[Bf-committers] point cache changes

Daniel Genrich daniel.genrich at gmx.net
Wed Mar 19 13:27:52 CET 2008


Hi Brecht,

I'm fine with the changes and also looking forward to see the 
memory/endian solution. :)

Daniel


Brecht Van Lommel schrieb:
> Hi,
>
> I started working on improving the point cache. There's still a bunch of problems I don't have a good solution for, but I'd like to make these changes to start with:
>
> - Abstract point cache file writing: rather than calling fread/fwrite/fclose directly, add functions to read/write n floats. This way it can do endian switching behind the scenes, and possibly write to memory instead of to disk.
> - Do point cache clearing from the dependency graph. Right now it's hard to find out if the point cache should be cleared when the physics system runs, and additionally this doesn't work if the modifier is disabled. So instead if an object changes it's point caches should be cleared. In the dependency graph code it should be easier to make a distinction between time changes or actual changes to the object.
>
> Another change I think we need is that the user should never see outdated results, unless they manually asked for this by pressing a protect button for example.
> - Change the particle system to never take time steps not equal to 1 frame. Right now if you skip ahead 10 frames, it will take on big step. However that means an invalid result is written to the point cache, since it's not the same as taking 10 smaller timesteps. That often leads to skips when playing back the animation.
> - Remove the cloth Autoprotect Cache feature. I think this is very confusing, since because of it you often see outdated results.
> - Replace the Protect button by a Bake button. In my opinion the concept of protecting is quite confusing, since for the user it's hard to know what is actually cached. A Bake button could still reuse cached results if they exist, and so pressing it wouldn't necessarily re-run the simulation, but at least make sure that all the specified frames are in fact in the cache and then protect them.
>
> Please give feedback if you agree with these changes?
>
> Cheers,
> Brecht.
>
>
>
>
> _______________________________________________
> Bf-committers mailing list
> Bf-committers at blender.org
> http://lists.blender.org/mailman/listinfo/bf-committers
>
>   



More information about the Bf-committers mailing list