The disassembly of the nrand48() routine (Solaris 10, AMD64) shows that it is protected with a call to the lmutex_lock routine. Replacing the nrand48() call with a call to my own random-number generator (with per-thread private state) shaves off 33% of running time just with two threads.
Since nrand48() uses an application-provided buffer, instead of a global shared buffer, is acquiring this mutex really necessary? This message posted from opensolaris.org _______________________________________________ perf-discuss mailing list perf-discuss@opensolaris.org
