NFS Square Root failed

[kurtb]

Thanks for the important information.

I'm trying to recreate the matrix with more row relations (so far 735M unique rels.)
The very large input-relations-files are now deleted for space reasons.
Kurt

[wreck]

From the binary name, which seems compiled by ATH,
it should be 64bit binary GMP. And when I search in the website , it says 64 bit programe could not use 32 bit library.

The second try after use remdup program has begin,
it is a little strange there are still some reading relation error when use msieve. I will ask Kurt to upload the related files to the web if this second try fail again. The data file is about 90GB, two check files each has about 10GB.

The 1st try full log and the 2nd try after remove error read log attached.
Perhaps it will give some useful information if msieve could write the read relation error line content to the log too or write it in msieve.error file, I don't know how to grep a given line in a big file, all text editor programe failed open such large file.

[wreck]

R507 has been factored at the second try successfully, after the following steps:
After msieve -nc2,
-WIN10 update
-new start
-nc3 without others programs.
also, the following difference between the 1st try,
1. Use a different binary which has been factored R343 successfully.
2. Use density 120, sieve another 32M new raw relations(with 50% duplicate), the purpose is to reduce the matrix size.
3. Kurt gets another 2TB external disk to save the backup files after -nc2.
4. The coefficients of R507's nc3 is 4430.16 million bits, still bigger than 4096Mb, but smaller than R1030's(c199,4621.15Mb).
5.Check all relations with remdups before -nc2, when I check the file badrels.txt, there are 19 lines content is "Hello", one of them is bellow,

4098140461411,6907:45d,96bc0f9,85374f5,294d517,1d2a9b,20b1f:4163d316868609662,3169:2f192f89,1192178b9,624133,1d12f,10133,c0b5:1dc55c90b,1b4487e8b,157c6b7f,1517,4a9,445,456e17,f14a1,1c037 Hello

there also another line contain many ? and "389T bad rels!!!"
I'm not sure why it is there and whether it would has some influence to msieve.
----------
Now we are factoring R459 snfs 306, nextly R1620L gnfs185.
The R507's post-processing log is attached.

[jasonp]

Congrats on a huge job complete!

The postprocessing has needed a lot of hardening because we have seen all kinds of weird stuff in relation output lines. The one you posted also has a very large initial algebraic factor that looks like many small factors without commas between them; possibly this is uninitialized memory from the sieving tools that was flushed to dusk when a sieve process crashed.

[fivemack]

"Hello" is very peculiar!

I have quite often seen corrupt lines with correct X,Y but where some of the factors have been concatenated with some hex digit rather than ',' as the separator; I wouldn't be surprised if that were the effect of one flipped bit in the .gz file confusing the dictionaries. I've poked at one or two for fun, but it didn't seem worth making a particular effort to automate recovering a few hundred relations out of a few hundred million.