[Bf-taskforce25] Blender spring cleaning
Nathan Vegdahl
cessen at cessen.com
Thu May 21 05:35:07 CEST 2009
After talking with Joe on IRC he pointed me to this paper:
http://www.vis.uni-stuttgart.de/~weiskopf/publications/eg03short.pdf
Indeed halfway shadow buffers do have artifacts that normal buffers
don't, namely self-unshadowing in some scenes with concave objects or
surfaces that aren't closed such as planes.
So I retract my proposal (with apologies to Joe) to remove normal
shadow buffers.
But I do think that halfway buffers (with a very small bias) should
be made the default. They generally "just work", unlike normal shadow
buffers which usually require a fair amount of tweaking on a
case-by-case basis to get rid of artifacts.
I also like Joe's idea of making the halfway buffers a checkbox on
normal shadow buffers. They are so closely related to each other that
it seems odd for them to each have a separate menu item. Then we can
just make the checkbox on by default.
--Nathan V
On Sat, May 16, 2009 at 4:29 AM, joe <joeedh at gmail.com> wrote:
> Perhaps it could become a little checkbutton ("Use Midpoint") that
> people can disable if they ever get the bad use cases. It could be
> kept amongst whatever other "advanced" options we end up with :) I
> think removing the option to ever use normal buffers would be a
> mistake, though.
>
> Joe
>
> On Fri, May 15, 2009 at 9:26 PM, Nathan Vegdahl <cessen at cessen.com> wrote:
>>>>> It's not always better; there are issues on 90 degree corners.
>>>>
>>>> I've been messing around for a while now trying to reproduce what
>>>> you're talking about. I'm not having any luck, and I've never seen
>>>> any issues with corners before. Could you send an example scene?
>>>
>>> Ah not really, I think I've seen it once or twice, but that's it. Not
>>> sure how to reproduce it either, heh.
>>
>> If that's the case then I would hazard a guess that it's either a bug,
>> or user error (too low or too high bias).
>> After drawing out some diagrams I'm still finding that halfway shadow
>> buffers always perform better. In cases where halfway shadow buffers
>> have artifacts, normal shadow buffers have the same artifact but
>> worse. And with halfway shadow buffers much less bias is necessary to
>> fix the artifact.
>>
>> --Nathan V
>>
>> On Fri, May 15, 2009 at 3:27 PM, joe <joeedh at gmail.com> wrote:
>>> On Fri, May 15, 2009 at 1:16 PM, Nathan Vegdahl <cessen at cessen.com> wrote:
>>>>>> - Can we kill classic shadow buffers? Classic-halfway is
>>>>>> universally superior as far as I know, and it doesn't seem to have any
>>>>>> noteworthy performance/memory penalties.
>>>>>
>>>>> It's not always better; there are issues on 90 degree corners.
>>>>
>>>> I've been messing around for a while now trying to reproduce what
>>>> you're talking about. I'm not having any luck, and I've never seen
>>>> any issues with corners before. Could you send an example scene?
>>>
>>> Ah not really, I think I've seen it once or twice, but that's it. Not
>>> sure how to reproduce it either, heh.
>>>
>>>>
>>>>> True oversampling is better, I think. Certainly easier then messing
>>>>> with resolution/softness settings.
>>>>
>>>> To be fair, true oversampling should be done via DSM. This was
>>>> just a hack to get around the lack of DSM at the time, because we
>>>> needed it for BBB. I'd much rather kill this in favor of your DSM
>>>> work (granted it's on the slow side, but it can be optimized over
>>>> time). In the mean time, larger buffer res + softness can do this in
>>>> productions that need it.
>>>
>>> DSM isn't always worth it though, it really is quite a bit slower.
>>> During BBB I tried very hard to get it working in time, but after a
>>> certain point it became clear that it was just too slow for you guys,
>>> so I eventually gave up on that goal.
>>>
>>>>
>>>>>> - B-bone Rest: kill this! It's an evil behavior kept only for
>>>>>> backwards compatability with older files! DIE!!!!
>>>>>
>>>>> As someone with such a rig, I would be sad to see it go. On the other
>>>>> hand, I'm sure my rig will need a fair amount of updating for 2.5
>>>>> anyway, so probably the least of my worres. :)
>>>>
>>>> I imagine several people out there have rigs like this. But it's
>>>> no longer the default behavior anyway, and for good reason: it's a
>>>> buggy behavior.
>>>> Best to get rid of it in 2.5, I think, leaving only the sane behavior.
>>>
>>> Yeah I agree.
>>>
>>> Joe
>>> _______________________________________________
>>> Bf-taskforce25 mailing list
>>> Bf-taskforce25 at blender.org
>>> http://lists.blender.org/mailman/listinfo/bf-taskforce25
>>>
>> _______________________________________________
>> Bf-taskforce25 mailing list
>> Bf-taskforce25 at blender.org
>> http://lists.blender.org/mailman/listinfo/bf-taskforce25
>>
> _______________________________________________
> Bf-taskforce25 mailing list
> Bf-taskforce25 at blender.org
> http://lists.blender.org/mailman/listinfo/bf-taskforce25
>
More information about the Bf-taskforce25
mailing list