- User Since
- Nov 28 2009, 10:22 PM (490 w, 2 d)
Technically, the cost of lookup consists of BLI_ghashutil_ptrhash()and BLI_ghashutil_ptrcmp()multiplied by the number of hash collisions. Both functions are rather cheap. and there should not that many collisions either.
Dependency graph might be NULL for the view layers which were never visible. So it is matter of adding NULL pointer check.
P.S. Not sure it's really a "Bug" task.
Fri, Apr 19
@Clément Foucault (fclem), we've discussed things here at the studio, and the solution was not related to the dependency graph changes.
Let me know if there is something else i can help you here with.
If nothing has changed, the cache will not be invalidated. This is an entire idea of cache.
I also don't see a difference in behavior with 2.79 and 2.79b releases. Not sure what you mean by previous blender builds, but is possible that it was a bug related on an unintentional invalidation.
Assigning to Campbell. We've looked into the behavior here at the studio.
Thu, Apr 18
Wed, Apr 17
Tue, Apr 16
@LazyDodo (LazyDodo), done!
Mon, Apr 15
Sun, Apr 14
Sat, Apr 13
Fri, Apr 12
Sounds very similar to T62521.
Will have a look next week.
If desired, I can also change it to make the sphere a double-sided emitter by switching to uniform sampling when the shading point is inside the sphere (this is what PBRT does).
Wed, Apr 10
@MACHIN3 (MACHIN3), ah, well. It worked by accident, by doing logic based on uninitialized variable. The issue is that the evaluated mesh is requested for a non-evaluated dependency graph. For some reason quick adjustments to the script didn't solve this. Looking deeper into it..
@Mikhail Rachinskiy (alm), agree. There is still a huge work to be done on documentation and release logs, and this indeed something what should be mentioned there.
Seems to be related to the degenerated nature of the intersection.
@Campbell Barton (campbellbarton) , mind having a look? Thanks!
@Clément Foucault (fclem), mind having a look? Thanks!
@Clément Foucault (fclem), seems to be drawing issue. If i assign instanced object to the particle system, it shows all correct in the viewport.
@lucas veber (lucky3), thanks for simplifying the test case!
@Campbell Barton (campbellbarton), since you're listed as an author of BVH addon, mind giving some input here? ;)
@Gaia Clary (gaiaclary), seems this slipped through the cracks. Is it still relevant for the 2.80 with the new instancing based on collections?
Any resolution here? Triage at some reasonable priority?
@Sybren A. Stüvel (sybren), this is a year old task, which has a patch and is still untriaged. Mind having a look?
Think it's all fine.
Can not confirm the issue yet, but see some assert failures which might be affecting on it.
Tue, Apr 9
In order to properly get mesh with modifier applies the dependency graph is to be evaluated first. What happens in this script is that it does modify the original object (by doing update_from_editmode), but it does not inform dependency graph about it. There are two ways to go about it:
This is quite weird that those cards were displayed in 2.79: the requirement back then was compute capability 2.0, but those cards are 1.1 .
Mon, Apr 8
That option did control GPU side tessellation. It is temporarily removed from 2.80 -- it needs slightly different integration into modelling workflow and requires re-write of the GPU bindings since the OpenGL upgrades and the new draw manager for viewport.
The modifier itself is always using OpenSubdiv. It still has some performance impacts in certain cases, but we will keep working on speed (the upcoming OpenSubdiv release should make us able to gain extra performance).
@MACHIN3 (MACHIN3), the issue is fixed by rBdf26f28835f, but this opens slightly different topic: the to_mesh function requires dependency graph to be evaluated. This means, constructing new scene and trying to do to_mesh from it will not give correct result.
Actually, re-opening as a stand-alone bug. It's somewhat related, but is not quite. Sorry for the noise.
@Sebastian Parborg (zeddb), not really, were too much changes in areas which might be affecting this, and quite some changes in the dependent libraries versions happened.
Sat, Apr 6
Binding can take some time to process the math, but several minutes for the file from this report is quite extreme.
Please compare the result on same mesh and cage in 2.7, and if the latest 2.8 is measurably slower make a new report.
Fri, Apr 5
Got vied by df99c54b5b8 :)
This givens me the following error on startup:
Update for Brecht's review
Thu, Apr 4
It is possible, but needs some surgeoning on operations and relations.
Fix unbinding of Laplacian Deform
Update for Brecht's review
@Kor Boerema (boemak), toggling sculpt mode creates all the internal structures like BVH. Also, when using dynamic topology this will also create a special BMesh representation of the object.
@Sebastian Parborg (zeddb), those ifdef-ed blocks are to be removed. The Bind operator is to pass evaluated object to the modifier stack, and copy bound data to original.