experience,
they take the ISP route of cutting off those people that use 100% of
the connection bandwidth 100% of the time by imposing some
ridiculously high limit that no one will ever hit. In theory. This fix
goes into viewer 1.23.
2) http://jira.secondlife.com/browse/VWR-13868 happens. I'm sure
everyone remembers the magic 172 sculpty prim necklace that consists
of over a quarter-million tris at full resolution. (I've got an old
rant about the acceptability of such an item, so I won't restate if
it's "right" or "wrong" to create an item like this.)
3) Somehow, LL manages to remain calm in the face of some of the ...
er... "bug confirmations" and admits that there was probably a better
way to do this. One Linden or another mention how it would be possible
to get in a rendering system to give priority detail to things closer
to the center of the camera, based on distance, movement speed, etc.
And how it actually wouldn't be that hard to put in. (Q? Qarl? Nyx? I
totally can't remember. Maybe it was Melinda?) Thus, they would get
their goal of a better, safer render, without breaking such "high end"
content as said necklace.
4) The train has moved over the horizon. This bug is now marked "Fix
Pending" but I remember hearing an argument that a new 1.23 will not
be released with the fix. I'm not sure if the fix is just taking it
out while they work on a different solution, or the other solution
already being in place -- and if it is, whether or not it's some fancy
system involving changing object detail based on location, distance,
and speed, and perhaps even current FPS.
I tend to be a bit more optimistic than I should be sometimes, so
don't take any of this as indicative of anything. I'd be interested to
hear the real resolution to that Jira if someone has it, though I'd be
a little sad if it wasn't what Mr. Inglewood already said. Since that
solution seems to make so much sense to me.
More information about the SLDev