[Coco] assembly questions?

[Kevin Diggs]

Hi,

	The speed of these probably isn't that crucial since there are
so few registers to transfer between (amongst?). If I ever find a job I
think I am gonna start with that 6809 VHDL core someone did and create
the 6909.  One thing I think I'll include is a set of shadow registers
and some special FAST tfr and exchange instructions to get at them. Kind
of like a high speed register cache. Probably also need a context save
instruction. And maybe some burst bus modes for some caches. And an
instruction cache. And a stack cache (to speed up stack operations). And
a 16-bit internal bus. ...

	I am working on an assembly port of the old X maze program. The
lack of registers has been ... constantly annoying.

					kevin
KnudsenMJ at aol.com wrote:
> 
> Thanks -- I myself have always wondered why TFR and EXG were so slow!
> Part of the problem seem s to be always treating them as 16-bit operations,
> so TFR A,B takes as long as TFR X,Y.
> 
> And the other is using that internal temp reg, which turns out not to be
> needed -- see below.
> 
> In a message dated 7/23/04 7:53:52 AM Eastern Daylight Time,
> jdaggett at gate.net writes:
> 
> > with the TFR instrtruction the third and fourth cycle write R1 to a temp
> > register internally. On cycle 5 and 6 the temp register is written to R2.
> 
> This sounds like an explanation I heard years ago, but below we see the temp
> wasn't needed. . .
> 
> >  With the EXG instruction the third and fourth cycle writes R1 to the temp
> > register.
> >  On the fifth and six instruction the contents of R2 is written to R1. On
> the
> > seventh  and eighth cycles the temp register is sritten to R2.
> 
> Since R2 was written directly to R1, the TFR instruction could have bypassed
> the intermediate register too.  But I guess this implementation simplified the
> control sequencing.  Remember, the 6809 was, and remains, the most
> sophisticated 8/16-bit micro ever made -- or darn close to it.
> 
> Thanks again for the details.  BTW, does the 6309 cut out any of these
> intermediate steps?  Maybe use a 16-bit internal bus?!?  --Mike K.
> 
> --
> Coco mailing list
> Coco at maltedmedia.com
> http://five.pairlist.net/mailman/listinfo/coco