phc-discussions - Re: [PHC] Non-temporal writes and uninitialized memory

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite for Android: free password hash cracker in your pocket

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <CAOLP8p6XY_nO91MZ=uhcZZwLce4tD-SDwFk-rDzforE6w0WuEQ@mail.gmail.com>
Date: Fri, 18 Apr 2014 23:50:00 -0400
From: Bill Cox <waywardgeek@...il.com>
To: discussions@...sword-hashing.net
Subject: Re: [PHC] Non-temporal writes and uninitialized memory

On Fri, Apr 18, 2014 at 9:33 PM, Solar Designer <solar@...nwall.com> wrote:

> >     _mm_stream_si128(p++, value);
>
> Yeah, I had tried that too.  No luck.

Your explanation makes sense.  The 0's must already be in L3 cache, so
there's less of a penalty when doing the read-modify-write.

> > TwoCats currently has no method for writing to previously initialized
> > memory, so it's no help to me.  Some of the other entries, like Yescript
> > and Lyra2 should be able to benefit from it, but only in the second loop,
> > not in the first.
>
> When YESCRYPT_RW is set, yescrypt's second loop writes only to the same
> V_j that has just been read, so it's already in cache.  When YESCRYPT_RW
> is not set, yescrypt's second loop only reads.
>
> In the first loop, each page being written to has just been zeroed by
> the kernel, so it's in cache.
>
> Alexander
>

That makes sense.  Now that I think about it, Lyra2 also reads both write
destinations, so they will already be in cache.  I assume this is no
coincidence.  You're other trials would have been slower, and in tuning
this scheme probably won.

I think both Lyra2 and Yescript should be able to beat TwoCats bandwidth
with this approach.  My test code achieved 15GiB/s bandwidth with the
non-temporal writes with the 10 iteration outer loop.  Isn't this a crazy
name?  How can we remember non-temporal vs temporal?  Grr... Intel makes
the PHC authors sound like naming geniuses :-)

Bill

Content of type "text/html" skipped