On 03/29/2014 09:12 AM, 陈梁 wrote:
> /* word at a time for speed */
> if (!res) {
> while (i < slen &&
> (*(long *)(old_buf + i)) == (*(long *)(new_buf + i))) {
> i += sizeof(long);
> zrun_len += sizeof(long);
> }
>
> /* go over the rest */
> //while (i < slen && old_buf[i] == new_buf[i]) {
> // zrun_len++;
> // i++;
> //}
> }
> i += sizeof(long);
> nzrun_len += sizeof(long);That does not produce a minimal compression (it treats all 8 bytes as different, even if 7 of them were the same). It might be a viable solution if the extra overhead of the additional bytes sent over the wire is less than the overhead saved by not re-checking the temporary variable to determine a better compression. But I'm not convinced - it seems that reading memory into a register, then doing multiple operations on that cached value, will be a win (that is, minimal compression matters, because time spent transmitting bytes over the network is slower than time spent calculating how to avoid bytes to transmit). -- Eric Blake eblake redhat com +1-919-301-3266 Libvirt virtualization library http://libvirt.org
signature.asc
Description: OpenPGP digital signature
