[m-rev.] for post-commit review: inst_var_sub_result

[Peter Wang]

On Fri, 05 Jul 2024 13:49:45 +1000 "Zoltan Somogyi" <zoltan.somogyi at runbox.com> wrote:
> For review by anyone.
> 
> Zoltan.

> Avoid overuse of higher order code.
> 
> compiler/inst_match.m:
>     Change how handle_inst_var_sub operates. Instead of taking as arguments
>     two predicates to call as continuations, simply return a structure
>     that tells its callers which of those two predicates to call, and
>     with what arguments. The new arrangement has the advantage of
> 
>     - separating handle_inst_var_sub's task from its callers' tasks,
>     - making the flow of control more visible and more easily understandable,
>     - allowing direct tail recursion in a situation where the old code did
>       tail recursion indirectly, through a higher-order call.
> 
>     The new arrangement does incur the cost of constructing a small cell
>     on the heap, but this cost should be similar to, or even smaller than
>     the cost of old arrangement's construction of closures.

That looks good.

Peter