tee has a simple main loop: while not EOF: (synchrounously) read up to BUFSIZE bytes from stdin into buf for each output file: (synchrounous

tee has a simple main loop: while not EOF: (synchrounously) read up to BUFSIZE bytes from stdin into buf for each output file: (synchrounously) write buf to file (source code here: http://git.savannah.gnu.org/cgit/coreutils.git/tree/src/tee.... http://git.savannah.gnu.org/cgit/coreutils.git/tree/src/tee....) All of the reads and writes are done through synchronous file apis. I don't see where you are getting one-thread-per-file (or any other kind of parallelism) there.