reset sp

[TomSon]

First off: hi to everyone on this forum - I am brand new to VCS programming and have extensive low-level hw programming

background, (also programming retro hw) but so far not on the 2600.

 

So, just now I stumbled upon a possible bug in Stella - the reset SP is according to this document:

 

http://www.pagetable.com/?p=410

 

at 0xFD but in Stella it is 0xFF.

 

Since every cart. on the planet seems to init. the stack right away anyways this is not a big deal but still..

 

p.s.

void M6502::reset()

SP = BSPF::containsIgnoreCase(cpurandom, "S") ?

mySystem->randGenerator().next() : 0xff;

 

Edited January 26, 2017 by TomSon

[TomSon]

I just verified that SP is indeed FD starting a test bin on a 2600 in a ROM so this is a bug.

It may even be a bug of the harmony cart. depending on how they init the sp before starting the

loaded cart.:

I do not have one myself to test this - and I don't know how to attach my test_sp.bin here (yet..).

[Thomas Jentzsch]

Either zip it, or renamed it, e.g. to bin.txt

 

Interesting finding. Please post this here too.

 

More evidence:

http://visual6502.org/JSSim/

http://www.youtube.com/watch?v=fWqBmmPQP40) (around 42:15)

Edited January 26, 2017 by Thomas Jentzsch

[Nukey Shay]

I wonder what it is when 7800 begins 2600 mode. Could become a shortcut for console type detection

[DirtyHairy]

This is a very interesting find I have already observed that some official ROMs fail to init S properly and may fail with a corrupted stack, depending on the value of S (Dark Chambers for example). This (1) proves that the value is determinstic and (2) also fixes the value. I'll open a ticket against Stella, and I'll also fix this in 6502.ts.

Edited January 26, 2017 by DirtyHairy

[+SpiceWare]

I do not have one myself to test this - and I don't know how to attach my test_sp.bin here (yet..).

You have to click on More Reply Options to get to the file attachment options, they'll appear below the text box:

 

Do note that if you have Flash installed the Attach Files options will look different. Also BIN files can be uploaded just fine, no need to zip or rename them.

[TomSon]

Quick test to see whether SP is correct after reset, displays either an orange screen if correct or black one if SP != FD.

 

 

ORG $F000

START_FLAG .byte 0

;##############################################################

;

; Reset handler - clears all mem. and regs and inits

; the static regs/mem

;

;##############################################################

Reset:

tsx

cpx #$fd

beq Reset2

CLEAN_START

reset_entry

STA COLUBK ;3

JMP frame_handler

Reset2

CLEAN_START

LDA #ORANGE ;2

jmp reset_entry

frame_handler:

LDA #$02 ; A = VSYNC enable

STA WSYNC ; Finish current line

STA VSYNC ; Start vertical sync

STA WSYNC ; 1st line vertical sync

STA WSYNC ; 2nd line vertical sync

ldy #37

LDX #$B0 ;2

LSR ; A = VSYNC disable

STA WSYNC ; 3rd line vertical sync

STA VSYNC ; Stop vertical sync

vblank sta WSYNC

dey

bne vblank

sta WSYNC

STA VBLANK ;3

SET_COLOR0

STX HMP0 ;3

STX HMP1 ;3

ldx #11

norom stx WSYNC

dex

bne norom

sta WSYNC

ldx #180

norom2 stx WSYNC

dex

bne norom2

LDA #$42 ;2

sta WSYNC

STA VBLANK ;3

ldx #27

overscan:

sta WSYNC

dex

bne overscan

beq frame_handler

frame_handler_end:

ORG $FFFA

.word Reset ; NMI

.word Reset ; RESET

.word Reset ; IRQ

 

tst_sp.bin

Edited January 26, 2017 by TomSon

[Nukey Shay]

Why a 4k rom? BTW test fails on the Flashback portable.

[TomSon]

Why a 4k rom? BTW test fails on the Flashback portable.

