 |
I've tested making the max number of timed samples 250 for each particle stage, instead of testing every frame of the particle's lifetime. Tested with 75 different particle stages. The size results match the older method perfectly. There's not going to be a big difference in map load time. It turns out that estimating particle sizes is disproportionately slower in a debug build, which I used for profiling. In NHAT forest for example, particle estimations take up 25% of map load time (22.6s) in a debug build, but only 5.5 seconds in a release build. Still, that 5.5s was reduced by 4s using the new method. It's only long-lived particles that take up a lot of time. I chose NHAT forest because it has leaves in it that last 5 minutes. The effect on maps with only short-lived particles will be negligible. |
|
|
Test results attached |
|
|
At rev 6543 /trunk/framework/DeclParticle.cpp |
- Anonymous
