[A89] Re: Quite in here... let's discuss something!
[Prev][Next][Index][Thread]
[A89] Re: Quite in here... let's discuss something!
> How does the DMA of the LCD work? And is there a chance for
waitstates(slow
> RAM)?
The DMA works differently on different HW types. On HW1 calculators, it does
slow down the calculator. This is because the lcd fetches the display data
directly from ram taking away precious bus cycles....
On HW2 calculators, the lcd dma doesn't affect the performance. This is
because all writes to the led are snooped, and copied to a special buffer
for the LCD (separate from normal ram). ie the dma doesn't have to
constantly fetch the lcd data from ram and use the bus.... which leaves it
free to do other stuff...
That's all i know :)
-Greg
GForce Programming - http://gforce.calc.org
----- Original Message -----
From: "Gijs Leegwater" <g_leegwater@hotmail.com>
To: <assembly-89@lists.ticalc.org>
Sent: Monday, April 29, 2002 11:14 AM
Subject: [A89] Re: Quite in here... let's discuss something!
>
> No, the read cycles to fetch instruction from RAM are included. The
> instruction move.l d0,d1 costs 4 clocks, just because it costs 4 clocks to
> do one read-cycle(the reading of the instruction). But maybe your version
> IS better, because bus is only used one time with add.l, and with LEA two
> times. With ADD it just has to do more internal things. BTW I have some
> questions about that, maybe someone knows the answer:
>
> How does the DMA of the LCD work? And is there a chance for
waitstates(slow
> RAM)? And what happens if the LCD driver is bus master? Does the processor
> just nothing then? What happens when a BR is requested between 2 read/
write
> cycles form one instruction? Does it just interrupt the execution of that
> instruction, let the requesting device do his bus work, and after that
> continue with the instruction? And when an instruction first reads and
then
> does some internals in the processor, can the processor do that internal
> things while another device is bus master after that read?
>
> Pffffff that are a lot of questions I hope someone can answer......
>
> Bye
> Cheiz
> ----- Original Message -----
> From: Samuel Nyall Stearley <sstear70@calvin.edu>
> To: <assembly-89@lists.ticalc.org>
> Sent: Monday, April 29, 2002 4:27 PM
> Subject: [A89] Re: Quite in here... let's discuss something!
>
>
> >
> > What I have always wondered is if those clock cycles given in various
> books
> > takes into account the time to fetch the instructions from ram. They
> might
> > just be how long they take to acomplish in the processor once the
> instruction
> > fetch is done. So if you can reduce the code in the loop, then you have
> > reduced the time in the instruction fetches, and so the loop runs
faster.
> So
> > then the moveq #30,dx would be an optomization.
> >
>
>
References: