[Coco] Re: [Color Computer] 68881/68882 & the CoCo

[Gene Heskett]

On Wednesday 08 February 2006 18:08, Mike Pepe wrote:
>Gene Heskett wrote:
>> Well, with all the setup, loads and readbacks being done at the
>> coco's bus's leasurely rate, I'm not sure it would be that much
>> faster.  If the FPU was running at its spec sheet clock speed, and
>> not loafing along at the coco's speed, it might be faster. I suspect
>> one would have to go ahead and do it before that question could be
>> answered with any authority though.
>
>The FPU would of course run at it's native rate. The 68k bus is
>asynchronous, and the setup and hold times of a CoCo at 1.8MHz are far
>above minimums specified in the data book.
>
>The setup time is an issue, but somehow I think it would still be a
> lot faster, especially on the trig functions.
>
>Natively, it supports 8, 16, or 32 bit data busses and the ability to
>function as a native co-processor or as a peripheral, so, like I said
>before, interfacing it to the CoCo bus appears almost trivial.
>
>> One things for sure, it would be an interesting experiment. :)
>
>Sure would
>
>> Just recall the 6309's integer divide (16 into 32, with 16 bit
>> remainders and dividends) operation is 39 cycles worst case, and its
>> wide mul operations is even faster, 29 cycles worst case IIRC. 
>> Unforch I don't think any of the math libraries have been optimized
>> for that. If I ever get curious, I'll do the trig library again, but
>> stop it at the assembly code stage and see if I can optimize it for
>> the 6309, I sense there might be a considerable speedup possible by
>> doing that.
>
>Well, You're talking integer. I'm talking about floating point. I
> think despite the effort needed to set the FPU up, it should still be
> tons faster, but we'll see.

Yes, but internally, the libraries use native integer with a final 
normalization conversion to the usual format at the output, in the 
printf function I think.  As described in the c compilers manual IIRC.  
The trig lib runs in double mode for most functions.  Since for some 
ops, the 6309 can handle doubles to ints and vice versa natively, there 
would be some potential speedups available there.
 
>-Mike

-- 
Cheers, Gene
People having trouble with vz bouncing email to me should add the word
'online' between the 'verizon', and the dot which bypasses vz's
stupid bounce rules.  I do use spamassassin too. :-)
Yahoo.com and AOL/TW attorneys please note, additions to the above
message by Gene Heskett are:
Copyright 2006 by Maurice Eugene Heskett, all rights reserved.