2 ms·
That's still true with REP MOVS/REP STOS today in x86, it runs in an internal loop (and reads/writes cacheline-sized blocks if it can) while the rest of the CPU
by userbinator 7y ago
That's still true with REP MOVS/REP STOS today in x86, it runs in an internal loop (and reads/writes cacheline-sized blocks if it can) while the rest of the CPU can continue executing other instructions in parallel. You can achieve similar results with vector instructions, but those tend to take up a lot of icache and also fetch bandwidth.