Nodes: Skip node tree update for unused sockets of muted node #109088

Iliya Katushenock · 2023-06-17T23:08:54+02:00

Iliya Katushenock commented

2023-06-17 23:08:54 +02:00

If node is mute, it sockets values update didn't really affect on something.
It possible to detect this and avoid extra update.
This make sense only for rna update calls, other update propagation kinds
can be related with tree topology, ... .
Data block sockets are questionable. Potentially, there is no problem
with ignoring them update.

If node is mute, it sockets values update didn't really affect on something. It possible to detect this and avoid extra update. This make sense only for rna update calls, other update propagation kinds can be related with tree topology, ... . Data block sockets are questionable. Potentially, there is no problem with ignoring them update.

👍 1

Iliya Katushenock added 1 commit 2023-06-17 23:09:00 +02:00

d10fcbc3e8 init

Iliya Katushenock requested review from Jacques Lucke 2023-06-17 23:09:10 +02:00

Iliya Katushenock requested review from Hans Goudey 2023-06-17 23:09:10 +02:00

Iliya Katushenock added this to the Nodes & Physics project 2023-06-17 23:11:52 +02:00

Jacques Lucke reviewed 2023-06-19 20:46:09 +02:00

source/blender/makesrna/intern/rna_nodetree.cc Outdated

						
				@ -2913,3 +2913,3 @@

				  bNodeTree *ntree = reinterpret_cast<bNodeTree *>(ptr->owner_id);

				  bNodeSocket *sock = static_cast<bNodeSocket *>(ptr->data);

				  const bool skip_tree_update = [=]() -> bool {

Jacques Lucke commented

2023-06-19 20:45:30 +02:00

Why use [=] instead of [&]?

Why use `[=]` instead of `[&]`?

Iliya Katushenock commented

2023-06-19 20:51:50 +02:00

Just socket/tree pointers, can be captured by value

Jacques Lucke commented

2023-06-19 21:00:00 +02:00

Yeah, generally prefer [&] if the lambda does not leave the current scope, unless there is a good reason for doing something else.

Yeah, generally prefer `[&]` if the lambda does not leave the current scope, unless there is a good reason for doing something else.

mod_moder marked this conversation as resolved

Jacques Lucke commented

2023-06-19 20:46:59 +02:00

Was there a bug report for this? It's not completely unreasonable to change this, but I do wonder what the motivation is. Why would someone change the rna values that don't affect the result?

Iliya Katushenock commented

2023-06-19 20:54:33 +02:00

This was suggested in chat by a user

Jacques Lucke commented

2023-06-19 21:00:26 +02:00

This was suggested in chat by a user

Then my question goes to that user :)

> This was suggested in chat by a user Then my question goes to that user :)

Iliya Katushenock added 1 commit 2023-06-19 21:31:07 +02:00

3a67f752ea change capture type

Jakub Jaszewski commented

2023-06-19 21:44:54 +02:00

First-time contributor

I suggested this in the chat. My motivation for this was that with more complicated and heavy node setups that take a lot of time to compute. Sometimes I would like to mute a node, change couple parameters together and unmute it. That way the node tree is calculated only once and not for every parameter separately.
I have couple heavy displacement node trees that can take couple minutes to compute, and every mistake or change costs time.

I also asked if I should report this as a bug but @mod_moder was faster with this pull request.
I attached trivial and real life examples.

ED: Right now to get over this limitation I do a bypass like in the linked image. Doing this repeatedly is monkey work and can be frustrating sometimes.

I suggested this in the chat. My motivation for this was that with more complicated and heavy node setups that take a lot of time to compute. Sometimes I would like to mute a node, change couple parameters together and unmute it. That way the node tree is calculated only once and not for every parameter separately. I have couple heavy displacement node trees that can take couple minutes to compute, and every mistake or change costs time. I also asked if I should report this as a bug but @mod_moder was faster with this pull request. I attached trivial and real life examples. ED: Right now to get over this limitation I do a bypass like in the linked image. Doing this repeatedly is monkey work and can be frustrating sometimes.

bypass.png

70 KiB

software_abuse.png

2.4 MiB

mute.png

272 KiB