At 12:53 AM 07/24/99 +0200, you wrote:
>And well, the only bad side I can find of Compass is really concerning
>to speed: in Turbo-R, when sourcecode is placed in a mapper slot other
>than primary (internal) one, assembling speed decreases a lot, almost to
>the Z80 speed level. But I suppose this can't be solved... or it can?
Very strange...
Code executes faster in the internal mapper of the turbo R, because of the
address select (I forgot whether it's row or column address select) of the
DRAMs. If the same block of memory is accessed as the time before, a cycle
is saved.
But since every instruction comes from memory, two accesses to data should
always be worst case, no matter which mapper is used.
Maybe I need an example to explain:
addr: instruction:
#1234 ld a,b
#1235 add c
#1236 ld d,a
These are all fetched from the same memory block, so it's faster in the
internal mapper.
addr: instruction:
#1234 ld a,(hl)
#1235 add c
#1236 ld (de),a
No matter if the first and the third instruction access the same memory
block, the block would be different from the block that contains the code
(different
mapper blocks). So the internal mapper won't benefit from the one step
shorter address selection.
Actually, assembling from and to an external mapper should be faster, if
Compass itself is located in the internal mapper, since reading the source
and writing the result doesn't undo the blocks selection.
Jon, do you have any idea why this happens?
Actually, I never used my external mapper in my turbo R, because GT's 512K
was always enough. So I never noticed this strange effect.
>> Anyone think about using COMPASS under PC? (((((((-:
Actually I'm still looking for a good command line Z80 assembler for
Windows or Linux. If anyone knows an URL...
Bye,
Maarten
****
MSX Mailinglist. To unsubscribe, send an email to [EMAIL PROTECTED] and put
in the body (not subject) "unsubscribe msx [EMAIL PROTECTED]" (without the
quotes :-) Problems? contact [EMAIL PROTECTED] (www.stack.nl/~wiebe/mailinglist/)
****
