UpdateFcn performance

[dragozarkovic]

Hi Author,

First of all, just to say how amazing work you have done in both of the projects. It is really helpful.

What I have noticed that UpdateFcn is slower than I expected. Orbiting in glViewer3D is perfect, I have tested the engine in much more triangles, and it works perfect. But, initial start seems to be more or less slow, at least that is my impression. Also, if I for example add update of coordinate origin on mouse movement, to update every frame, it is approximately 200-300fps, on my Ryzen 5600H, GeForce RTX 3050.

Is there any space for improvement in updating of buffer process?

Thank you verry much for the response. I would like to get in touch with you regarding the topic.

Best regards,
Drago

[fr0nkk]

Thank you for the kind comments!

For the performance issue, is it with the matogl toolbox while making your own renderer? Or is it while using FastVisualization?

In any case, it would really help if you could provide a working sample in which the performance issue is present. It would enable me to either fix some slower parts in my code or even suggest improvements to make your code run faster.

Have a nice day!

[dragozarkovic]

Thank you for your prompt response.

I want to achieve fluent cursor interaction over the graphic, with certain measures or IDs of the objects appearing in the 3D text and highlighting the objects below the cursor. Similar to the video below.
https://youtu.be/SRKRllM5bGY

For the sake of simplicity, I have modified example 5 of the Matogl library to present a number of points, one of which follows the cursor position, and the others are static.
So, depending on the size of the buffer, the point following the cursor is perfect for a number of points, something like 100k, but quite slow for 1M of points. You can edit the MAX_POINTS_NUM at line 73 of the file.

Profiling the problem pinpointed the "problem" to be in the javabuffer function. But I am sure you can confirm that easily.

I would appreciate it if you were open to a discussion on the topic, as I am convinced that you are the best person to provide advice on this.
You can find my email contact in the profile.

Best regards, and thank you in advance.
Drago

glExample5_modified.zip

[fr0nkk]

You are right, javabuffer is the culprit. In this context, there is no real workaround to transfer a large amount of data into a OpenGL buffer at each frame update.

If it is possible, what you could do is to divide the data into multiple buffers, and update only the necessary ones.

If the data does not need to be refreshed at each frame, you can write to the buffer only once to avoid the overhead of writing the same data at each frame.

It may also be possible to write a shader to handle the render of the data, as updating shader uniforms is a lot faster than updating buffers.

This toolbox is made to interface directly with OpenGL. If you are less familiar with OpenGL, I suggest (if you have not already) looking into my other toolbox FastVisualization. It is built from matogl but is a lot more user friendly to visualize and interact with all kinds of 3D data!

What kind of render are you trying to make? Maybe I could make a quick working example for your needs in FastVisualization

Best,