You've already forked torvalds-GuitarPedal
mirror of
https://github.com/torvalds/GuitarPedal.git
synced 2026-08-13 12:36:34 +00:00
f478e2de89
process_output() returns immediately when "USB L/R Out" is None, and otherwise stores a frame into a ring with a release barrier - every sample, on the audio core, whether or not a host is listening. Measured across the four modes inside one boot, empty chain, nothing else moving: None 9.263 % Wet 9.605 % +0.343 % Dry (default) 9.569 % +0.306 % Wet/Dry 9.611 % +0.348 % Small, and still big enough to be mistaken for something else. Two runs an hour apart differed by 0.37% of the sample period and got as far as a filed issue guessing at browsers and USB polling; what had actually changed was that this setting had been moved in the web app in between. The earlier session reads as Dry and the later one as None, which is the whole of it. A difference between two effects cancels the tap. An absolute load does not, and comparing absolutes across firmwares is what this tool is for - so a run pins it to None and lets the closing program change put the scene back, exactly as it already does with the routing. What gets printed is therefore what the chain costs with the USB tap deliberately excluded, and the header says so. Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>