[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Fri, 14 Sep 2012 19:21:27 +0200
From: magnum <john.magnum@...hmail.com>
To: john-dev@...ts.openwall.com
Subject: Re: intrinsics: speed up for linux-x86-64-native
On 14 Sep, 2012, at 15:08 , Aleksey Cherepanov <aleksey.4erepanov@...il.com> wrote:
> Looking over sse-intrinsics.c I noticed weird thing: multiple
> MD5_PARA_DO cycles when it is possible to write one cycle over
> everything and avoid use of tmp variable. I tried to avoid some cycles
> and got a speed up. But when I merged them into one cycle per MD5_STEP
> I got a significant slowdown.
It's not very intuitive but AFAIK it was made that way on purpose: The intention is to get code that hides latency, much like GPU coding. This is pretty compiler-dependant and that is why icc can make such a great difference.
That is about all I know so I'll leave the details to the experts. I hope someone can give a more thorough explanation - I'd read it with interest.
magnum
Powered by blists - more mailing lists
Confused about mailing lists and their use?
Read about mailing lists on Wikipedia
and check out these
guidelines on proper formatting of your messages.
