GPencil: Multiple lattices reduces FPS #89349

New Issue

raylee · 2021-06-22T04:17:09+02:00

raylee commented

2021-06-22 04:17:09 +02:00

System Information
Operating system:Windows 10
Graphics card:RTX6000
CPU:intel xeon platinum 8179m

Blender Version
Broken: 2.93

Short description of error
When version 2.93 is updated, you can add multiple lattices to the same Greasepencil, you have a new one problem
Adding multiple lattices to the same Greasepencil reduces the frame rate to 3 fps
CPU utilization is very poor,mainly focused on the use of a single

Step to Reproduce the error

Open .blend file
Select lattice then enter edit mode
Select vertices and move them with key G
Notice the FPS value
In object mode either add or remove lattice modifiers
repeat steps 2-3
Check the FPS value

Test File:

#89349.blend

Short video:

**System Information** Operating system:Windows 10 Graphics card:RTX6000 CPU:intel xeon platinum 8179m **Blender Version** Broken: 2.93 **Short description of error** When version 2.93 is updated, you can add multiple lattices to the same Greasepencil, you have a new one problem Adding multiple lattices to the same Greasepencil reduces the frame rate to 3 fps CPU utilization is very poor,mainly focused on the use of a single **Step to Reproduce the error** - Open .blend file - Select lattice then enter edit mode - Select vertices and move them with key `G` - Notice the FPS value - In object mode either add or remove lattice modifiers - repeat steps 2-3 - Check the FPS value Test File: [#89349.blend](https://archive.blender.org/developer/F10208021/T89349.blend) ![QQ截图20210622100256.jpg](https://archive.blender.org/developer/F10187307/QQ截图20210622100256.jpg) ![QQ截图20210622100238.jpg](https://archive.blender.org/developer/F10187306/QQ截图20210622100238.jpg) Short video: ![#89349.gif](https://archive.blender.org/developer/F10208550/T89349.gif)

raylee commented

2021-06-22 04:17:09 +02:00

Added subscriber: @RayLee

Pratik Borhade commented

2021-06-22 05:34:24 +02:00

Added subscriber: @PratikPB2123

Pratik Borhade commented

2021-06-22 05:34:24 +02:00

Changed status from 'Needs Triage' to: 'Confirmed'

Pratik Borhade commented

2021-06-22 05:34:24 +02:00

Adding multiple lattice modifier drops the FPS almost to 6-12.

Confirming the report but Grease pencil module team can answer this in better way.

System Information

Graphics card : AMD Radeon(TM) 535 ATI Technologies .
Blender version: 2.93 LTS```

Adding multiple lattice modifier drops the FPS almost to 6-12. Confirming the report but `Grease pencil` module team can answer this in better way. --- **System Information** ```Operating system : Windows-10-10.0.18362-SP0 64 Bits Graphics card : AMD Radeon(TM) 535 ATI Technologies . Blender version: 2.93 LTS```

Antonio Vazquez commented

2021-06-22 16:20:26 +02:00

Added subscriber: @antoniov

Antonio Vazquez commented

2021-06-22 16:20:26 +02:00

@RayLee Could you provide a simple test file to profile where is the bottleneck?

Antonio Vazquez changed title from ~~【Greasepencil】Multiple lattices cause Cottonwood~~ to GPencil: Multiple lattices reduces FPS

2021-06-22 16:21:08 +02:00

Antonio Vazquez commented

2021-06-23 17:40:10 +02:00

Changed status from 'Confirmed' to: 'Needs User Info'

Pratik Borhade commented

2021-07-01 13:02:47 +02:00

Hi @antoniov , initially I confirmed the ticket, now unsure if there is actual fall in FPS.

Could you provide a simple test

Tried with default 2d animation scene.

