Re: A86: 5MHz

[Prev][Next][Index][Thread]

Re: A86: 5MHz


While writing a mouse routine for several different programs, I found that
certain places in the screen would not always refresh at the right time, so that
the mouse would virtually disappear until you stopped moving it. I tried it with
several sprite routines, all with the same result.
An example:
KeyLoop:
 ;clear the sprite
 ;check a bunch of keys (ports)
 ;draw the sprite
 ;save coords
 jr KeyLoop
The key checking was quite extensive, but the routine seemed to be extremely
fast. If i put a crapload of halts in the loop, then the mouse disappearace went
away.

Dont know if this has any relevance, but I thought I would share it in case it
does.

Later,
Chicane
Aaron Curtis wrote:

> I'm not going to claim to know anything about hardware, but I'm pretty
> sure the CPU doesn't pause during a screen refresh.  A while back I
> tried some code that was set up like this:
>
> loop:
>         ;draw a sprite
>         ;halt
>         ;erase it
>         ;minor amounts of code...
>         jr loop
>
> If the sprite was near the top of the screen, you could see it; if it
> was near the bottom, you couldn't.  The screen update occurs when the
> halt is broken by an interrupt, I think we can agree on that...  So what
> I figure was going on, was that the top of the screen was refreshed
> while the sprite was still in video mem, but the bottom was done after
> the sprite was erased.
>
> David Phillips wrote:
> >
> > My guess is that the answer to your question, and the solution to Kirk's
> > "problem", would be that the calc uses the chip that generates interrupts to
> > do the screen refreshes.  Even if interrupts are disabled, the chip still
> > runs at the same frequency, which is why the screen still works.  This is
> > supported by the fact that when you change the interrupt rate, the screen
> > gets distorted.  It probably uses DMA to transfer the data, which is why the
> > screen must start on an 256 byte boundary.  Since the memory is not dual
> > port, meaning only one thing can access it at a time, the CPU is paused
> > during the DMA transfer.
Follow-Ups: References: