Zach's Projects

Not sure, but maybe something like ASL or DEC VBLANK might work too.

Not sure, but maybe something like ASL or DEC VBLANK might work too.

Thanks for the idea, Thomas! I did try DEC, and while it did clear VBLANK it seems to do it a cycle earlier than SHX, or even a regular LDA #0, STA VBLANK. I've attached a screenshot on top so you can see what I'm talking about. I don't think I can afford to delay DEC VBLANK for a cycle.

 

I also tried INC, ASL, LSR, ROL, ROR, DCP, and ISB, and all of them seem to change VBLANK early.

I also tried INC, ASL, LSR, ROL, ROR, DCP, and ISB, and all of them seem to change VBLANK early.

Weird.

What's even weirder is the behavior of INC and some other opcodes. They seem to change VBLANK twice, first clearing it for a cycle, and then setting it again. See above. I suspect that this behavior has been studied before, based on the fact that both z26 and Stella appear to emulate it accurately.

What's even weirder is the behavior of INC and some other opcodes. They seem to change VBLANK twice, first clearing it for a cycle, and then setting it again. See above. I suspect that this behavior has been studied before, based on the fact that both z26 and Stella appear to emulate it accurately.

The read-modify write instructions write twice. The first write occurs on cycle 4 and just writes back what was read from the location, and then it does the operation and writes the result on cycle 5.

 

INC VBLANK would first write a $01, then a $02 one cycle apart, as would ASL and ROL, but DEC, LSR and ROR would write a $01 and then a $00.

The read-modify write instructions write twice. The first write occurs on cycle 4 and just writes back what was read from the location, and then it does the operation and writes the result on cycle 5.

Hm, I wonder if eventually someone finds a good use for that.

The read-modify write instructions write twice. The first write occurs on cycle 4 and just writes back what was read from the location, and then it does the operation and writes the result on cycle 5.

Hm, I wonder if eventually someone finds a good use for that.

I was just thinking this morning, could you use INC ENAM0 to create a three-pixel long missile? (and would that ever be useful?)

The read-modify write instructions write twice. The first write occurs on cycle 4 and just writes back what was read from the location, and then it does the operation and writes the result on cycle 5.

Hm, I wonder if eventually someone finds a good use for that.

I was just thinking this morning, could you use INC ENAM0 to create a three-pixel long missile? (and would that ever be useful?)

I just found this posted by Supercat on [stella] but I don't know when it was posted since my email is set to GMT and I never bothered to fix it.

[stella] Read/modify/write VBLANK, ENAxx, etc.

 

The following is of particular relevance to a "Chitry" thread in the

homebrew forum on AtariAge, but for some reason attempts to post there

always time out. Maybe someone can forward this information.

 

Read/modify/write instructions (absolute mode) take five cycles, performing

writes on cycles 4 and 5. The first value written will be the value just

read; the next value written will be the new value.

 

Thus:

 

ASL VBLANK+$0300 will turn on VBLANK in cycle 4

 

INC VBLANK+$0300 will turn on VBLANK in cycle 4 and turn it off in cycle 5

 

ASL VBLANK+$0100 will turn on VBLANK in cycle 5; it will most likely turn

it off in cycle 4, but on 7800's it may turn on then (it will be on in

cycle 5 regardless)

 

INC VBLANK will probably turn off VBLANK in cycle 4, and if that fails

(some 7800's) it will probably turn it off in cycle 5. This is probably

the least reliable one.

 

The status of the upper bits of VBLANK should be considered indeterminate

after the above operations, so do not use them with paddles and be sure to

write "0" to them before trying to read the fire buttons.

Hitting HMOVE twice might also give some interesting results.

I just found this posted by Supercat on [stella] but I don't know when it was posted since my email is set to GMT and I never bothered to fix it.

 

Thanks for reposting it, though I flubbed by a cycle the timings for all "RMW abs" instructions.

 

I would consider "dec vblank" a reliable means of turning off VBLANK on cycle 4 of a 5-cycle instruction; "asl vblank" would probably turn it off at cycle 4 and on at cycle 5, but on a 7800 it might turn it on at cycle 4 (leaving it on at cycle 5).

 

Getting ten flicker-free colors on a line is a nice trick, though I don't know if the fill-rows-with-any-color-blocks game is the best use of it. Some other styles of color-grouping games that don't involve rotation might be better since aspect ratio would not be a problem. Alternatively, it might be possible to use a taller aspect ratio and a squashing or scrolling screen. The squashing screen might be particularly interesting since the game is usually focused on the falling piece and the top four non-empty rows. Other rows are far less important.