Re: [libvirt] [PATCH] Replace hashing algorithm with murmurhash
On 01/25/2012 09:55 AM, Eric Blake wrote:
>>> +
>>> + return le32toh(r);
>>
>> <endian.h>, and thus le32toh(), is not yet standardized (although POSIX
>> will be adding it in the future), nor is it currently provided by
>> gnulib. We'd have to get that fixed first.
>
> The le32toh call was only here because the code I copied wanted to be
> endian neutral. I don't think libvirt really cares if its hash codes
> are endian neutral, so I trivially just removed the le32toh call and
> avoid the problem of endian.h
Agreed - we aren't sharing hash values over the wire, so all hash values
within a particular libvirtd process will be the same endianness,
without having to waste time on swapping bytes around.
Actually, the more I think about this, the more I have to wonder: Does
the incoming alignment affect the output hash? That is, if I do
int i;
char array[12];
for (i = 0; i < 4; i++) {
strcpy(array + i, "12345678");
printf("%x\n", (int) virHashStrCode(array + i, 0));
}
do I get the same values for all four iterations, on both little- and
big-endian architectures? If not, then the byte-rearranging really is
important to the algorithm (that is, the algorithm is operating on
4-byte quantities, but must build up those quantities from 1-byte
quantities regardless of starting alignment, so endianness looks like it
plays a role in doing the conversion correctly).
--
Eric Blake eblake(a)redhat.com +1-919-301-3266
Libvirt virtualization library http://libvirt.org
Attachments:
- signature.asc (application/pgp-signature — 620 bytes)