[Haskell-cafe] Re: Atom bug(?)

[Tom Hawkins]

On Tue, Jan 19, 2010 at 10:42 PM, Lee Pike <leepike at gmail.com> wrote:
> Tom,
>
> Is this a bug?  The following program compiles, but the rule is scheduled at
> period 1 (rather than 0).  I wouldn't have thought to have an assignment
> outside of an atom until another engineer here wrote it.  In any event, I
> would have expected that the rule should have period 3.
>
> Lee
>
> example :: Atom ()
> example = period 3 $ do
>
>  a <- word32 "a" 0
>
>  a <== 21

The top-most rule is implied with the compile command.  This makes it
appear as if the assignment doesn't belong to a rule, when if fact it
belongs to the implicit top level rule.  The top level rule always has
a period of 1.

Period constraints are only applied to sub-rules, not the current
rule.  This prevents ambiguous situations, such as:

atom "something" $ do
  period 1 (a <== 1)  -- Does 'something' execute every 1 or ...
  period 2 (b <== 2)  -- 2 cycles?

So in your example, 'period 3' has no sub rules that it applies to.

-Tom