[Soplex] Reporting results for soplex

Thu Mar 6 15:44:40 CET 2025

Hi Ambros,

Yes, if I run the tool as you said, I indeed get the same result, which is what I needed.
What confused me is the fact that, when I used soplex as a c++ library, calling `numRefinements` would give me completely different results. Now that I have confirmation something is wrong with my code, going over it again I think I found the root cause. I was setting `RATFAC_MINSTALLS` to 0, while the default (and what I should use) is 2. Now it seems like the output is consistent.
In any case, I will just use the metrics I get with the `-q` flag, as you suggested.

Cheers,
Ernesto Casablanca
________________________________
From: Ambros Gleixner <ambros.gleixner at HTW-Berlin.de>
Sent: 06 March 2025 13:45
To: Ernesto Casablanca (PGR) <E.Casablanca2 at newcastle.ac.uk>; soplex at zib.de <soplex at zib.de>
Subject: Re: [Soplex] Reporting results for soplex

⚠ External sender. Take care when opening links or attachments. Do not provide your login details.

Hi Ernesto,

Good to hear.  numRefinements should not be always 0, e.g.

bin/soplex --readmode=1 -f0 -o0 ../../check/instances/etamacro.mps  -q

gives me

Refinements         : 3
   Stalling          : 2
   Pivoting          : 1
   Feasibility       : 0
   Unboundedness     : 0

Can you confirm that?

If you have a particularly easy test set, it may of course be true that
no refinements are needed.

