Date   : Sat, 31 Jul 2004 10:10:45 +0200
From   : John Kortink <kortink@...>
Subject: Re: GoMMC and sideways RAM systems

On Fri, 30 Jul 2004 23:23:37 +0100, Sprow <info@...> wrote:

>In article <c0snf09e1r70umld3kghmnsggaeojtpvod@...>,
>   John Kortink <kortink@...> wrote:
>
>> The GoMMC code writes to the sideways ROM space it's in. Much
>> like anything could, were it in a 'standard' RAM bank (as in :
>> selected exclusively by ROMSEL).
>>
>> This presents no problems if ROMSEL is always used to point
>> to the read-enabled as well as write-enabled sideways RAM/ROM
>> bank. Simple SWRam systems and the Master SWR system use this.
>
>[snip]
>
>> Those who can't live with these two caveats, please steer clear
>> from GoMMC. No fix can and/or will be provided.
>
>I can't resist throwing two ideas in
> a) The GoMMC firmware could supply a SRREAD/SRWRITE command (compatible
>    with the Master) which does the sideways RAM write properly.
>
>    I can only think that boards like the excellent ATPL allowed the
>    ROMSEL bank to differ from the write bank so that they didn't have
to
>    waste a ROM slot with firmware/supply easy to lose disc programs.
>    Since GoMMC uses a slot anyway it's no loss to pop a couple more
commands
>    into the ROM.
>
> b) Implement some crazy R-W-W sequence in the CPLD on GoMMC where the 
>    driver reads the old contents, writes the byte to the MMC then
replaces
>    the original byte.
>
>    Oh, hang on, that wont work for a multitude of reasons I've just 
>    thought of. Bah.

Bah indeed. The whole idea of leaving some RAM bank wide open for
random corruption by external factors, just to support directly
*LOADing into slots, is sloppy design. Kind of like leaving a file
open for writing, just in case you want to write to it again,
regardless of everything that might screw it up in the meantime.

Anyway, like I said, I'll not be pursuing any kind of fix for the
consequences, to some SWRam systems, of writing to my own ROM slot.
Users will simply need to ensure no sensitive RAM bank remains
'selected'. *LOADing straight into a RAM bank is fixable by some
horrible CPLD logic patch, but I've doubts about its chances to
end up in the final version.


John Kortink

-- 

Email    : kortink@...         
Homepage : http://www.inter.nl.net/users/J.Kortink