No subject

[Leo Caves]

Just to report our performance with Steve Huang's Lite On patches
for the tulip driver.
We used tulip.c v0.89H as this came with our RH5.2 boot disk,
and the patches (which came as context diff's only) seemed applicable.
(I took at look at current version 0.90 drivers and they seem to have
diverged from the code against which the diffs were generated and I
didn't feel that lucky)

Anyway to cut a long story to the short I performed the
netperf tests that were posted by Doug Eadline and we got
the following results.

I would like to add that prior to our fix, we had the ~70% of
"real" tulip performance as others have indicated.  We are now
up to something "respectable" and this should do us until Donald
Becker incorporates such changes into "standard" driver.

we have been happy with the 0.89H driver and I hope that this stability
is maintained with "the Fix"

Just thought I would report our findings.
Leo


NetGear FA310TX Lite On PNIC 169 revision
RH5.2/2.2.0
netperf-2.1pl3 (default build) 
tulip.c v0.89H
switched via Baystack 450
(Celeron 300A's running at 450)

roc9 netperf-2.1pl3: ./netperf -t UDP_STREAM -p 12865 -n 2 -l 60 -H roc10 -- -s 65535 -m 1472
UDP UNIDIRECTIONAL SEND TEST to roc10
Socket  Message  Elapsed      Messages                
Size    Size     Time         Okay Errors   Throughput
bytes   bytes    secs            #      #   10^6bits/sec

131070    1472   60.00      468861      0      92.03
 65535           60.00      468861             92.03

roc9 netperf-2.1pl3: ./netperf -t TCP_STREAM -p 12865 -n 2 -l 60 -H roc10
TCP STREAM TEST to roc10
Recv   Send    Send                          
Socket Socket  Message  Elapsed              
Size   Size    Size     Time     Throughput  
bytes  bytes   bytes    secs.    10^6bits/sec  

 65535  65535  65535    60.01      90.53   

Protein Structure Group  v: (+44) 1904 434521
Department of Chemistry  f: (+44) 1904 410519
University of York       e: caves@yorvic.york.ac.uk
YORK, YO1 5DD, UK