Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Wed, 12 Aug 2015 01:32:21 +0200
From: magnum <>
Subject: Re: Lei's weekly report #15

On 2015-08-11 20:39, magnum wrote:
> On 2015-08-10 18:27, Lei Zhang wrote:
>> I'm still struggling to make episerver work with the SIMD SHA1/2
>> functions. Magnum, could you take a look at my code?
>> <>
>> I've been tweaking this file hard, but there's still something wrong.
>> Maybe you can notice some obvious mistake in it (?)
> Attached is a diff against your repo that fixes the problems. There were
> a few minor (but very effective) bugs and I did a few other changes for
> clarity. Try to understand the bug fixes fully.
> Also, now you should move the "write salt to SIMD buffer" from
> crypt_all() to set_salt(). I believe that will boost the "Many salts"
> benchmark figure a lot.

BTW some other problems remain:

1. You use NBKEYS assuming that SIMD_PARA_SHA1 and SIMD_PARA_SHA256 are 
the same value. They may not.

2. The output of SIMDSHA1body() and SIMDSHA256body() differs in size, 
but your cmp_* and get_hash_* functions assume SHA-256. You either have 
to re-map SHA-1 to SHA-256 format after calling the function, or make 
these functions act differently depending on cur_salt->version.

3. You should set ALGORITHM_NAME to reflect use of SIMD.


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.