Re: [Simulavr-devel] two things

[Theodore A. Roth]

On Thu, 20 Nov 2003, John Regehr wrote:

> What would you think of code like this in intvecs.h?  The idea is that
> each interrupt vector is associated with two bits: a flag bit and an
> enable bit.
>
> typedef struct _IntBit IntBit;
>
> struct _IntBit {
>     int addr;
>     int bit;
> };

I'd change that to look like this:

 struct _IntBit {
     int addr;
     uint8_t bit;
 };


>
> typedef struct _IntVect IntVect;
>
> struct _IntVect {
>     char *name;                 /* The name of the interrupt. */
>     int   addr;                 /* Where to vector to when interrupt occurs. 
> */
>     uint32_t can_wake;          /* If the interrupt occurs while in sleep
>                                    mode, can it wake the processor? Each bit
>                                    represents a different sleep mode which the
>                                    processor might be in. */
>     IntBit enable;
>     IntBit flag;
> };
>
> The semantics, as I understand things, are these:
>
> There are two kinds of interrupts (I get this from ATmega128(L) docs page
> 14).
>
> The first has a flag bit, which becomes set when the interrupt is
> triggered (or it can be set programmatically -- who sets the bit is
> irrelevant).  The interrupt fires any time the flag bit is set && the
> enable bit is set && the global enable bit is set.
>
> The second kind of interrupt does not necessarily have a flag bit and if
> the condition disappears before the CPU has a chance to handle the
> interrupt, then the interrupt is lost.

That sounds logical to me. The naming of enable and flag adds a bit of
confusion to your description though, but that can be cleaned up a bit.

Ted Roth