Linear algebra reservations, progress and results

[VictordeHolland]

Originally Posted by YuL

C189_139_46 is factored.

Fri Oct 30 12:07:40 2015  p49 factor: 1642812652174511413643627817470235409676558815837
Fri Oct 30 12:07:40 2015  p62 factor: 31142370957270822179830901941415049974944773031624240161640067
Fri Oct 30 12:07:40 2015  p80 factor: 14213880718999371928288929659174388924378336829981761318130804503479990602736603

~64.7h of LA on 10.8M matrix, td=116, 2x -t 6 on dual Xeon E5-2620GHz.

Hmm, an ECM miss, the 2t50 should have caught the p49.

[swellman]

Reserving GC_5_342.

[swellman]

Originally Posted by swellman

Reserving C200_146_37.

prp89 factor: 80136501012527292470877827183094885729916121103823725480706052576069752498770052345491721
prp112 factor: 1241192164743127677811288336403214025740355182046736326318112546111395026881278017004862286455457594718387708061

[Xyzzy]

Originally Posted by VBCurtis

The higher the target density, the smaller and faster the matrix, but more relations are needed. Since relation-gathering is highly parallel while matrix-solving is not, NFS@home usually gathers enough extra relations to make 112 or 120 density work, in order to use the post-processor machines more efficiently. Matrix-solving rises roughly with the square of matrix size, so a 10M by 10M matrix will take about 4 times as long as a 5M by 5M matrix.

As an experiment we tried to run one of our current jobs (F1415) with target densities greater than 120. We tested 124, 128 and 132. All of those worked and we killed them after they successfully entered the linear algebra stage. However, the log files show that all of those tries had the same matrix "output" as our reference run at 120. Does that sound right?

[fivemack]

That worries me slightly: could you check that the log files actually have a line like

Sat Jun 20 00:27:20 2015  setting target matrix density to 110.0

The command-line parser has been known to misbehave and not pick up the density instruction if the parameters are in an order it didn't expect.

[Xyzzy]

They all had the appropriate line indicating that they were using the specified target density.

[VictordeHolland]

I'd like to do post-processing of F2145 once it has finished sieving.

[swellman]

I'd like to reserve 142^141+141^142.

On the same subject

Originally Posted by wombatman

It'll be a little while, but I'd like to go ahead and reserve C184_HP2_4496_294. Thanks!

[Xyzzy]

We had to run the work over so apologies for the delay.

These two files are for F1415. One is with a target density of 120 and the other is 144.

[Xyzzy]

Originally Posted by VictordeHolland

I'd like to do post-processing of F2145 once it has finished sieving.

Originally Posted by swellman

I'd like to reserve 142^141+141^142.

We have permission to update reservations.

We will attempt to process all reservations in this thread.

If we mess up or forget a reservation just PM us.

[fivemack]

Fri Nov  6 01:57:00 2015  p74 factor: 94037914744837516365792202572404862737090804917962567209076939374883165797
Fri Nov  6 01:57:00 2015  p140 factor: 72467356675331713119814628540745991940365279476850029500180554255521429798613354156977015297427214588300165871252308900733837057376096318307

37.5 hours for 10.1M matrix (density 110 - not enough relations for 120) on eight cores Xeon E5-2650v2

Taking GW_3_501 next