Pylon::Write::Writer(F, S)
Constants
DEFAULT_PARALLELISM = Platform.select do
macos do 2 end
linux do System.cpu_count.to_i * 2 end
end
APFS contends on staged writes past ~4 workers, 2 was the measured as being the most optimal. ext4 on Linux didn't seem to have the same problem so just leaving it to scale by CPU count for now.
PARALLEL_THRESHOLD = 16