Something about the way node relationships are tracked and recalculated was 
changed in Nuke 9 (I think... maybe it was 8), which has resulted in a lot of 
of strange performance hits from nodes that are out of the evaluation tree. One 
of the most obvious side-effects of this is that the `Node.dependent()` Python 
method is now orders of magnitude slower, to the point that it's actually 
faster to call `.dependencies()` on every single node in the script and 
intersect their inputs with a given node to find downstream nodes.

I've also got an outstanding bug related to all of this performance stuff: Bug 
50719 - 3D system causing UI lockups on upstream nodes

-Nathan



From: Daniel Hartlehnert 
Sent: Monday, March 07, 2016 4:04 AM
To: Nuke user discussion 
Subject: Re: [Nuke-users] Nuke locking up with 1 cpu busy

I understand. But i have seen it when opening standard grade or merge nodes. 
These definately didn’t have expressions on them.


> Am 07.03.2016 um 09:48 schrieb Ben Dickson <[email protected]>:
> 
> The expressions can be in unobvious places, like inside gizmos.
> 
> The bug Elvis Au mentioned sounds similar to "Bug 35703. Expressions force 
> tree to validate unnecessarily, when changing unrelated parts of the tree", 
> which I reported as such:
> 
> On 06/05/13 20:24, Ben Dickson wrote:
>> Attached another example, which doesn't involve the RotoPaint:
> >
>> The ImagePlane gizmo (available from Nukepedia) contains an "width"
>> expression in a crop. This combined with the motionblur option in
>> the Card3D causes the same UI sluggishness
> >
>> Adjust the slider in the Keyer node, or try and type something in a
>> label - it is noticably laggy.
> >
>> If the motion blur is setting is set to 0, it is much quicker.
> >
>> The UI slowness gets progressively worse the more copies of the
>> ImagePlane there is, to the point the comp was not workable..
> >
>> I haven't tested in older versions, but this behaviour would very
>> much explain many of the slowdowns we have experienced in the past..
>> Especially because we automatically create a Crop node above our
>> Write nodes, and this contains input.width and input.height
>> expressions (to prevent writing out EXR's with oversized bbox)
> >
>> It seems like a relatively serious bug, as it will exacerbate any
>> inefficiencies in any nodes - worse as it can be very hard to track
>> down as the cause of a slow script..
> 
> Attachment was essentially this script:
> https://gist.github.com/dbr/2f3faff433d59f58be7d
> 
> Reported for 7.0v5 in 2013 - but remains same in all current Nuke versions.
> 
> Some variant of this is usually the cause of a majority of "my script is 
> slow" complaints I've seen here - some slow node is upstream of a time-based 
> node (TimeEcho or lots of TimeOffset's, transform with the "motion blur" 
> setting enabled, ScanlineRender with many motion subsamples etc)
> 
> The only way I have found to narrow it down is to delete half the script, see 
> if it speeds up or not, then repeat (using the Keyer slider as a good way of 
> checking - it's easy to see if the slider control tracks quickly with the 
> mouse cursor or not)
> 
> This doesn't exactly sound like the original problem Daniel described 
> (particularly "i can sometimes work in the same script for hours, then 
> suddenly the symptons start to show up"), although it can be made sometimes 
> seem quite random if the slowness is in part due to network file access 
> slowdowns.
> 
> Something similar also happened with a script containing a lot of ReadGeo 
> nodes loading Alembic .abc geometry with a relatively complex scene hierarchy.
> 
> 
> On 07/03/16 18:34, Daniel Hartlehnert wrote:
>> @Lucy:
>> another question though: if it really is the expressions that are
>> causing this. how is it possible that nodes that do not have expressions
>> in them are causing this behavior?
>> Thats the way it was happening for me. I have scripts with many
>> expressions in them, but even opening nodes that have none may cause the
>> freeze.
>> Oh and sending you a scrlpt is unfortunately not possible. NDA reasons
>> and its not really reproducable.
>> 
>> Daniel
>> 
>>> Am 04.03.2016 um 17:39 schrieb Lucy Wilkes <[email protected]
>>> <mailto:[email protected]>>:
>>> 
>>> Hi Daniel,
>>> 
>>> I've answered on your forum thread. I also wondered whether
>>> expressions might be causing this
>>> (http://community.thefoundry.co.uk/discussion/post.aspx?f=189&t=120025&p=1008064).
>>> 
>>> If you have a script and repro steps, please do send them in to our
>>> Support team so that we can take a look.
>>> 
>>> Regards,
>>> Lucy
>> 
>> 
>> 
>> _______________________________________________
>> Nuke-users mailing list
>> [email protected], http://forums.thefoundry.co.uk/
>> http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users
>> 
> 
> -- 
> ben dickson
> 2D TD | [email protected]
> rising sun pictures | www.rsp.com.au
> _______________________________________________
> Nuke-users mailing list
> [email protected], http://forums.thefoundry.co.uk/
> http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users

_______________________________________________
Nuke-users mailing list
[email protected], http://forums.thefoundry.co.uk/
http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users
_______________________________________________
Nuke-users mailing list
[email protected], http://forums.thefoundry.co.uk/
http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users

Reply via email to