[dpdk-dev] cuckoo hash in dpdk

Pragash Vijayaragavan pxv3620 at rit.edu
Wed Aug 23 22:39:46 CEST 2017


And yea, mod will take only 1 Clock cycle that way :)

Also if divisor is a power of 2,  a / b  can be done like

 ( Given a is unsigned, a > b and since our hash values are greater than 0,
a is our hash value, b the power of 2)

while(b != 1) {
   a >>= 1;
   b >>= 1;
}

 which is less clock cycles than normal division i guess :)




Thanks,

Pragash Vijayaragavan
Grad Student at Rochester Institute of Technology
email : pxv3620 at rit.edu
ph : 585 764 4662


On Wed, Aug 23, 2017 at 4:20 PM, Pragash Vijayaragavan <pxv3620 at g.rit.edu>
wrote:

> Hi ,
>
> The performance will depend on the time taken for calculating the hash1
> and hash2 values for each lookup.
>
> Can i know which hash functions are used to calculate the hash values for
> each incoming key. We use CRC32 which uses xxhash i guess.
> I could not find the implementation of the hash functions.
>
>
> Thanks,
>
> Pragash Vijayaragavan
> Grad Student at Rochester Institute of Technology
> email : pxv3620 at rit.edu
> ph : 585 764 4662 <(585)%20764-4662>
>
>
> On Wed, Aug 23, 2017 at 2:28 PM, Dumitrescu, Cristian <
> cristian.dumitrescu at intel.com> wrote:
>
>>
>>
>> > -----Original Message-----
>> > From: dev [mailto:dev-bounces at dpdk.org] On Behalf Of Pragash
>> > Vijayaragavan
>> > Sent: Wednesday, August 23, 2017 3:16 PM
>> > To: dev at dpdk.org
>> > Cc: Minseok Kwon <mxkvcs at rit.edu>
>> > Subject: [dpdk-dev] cuckoo hash in dpdk
>> >
>> > Hi,
>> >
>> > I got the chance to look at the cuckoo hash used in dpdk and have a
>> query.
>> >
>> > would using division and modulo operations be slower than bitwise
>> > operations on RTE_HASH_BUCKET_ENTRIES, specially since
>> > RTE_HASH_BUCKET_ENTRIES is a power of 2.
>> > For example, to do a modulo we can do a "AND" operation on
>> > (RTE_HASH_BUCKET_ENTRIES - 1), which might be faster. We did a cuckoo
>> > filter for VPP and doing this gave a slight improvement in speed.
>> > Is there any particular reason its done this way.
>> >
>> > Sorry if i am being wrong in any way, i was just curious.
>> >
>> > Thanks,
>> >
>> > Pragash Vijayaragavan
>> > Grad Student at Rochester Institute of Technology
>> > email : pxv3620 at rit.edu
>> > ph : 585 764 4662
>>
>> Bitwise AND typically takes 1 cycle on any CPU, while modulo takes dozens
>> of cycles.
>>
>
>


More information about the dev mailing list