From my observation: FPS increased when adding Lattice modifier in stack but FPS reduces when I stopped to add them (I think its obvious)

Hi @antoniov , initially I confirmed the ticket, now unsure if there is actual fall in FPS. >Could you provide a simple test Tried with default 2d animation scene. From my observation: FPS increased when adding Lattice modifier in stack but FPS reduces when I stopped to add them (I think its obvious)

Antonio Vazquez commented

2021-07-01 16:49:44 +02:00

@PratikPB2123 the point in your comment is From my observation... I have tested in my PC and I don't see anything, so maybe my example is not the best or I'm doing something different, this is the reason I would like to have a test file.

You can understand that for us is a time consuming task to prepare a test file for each bug report, so if you provide a test file we can go to the problem itself, instead to try to reproduce it first creating a test file.

Thanks for your understanding.

@PratikPB2123 the point in your comment is `From my observation`... I have tested in my PC and I don't see anything, so maybe my example is not the best or I'm doing something different, this is the reason I would like to have a test file. You can understand that for us is a time consuming task to prepare a test file for each bug report, so if you provide a test file we can go to the problem itself, instead to try to reproduce it first creating a test file. Thanks for your understanding.

Pratik Borhade commented

2021-07-01 17:58:46 +02:00

@antoniov , sorry to keep you on wait. Attaching the file. (Also updating the task description)

is a time consuming task to prepare a test file for each bug report

Yes, I understand.

#89349.blend

@antoniov , sorry to keep you on wait. Attaching the file. (Also updating the task description) > is a time consuming task to prepare a test file for each bug report Yes, I understand. [#89349.blend](https://archive.blender.org/developer/F10208021/T89349.blend)

Antonio Vazquez commented

2021-07-01 18:07:18 +02:00

Thanks @PratikPB2123 for the file. I will do a profile test to see what is going on.

Antonio Vazquez commented

2021-07-01 18:59:52 +02:00

In the first test with your file, I don't see anything below 21 FPS (only in one frame) and later keeps 25 FPS all the time. I have a i9-9900K CPU with NVIDIA RTX2080 TI using Windows 10.

Could you record a small video with the exact steps you do?...maybe you do something in a different way than me and this is the reason.

Also, have you tested in a different PC with different hardware setup?

About threading, the modifiers evaluation is done in a single thread because need to do sequential, so don't see a way to make it multithread.

In the first test with your file, I don't see anything below 21 FPS (only in one frame) and later keeps 25 FPS all the time. I have a i9-9900K CPU with NVIDIA RTX2080 TI using Windows 10. Could you record a small video with the exact steps you do?...maybe you do something in a different way than me and this is the reason. Also, have you tested in a different PC with different hardware setup? About threading, the modifiers evaluation is done in a single thread because need to do sequential, so don't see a way to make it multithread.

Antonio Vazquez commented

2021-07-01 19:19:19 +02:00

Second test with your file. I'm able to get sustained 40 FPS

Pratik Borhade commented

2021-07-02 07:45:00 +02:00

Could you record a small video with the exact steps you do?

Uploaded the video

have you tested in a different PC with different hardware setup?

Not yet. I will check with other laptop on weekend

>Could you record a small video with the exact steps you do? Uploaded the video >have you tested in a different PC with different hardware setup? Not yet. I will check with other laptop on weekend

Antonio Vazquez commented

2021-07-02 10:45:12 +02:00

Looking at your video I see why I could not reproduce, we were looking at different things. The modifiers are optimized to keep fast FPS playing animation.

In you video you are editing the data and you are looking at the FPS of the viewport refresh. In this case, the system precalculate a lot of things every time you edit a point. This is done to get fast play, but the cost is the edition is slower.

Every time you edit the stroke, the internal geometry of several things of the stroke is calculated and saved to use later and this is what you get. Also, this is not related to modifiers only, modifiers just add more complexity to the calculation.

I have tested in my system and I don't notice any lag editting the lattice, but I did not made a measure with a tool. Of course edit with lattice modifiers is slower, but this is normal and don't make edit unusable.

I don't know what we can do here to improve speed because the problem is if we delay updates in edit, the play speed will be worst.

