[stella] Understanding the PIA

[bob.montgomery at thomson.com]

Hi,

>If you get a chance, could you also test those  (ie, let me know what
*is* 
>happening, and what *should* be happening in terms of output)?

Ok, did some testing.  All with emulators.

For TIM8T, after the timer passes zero Stella begins (correctly)
counting down once every cycle UNTIL INTIM is read, then the timer
begins counting down every 8 cycles again:

	lda #3
	sta TIM8T
	nop
	nop
	lda INTIM		;A holds 2
	nop
	nop
	lda INTIM		;A holds 1
	nop
	nop
	lda INTIM		;A holds 0
	nop
	nop
	lda INTIM		;A holds (correctly) F8
	nop
	nop
	lda INTIM		;A should hold F0, but in Stella has
A=F7

For TIM1T, things are really weird:

	lda #4
	sta TIM1T
	lda INTIM		;A holds 0
	lda INTIM		;A should be FC, Stella has A=F9
(?!)(counting double starting with FF?  Or?)
	lda INTIM		;A should be F8 (FC-4), instead Stella
has A=F5 (F9-4)	(counting down by 1s again)
	lda INTIM		;Stella: A holds F1	(etc.)
	lda INTIM		;Stella: A holds ED

For TIM64T, seems to be the same as TIM8T, only using a counter value of
64 rather than 8.  I.e., it counts down normally to zero, holds zero for
64 cycles, then counts down once very cycle until you read INTIM, then
it starts counting down once every 64 cycles.

Haven't tested TIM1024T yet.

Seems like two issues in Stella: 
1st: timer goes crazy after passing zero when using TIM1T
2nd: reading INTIM after timer has passed zero improperly resets the
count-down speed 

-Bob

P.S.  Additionally, undocumented in the programmer's guide but the same
in z26 and Stella:  The timer counts down by one immediately and *then*
begins counting down as you would expect (i.e., by 1s, 8s, 64s, 1024s).
This has no effect on TIM1T, but with TIM8T, for example, counting by
cycles:
Cycle	Timer (TIM8T stored with 8)
1	7?
2	7?
3	7?
4	7	tested in z26
5	7	tested in z26
6	7	tested in z26
7	7	tested in z26
8	7	tested in z26
9	6	tested in z26

Cycle Timer (TIM1T stored with 8)
1	7?
2	6?
3	5?
4	4	tested in z26
5	3	tested in z26
6	2	tested in z26
7	1	tested in z26
8	0	tested in z26
9	FF	tested in z26
10	FE	tested in z26
11	FD	tested in z26

I have a very vague memory of this being discussed in [stella] before...