Date   : Sun, 18 Dec 2011 07:57:04 +0100
From   : rick@... (Rick Murray)
Subject: 32016 + 32082

On 17/12/2011 23:00, jgh@... wrote:

> and has lots of code to work around problems with other systems
> trampling over resources they shouldn't.

Are you going to name & shame any of the offenders? :-)

[albeit shame two decades later is a hard sell]


> I'm currently carefully going through the code doing some tidying up
> and optimisations, and better commenting it.

Plus support for BigStuff which may have been rather less of a factor 
when HADFS was first devised. After all, a fully loaded FileStore can 
only hold 240Mb (four E60S's), so back in the day 16Mb would have been 
plenty.
In fact, the only reason I'm looking to set up a BigStuff partition on 
my GoMMC (when I install it) is because a tiny partition on a 2Gb card 
is... a bit ridiculous. I can't actually think I'd have anything even 
remotely near enough data to fill a big partition.


> I'm about halfway through, but the slowness is that I'm testing it
> for about a week for each change I make.

Indeed, that's the fun of writing a filesystem. Do you have a test suite 
to help verification go more rapidly?


> but the i86 MOV is actually LD.

Yes, I noticed. But on the other hand the ARM is a load/store processor, 
so it would make a specific distinction here.


> And who on earth thinks to use CMP #0,R0 instead of CMP R0,#0 ???

Well, I guess you are comparing zero with R0, instead of R0 with zero? 
It might have more effect in non-zero comparisons with how the flags are 
set - is R0 greater than #42? Is #42 greater than R0?


Best wishes,

Rick.

-- 
Rick Murray, eeePC901 & ADSL WiFI'd into it, all ETLAs!
BBC B: DNFS, 2 x 5.25" floppies, EPROM prog, Acorn TTX
E01S FileStore, A3000/A5000/RiscPC/various PCs/blahblah...