parallel

#!/bin/bash

parallel() {
  local workers=$1 handler=$2 w i
  shift 2
  local elements=("$@")

  # loop once for each worker to launch
  for ((w=0; w<workers; w++)); do
    # fork worker with 1/5 of the total elements to handle
    for ((i=w; i<${#elements[@]}; i+=workers)); do
      # as each element for the worker completes, launch the next
      "$handler" "${elements[i]}"
    done &
  done

  # wait for all workers to finish
  wait
}

: <<'EOC'
say you have parallel 5 md5 *.txt, and 20 *.txt files
it will launch 5 loops SIMULTANEOUSLY. the first will handle files:
1, 6, 11, 16
second:
2, 7, 12, 17
third:
3, 8, 13, 18
fourth:
4, 9, 14, 19
fifth:
5, 10, 15, 20

each worker handles those IN ORDER. if file 1 finishes first, worker1 will
immediately start file 6. if file 4 finishes second, worker4 will then
immediately start file 9. each whole worker is launched in the background.
after all workers are launched, they are all waited on.
EOC