coq-club AT inria.fr
Subject: The Coq mailing list
List archive
- From: Jonathan Leivent <jonikelee AT gmail.com>
- To: Coq Club <coq-club AT inria.fr>
- Subject: [Coq-Club] auto performance vs. match (was: hint cost vs. order)
- Date: Tue, 20 Oct 2015 11:23:10 -0400
- Authentication-results: mail3-smtp-sop.national.inria.fr; spf=None smtp.pra=jonikelee AT gmail.com; spf=Pass smtp.mailfrom=jonikelee AT gmail.com; spf=None smtp.helo=postmaster AT mail-qk0-f175.google.com
- Ironport-phdr: 9a23:AiB5GhBcdE+t1UDKxXDPUyQJP3N1i/DPJgcQr6AfoPdwSP78pcbcNUDSrc9gkEXOFd2CrakU1qyO6+u+AiQp2tWojjMrSNR0TRgLiMEbzUQLIfWuLgnFFsPsdDEwB89YVVVorDmROElRH9viNRWJ+iXhpQAbFhi3DwdpPOO9QteU1JTqkbjqsMOIKyxzxxODIppKZC2sqgvQssREyaBDEY0WjiXzn31TZu5NznlpL1/A1zz158O34YIxu38I46Fp34d6XK77Z6U1S6BDRHRjajhtpZ6jiR6WZgyWrlAYT29exhFPGk3O6Azwdpb3qCrz8ORnjnq0J8rzGJIzXz2+765tADvliTkKMSJxpGPQjM1zgaZWrTquohV+x8jfZ4TDZ6k2Rb/UYd5PHTkJZc1WTSEUR9rkN4Y=
On 10/15/2015 07:10 PM, Jonathan Leivent wrote:
...
Interestingly, it looks like using auto with nocore and a small db can outperform combining the hints from that small db into a single recursive match-based tactic. However, I'm finding it difficult to use hint cost to control the order in which hints are tried.
Some more about this: The cost-vs-order for hints bug (3199) has been fixed in the latest 8.5 builds, but the performance advantage of [auto with nocore mydb] with a set of Extern hints in mydb still exists over a recursive Ltac tactic where each of those "hints" is a branch of a single match construct. In one case I have, its 464 seconds vs. 579 seconds - about a 20% improvement.
Of course, there are limitations to using auto as a replacement for recursive match - the obvious one being that auto must either solve all goals or leave them unchanged.
-- Jonathan
- [Coq-Club] hint cost vs. order, Jonathan Leivent, 10/16/2015
- Re: [Coq-Club] hint cost vs. order, Jason Gross, 10/16/2015
- Re: [Coq-Club] hint cost vs. order, Jonathan Leivent, 10/16/2015
- Re: [Coq-Club] hint cost vs. order, Jason Gross, 10/16/2015
- Re: [Coq-Club] hint cost vs. order, Jonathan Leivent, 10/16/2015
- Re: [Coq-Club] hint cost vs. order, Jason Gross, 10/16/2015
- Re: [Coq-Club] hint cost vs. order, Jonathan Leivent, 10/16/2015
- [Coq-Club] auto performance vs. match (was: hint cost vs. order), Jonathan Leivent, 10/20/2015
- Re: [Coq-Club] hint cost vs. order, Jason Gross, 10/16/2015
Archive powered by MHonArc 2.6.18.