[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CAKtfLct5dhBK8ccFR9n2uUvexVtERnLrHXa-pZMa-xWbru_xxQ@mail.gmail.com>
Date: Tue, 27 Mar 2012 07:55:27 -0700
From: Alain Espinosa <alainesp@...il.com>
To: john-dev@...ts.openwall.com
Subject: Re: Research ideas.
On 3/22/12, Solar Designer <solar@...nwall.com> wrote:
> How does your code compare (e.g., in terms of instruction count) to
> what's found in x86-64.S and x86-sse.S? These were generated with my
> Perl scripts, then hand-edited a bit. For x86-64, gcc 4.3.x (only these
> versions) generate very decent code as well (maybe even slightly better
> than mine due to instruction scheduling across S-box boundaries), but
> for 32-bit x86's SSE2 and MMX so far the Perl-generated and hand-edited
> code is the best I have seen.
For x86 i let compiler do the job. For x64 there is the numbers:
----------------------------------------------------------------------
HashSuite_old HashSuite_new john-x64
----------------------------------------------------------------------
s1: 91 69(24%) 73(5%)
s2: 79 62(22%) 64(3%)
s3: 84 63(25%) 68(7%)
s4: 62 50(19%) 52(4%)
s5: 85 65(24%) 68(4%)
s6: 90 64(29%) 69(7%)
s7: 76 64(16%) 67(4%)
s8: 78 58(26%) 59(2%)
-----------------------------------------------------------------------
all:645 495(23%) 520(5%)
Also to note is that there is 1 xmm register free globally.
saludos,
alain
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.
