Sector interleave (was: H89 CP/M hard-secored boot diskette

From: Dwight K. Elvey <dwightk.elvey_at_amd.com>
Date: Thu Apr 3 18:35:35 2003

>From: "Fred Cisin (XenoSoft)" <cisin_at_xenosoft.com>
>
>> > I always wondered, if it was so easy to change the sector interleave on
>> > disks to make them load faster, why didn't the original
>> > programmers/designers optimize the interleave in the first place?
>On Thu, 3 Apr 2003, ben franchuk wrote:
>> It depends on your disk hardware and the speed of your computer.
>> 1) Computers got faster 2) Disk i/o got dma rather than poiled
>> i/o.
>
>It also depends on what kind of processinbg is being done to it during
>loading. For example, dumping file contents into RAM can be done a lot
>faster than loading a document into a word-processor.
>
>The fastest interleave occurs if the system can handle the next sector
>immediately after the previous. In that case, the fastest sequence is
>plain sequential, and a track can be read in a single revolution. But
>when you add extra processing of the data during reading, (such as a word
>processor loading a document), you could end up with NOT being ready for
>the next sector in time, resulting in the read taking as many revolutions
>of the disk as there are sectors per track. Alternating sectors might
>reduce that to two revolutions (or skipping 2 sectors, etc.), but that can
>easily be twice as long as necessary for a task that did NOT need the
>extraneous processing. Most system software developers DO think that
>they have optimized it. But they have optimized it for the type of file
>access that they EXPECT you to be doing (usually whatever THEY do most).
>
>BTW, it can be done either by placing the physical sectors in various
>orders (such as 1,3,5,7,9,2,4,6,8,10) OR can be done by a translation
>between logical sector and physical sector (physical sectors numbered
>1,2,3,4, . . . , but data that doesn't finish in sector 1 continues in
>sector 3, etc.)
>

Hi Fred
 As I recall, most things worked well with just 2 to 1 interleaving.
The major problem was how the H89 used the hard sectored. It was
guaranteed to wait one extra sector hole, regardless of the speed of
the software. This meant that consecutive sectors took an extra
revolution+ to find the next sector.
 I always figured that Heathkit continued to sell
the formatter with the 1 to 1 because it made the soft sectored
look much better when there really wasn't much difference between
the two. This was one of the big claims for upgrading to the soft
sectored. As I recall, there was formatter from HUG that was
like the one I had.
 I wrote my Forth because I couldn't justify a separate purchase
of HDOS( which I later acquired ). I wrote my own disk interface
to FIG's block access. That was when I noticed the slow access
and rewrote my formatter. It wasn't until later that I found
that all of the HDOS disk had the slow access as well. My
H89 was my second computer, after my Poly88.
 I did find that some BASIC programs worked a little better
with a 3 to 1 formatting but that wasn't typical. Even then,
writes and read were different.
Dwight
Received on Thu Apr 03 2003 - 18:35:35 BST