🚀 go-pugleaf

In article <1993Jan8.063123.5194@qualcomm.com>, karn@servo.qualcomm.com (Phil Karn) writes:
|> In article <1ihdcmINNslg@fbi-news.Informatik.Uni-Dortmund.DE> muenx@heike.informatik.uni-dortmund.de (Holger Muenx) writes:
|> >  Is it possible for a software solution to reach an encryption rate
|> >of about 100kB/sec (read 100 * 1024 Byte not 100 * 1024 Bit per second)
|> >on a typical PC (say, Intel 386DX 33MHz)?
|> 
|> Well, my DES code (ucsd.edu, /hamradio/packet/tcpip/crypto/des.tar.Z)
|> when compiled and run on a 486DX-50 using GCC under 386BSD encrypts 1
|> million 64-bit blocks (8 megabytes) in 90.5 seconds of CPU time.
|> That's not quite the rate you want, and on a faster machine too.  Oh
|> well.
    ~88k/sec on 486DX-50

The DES library/application I had posted in comp.sources.misc a while
back could do 67k/s on a 386/33 using a non gcc compiler (but still
in 32bit mode - I did not do the testing).
This package uses standard boring C with few evil tricks (it compiles under
16bit, 32bit and 64bit machines with no problems or endian dependencies).
With 386 assembler it would probably be possible to beat 100k/s on a 33Mz
machine.

In 16bit MS-DOS the library does 27k/sec (Turbo C v 2.0 :-).
There was a DES encryption program posted
to comp.binaries.ibm.pc that was twice as fast as my code (~50k/sec).
DES can be fast if you have either a good C compiler or resort to
assembler.

The most recent version of the library can be found in pub/DES/libdes.tar.92-10-13.Z on ftp.psy.uq.oz.au (130.102.32.1)
The main difference between this version and the one in c.s.m is that
it will compile on 64bit machines.

BTW a DEC Alpha 500 APX I played with does 830kbytes/s in cbc mode.
From what I remember seeing in this newsgroup, DES hardware often does
~20mbit/sec.  The top end Alpha, with assembler should be >= 8mbit/sec.

Hardware giving only a 2.5 speedup?
I get the feeling it should be possible to produce much faster DES
chips.... :-)

eric