Does it help? 2.5d view on OpenGL 3.0

[klayoutmatthias]

In order to deal with a number of reports I received with respect to non-functional 2.5d views, I ported the view to GLSL 1.20 and removed the geometry shader. It should be compatible with OpenGL 3.0 now.

I have some hope this fixes the issue on MacOS.

The code is in the d25view-fixing branch (https://github.com/KLayout/klayout/tree/d25view-fixing).

It could use some memory optimization, but first I would like to have some feedback.

If you want to help, please give it a try and let me know if your 2.5d problem is fixed.

Thanks and best regards,

Matthias

[Kazzz-S]

1. macOS

Hello @klayoutmatthias,

Thanks a lot for trying to fix this issue.
I built the d25view-fixing branch on macOS Monterey (Intel Mac).

First, I got a syntax error relating to ";" but you have already fixed it.

Next, I got some compilation errors at runtime.

I've modified the code as below.

Then, I could successfully get the 2.5D view. (Also refer to #1569 and #1666)

T1700.mp4

Best regards,
Kazzz-S

[Kazzz-S]

cont.

The above application bundle built on Monterey can also run on Ventura and Sonoma.

Kazzz-S

[Kazzz-S]

2. Linux Mint 21.3

Using the same source code as macOS...

Kazzz-S

[stefanottili]

Works on M1 macOS Sonoma too.

[klayoutmatthias]

Very good. Thanks for the patch (interesting, how different the GLSL implementations are) and the feedback!

I am happy to hear it works on MacOS too - OpenGL really is weird.

I will include the new version in the next 0.29.1 release (tomorrow, hopefully). From the memory perspective, I conducted an experiment with about 20M shapes (boxes). During 2.5d rendering with some booleans and mesh lines, memory goes up to 14G, but I hope that is an extreme case. As I was removing the geometry shader, I now need to supply normal vectors for the lighting and this doubles the vertex memory. It is possible to reduce the footprint by using indexed buffers, but I first need to learn how to use them properly. I spare that for a future release :)

Best regards and thanks again,

Matthias

[klayoutmatthias]

My 20M rectangle test case (sluggish ...)

[ejprinz]

Tested on RedHat EL8. Works fine.

[klayoutmatthias]

Closing, as it is released. Thanks for your support!