compiler construction - How to speed up compilation time in linux -

while compiling under linux use flag -j16 have 16 cores. wondering if makes sense use sth -j32. quesiton scheduling of processor time , if possible put more pressure on particular process other way (let have pararell compilations each -j16 , if 1 -j32?). think not make sense not sure not know how kernel solves such things.

kind regards,

i use non-recursive build system based on gnu make , wondering how scales.

i ran benchmarks on 6-core intel cpu hyper-threading. measured compile times using -j1 -j20. each -j option make ran 3 times , shortest time recorded. using -j9 results in shortest compile time, 11% better -j6.

in other words, hyper-threading little, , optimal formula intel processors hyper-threading number_of_cores * 1.5:

chart data here.