On Mon, May 15, 2023 at 5:29 PM Avichal Rakesh <arakesh@xxxxxxxxxx> wrote: > > TL;DR: Naively queueing up 0-length requests as fast as possible has the > consequence of delaying the entire stream by ~8ms, but does not seem to be > delaying frames individually any more than before. Changing the number of > 0-length requests has very little consequence on the stream, but increases > the chances of missed ISOCs. > Dan and Laurent, could you review https://lore.kernel.org/20230508231103.1621375-1-arakesh@xxxxxxxxxx/ when you get the chance? Thank you! Avi
