i was wondering about the thread->block->grid thing, and the amount of parallel processed data.
assume i would know how many datapoints i have to process in a specific time. ( in my case a real time vst-plugin for audio processing)
so it would be something like every ~23ms (in case of 44,1KHz and a Blocksize of 1024 samples) i got new data.
also i have 2 channels. -> 2*1024 datapoints in <23ms.
based on that i want to know which GPU i need. (or any other case, its just an example)
you know what i mean??? or am i completly goin the wrong direction?