Hi, > This is horribly non-portable. How much of a performance difference > does it make? How does it do on ppc64 systems where the cacheline > size is not 32? Yes, if anything we should catch cacheline aligned, multiple cacheline sized zeroing in memset. Anton