@neauoire it’s usually solved with a thread pool. Typically you’d spawn 1 thread per core at the start of your program and have them ready to go. Then for your map you divide the work into chunks.

If every item takes the same effort then for N cores you give each core 1/N items, and each thread does their share in a for loop. If the work is in uneven, you divide the work into 2-3x smaller chunks on a queue and the threads will grab as the go.