[m-rev.] diff: alternative fix for bug 264 test case
mark at mercurylang.org
Mon Nov 23 01:09:59 AEDT 2015
On Mon, Nov 23, 2015 at 12:23 AM, Zoltan Somogyi
<zoltan.somogyi at runbox.com> wrote:
> On Sun, 22 Nov 2015 17:11:58 +1100, Mark Brown <mark at mercurylang.org> wrote:
>> I've now checked these. Abstract unify is not a problem, but merge had
>> a similar bug to the ones in inst matching; see the attached diff,
>> which has been committed.
> Have you tested the performance effect of this change, especially
> on unusually large predicates? I suspect that the additional traversals
> over insts could cause significant slowdowns. A possible way to avoid that
> would be to add a field to inst_results to record whether the inst
> contains what the new tests look for.
Yes; I saw the comment you wrote in prog_data.m.
Given where the test is performed now, I think it will be at least as
significant as inst_result_contains_any and therefore should be added.
What is a good example to measure on?
More information about the reviews