[Coco] Implications of the Speed Up Poke

[Alex Evans]

the thing that is given first (row?) determines what part of memory is
refreshed. IIRC on most 8-bitters, the least significant part of the
address is generally passed first, so if we assume that our memory is
using memory with 8-bit row addresses, then accessing 256 consecutive
addresses will refresh all of memory.  My experience with the highest
speed poke on a CoCo2 is that I never had any detectable memory loss
even with very long runs (hours). Of course the fact that the part of
memory your program is actually using will definitely be refreshed
reduces the change that you are going to notice the effects.

On Sat, Feb 1, 2020 at 1:58 PM William Astle <lost at l-w.ca> wrote:
>
> On 2020-02-01 10:33 a.m., Allen Huffman wrote:
> > Programs like Musica 2 on a CoCo 1/2 had the option to play music in the fast mode, losing video while the audio played at better quality. I never knew there was any RAM concern as you could play long songs with no issues.
> >
> > What is the source of the RAM refresh concern?
>
> The source of the RAM refresh concern is the SAM data sheet which
> explicitly says that in the "Fast" mode, RAM refresh and VDG RAM access
> are both disabled because it gives the refresh/VDG part of the machine
> cycle to the CPU.
>
> As far as things running a long time without problems under Fast mode,
> most DRAM holds its contents for a som time longer than the refresh
> interval. Refresh intervals are actually quite conservative because you
> don't want your RAM losing it's contents. For added fun, simply reading
> from memory refreshes the entire row/column (can't remember which) that
> address is in so the code running the music player in your example would
> likely be accessing the memory that implements the music player
> sufficiently often to keep the DRAM refreshed. Well implemented Music
> players would probably also be programmed to ensure that the entire RAM
> space was covered by strategic reads to prevent any of it from decaying.
>
> --
> Coco mailing list
> Coco at maltedmedia.com
> https://pairlist5.pair.net/mailman/listinfo/coco