|
Post by dshadoff on Feb 12, 2020 1:18:10 GMT
I was wondering whether anybody knows much about the slow speed CPU clock...
Specifically, if the CPU is in slow mode, I understand that instructions are executed at 1/4 of high speed (1.78MHz instead of 7.16MHz), but: - Is the base clock feeding the TIMER interrupt affected ? (it's on the Hu6280 physical package) - Is the base clock of the PSG affected ? (it's on the Hu6280 physical package)
Are there any other potential candidates which might be affected ? (I don't expect anything would be affected except things on the Hu6280 package itself.)
|
|
|
Post by dshadoff on Feb 12, 2020 14:47:24 GMT
Accroding to the PCEdocs contributions by Charles MacDonald, PSG and TIMER are unaffected. But I know that these documents - in many places - were gleaned from other documents, and were not tested.
I'll test the PSG today - it's easy enough - but I'm not sure whether I can easily find a TIMER test.
|
|
|
Post by dshadoff on Feb 12, 2020 23:03:52 GMT
OK, I altered Alien Crush to change the CSH to CSL, and the sound was the same.
So PSG is not driven from the CPU divider. I'm not clear where there is a test sample for TIMER, so perhaps I'll assume that it is also not dependent on CPU divider, until proven otherwise.
|
|
|
Post by turboxray on Feb 14, 2020 16:15:21 GMT
I remember a few games that switch into low speed mode while waiting for vblank. I just assumed they did that for battery usage on TG-Express. But that's just speculation. IIRC, it's one or more of the bomberman titles that does this.
Everything from Charles' documents are from him testing it first hand, but one or two things he references from the official documentation (still verified though).
Off-topic: He never finished investigating unmarked data access in the VDC 8 slot dot clock diagram. I suspect it's a redundant BAT read that ORs against the first, because when I was racing the beam for palette updates on the BAT, I saw what looked like that as a graphic artifact. But we never confirmed it.
|
|
|
Post by dshadoff on Feb 14, 2020 16:38:25 GMT
In my original review of his document, only one of the two items was explicitly listed; the other was merely implied.
I recall working with him on some of the tests. And I saw conversations between Ryphecha and him about contradictory evidence to one or two of his results - just saying that it's always better to doubly-confirm, in case the original test was not comprehensive enough, or if the behavior is rare enough. And CSL usage seems very rare...
|
|