From: Gideon Zweijtzer (gideonz_at_dds.nl)
Date: 2002-05-02 13:11:15
|> Hmmm, wait a minute..... Not a good idea afterall: this would mean that
the
|> SDRAM is occupied at least 50% of the time only because the VIC needs to
|> read data. (or can you circumvent this, Gideon ???)
|
|Since the VIC does only need 8 bit data being fetched/present during that
|time, why not wasting 8 (internal) flipflops on them... That way the SDRAM
|would only be occupied the minimum possible time where VIC can be as slow
|as it wants to.
Yes, using a read-buffer (one byte deep) already gives a great performance
gain. A read-buffer of four-deep (to match the 32 bit SDRAM width) would
even be better. Most VIC accesses are sequential anyway.
So in other words: Whether we use TRUE dpram or we implement pseudo dpram
(which has a VIC port that maps to the SDRAM), the solution of replacing the
on board memory is always faster that queuing the writes to video-memory
into the on board DRAMs.
IMHO it IS possible to get the VIC address out of the expansion port, so if
we disable or remove the DRAMs from the main board, we can actually provide
the VIC with the right videodata through the expansion port.
Does anyone know if A14 and A15 on the expansion connector also represent
the state of the CIA bank select signals during the VIC access phase?
Gideon
Message was sent through the cbm-hackers mailing list
Archive generated by hypermail 2.1.4.