[Coco] Orchestra 90CC

John Kowalski sock at axess.com
Sun Jan 28 00:00:39 EST 2007


At 08:20 PM 27/01/2007 -0800, Kevin Diggs wrote:
>	With all the recent discussion of the 90cc, I thought I'd ask:
>
>If this thing had had a FIFO and some logic to send samples out at 
>regular intervals how much would that have aided performance? To ask 
>another way, how much of a burden was it that the 6809 not only had to 
>compute the samples but also had to deal with sending them out at the 
>right time?

If the sound has to run in the background (like game sound effects/music),
the overhead just from servicing the interrupts adds up to quite a bit.

The FIRQ interrupt, running at say 12000 times per second will typically eat
up 372000 cycles, or 20% of the CPU just in overhead and register stacking.
If the program were instead able to simply push blocks of sound data
directly to the buffer without timing out each sample with the FIRQ, all
those cycles would be saved.

                                         John Kowalski (Sock Master)
                                         http://www.axess.com/twilight/sock/




More information about the Coco mailing list