Rijndael in Assembler for x86?

[Sidney Markowitz]

A little over a month ago Perry Metzger asked about free assembler
language implementations of Rijndael for x86. Helger Lipmaa, whose
commercial assembler language version seems to be the fastest,
mentioned Brian Gladman as having the best free C implementation.
Gladman's web page now says that he has a free assembler language
version. For comparison, Lipmaa says that his runs at 230 cycles per
block, Gladman's C version runs at 360 cycles per block, and Gladman's
assembler language version runs at 300 cycles/block. It is not yet a
complete implementation. Here's a quote from the website:

"Here is a preliminary version of this code in assembler for the
Pentium family with MMX (Pentium II/III/IV). This only implements the
standard block size of 128 bits but is 15-20% faster than the C code.
It achieves a maximum speed with a fully primed processor cache of
about 300 cycles/block, which is around 50 Mbytes/second on a 1GHz
processor.  This version has not been extensively tested so please be
aware that there may be bugs in it. Note also that it uses the
Microsoft VC++ register saving conventions and may hence need changes
if used with other C/C++ compilers."

The URL is
http://fp.gladman.plus.com/cryptography_technology/rijndael/


 -- sidney





---------------------------------------------------------------------
The Cryptography Mailing List
Unsubscribe by sending "unsubscribe cryptography" to majordomo at wasabisystems.com