Looking at your video I see why I could not reproduce, we were looking at different things. The modifiers are optimized to keep fast FPS playing animation. In you video you are editing the data and you are looking at the FPS of the viewport refresh. In this case, the system precalculate a lot of things every time you edit a point. This is done to get fast play, but the cost is the edition is slower. Every time you edit the stroke, the internal geometry of several things of the stroke is calculated and saved to use later and this is what you get. Also, this is not related to modifiers only, modifiers just add more complexity to the calculation. I have tested in my system and I don't notice any lag editting the lattice, but I did not made a measure with a tool. Of course edit with lattice modifiers is slower, but this is normal and don't make edit unusable. I don't know what we can do here to improve speed because the problem is if we delay updates in edit, the play speed will be worst.

raylee commented

2021-07-06 10:19:30 +02:00

In #89349#1180793, @antoniov wrote:
@RayLee Could you provide a simple test file to profile where is the bottleneck?

@antoniov I’m sorry I didn’t get the message until now. Here’s an interesting question: 1. When I have a character that has 10 lattices, FPS is very low.

When I split the characters into different GP, adding lattices to different GP, they did well.

Also, I don’t know if you’ve noticed, but the CPU only uses a single thread, and the utilization is very low.

Check out the video for details
https://youtu.be/RmgKGsnJlXA

> In #89349#1180793, @antoniov wrote: > @RayLee Could you provide a simple test file to profile where is the bottleneck? @antoniov I’m sorry I didn’t get the message until now. Here’s an interesting question: 1. When I have a character that has 10 lattices, FPS is very low. 2. When I split the characters into different GP, adding lattices to different GP, they did well. Also, I don’t know if you’ve noticed, but the **CPU only uses a single thread, and the utilization is very low.** Check out the video for details https://youtu.be/RmgKGsnJlXA

Pratik Borhade commented

2021-07-13 13:09:20 +02:00

Changed status from 'Needs User Info' to: 'Needs Triage'

Philipp Oeser commented

2021-09-27 14:47:37 +02:00

Added subscriber: @lichtwerk

Philipp Oeser commented

2021-09-27 14:47:37 +02:00

Changed status from 'Needs Triage' to: 'Needs Developer To Reproduce'

Philipp Oeser commented

2021-09-27 14:47:37 +02:00

Thx for the report!

Reading all this makes me think this is more of a request for performance improvement than a real bug.
Even if we run into a suboptimal, non-linear scaling of performance with multiple lattices or seeing the fact that multithreading is not utilized well, the tracker would still be the wrong place to handle this I think.
Unless there is something obviously wrong @antoniov? Usually this would be closed due to the following policy (but will let @antoniov decide):

Thanks for the report. While we do continue to work on improving performance in general, potential performance improvements are not handled as bug reports.
To improve performance, consider using less complex geometry, simpler shaders and smaller textures.

Thx for the report! Reading all this makes me think this is more of a request for performance improvement than a real bug. Even if we run into a suboptimal, non-linear scaling of performance with multiple lattices or seeing the fact that multithreading is not utilized well, the tracker would still be the wrong place to handle this I think. Unless there is something obviously wrong @antoniov? Usually this would be closed due to the following policy (but will let @antoniov decide): ``` Thanks for the report. While we do continue to work on improving performance in general, potential performance improvements are not handled as bug reports. To improve performance, consider using less complex geometry, simpler shaders and smaller textures. ```

Antonio Vazquez commented

2021-09-27 20:20:19 +02:00

Changed status from 'Needs Developer To Reproduce' to: 'Archived'

Antonio Vazquez closed this issue

2021-09-27 20:20:19 +02:00

Antonio Vazquez commented

2021-09-27 20:20:19 +02:00

We can close it because this is a feature request..not a bug.

Sign in to join this conversation.

No Label

Download

What's New

Blender Studio

Manual

Developers Blog

Documentation

Benchmark

Blender Conference

Development Fund

One-time Donations

GPencil: Multiple lattices reduces FPS #89349