Update livesplat_realsense.py

[ardabck]

optimize matrix assembly by removing redundant reshaping and copying

[jasonjmcghee]

Was this AI generated? Nothing is being optimized here. The code is equivalent.

[ardabck]

haha no im a real person.
yeah functionally is equivalent but it was creating redundant arrays. which could be a problem with low spec devices.
this is not for CPU optimization. it is for memory optimization.

when gpus are better this could be a bottle neck at high FPS. so why not change it now?

edit: after checking who you are, i decided to do a real test to be 100% sure. it is way more probable that you are right :)

[jasonjmcghee]

I haven't memory profiled the code but wrapping in a np.array should be a no-op and the same reshape is happening. Afaict there's not a copy being replaced with an in-place operation.

Admittedly it's been a few years since I've worked with numpy but I would be blown away if inlining that variable (the core of the change) would reduce memory. It should just be a pointer to the same memory.

I've just been getting / seeing a number of ai generated prs that don't do what they claim so maybe I'm touchy / overstepping.