Given the vagaries of multiprocessing jobs, it is likely that this post will only be useful to me. But here it is any way.
I tested compiling the 4.20 kernel using a range of processes. That is, the
-j flag for
make. Here are the results of the compilation time.
Although the advice is normally to compile with one more process than the number of cores you have, my compilation times bottomed-out when the number of processes was the same as the number of cores.
I also tested compilation with
-j 4 on a ramdisk, but there was (surprisingly?) no significant difference in time.