On Thu, 2009-04-09 at 16:21 +0800, Herbert Xu wrote:
> On Thu, Apr 09, 2009 at 03:50:21PM +0800, Huang Ying wrote:
> > Hi, Herbert,
> >
> > I am working on GCM acceleration with Intel new PCLMULQDQ instructions
> > now. Can you tell me how to do GCM benchmark?
>
> We need to add the infrastructure for testing AEAD speed. It can
> be modelled after the cipher/hash speed testing code.
One issue here is that blkcipher testing infrastructure uses
local_bh_disable()/local_irq_disable() to get more accurate result. But
AEAD uses asynchronous interface, something like
wait_for_completion_interruptable() will be used in test path. How to
get more accurate result for testing asynchronous interface? Maybe it
can be: test many times, select the one with best performance. What's
your opinion?
> > BTW: Similar with AES-NI support, I use cryptd to defer requests in soft
> > IRQ context. Now, cryptd uses crypto_hash instead of crypto_shash as
> > child. I think we should change that. If you have not done that, I will
> > do that.
>
> We need to complete the conversion of all hash algorithms to shash
> first.
Yes.
Best Regards,
Huang Ying