At 1:40 Uhr +0200 09.06.1998, Ken Nakata wrote:
>On Mon, 8 Jun 1998 21:47:36 +0200,
>Hauke Fath <hauke@Espresso.Rhein-Neckar.DE> wrote:
>> At 2:19 Uhr +0200 08.06.1998, Ken Nakata wrote:
>> >On Mon, 8 Jun 1998 00:30:23 +0200,
>> >Hauke Fath <hauke@Espresso.Rhein-Neckar.DE> wrote:
>>
>> I used the term IOP somewhat loosely in the sense of "doing GCR decoding
>> on-chip". In fact, I have just re-read a set of snippets from technotes
>> Apple has online. They said about the 840AV's floppy controller...
>
>You mean IOP *not* "doing GCR decoding on-chip"?  The New Age (AV's
>floppy controller chip) just doesn't seem to have hardware GCR
>encoder/decoder, from what I can read below.

Not quite. As I read it, they say the AV floppy controller is slower than
previous Macs wrt. GCR because it does not have the CPU horsepower for an
algorithm that was designed to save hardware by spending CPU cycles. The
.Sony read/write routines consist of highly optimized hand-assembled code
and use every register they can get hold of (sp/a7 pointing to the buffer
for incoming SCC data). The code was designed for a 7.8 MHz 68000 CPU with
an effective memory access timing of ~4MHz, after all...

>> ---------------------------------------------------------------------------
>> >[Macintosh AV Series: Floppy Drive Performance (5/94)]
>> >
>> >This microprocessor (the new integrated floppy controller), while taking
>> >over all of the timing-critical floppy control functions,
>> >does not have the computational
>> >speed of the main processor, or the speed of the 68000 used in original
>> >Macintosh systems. The result is that for GCR formatted disks (800K), the
>> >additional computation used in GCR coding and decoding makes accesses
>>to and
>> >from an 800K floppy slower than on previous macintosh systems. Our
>>tests show
>> >that it usually takes about 50% longer to read and write to an 800K
>>floppy on
>> >the AV systems as compared to other Macintosh systems.

[...]

>I tried to get my hands on whatever on-line docs available on AV Macs,
>but I anyhow missed this one...  The New Age on my Q840AV's logic
>board is in fact an NEC uPD72070.  That's why I called their local
>branch for a copy of datasheet.  Looks like MFM format (1.4M) will be
>much easier to implement than GCR formats on AV's.

If they do GCR on-chip (that's what I understand), there should be no
difference. If we have to have the CPU do it, the code is here.  =8)

	hauke



--
"It's never straight up and down"     (DEVO)