[sldev] Re: [VWR] [PATCH] particles VWR-418 VWR-983 VWR-2164 (Tofu
Linden)
Argent Stonecutter
secret.argent at gmail.com
Mon Sep 17 07:54:30 PDT 2007
> Dirk Moerenhout wrote:
>> VWR-418:
>> This focuses on particle generation. This should fix most of the
>> issues where the particle system is overloaded. Note that the system
>> is adaptive. Depending on the amount of particles vs max particles a
>> reference variable is updated. This variable is in turn used to limit
>> particle generations. Some of the limits that'll be set:
>> - Particles that are too small to be useful are not generated at all
I didn't catch this before, or parsed it as "particles that are too
small to be seen".
Clouds of many small particles, particles that may individually be
"too small to be useful", are common for things like smoke effects.
You need to make sure that this case doesn't break them.
I tried a number of variations of this kind of thing and it's hard to
get right... I finally sent one to Nicolaz with a relatively simple
algorithm that at least did no harm.
>> VWR-983:
>> This focuses on particle updates. This should fix the issues where
>> particles go out of sync and end up where they should not be
>> according
>> to timing. This is done by keeping accurate timing for particles that
>> are invisible (and hence get updated only every 8 grames). And a bug
>> is fixed that caused particles not to be updated every time another
>> particle was moved to another group or deleted.
Does this fix the "flashing" bug where particles that have follow-
source set are generated as if they didn't have follow-source set
then *snap* to the source the first time they're updated? That one is
really frustrating.
This doesn't just effect particles that have mistakenly set burst
radius and follow source together (that was where it was first
discovered, and that case could (as Nicolaz suggested) be fixed by
killing burst radius when follow source is set) unfortunately... it
also makes particles from moving emitters seem to "shed" in a
flickering cloud behind them. The underlying bug needs to be addressed.
More information about the SLDev
mailing list