 

hehe, every byte counts, I know But that's actually a good hint - I myself don't have a rom < 2k and have not checked whether they

exist and how they are treated in Stella. I assume it's just mirrored in the entire 4K and a 4K mapper used.

 

Very interesting that it fails on the flashback portable. I'd be also interested to know whether the Harmony does the same thing

(I really have to order one asap myself). I am investigating bank switching in general and initializations in particular in the moment

which is why this came up - the FE one is automatically set to bank 1 by the reset mechanism of the 6502 (the reset vector fetch

looks just like a jsr fffc when just snooping the cart. bus) which makes this 8k bank switch even more elegant in my eyes.

[Thomas Jentzsch]

I am pretty sure it fails on Harmony, since it doesn't care for registers, flags or ZP-RAM. It relies on the ROM initializing correctly.

[alex_79]

By looking at the 7800 bios source, the Stack Pointer is set at startup, and its value is $FF when starting a 2600 cart.

 

I used a simple test rom that displays the initial value of SP (loaded into the Harmony Cart in single-cart mode) which confirms that. On a 2600 the value is indeed $FD.

Loading the rom from the Harmony menu (firmware NTSC 1.06 beta), the SP value seems to always be $07.

 

SP.bin

[alex_79]

I forgot about this test rom posted by Omegamatrix a while ago:

https://atariage.com/forums/topic/230219-startup-values-of-the-2600/

 

It shows the startup values of the entire ram and processor registers

[+stephena]

This is now fixed, and will be part of Stella-5.0-pre5.

[Nukey Shay]

Is it $FD in the various 2600 knockoffs and adaptors, as well?

[alex_79]

I'm resurrecting this thread, as further testing that I've done recently actually contradicts the fact that the Stack pointer has a definite value at startup.

 

I tested 5 different consoles, with CPU manufacturing dates ranging from 1980 to 1991 and SP is not constant. A couple of consoles (including the one I briefly tested in 2017 and that initially seemed to confirm the theory) return $FD more often, sometimes for several power cycles in a row, but eventually will result in different values as well. The other 3 consoles return more random values, changing on each power cycle.

 

All the sources linked in this thread are based on the Visual6502 simulator, which happen to start always in the exact same state. Not only for SP, but all the registers and flags have always the same value there.

 

The MOS 6500 Programming manual, on the other hand, clearly states:

 

While the reset line is in the low state, it can be assumed that
internal registers may be initialized to any random condition; therefore,
no conditions about the internal state of the microprocessor are assumed
other than that the microprocessor will, one cycle after the reset line
goes high, implement the following sequence[...]

 

The reset sequence decrements SP 3 times and will result in SP=$FD only if the register was 0 initially. Which is possible, and might seem to happen consistently on individual CPUs, but cannot be assumed as a general behavior.

 

--------------------------

 

P.S. the test rom I attached at the time a few posts above has  a bug: it doesn't clear the decimal flag. If D is set (and happened a few times in my tests, as it's undetermined after a reset too) the result is garbled gfx.

Here is a fixed rom:

SP_startup_value.bin

Better yet, use this rom by Omegamatrix, that displays startup values for zeropage ram and all CPU registers

Edited July 30, 2019 by alex_79

[Omegamatrix]

Well that is good to know. I wouldn't have depended on the SP being at $FD for a normal game, but I did with the 32 byte demos we did last year

[+SvOlli]

The 32 bytes demos was what came to my mind when I was toying around with the "Startup Values" cart on different hardware. But since we used that trick not do clean up RAM (that wasn't used anyway), but only to clean up the TIA, it would have worked in far more than 99% of my tests, since the SP would have to be less than $27 and A would have to have the top bit set to leave an unexpected mark. And even in this case you would just wonder and turn it off and on again.

 

But the only thing that I found "stable" after a reset was that the interrupt disable flag was set.

 

BTW: Omegamatrix, would you care to update "Startup Values" with dumping out RIOT's I/O as well?

Edited August 7, 2019 by SvOlli
Name typo