Best,
Ambros

Am 06.03.2025 um 14:25 schrieb Ernesto Casablanca (PGR):
> Hi Ambros,
>
> Thank you for the insight! The statistics reported with the `-q` are
> perfect for my use case, so I will be using it.
>
> If I compare the results I get with the ones in https://static-
> content.springer.com/esm/art%3A10.1007%2Fs10107-019-01444-6/
> MediaObjects/10107_2019_1444_MOESM1_ESM.pdf <https://static-
> content.springer.com/esm/art%3A10.1007%2Fs10107-019-01444-6/
> MediaObjects/10107_2019_1444_MOESM1_ESM.pdf>, I can still see some
> mismatches (e.g.,  app1-2 or atlanta-ip), but I assume it is due to the
> parameters I'm using and the fact that Soplex's version has changed.
> The vast majority are the same.
> Also, it may be worth checking the method `numRefinements`: unless I'm
> misusing it (which is likely), I think the number reported is incorrect
> (it seems to always be 0).
>
> Cheers,
> /Ernesto Casablanca/
> ------------------------------------------------------------------------
> *From:* Ambros Gleixner <ambros.gleixner at HTW-Berlin.de>
> *Sent:* 05 March 2025 11:33
> *To:* Ernesto Casablanca (PGR) <E.Casablanca2 at newcastle.ac.uk>;
> soplex at zib.de <soplex at zib.de>
> *Subject:* Re: [Soplex] Reporting results for soplex
> ⚠ External sender. Take care when opening links or attachments. Do not
> provide your login details.
>
> Hi Ernesto,
>
> have you tried the -q option on the command line?  Are the statistics in
> in this table sufficient?
>
> Best,
> Ambros
>
> Am 03.03.2025 um 17:43 schrieb Ernesto Casablanca (PGR):
>> Hi all,
>>
>> I am running some LP benchmarks, and I would like to compare
>> SoPlex's performance with other approaches only considering exact solutions.
>> For this reason, I'm launching the tool with the following command line
>> frags: `--readmode=1 -f0 -o0`.
>> Then, from the output I need to collect data about the result and, most
>> importantly, some collateral information such as number of refinements
>> and the precision the solver operated at last (if precision boosting
>> occurred).
>> Comparing the output I get from the standalone tool against the
>> information I collect using SoPlex as a c++ library, I noticed some
>> mismatches that may arise from the fact I'm not interpreting the output
>> correctly.
>>
>> Since I do not want to miss-represent your work, what do you suggest
>> would be the best way of recording those two metrics or any other metric
>> you think is relevant given the configuration?
>> Can I collect them from the standard output, or do I necessarily need to
>> use SoPlex as a library?
>>
>> I have attached an example output I would like to collect the metrics from.
>>
>> Thanks,
>> /Ernesto Casablanca/
>>
>> _______________________________________________
>> SoPlex mailing list
>> SoPlex at zib.de
>> https://eur03.safelinks.protection.outlook.com/?
> url=https%3A%2F%2Flistserv.zib.de%2Fmailman%2Flistinfo%2Fsoplex&data=05%7C02%7CE.Casablanca2%40newcastle.ac.uk%7C0b0384d46d43465336f808dd5bd98f37%7C9c5012c9b61644c2a91766814fbe3e87%7C1%7C0%7C638767713185248279%7CUnknown%7CTWFpbGZsb3d8eyJFbXB0eU1hcGkiOnRydWUsIlYiOiIwLjAuMDAwMCIsIlAiOiJXaW4zMiIsIkFOIjoiTWFpbCIsIldUIjoyfQ%3D%3D%7C60000%7C%7C%7C&sdata=yM%2FXSw0oaF%2B9rbWYDkspW4qCd4GZL7WwxJfZ%2Buc1KWk%3D&reserved=0 <https://eur03.safelinks.protection.outlook.com/?url=https%3A%2F%2Flistserv.zib.de%2Fmailman%2Flistinfo%2Fsoplex&data=05%7C02%7CE.Casablanca2%40newcastle.ac.uk%7C1c8fc2eb60e940da959d08dd5cb52d85%7C9c5012c9b61644c2a91766814fbe3e87%7C1%7C0%7C638768655975659203%7CUnknown%7CTWFpbGZsb3d8eyJFbXB0eU1hcGkiOnRydWUsIlYiOiIwLjAuMDAwMCIsIlAiOiJXaW4zMiIsIkFOIjoiTWFpbCIsIldUIjoyfQ%3D%3D%7C80000%7C%7C%7C&sdata=koLCOI6J4L4i9h1He7Wi5qhZHmzGJz9Vd8X9cnfKdu8%3D&reserved=0<https://listserv.zib.de/mailman/listinfo/soplex>>
>
> --
> Prof. Dr. Ambros Gleixner - HTW & Zuse Institute Berlin
> https://eur03.safelinks.protection.outlook.com/?
> url=https%3A%2F%2Fwww.htw-
> berlin.de%2Fhochschule%2Fpersonen%2Fperson%2F%3Feid%3D12822&data=05%7C02%7CE.Casablanca2%40newcastle.ac.uk%7C0b0384d46d43465336f808dd5bd98f37%7C9c5012c9b61644c2a91766814fbe3e87%7C1%7C0%7C638767713185267848%7CUnknown%7CTWFpbGZsb3d8eyJFbXB0eU1hcGkiOnRydWUsIlYiOiIwLjAuMDAwMCIsIlAiOiJXaW4zMiIsIkFOIjoiTWFpbCIsIldUIjoyfQ%3D%3D%7C60000%7C%7C%7C&sdata=hGrGuljlaJ%2FzzCOS47vK20AF5CkPOySydhevdGRvXa0%3D&reserved=0 <https://eur03.safelinks.protection.outlook.com/?url=https%3A%2F%2Fwww.htw-berlin.de%2Fhochschule%2Fpersonen%2Fperson%2F%3Feid%3D12822&data=05%7C02%7CE.Casablanca2%40newcastle.ac.uk%7C1c8fc2eb60e940da959d08dd5cb52d85%7C9c5012c9b61644c2a91766814fbe3e87%7C1%7C0%7C638768655975673241%7CUnknown%7CTWFpbGZsb3d8eyJFbXB0eU1hcGkiOnRydWUsIlYiOiIwLjAuMDAwMCIsIlAiOiJXaW4zMiIsIkFOIjoiTWFpbCIsIldUIjoyfQ%3D%3D%7C80000%7C%7C%7C&sdata=OCZ42o9n2K9%2B0CzBZMnQSWq5eWg8u9Kj%2BIqjzRosPw8%3D&reserved=0<https://www.htw-berlin.de/hochschule/personen/person/?eid=12822>>
>

--
Prof. Dr. Ambros Gleixner - HTW & Zuse Institute Berlin
https://eur03.safelinks.protection.outlook.com/?url=https%3A%2F%2Fwww.htw-berlin.de%2Fhochschule%2Fpersonen%2Fperson%2F%3Feid%3D12822&data=05%7C02%7CE.Casablanca2%40newcastle.ac.uk%7C1c8fc2eb60e940da959d08dd5cb52d85%7C9c5012c9b61644c2a91766814fbe3e87%7C1%7C0%7C638768655975682398%7CUnknown%7CTWFpbGZsb3d8eyJFbXB0eU1hcGkiOnRydWUsIlYiOiIwLjAuMDAwMCIsIlAiOiJXaW4zMiIsIkFOIjoiTWFpbCIsIldUIjoyfQ%3D%3D%7C80000%7C%7C%7C&sdata=WOiZ8v6w7T4kDZHkruDE1DCpnglKY%2FqX0oMG%2FAKrCYI%3D&reserved=0<https://www.htw-berlin.de/hochschule/personen/person/?eid=12822>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.zib.de/pipermail/soplex/attachments/20250306/90f2942b/attachment.html>