Ascend Archive
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: (ASCEND) Software compression ?? (fwd)



On Wed, Apr 15, 1998 at 12:55:43PM +0300, Vesa Halkka wrote:
> 
> > Seriously, the Ascends are doing LZS in hardware using the appropriate
> > Hi/Fn (Stac) chips. The Win95 DUN implementation is pure software, so
> > they call it "software compression".
> 
> DUN uncompression is reasonably well coded, and a Stac chip would be an
> overkill in workstation use. 

Of course. LZS decompression is easy, I've written one myself. The hard
part is LZS compression. But the typical PC processor today is easily
capable of doing some such streams in parallel. The hardware is only
necessary on machines which do a lot of parallel streams and have a CPU
not suited for this (Max) or have a very limited CPU anyway (P50). The
Stac performance of a P50 (uses HiFn chip) is much better than of a Cisco
1003 (same processor, but uses software for Stac).

> My old 60MHz Pentium uses about 20% CPU to a compressed 128k ISDN ftp
> download. Passive Siemens chip ISDN card (Teles). 

I'd even guess that the 20% are mostly not going to de-LZS but rather
to Win95s immanent overhead of whatever it is constantly doing. The
more interesting thing IMHO is,

1) Does DUN actually do LZS _compression_ ?
2) If yes, how well ?
3) If well, how much CPU eats that ?

I'd guess it does and even does well because M$ just licensed the Code
from HiFn (Stac) and probably even took the "hand optimized x86 assembly
code" HiFn provides (see www.hifn.com). But would still be interesting
how much cycles _this_ eats ;)

-- 

Kanther-Line: PGP SSH IDEA MD5 GOST RIPE-MD160 3DES RSA FEAL32 RC4

+-o-+--------------------------------------------------------+-o-+
| o |               \\\- Brain Inside -///                   | o |
| o |                   ^^^^^^^^^^^^^^                       | o |
| o | Andre' Beck (ABPSoft) beck@ibh-dd.de XLink PoP Dresden | o |
+-o-+--------------------------------------------------------+-o-+
++ Ascend Users Mailing List ++
To unsubscribe:	send unsubscribe to ascend-users-request@bungi.com
To get FAQ'd:	<http://www.nealis.net/ascend/faq>


Follow-Ups: References: