Re: IRQ's in Amiga VSTa

>
> Did I make the P-code idea clear? I think a threaded interpreter would be
> overkill, and also most of the fast TIL (Threaded Interpretive Language)
> interpreters I've worked with would be difficult to make position
> independent.
>
> Here's some off-the-cuff code ideas.
>

maybe i'm missing the point of all this, but a couple of things come to
mind...

it seems to me that for performance, you'd want to "compile" the code at
driver registration time, rather than do the interpretation at interrupt
time.

but i'd think for flexibility, since you can't _really_ predict what
a specific driver will need to do in terms of masking and checking and
restoring state etc, you should let the driver-writer simply provide a
block of assembly code. this is more dangerous from a security point of
view, but then, installing a driver at all is somewhat dangerous from
_that_ point of view, and shouldn't be possible unless you're "trusted"
to some extent, right? the kernel coulld in principle do some address and
opcode checking of the passed-in assembly code to be sure they're not doing
anything dumb, like jumping to or overwriting location zero, or not doing
the proper returns, or whatever. (i'd think you'd want to do that sort of
sanity check in the p-code interpreter you proposed, in any case, to catch
the same kinds of problems.)

paul
---------------------
    paul fox, pgf@foxharp.boston.ma.us (arlington, ma)
Received on Fri Nov 11 11:03:59 1994