Re: EE Times on IDEA

[Eric Young <eay@mincom.oz.au>]

On Sat, 28 Oct 1995, Anonymous wrote:
> Electronic Engineering Times
> Oct 23, 1995 p. 66
> 
> T.R MADHUSUDAN SASTRY,
> T. GANESAN, B.MADHUKAR
> AND N. SRINIVASA

> involved. The current implementation of Idea in ECB on Motorola's DSP
> 56166 running at 60 Mhz supports up to 625 kbits/second in full-duplex
> mode. That is 3.6 > times faster than using the DES algorithm. 

That last statement is rubbish.  For most CPU's I've seen, IDEA is either 
the same speed as DES or a bit slower.  There is no way that it is 3 
times faster that single DES.

		DES cbc		IDEA cbc
DSP 56166	 ???		  78 k/s	(625 kbits/sec)
486DX/50	 274 k/s	 177 k/s 	Solaris 2.4
Pentium 90	 703 k/s	 605 k/s	Linux
'old' R6000 box  295 k/s	 283 k/s
'old' Alpha box	1300 k/s	 690 k/s	(no 64 bit optimisations)
88100		 558 k/s	 526 k/s
PA-RISC		1072 k/s	 616 k/s	(very compiler dependant)
R4400 200mhz	1600 k/s	1290 k/s
Sparc 10	 695 k/s	 679 k/s

All speeds are in bytes/second.  The DES implementation is a 'standard'
fast DES (there are at least 2 DES packages on the net with this type of
speed) and the IDEA implementation is a 'standard' IDEA implementation
(similar speed to the reference versions) Both written in C.  I have
looked at how I could optimise IDEA but it does not lend it's self to vast
improvements the way DES does.  It may be possible to speed up IDEA by
%20 with good C compilers but that will still only be the speed of DES, I
do not think IDEA should be 'sold' as being faster than DES is software. 
It is not.  It has over twice the key length (128 vs 56) and that is it's
main benefit.

eric (having a rant at mis-information).
--
Eric Young                  | Signature removed since it was generating
AARNet: eay@mincom.oz.au    | more followups than the message contents :-)