Strange issue on real HW

Page 20/20
13 | 14 | 15 | 16 | 17 | 18 | 19 |

By Grauw

Enlighted (8076)

Grauw's picture

16-06-2019, 20:46

PingPong wrote:
ARTRAG wrote:

I have found the pages with your experiments
https://www.msx.org/forum/development/msx-development/about-...
BTW now I am curious about what an invalid command code would do while a copy command is executed

invalid command code ? what you mean with invalid command code ? example?

Command codes 1, 2 and 3 are “invalid” (see V9938 manual p. 54). Actually I wonder if they do anything in the first place? (Or behave like STOP?)

On that note, I wonder what logical operations 5-7 and 14-16 do as well. NAND / NOR / NEOR and TNAND, TNOR, TNEOR? It would be a logical extension if I fill in the blanks, but I haven’t seen anybody ever discuss them.

By ARTRAG

Enlighted (6181)

ARTRAG's picture

16-06-2019, 21:04

Yes, those above

By PingPong

Prophet (3297)

PingPong's picture

17-06-2019, 14:01

I forgot that not all bit configurations where used. Sorry.
Anyway to make some test one would need a real thing. I've no longer access to it, so i cannot try.
Another thing that should be investigated is if there is a safe moment where a change on R18 does not corrupt the command.
Expecially now that we know precise VRAM timings.
If it was possible, this solve the problem of breaking a command in multiple chunks.

anyone can do some tests?

By Grauw

Enlighted (8076)

Grauw's picture

17-06-2019, 14:04

What I think may happen with R18 is that, it changes the moments the access slots occur, so whatever is using one right at that moment gets a corrupted access.

So it may also be that horizontal offsets cause a problem, but vertical offset changes do not (at least not during active display more than 8 lines away from the border).

By ARTRAG

Enlighted (6181)

ARTRAG's picture

17-06-2019, 15:25

Ok but once you have disabled the screen the vdp engine is the sole to access to the ram
Shouldn't be sufficient to blank the screen to avoid the effect you say?

By PingPong

Prophet (3297)

PingPong's picture

18-06-2019, 08:46

hint: someone told the corruption only appear with sprites disabled, even if i've observed even with sprites enabled.
maybe that the corruption gets even worse with sprites disabled for no apparent reason? (greater amount of access slot)
if this is true for the same reason on vblank should be the most visible effect.

Maybe that R18 is always involved when calculating every access slot ? and the greater availibility of access slot paradoxically increase the problem?

Page 20/20
13 | 14 | 15 | 16 | 17 | 18 | 19 |