Subject: Re: hifn accelerator slows things down?
To: der Mouse <mouse@Rodents.Montreal.QC.CA>
From: Nathan J. Williams <firstname.lastname@example.org>
Date: 06/13/2005 15:57:22
der Mouse <mouse@Rodents.Montreal.QC.CA> writes:
> At first sight this seems surprising - an accelerator makes things
> slower? - but on reflection it's not all that impossible; presumably
> the syscall overhead cost is swamping the crypto speedup. (I assume
> the hifn would actually help something like IPsec that's entirely
> in-kernel and hence doesn't have to pay a syscall cost to talk to the
> accelerator chip.)
In addition to the usual syscall overhead, the /dev/crypto
implementation has to pay to copy the data into the kernel, perform
the operation, and then copy the result back out.