Received Framebuffers of two modes are the same moment?

[Counterfeiter]

Hello,

I try to process raw frames of the IMX290 sensor as fast as possible on a raspberry PI 4. I use the python picamera2 wrapper for this task.
The fastes processing I could get at the moment:
Record a raw video, read the video file back and send the single raw frames over multiprocessing queus to some worker processes, while a new Video is recorded. I get about 6 fps on 4 CPU cores with some extra blur checking processing.
I could get slidly more speed if I register two image formats: full raw and main with reduced resolution size and only 8 bit coding. So the processing could be done on the "main" image.

But how can I be sure that I have exaclty the same exposure moment? Are there any drawbacks and drops if I reach a limit?
I talking about this implementation:
def capture_buffers(self, names=["main"], wait=None, signal_function=None):

My current script with the raw video buffer version is here:
https://github.com/Counterfeiter/NotSoLuckyImaging

Best regards

Sebastian

[kbingham]

Hi Sebastian.

As this example is using picamera2, I suspect it might be best to ask questions over there. (https://github.com/raspberrypi/picamera2/issues)

Any pair of streams that complete in the same request are guaranteed to be of the same 'exposure time'. But what you receive depends on how picamera2 is implemented.