|
 |
Am 26.11.2016 um 19:12 schrieb [GDS|Entropy]:
> clipka <ano### [at] anonymous org> wrote:
>> Am 25.11.2016 um 17:51 schrieb Mike Horvath:
>>> You can see in the image there is a jagged line where the isosurface
>>> intersects the cylinder. Is it possible to get rid of this completely?
>>>
>>> I have set accuracy very low to 0.0001 and max gradient very high to
>>> 10000, but I'm afraid it will never go away completely.
>>
>> You might try changing the overall scale of the entire scene.
>>
>> An overly high max_gradient will not help at all, and will only increase
>> render time.
>>
>>
>
> Out of curiosity, how close to "as good as it gets" is isosurface
> performance? I love the things but they are prohibitively slow most of the
> time...at least for the things I do with them, which is probably all sorts
> of backwards and wrong to begin with. :)
There are certainly still ways to optimize, at least for special cases.
One such example would be functions whith lots of subexpressions, each
of which affect the result only in a well-defined region of space, like
the components in a blob.
Other approaches would involve preprocessing the isosurface prior to
rendering -- whether it is conversion to a mesh, pre-computing of
samples in a spatial tree (with the extreme being conversion to voxel
data), or other such fancy stuff.
And last not least there's the execution speed of the user-defined
functions. At present they are executed by a homegrown virtual machine,
which is quite simple and therefore presumably quite fast, but most
certainly no match for just-in-time-compiled code.
Post a reply to this message
|
|
