CIA TOD Rollover

[Ernst Blofeld]

Editing post as my problem has now changed.

Can anyone explain why I get occasional silly values when reading the CIA B TOD clock?

When reading it at the start and end of a function, I occasionally get bad values, with the finish time being less than the end time, or even the finish time being zero sometimes.

More details in post 4 below.

[a/b]

Read once in this order: high (latch activated), middle, low byte (latch deactivated).

[Ernst Blofeld]

Originally Posted by a/b

Read once in this order: high (latch activated), middle, low byte (latch deactivated).

Ooooh... that explains why I couldn't find anything when I searched.

Thank you.

[Ernst Blofeld]

It's strange though, that I'm occasionally getting start times that are greater than my end times...

In Profiler_EndFunction: startTime 65532 > finishTime 0
In Profiler_EndFunction: startTime 192250 > finishTime 192000
In Profiler_EndFunction: startTime 258555 > finishTime 258304

In fact, how can I have a finish time of zero?

(edit - the low byte of all the finishTimes is zero, if that's a clue)

ULONG GetCurrentTime(void) {
    ULONG hi = ciab->ciatodhi;		// latch activated
    ULONG mid = ciab->ciatodmid;
    ULONG low = ciab->ciatodlow;	// latch deactivated

    return (hi << 8 | mid) << 8 | low;
}

ciab is

volatile struct CIA * ciab = CIAB;

where CIAB is

#define CIAB ((volatile struct CIA *) 0xbfd000UL)

The generated assembly looks sensible, to me:

00002c0a <GetCurrentTime>:
ULONG GetCurrentTime(void) {
    2c0a:	move.l d2,-(sp)
    ULONG hi = ciab->ciatodhi;		// latch activated
    2c0c:	move.b bfda00 <gcc8_c_support.c.62367fe7+0xbdadad>,d0
    ULONG mid = ciab->ciatodmid;
    2c12:	move.b bfd900 <gcc8_c_support.c.62367fe7+0xbdacad>,d2
    ULONG low = ciab->ciatodlow;	// latch deactivated
    2c18:	move.b bfd800 <gcc8_c_support.c.62367fe7+0xbdabad>,d1
    ULONG hi = ciab->ciatodhi;		// latch activated
    2c1e:	andi.l #255,d0
    return (hi << 8 | mid) << 8 | low;
    2c24:	lsl.l #8,d0
    2c26:	or.b d2,d0
    2c28:	lsl.l #8,d0
}
    2c2a:	or.b d1,d0
    2c2c:	move.l (sp)+,d2
    2c2e:	rts

[Toni Wilen]

Do you have OS running? TOD is reset periodically by system.

[Ernst Blofeld]

Originally Posted by Toni Wilen

Do you have OS running? TOD is reset periodically by system.

No, I've taken over the system, I believe correctly. The only interrupts that should be running are a copper one for my screen flipping and a ports one for the keyboard, which I believe only uses cia a.

void TakeSystem(void) {
	Forbid(); // forbid multitasking

	ActiView = GfxBase->ActiView; // save current view

	// Save current interrupts and DMA settings.
    systemADKCON = custom->adkconr;
    systemINTENA = custom->intenar;
    systemDMACON = custom->dmaconr;

	// The VBR is stored in a different place on 68010s and above.
    if ((SysBase->AttnFlags & AFF_68010)) {
		UWORD getVBR [] = {
			0x4e7a, 0x0801,     // movec.l vbr, d0
			0x4e73              // rte
		};

		VBR = (APTR *) Supervisor((void *) getVBR);
	}

    // Save system interrupt handers.
    for (int i = 0; i < 6; i++)
        systemInterruptHandlers[i] = VBR[25 + i];

	WaitForVerticalBlank(); WaitForVerticalBlank();

	custom->dmacon = DMAF_MASTER | DMAF_BLITHOG | DMAF_ALL; // clear all DMA channels

	// Set all colours to black.
	for (int a = 0; a <= 31; a++)
		custom->color[a] = 0x000;

	LoadView(NULL); // has side effect of enabling VERTB

	WaitTOF(); WaitTOF();

	WaitForVerticalBlank(); WaitForVerticalBlank();

	OwnBlitter();
	WaitForBlitter();

    custom->intena = INTF_INTEN | INTF_ALL; // disable all interrupts
    custom->intreq = INTF_INTEN | INTF_ALL; // clear any interrupts that were pending
}

[Ernst Blofeld]

At the moment I don't believe this is a problem with my code, because when the problem occurs the low byte of the is always zero. I could imagine something overwriting the variable on the stack, but only the lowest byte?

I've changed the code that detects the problem to the following:

    if (startTime > finishTime)
        * (WORD *) 0xdff180UL = 0xfff;

So that you get a white flash on the screen when it happens. On my set up this happens on around the 12th and 16th frames.

I've attached the executable in case anyone wants to have a look.

[Toni Wilen]

ALARM bit is set (CRB bit 7).
Reading TOD when ALARM bit is set won't latch the TOD but returns "live" TOD. (EDIT: which is undocumented feature)

[ross]

I had just found this: https://eab.abime.net/showpost.php?p...8&postcount=14
But Toni already posted

[Ernst Blofeld]

Originally Posted by Toni Wilen

ALARM bit is set (CRB bit 7).
Reading TOD when ALARM bit is set won't latch the TOD but returns "live" TOD. (EDIT: which is undocumented feature)

Thanks Toni.

I'm now clearing the ALARM bit,

    ciab->ciacrb = CIACRBF_ALARM;

in my init code, but not seeing any difference. Am I doing it right?

[Ernst Blofeld]

Like this?

    ciab->ciacrb &= ~CIACRBF_ALARM;

This seems to work, but it's hard to find good documentation and examples. This control register is not like the interrupt control registers, it should be read then a modified value written back to it?

[Toni Wilen]

Most CIA registers are normal read/write. Interrupt register is special.