:: 27-Dec-1998 02:48 (Sunday) ::

Well, not much is new code wise. I wrote a little snippet so that the win32
“cli” client minimizes to the task bar, and (also a “RuleR” suggestion)
added long block benchmarking as a right-click option.

:: 06-Dec-1998 05:27 (Sunday) ::

Another week, another …, hmm, … er, um, never mind.

After a flurry of activity last weekend, including a from-scratch rewrite
of much of buffwork and quite a bit of bufupd, I bumped the version to 421.
Two more modules that are much cleaner (and saner) now.

I spent the first half of today doing things I hate to do but have to be
done anyway, and most of the rest of the day restructuring the flush code.
To reduce network traffic (theoretically to an eighth of what it used to be,
in RealLife(tm) to about a quarter), flush now coalesces blocks before it
sends them out.

Bovine is pushing me to get dynamic core selection working. Its a real pain
in the you-know-what as long as we are stuck with that ‘orrible bryd stuff.
Besides, I’m strongly against doing clib stuff from child threads (thats my
NetWare background thats talking there). Well, the infrastructure is in
place now – no limits on the number of threads, isolated thread management
routines, one-block-per-cruncher handling, clearly defined lines between
disk and network i/o, and so on and so forth. What it all boils down to
is simply having to knuckle down, buckle down, (and) do it, do it, do it…

More news from this member of the buffalo herd … next week…
Don’t go away…
We’ll be rrrright back…

:: 22-Nov-1998 16:31 (Sunday) ::

moo. Finally got around to cleaning up cliconfig. What a humdinger. I haven’t
implemented the changes in the problem loader or buffer handler yet, but will
do so later tonight.

-runoffline, -runbuffers, -frequent, -run, -n (blockcount=) are among the
options that have changed meaning or (as is the case with -run) are simply
shorthand combinations of other options.

The changes, from the users perspective, are minor:
a) blockcount can now be -1 which indicates ‘exit when buffers are empty
AND cannot be refilled’. This is similar to the behavior of the old
-runbuffers switch with the exception that it can also be used when
networking *is* enabled. In other words, blockcount=-1 is equivalent to
saying, ‘don’t ever generate random blocks’.
b) -runbuffers is now exactly the same as blockcount=-1. It does not
touch online/off-line mode.
c) -runoffline is now a pure network switch and won’t spontaneously combust
just because runbuffers happens to be on the command line as well.
-runoffline can can now be counterbalanced by -runonline.
d) -frequent has changed only in its description. What ‘frequent’ does is
not go online frequently as it had been previously described, but check
buffers frequently (to ensure that they are never totally empty). This
option is meaningless when running offline.
e) -run is obsolete and implies -runonline and -n 0. As you see, this is
not exactly the same as it was before inasfar as the -n switch is
concerned. However, the *effect* is the same since -run continues to
counterbalance the effect of runoffline=1 and runbuffers=1.

:: 20-Nov-1998 00:44 (Friday) ::

Rewrote -until, -h and hour handling today. All of them are ‘:’ now aware.
Whoever wrote that stuff originally was one heck of a lazy SOB. -until works
again (obviously).

I rewrote Do…/UndoCheckpoint() and some buffer code as well. The client
only uses one checkpoint file (for all contests together) now. I think I
forgot to mention (last week?) that the checkpoint time interval option
is now obsolete. The interval is now computed dynamically and is the greater
of one percent average change and one minute.

Fetches and flushes invoked from the command line are now correctly logged
to file/mail now. I’m sure that those odd-ball users who don’t trust stats
will appreciate this.

Time stamps are now tagged as being “UTC” rather than “GMT”. Besides being
technically correct, I expect that this will greatly expand the vocabulary
of the people who thought GMT stood for “General Mountain Time” (yes, that
is a real quotation).

I haven’t heard from dbaker regarding the potentially-new euro proxy yet.
From what I hear, he’s very busy counting finger requests.

Davehart has kindly accepted to redesign the buffer file format. Although
we won’t (technically speaking) need it for implementing POW, its an ideal
opportunity I’d rather not pass up.

:: 15-Nov-1998 22:44 (Sunday) ::

Didn’t do anything today. Everything seems to be working fine. I’m
thinking about releasing the DOS and Win16 clients tonight.

Yesterday’s news: Fixed the “cannot create console” problem some
win32 clients were reporting at 1024×768 and 640×480. Fixed the win32
restart bug. Speeded up the shutdown procedure a bit. Oh, yeah. Fixed
(what Silby termed a “minor bug”!) the ^2 of the time slice for DES
cores. That was real fun to watch on my NetWare machine.

Friday: worked with chrisb (rather he worked while I just lounged
around) on the x86 coprocessor board support he’s doing for RISC OS.
We giggled half the afternoon away until we finally decided that the
best solution would probably be to stop building RISC OS
clients and instead insist that RISC PC users buy x86 option boards
to run a win32 client running under a WINE binary built for Linux
under FreeBSD which would be running under BOCHS in a DOS emulator
in a Mac emulation which would have been previously spawned by an
XFree session under EMX that had been started by command.com run
by winoa386.mod in a winos2 session running under OS/2.

:: 13-Nov-1998 00:41 (Friday) ::

Completely restructured Client::Main() and ParseCommandLine() so that
the client object is completely initialized before anything else is done.
This has several advantages, two of which I’ve made use of right away:
(a) win32 and win16 clients know in advance whether they will be running
“modes” or not. This allows them to conditionally allow multiple
instances, ie any number of instances is permitted as long as only
instances is actually crunching.
(b) the implementation of -hide/-quiet for unix’ish clients. User use of
‘>/dev/null &’ is deprecated since the client may actually have
something to say when it starts up (such as choking on a bad command
line option). With -hide/-quiet, having the client say nothing is a
good sign and indicates that the client has started successfully.

I also rewrote much of rc5des.txt today. Specifically, I took out all the
platform specific stuff (now in separate readmes), took out all 10K of
the (mostly obsolete) FAQ, and created a “Help!” section with pointers to
the various sources of assistance available online or already built-in
to the client.

:: 08-Nov-1998 22:06 (Sunday) ::

I’ve been a busy beaver lately – too busy in fact to allow myself to be
frustrated by the packet loss problems I’ve been having.

The biggest changes I have implemented in the last days have been:
a) I drop-kicked the maximum number of CPUs/threads limit.
b) the client only runs one buffer per thread.
c) the client now displays the number of 2^28 blocks in the buffer files.

:: 28-Sep-1998 17:18 (Monday) ::

Back, after a trying tryst with an unyielding cast, this hitherto
.planless coder is not surprised to find much afoot at distributed.net.

…join us next week in an all-new episode as our heartful heroes
move on from the realm of ciphers and secrets to rulers and radians.