2014-12-04 15:56:47

by Herbert Xu

Subject: Re: [PATCH net-next] arch_fast_hash: avoid indirect function calls and implement hash in asm

On Thu, Dec 04, 2014 at 02:08:50PM +0100, Hannes Frederic Sowa wrote:
> By default the arch_fast_hash hashing function pointers are initialized
> to jhash(2). If during boot-up a CPU with SSE4.2 is detected they get
> updated to the CRC32 ones. This dispatching scheme incurs a function
> pointer lookup and indirect call for every hashing operation.

Just curious, is jhash actually faster than generic C CRC32 on
common platforms?

Cheers,
--
Email: Herbert Xu <[email protected]>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt

2014-12-04 16:37:15

by Hannes Frederic Sowa

Subject: Re: [PATCH net-next] arch_fast_hash: avoid indirect function calls and implement hash in asm

On Do, 2014-12-04 at 23:56 +0800, Herbert Xu wrote:
> On Thu, Dec 04, 2014 at 02:08:50PM +0100, Hannes Frederic Sowa wrote:
> > By default the arch_fast_hash hashing function pointers are initialized
> > to jhash(2). If during boot-up a CPU with SSE4.2 is detected they get
> > updated to the CRC32 ones. This dispatching scheme incurs a function
> > pointer lookup and indirect call for every hashing operation.
>
> Just curious, is jhash actually faster than generic C CRC32 on
> common platforms?

Yes, jhash always beats crc32 in software on x86_64 and ia32.

Bye,
Hannes