[rc5] Mac v2 client confusion
Rebecca and Rowland
rebecca at astrid.u-net.com
Wed Jul 9 15:56:46 EDT 1997
>Date: Tue, 8 Jul 1997 08:29:36 -0700 (PDT)
>From: Corin Anderson <corin at cs.washington.edu>
>Subject: Re: [rc5] Mac v2 client confusion
>On Tue, 8 Jul 1997, Rebecca and Rowland wrote:
>> What's all this business about buffering and flushing blocks, and how do
>> you do it?
>My gut feeling on the Mac client is that it takes care of the block
>buffering for you. You get to tell the client how many blocks you want to
>buffer, and it takes care of the rest. I don't think that there's any way
>to run a second client with the equivalent of a -update flag. I also
>don't think that there's any _need_ to do so, either.
Thanks for that - but what do `buffering' and `flushing' mean?
>> Can I quit the client half-way through a run and not lose any work?
>It appears so. I quit the client, and it looks like it saved the
>partially completed block to disk.
>> What happens if the Mac crashes half-way through a run? Does the client
>> save work on disc?
>I don't know. Here's a test you can run, though: Run the client, and
>look to see if it creates and updates periodically the rc5-buff.in file.
>If it does, then it's checkpointing its work, and a crash wouldn't be
>catastrophic. If the .in file is created/updated only when you quit the
>client, then you're likely hosed in a crash.
It appears the .in file isn't updated periodically.
>> How do I tell the client to run a speed check, or execute any of the
>> documented run-time options?
>When you start the client, it runs a selftest, and calculates the key rate
>at the same time. Click once on the about window, and the little progress
>meter should show up. Also, while munching on keys, the key rate is
>displayed in the window.
The thing about this is that the key-rate displayed while it was doing its
self-test flitted between 5k and 7k keys/sec; never 6. It doesn't inspire
confidence. Is there any way to get an exact statement of the key rate as
is possible with the -m option using the vl1 client?
>As for the other run-time options, you can set these in the various
>settings dialog boxes (email address, port, etc).
>> And has anyone else had trouble with some of the screens taking a minute or
>> more to draw?
>Yes, I noticed this as well. If I select the Key Rate window, it would
>take about 30 seconds to get a response. No idea here.
Odd - is there an email address for bug and problem reports for the Mac
client? I'd quite like to suggest to the developers that they'd do well to
provide comprehensible documentation on how the clients work - I've been
waiting until I had a decent understanding of how the Mac v2 client works
and how to control it before writing my own attempt at documentation, but
I'm not sure I'll ever manage it.
>> FWIW, I'm running on a 68LC040 Performa 475, and the v2.000 client appears
>> to be no faster than the v1 client - the initial check showed a speed of
>> either 5k/sec or 7k/sec during the check, but didn't stop and given me an
>> overall exact rate at the end. The speed I got from the v1 client was
>I'm on a 7200/80. I saw a speedup from 94Kk/sec to ~143Kk/sec. Maybe the
>optimizations are only for the PPC chip?
It seems so - there's a report of someone using a 68040 Mac who's found
that the v2 client runs at about half the speed of the v1 client. I've
suffered a drop in speed from 6400 to 5400 keys/sec.
I'm also wondering if there's any way to tell if the client's reported back
that it's checked a block - I've had a look in the log file and seen
reports of fetching, starting, and completing blocks, but nothing about
sending the information back. Is this normal?
And I've not managed to figure out what the Mac client settings mean...
You know where it says:
Amount of time to spend working during each
... when the client is in the
... when the client is in the
What does it mean?
Thanks for your help so far,
To unsubscribe, send email to majordomo at llamas.net with 'unsubscribe rc5' in the body.
More information about the rc5