[Coco] System gen problem - SOLVED!!

[Steven Hirsch]

On Thu, 11 Feb 2010, Aaron Wolfe wrote:

> Just to be clear, there is no reason anyone cannot continue to use the
> version they have now if they prefer it.  DriveWire 4 does not enhance
> the disk drive, RTC or printing capabilities found in DW3.  If the
> other capabilities added in dw4 are not interesting to someone, I
> strongly suggest they stick with the software they have now and for
> the foreseeable future.  DW3 works very well.

Great - thanks for the info.

>
>> Every time I've coded anything in Java, it just seems to take five pages of
>> source to do even the simplest things.  For portability, I vastly prefer
>> Perl and have written some fairly complex applications in that language that
>> truly are "write-once, run everywhere".  But that's just me, I guess, and it
>> doesn't diminish my admiration for the work you guys have done one iota!
>>
>
> I like Perl quite a lot myself and have used it in some large
> projects.  However getting Perl working on Windows would be more work
> than I think most Windows folks are willing to do just for drivewire.
> Unless you're planning on contributing to our work, I think the
> verbosity of Java source won't be an issue in this case.

It's neither here nor there, but ActiveState Perl is simple to install.

At any rate, pay no attention to my rantings about Java.  They are perhaps 
based on compatiblity issues that are years in the past, but which left 
such a bad taste in my mouth that I've avoided it ever since.  I earn my 
living writing scientific and engineering applications that couldn't 
possibly be written in Java for performance and memory-footprint reasons, 
so I've never had any reason to go back to it again for a second look.

Our code is so finely tuned for performance, that we've actually gotten 
down to looking at the generated ASM code to figure out what the compiler 
was doing with certain constructs!  Memory footprints in excess of 200GB 
are not at all uncommon, and you really do need to fret over all those 
things about which they tell you "oh, just let the compiler deal with 
it". Cache hit patterns and TLB thrashing can make or break you on huge 
data sets.  Not to digress, but there is a great presentation on this 
subject written by one of the folks involved in optimization of JIT Java 
runtimes.  Wish I had the URL handy.  It's more of a problem than folks 
often realize.

>> Aaron sent me some input on getting it going, so I'll give it a try.
>>
>
> side note, sorry: did you receive my email containing the text of all
> the messages I tried to send before?  something seems to be broken in
> your mail flow, just wanted to make sure you're getting my messages.
> send me a note off list either way.

Yes, thanks!  Don't have any idea what happened to the earlier traffic.

Steve


--