2018 14e post processing reservations and results

[pinhodecarlos]

Fri Jun 01 03:08:20 2018  p67 factor: 2753638177475221936289540926431715709718268228057683656745239296463
Fri Jun 01 03:08:20 2018  p126 factor: 868089586342800554478400027452634546066331816298590736409139756490987251040504996644207750674345753871018025603989237066929183

https://pastebin.com/M13zzvd7

[pinhodecarlos]

Let me try C242_14831_59, hope it fits 16GB. Which td should I use, 130?

[fivemack]

Originally Posted by pinhodecarlos

Let me try C242_14831_59, hope it fits 16GB. Which td should I use, 130?

Try 130; if that doesn't work try 100, if that does work try 115. Binary search :)

(I have a script which does the binary search to get to the nearest 2, but I think it almost always isn't worth going that far - except for truly enormous jobs, the time saved by optimal choice of density is less than the cost of even one extra filtering run)

[fivemack]

M127_k24 is entirely my fault for having taken a good GNFS polynomial then submitted it for sieving on the rational rather than the algebraic side. As penance, I'll do the post-processing, starting in four days when my next post-processor becomes free.

[pinhodecarlos]

Originally Posted by fivemack

Try 130; if that doesn't work try 100, if that does work try 115. Binary search :)

(I have a script which does the binary search to get to the nearest 2, but I think it almost always isn't worth going that far - except for truly enormous jobs, the time saved by optimal choice of density is less than the cost of even one extra filtering run)

In fact I’ve set td to 140 and it is still on filtering phase for the last 3-4 hours.

[pinhodecarlos]

C242_14831_59 failed with td=140 and td=110, trying 100.

[pinhodecarlos]

Originally Posted by pinhodecarlos

C242_14831_59 failed with td=140 and td=110, trying 100.

Failed with td=100, needs more relations. Unreserving C242_14831_59.

[debrouxl]

Alright, but the upper end of the range is already at 400M, whereas the management page recommends an ending Q of only 362M

[RichD]

Originally Posted by pinhodecarlos

C242_14831_59 failed with td=140 and td=110, trying 100.

The TD=110 log shows:

Sat Jun 02 05:40:32 2018  found 25395886 cycles, need 17115497

Something is not quite right here!

[fivemack]

That is badly over-sieved, I would cut off the last hundred million lines and try the filtering again.

[pinhodecarlos]

Originally Posted by fivemack

That is badly over-sieved, I would cut off the last hundred million lines and try the filtering again.

I’m already running gap on this machine for the next 10 days.

Why wasn’t this composite sieved with 31 bits, aiming for ~210M unique relations? It was possible in the past. Something is escaping me since all the 32 bits on 14e need relations to be cutoff during the filtering stage. Will the 32 bits (14e siever) be quicker on post-processing?