>Ok, I called NCR's SCSI division, got some literature on the 53c90
>series, saw that the 53c90 looks like what we wanted, and talked to
>the local NCR distributor, Electrodyne (303)-695-903 for pricing
>information.
Good one. I believe that the 53c94 may actually be a slightly improved
version of the '90, with some relaxed timing requirements and miscellaneous
bug fixes. Do you have any info on the '94?
>The host/DMA interface is 8 bits wide, so it's best on an
>8 bit bus of the 3730.
Some friends of mine were did an ARM-based machine similar to what we're
aiming at here, and they used some variant of this NCR chip. A neat trick
they were able to use was to attach the DREQ/DACK signals to the regular
bus I/O signals, and use the LDM/STM (load/store multiple registers)
instructions to do "fake" DMA. The ARM only has 16 registers, so they could
do a maximum of 32 (16 x 2) bytes at once, but it was still significantly
faster than doing it across the regular 8 bit bus.
Sorry I can't tell you about the 3730's DMA interface. Seeing as the TLB is
on the CPU, it probably doesn't support any kind of scatter-gather. I know
DEC used the 53cf94 in some of their MIPS-based machines - anyone know how
they dealt with the problem there? Personally, I don't think it's such a
problem. I had first hand experience of the Seagate scatter-gather
performance improvement (thanks Drew <grin>), but suspect that it wouldn't
be so drastic with this chip (much more intelligent/faster setup phases).
So, I think the question now is: do we try and attach it to a DMA channel,
or do we just make it a PIO device? If we have a dedicated I/O controller
(3041), there's no real choice to be made. If we use a 3730, however, I
think it's an important point to consider.
Pat -- "There's only one thing left to do Mama, I got to ding a ding dang
my dang a long ling long" (Jesus Built My Hotrod -- Ministry)
GCS d* -p+ c++ l++ m--- s+/- !g w- t- r
|