[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Fri, 11 Nov 2011 16:31:17 -0600
From: "jfoug" <jfoug@....net>
To: <john-dev@...ts.openwall.com>
Subject: RE: more targets using sse-intrinsics.S
I will make a thin for pix, and start on HMAC MD5.
Older revisions of raw-md5 are pretty good places to look, along with the raw-md5.c in the original intrinsics patch. Most of it can be done in the GETPOS macro, and require very little 'code' changes. Md5_gen already had arrays of MMX_COEF items, since I found it was faster to load a dozen or so of them (actually, 32 or 64 for mmx_coef=2 was a good count). Since that double dim array was already setup, it made hooking in the intrinsic code very easy to do. It is a little slower, in loading.
The way the raw-md5 (in the intrinsics patch) did it, was to have a single array. That array was MMX_COEF*SSE_PARA in size, and the GETPOS tied it all together. For formats which do not have a double array, I think going back to that design, is likely the best way.
Jim.
>From: magnum [mailto:john.magnum@...hmail.com]
>
>2011-11-11 22:05, jfoug wrote:
>> The big difference, is we will likely need to change to an array of
>> SSE buffers, vs a single buffer of MMX_COEF. We need SSE_PARA buffers
>> of MMX_COEF. The changes are not too hard, but DO take some time and
>> finesse. One of the formats (pix-md5), we should simply make thin,
>> and link to dynamic_19, which already has x86, SSE2 (mmx-sse) and
>> SSE2i (sse-intrisics). The other formats will have to be tweaked to
>> properly use sse-intrisics.
>
>OK... so let's do them! Piece of cake (lol). I can start with the
>MS-SQL's as I am aquainted to them from the Unicode work.
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.
