> >Cache makes your drive much slower. My Sunrise IDE:
> >w/o cache: 230 kbps
> >w/ cache: 50 kbps
>
> How did you test this? By just testing disk I/O time?
Yes. And you should have quoted (or at least read) the rest of my message
too.
> Well then you probably
> don't understand caching: (I can hardly believe it, me explaining you
> something :-) )
I already know how cache works. Otherwise I wouldn't use it, don't you
think?
> present in the cache - is needed again, cache is just a waste of time.
> Cache doesn't make I/O faster, but reduces the times the slowest I/O is
> needed. So in practice you will gain speed.
> How larger the memory used for cache, how higher the chance used data is
> present in the cache, so how faster disk acces becomes.
I know, but as I said the gain in speed reading cached data is very small
compared to when using non-cached data only. And the loss in speed when
reading non-cached data is huge.
Let me explain you something about how the Sunrise IDE-interface works (and
I think the MegaSCSI also works this way). The interface uses a
memory-mapped I/O system, hence it doesn't use any I/O ports, but memory
addresses instead. It is designed in a very smart way, because the sector
read from the drive is DIRECTLY loaded into a dedicated area of the
IDE-interface its memory. The only thing which has to be done is using an
LDIR to move it to the appropriate location.
When caching the data (first time scenario), an additional LDIR is issued to
move it to the cache-memory, so that makes 2 LDIR's per sector read, instead
of one without cache. When reading cached data, it is LDIRed from the
cache-memory, but when reading from disk also only one LDIR is used, so in
that case there is no gain either (but also no loss). In theory there still
is a small gain in the last scenario, because the cache can LDIR the cached
data in larger chunks, and it doesn't have to wait for the harddisk (however
you won't really notice it on MSX with new harddisks). But in practice, this
gain is very little and absolutely NOT worth 50% of my 2MB mapper and the
slow first-time loading.
On my old BERT (I/O-based) SCSI-interface, the non-cached transfer rate was
about 50kb/s. So on that one, caching really did make a difference.
~Grauw
--
>>>>>>>>>>>>>>>>>>>>>>>>><<<<<<<<<<<<<<
email me: [EMAIL PROTECTED] or ICQ: 10196372
visit my homepage at http://grauw.blehq.org/
>>>>>>>>>>>>>>>>>>>>>>>>><<<<<<<<<<<<<<
****
MSX Mailinglist. To unsubscribe, send an email to [EMAIL PROTECTED]
and put "unsubscribe msx [EMAIL PROTECTED]" (without the quotes) in
the body (not the subject) of the message.
Problems? contact [EMAIL PROTECTED]
More information on MSX can be found in the following places:
The MSX faq: http://www.faq.msxnet.org/
The MSX newsgroup: comp.sys.msx
The MSX IRC channel: #MSX on Undernet
****
