Re: Release 1.1 (beta) of my AES implementation

From: Mok-Kong Shen (mok-kong.shen_at_t-online.de)
Date: 06/29/03 

Date: Sun, 29 Jun 2003 22:36:00 +0200

Russ Lyttle wrote:
>
> Try using Cygwin, free from Redhat, on your Windows machine. You can then
> test your code using both VC and gcc. I usually test everything under gcc
> with at least -W -Wall -Wstrict-align warnings. Final released code should
> compile with no warnings.

I have gcc downloaded from DJGPP and VC++ from Microsoft.
On my gcc, I don't have -Wstrict-align. (Do you perhaps
have a newer version of gcc that has that option?) From
what others have suggested in this thread, I now use
-Wall -ansi -pedantic. Would that be o.k.?

> Your code does raise an interesting question. If it were optimized, then
> endian and other platform issues would be a bigger problem than they are.
> Most of us, I think, don't have unrestricted access to all the hardware
> necessary to properly test code. Posting to sci.crypt seems the best way to
> get code tested. But long discussions of code writing issues don't seem
> helpful here.
>
> I raise this issue because I'm revising my long neglected web page. I think
> I'll use one of my e-mail addresses to collect comments on code/language
> issues and ask that algorithm issues be posted to sci.crypt. Do you think
> that will work?

On the level of C/C++, I don't yet see any way of optimizing
my code thru using stuffs that are sensitive to endian-ness.
For the scheme of Daemen and Rijmen doesn't inherently
require shifts of words. One could certainly do shifts
and masking to get the individual bytes that are contained
in a word (this is endian sensitive), but directly
addressing the bytes is apparently better in my view
(anyway in the style with which I implement the scheme).
Perhaps you could elaborate your point a bit more,
if you don't agree with this.

Algorithmic issues that are of relevance to implementation
of crypto algorithms are certainly of interest to the
group. If you yourself have something concrete to discuss,
it's clearly fine if you would post it to the group.
Otherwise, your webpage could certainly have statements
soliciting comments/contributions in the sense you mean
in my view.

M. K. Shen

Relevant Pages

  • Re: linking c++ code with c library
    ... Or too many people who ignore facts which are inconvenient... ... I pointed out some cases that your algorithm would not cover. ... one posted that some obscure option of gcc would warn about this ... If you feel your original question hasn't been answered, ...
    (comp.lang.c)
  • Re: grumblegrumblegrumble *BLAM*
    ... It would depend how much had to be changed to make the algorithm work, ... how to unpack the tarball of the gcc source. ... David Cameron Staples | staples AT unimelb DOT edu DOT au ...
    (alt.sysadmin.recovery)
  • Re: my rwmutex algorithm for Linux...
    ... it should compile on any ... the native PThread read/write mutex vs. my algorithm. ... which was designed specifically for windows, is getting extremely similar performance numbers on most recent versions of Linux/Solaris/OpenSolaris with most recent version of GCC indeed! ... I test on Fedora 10/Most Recent OpenSoalris/Solaris/Windows XP Most Recent GCC.... ...
    (comp.programming.threads)
  • Re: Release 1.1 (beta) of my AES implementation
    ... >> Try using Cygwin, free from Redhat, on your Windows machine. ... >> test your code using both VC and gcc. ... >> endian and other platform issues would be a bigger problem than they are. ... Russ Lyttle ...
    (sci.crypt)
  • Re: /dev/urandom uses uninit bytes, leaks user data
    ... by using shifts instead of three divides, such as (gcc 4.1.2): ... There ought to be a warning about this sort of thing. ...
    (Linux-Kernel)
Quantcast