Efficiency of simple GPU matrix readback
Hello
I have the patch below, which I've obviously hashed together from bits of tutorials.
I've noticed that with a 1sec movie loaded into both sides, the slab/videoplane works fine, until I want to record my filtration.
When I connect the slab to the matrix in order to record the two movies filtered together, the result in the window visibly suffers. It's only slight, but noticeable and I have a great deal more I want to do, and then record.
I'm on a mac pro with 8 cores and 16gb of memory. I shouldn't be troubled by this but I am and it means that there is something with the patch/max performance setup which I'm not doing, but as I cannot record on the GPU, it seems as though as soon as I drag the result of a GPU op back to record, performance goes down.
Just had a thought as I typed this which I'll check tomorrow - it's not cos I'm not using uyvy on the matrix is it? Or maybe because I'm using a window/render/videoplane? In the latter would I get any performance benefit by flying blind and checking the result?.
if you want to record the output of your opengl scene you should be using jit.gl.asyncread.
imho, a better solution is a screengrab software like iShowU or Fraps.
or use an external device.
Robert thx v much for your post. I'd not seen asyncread in the module a-z in the manual cos it's not there... works a treat. although like you, i think ishowu is prob the way forward - thanks for that too - i'd not heard of it before, i was using a much worse equivalent.
efe - i'm interested to know what h/w you'd had in mind.
thanks v much!
Hello - I've put a patch together and noticed the result of the asyncread -> record produces a movie running at twice the speed - why is this?
Got the same problem as strontiumDog with jit.gl.asyncread, the .mov is a little bit faster than the original so it's impossibile to sync with the audio track.
Codec: photojpg
fps: 25
resolution: 720 480
any issue