| |
chatGPZ
Registered: Dec 2001 Posts: 11386 |
what is your RAM init pattern?
since i stumbled about yet another case of some ancient demo that relies on uninitialized RAM - and which would only work if you change the default VICE uses currently - i think its about time to collect some hard data about this phenomenon and perhaps at the end add proper configuration for it to VICE (or even choose a better default).
background: when you power on your C64, the RAM will not be zero, instead (about) half of the RAM cells will be 1, and the other half will be 0. this results in some kind of pattern (mostly) consisting of $00 and $ff. for example if you start VICE, then enter the monitor and look at a page of RAM it shows this:
(C:$e5d4) m c000 c0ff
>C:c000 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
>C:c010 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
>C:c020 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
>C:c030 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
>C:c040 ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ................
>C:c050 ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ................
>C:c060 ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ................
>C:c070 ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ................
>C:c080 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
>C:c090 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
>C:c0a0 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
>C:c0b0 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
>C:c0c0 ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ................
>C:c0d0 ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ................
>C:c0e0 ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ................
>C:c0f0 ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ................
this is the default pattern, which could be shortened to "64 times $00, 64 times $ff, repeat"
now i'd like to know how it looks like on your (real) C64... if you have some cartridge with ML monitor, all you need to do is power on your c64, check some memory page, and post the pattern here (you can ignore that some values wont actually be 0 or $ff - what matters is the general trend). additionally it would be extra cool if you could look into your C64 and post the type of RAM chips that it uses as well (since the pattern most likely depends on them) - they are labelled 4164 or 4464 or 41464 (please post everything written on it, the exact type and manufacturer is interesting here). Also please post the ASSY NO of the board, as the way RAM is wired up matters as well.
thanks for your help!
PS: if you are curious, Typical is the demo in question... if you can name other examples ("works in CCS64 but not in VICE is often an indicator for it), please post them too! |
|
... 47 posts hidden. Click here to view all posts.... |
| |
Flavioweb
Registered: Nov 2011 Posts: 463 |
What make ram cells have the values they have after a cold start?
Is a "internal ram chip" question or some external factor can influence them?
Maybe bus 'activity' or other chips init?
Reading this Gpz sentence:
Quote:repeating ff,ff,00,00,00,00,ff,ff pattern, every $4000 bytes the pattern seems to be inverted (for another $4000 bytes). occasional random bytes
i started thinking if Vic init-activity can influence ram values...
FF/00 are values usually "leaved" on bus by vic during normal operations... |
| |
chatGPZ
Registered: Dec 2001 Posts: 11386 |
IAN: what an epic thread that is - we've come a long way since then :)
others: if you post dumps, try to post a page at least, that makes it much easier to see patterns. if you dont want to type it all out, you could also just save c000-d000 in monitor and send the resulting file to me via email. (in this case please put assy no and ram type in the filename AND add the same info in the mail itself) |
| |
chatGPZ
Registered: Dec 2001 Posts: 11386 |
Quote:What make ram cells have the values they have after a cold start?
Is a "internal ram chip" question or some external factor can influence them?
Maybe bus 'activity' or other chips init?
its simple law of physics - at poweron the internal flipflops will flip to either 0 or 1 randomly, roughly with a 50:50 chance. now the actual pattern you see results from the internal organisation of the RAM chips, for example neighboured cells may more likely flip into the same state. additionally *some* RAMs apparently contain reset circuitry that actually makes them start up in a defined way (i dont think this is a case for those old RAMs as found in C64s though).
other chips having an impact on the behaviour is unlikely IMHO - because for that to happen the write line of the RAM would have to be pulled low actively from somewhere, which shouldnt happen at poweron. |
| |
tlr
Registered: Sep 2003 Posts: 1790 |
It's DRAM we are talking about so the memory cells do not consist of flip-flops. They are basically caps.
I agree that the pattern probably arises from the organisation of the cells. Perhaps half of the cells are referenced to VDD? |
| |
chatGPZ
Registered: Dec 2001 Posts: 11386 |
ok perhaps i should have written "RAM cells" :) |
| |
tlr
Registered: Sep 2003 Posts: 1790 |
An important point about the cells being caps is that they aren't going to have 50:50 to land in any state. The most likely state is uncharged, which will mean 1 or 0 depending on if it is referenced to VDD or GND. Upon a power cycle all cells will (surprisingly slowly) fade towards the uncharged state. A quick power cycle will likely keep most contents.
The source of the spurious bits in the initial state could be due to how the refresh logic is implemented perhaps? E.g an initial glitch that injects some random bits. Remember that there are no resets to the ram chips so some strange startup behaviour is likely. |
| |
chatGPZ
Registered: Dec 2001 Posts: 11386 |
sounds convincing :)
infact, now that i checked all my C64s (updates in the repo coming soon) i can say that even the one that first showed a seemingly random pattern (like TMR posted) dropped to mostly $00 and $ff after powering it off for several minutes. |
| |
Flavioweb
Registered: Nov 2011 Posts: 463 |
...mmm...
What happen if you fill all ram with 00 or FF and power-cycle your C64?
If is only a question of caps charge,at least in one case,you should get the perfect 00/FF pattern back...
No? |
| |
chatGPZ
Registered: Dec 2001 Posts: 11386 |
i actually tried that on the C64 that retains its RAM content for very long... no, it doesnt improve the situation, you really have to power off for some (!) minutes (!).
i updated the test program and also the readme with results from my own C64s |
| |
tlr
Registered: Sep 2003 Posts: 1790 |
btw, I think the presence of randomness is an important emulation aspect. Some programs (originals) check for the presence of randomness in the start up pattern as a means of detecting a prefilled memory vs a normal "dirty" startup.
DarkStar BBS V3.1 did this, but I'm sure it was fairly common.
I guess it has to be pseudo random in some smart way, not to violate the reproducability of the startup sequence. |
Previous - 1 | 2 | 3 | 4 | 5 | 6 - Next |