Best msieve poly scores

[swellman]

Originally Posted by VBCurtis

RSA-240, from the paper summarizing the job as scored by cownoise:
221784.77118 2.95669394e-17

Edit: RSA-250, as scored by cownoise:
382126.36029 7.00754597e-18

A nice discussion of parameter choice and the polys for both these jobs can be found at https://link.springer.com/chapter/10...-030-56880-1_3

Nice addition. I entered this value in the 2021 table.

[charybdis]

If we're adding scores for enormously large numbers, there's also this paper which gives a polynomial for RSA-768 (232 digits) with score 7.690e-17, better than the original poly used for this number.

RSA-230 and RSA-232 have both been factored but I couldn't find the polynomial for either.

[swellman]

Originally Posted by charybdis

If we're adding scores for enormously large numbers, there's also this paper which gives a polynomial for RSA-768 (232 digits) with score 7.690e-17, better than the original poly used for this number.

RSA-230 and RSA-232 have both been factored but I couldn't find the polynomial for either.

I contacted Dr. Sam Gross, the person who factored RSA-230. He provided the polynomial he used (e-score 8.065e-17) and I have updated the table above accordingly. See this thread for more details.

Notes in the announcement of RSA-232 being factored mention plans to publish the details in a “forthcoming scientific paper”, so we will just have to wait on that.

[swellman]

The recent interest in older RSA GNFS polys has caused me to relook at some of the older more conventional factorization records.

I located one job involving a C212 cofactor of 10,770M, factored by frmky back in Sep 2013, though the deg 5 poly itself was found in June 2013 by Gimarel per this post. Gimeral’s poly was used by frmky in the factorization. The confirmed e-score is 6.698e-16, considerably better than the current listed deg 5 record for a c212 (6.435e-16 in Nov 2015 on another factorization job seemingly by frmky?).

Unless I’m mistaken here, the c212 record score needs to be updated once someone confirms my findings.

[swellman]

More searching through old posts of the forum turned up another high poly score.

Gimarel found a poly score of 2.263e-13 on a C175 back in Dec 2015. This is the record score for C175. The table should be updated - can someone else confirm this finding? TYIA.

[swellman]

Found yet another. The best C195 score seems to be here. Escore of 1.044e-14, found by RichD in March 2016. It appears to be the record for a C195.

[VBCurtis]

What you're doing is what I did to start this table- just make the updates with any poly you find. I don't doubt that I made some mistakes!
The only ones I intentionally omitted were when someone 'spun' a poly after the factorization job was run- I don't think setting new records on previously run jobs is a good idea. Though for RSA numbers, I think new poly searches are a worthy way to measure current tools / techniques against previous state-of-the-art.

[swellman]

Originally Posted by VBCurtis

What you're doing is what I did to start this table- just make the updates with any poly you find. I don't doubt that I made some mistakes!
The only ones I intentionally omitted were when someone 'spun' a poly after the factorization job was run- I don't think setting new records on previously run jobs is a good idea. Though for RSA numbers, I think new poly searches are a worthy way to measure current tools / techniques against previous state-of-the-art.

My request for verification is meant only as a form of quality control. Keeps me from editing the table with an erroneous value and tossing aside the old record with no obvious means of backup. Not the end of the world, and we could possibly recreate the old entry, just trying to be careful is all.

Sometimes forum members would alert me to errors in table, though now not for many months.

[jwaltos]

https://members.loria.fr/PZimmermann/talks/unsw.pdf

Although not containing specific values generated by msieve perhaps this PPT presentation may provide an assist in reviewing prior approaches and developing new ones. There are other papers of this kind and quality but this one is succinct.

[charybdis]

Originally Posted by jwaltos

https://members.loria.fr/PZimmermann/talks/unsw.pdf

Although not containing specific values generated by msieve perhaps this PPT presentation may provide an assist in reviewing prior approaches and developing new ones. There are other papers of this kind and quality but this one is succinct.

This presentation was based on the work in this paper, and the methods are implemented in CADO. I don't think they've been implemented in msieve yet.

[jwaltos]

Originally Posted by charybdis

This presentation was based on the work in this paper, and the methods are implemented in CADO. I don't think they've been implemented in msieve yet.

Correct on both counts. The methods outlined in the paper cited aren't implemented in msieve but they provide a guideline on how to view msieve results comparatively. The PPT presentation sketches out everything necessary and includes a result from one of the forum moderators.

I'm working on a promising alternative method where such published work provides some solid guidelines which can be transformed into other forms and the results compared accordingly. Any possible source of insight to a vexing question is always welcomed.