40 column video memory mirroring not working

[rhester72]

Please observe the following VICE behavior when directly manipulating screen memory.

In 40 column mode, the start of screen memory (at $8000) is mirrored into the 80 column region, such that POKEing a memory location changes the same location in the 80 column half of RAM:

which obviously doesn't (and can't) occur in 80 column mode:

This mirroring does not happen currently in the core, which breaks 80-column detection routines like those used in later issues of CURSOR magazine.

Proper mirroring would be:

$8000-83FF -> $8400-87FF (per testing with the 4032 in VICE 3.7)

[sorgelig]

i've added mirroring in new commit

[rhester72]

Confirmed as fixed, thanks very much!

[rhester72]

Sincere apologies - now that I know the real-world target is a 2001-32N, a bit more mirroring is needed if historical accuracy is desired.

$8000-83FF (on the 2001 only!) mirrors to three ranges:

$8400-87FF
$8800-8BFF
$8C00-8FFF

[sorgelig]

I'm not familiar with PET, but according to VICE we can emulate also 3008,3016,3032 models without much changes. Are those mirroring are really required to make some software working? Since we have custom ROM loading now, we can use 8800-8FFF for some ROM versions.