[Larceny-users] Parallel assignment optimization and complicatedness.

[William D Clinger]

Jed Davis wrote:

> This last was a problem because complicated? considers any non-primitive
> call to be complicated, and would also give up early due to the
> expression's size.  So, for the sake of experiment, I adjusted
> complicated? to consider a call of a lambda with an uncomplicated body
> the same way as a primitive (which may well not be the right thing in
> general, but this is just a heuristic) and raised the budget (and made
> it tunable with set!).  And that fixes that (and should be attached to
> this message), though I haven't yet tried to measure what if any effect
> it has on actual performance.

Thanks!  That looks like a simple improvement that
yields large dividends.  Just the kind of thing we
like.

We'll run some benchmarks and, if the results go as
expected, add your improvements to Twobit.

Will