[QUOTE=VBCurtis;418854]Dubslow-
Both CADO and msieve have A1 as a user-controlled parameter/search range. If you mean that CADO's best poly in a range often has higher A1 than msieve's best, I'm not sure- I've only used CADO for one poly search, and I ran a really wide range (like 0-20M).[/QUOTE]

I figured the former, and I'm reasonably certain I mean the latter. Perhaps I'm imagining it.

[QUOTE=VBCurtis;418856]Yes, though what is "better" is the issue. My measure of value is greatest number of -nps hits of norm better (sorry, lower) than {pick a number} per day. What stage1norm produces the highest rate of quality hits? Note I'm not talking about actual poly E-score; that's a crapshoot only lightly correlated with settings of stage1/stage2.

setting stage2norm merely alters the output that is kept for root-opt. It's not valuable to run root-opt on thousands of candidates, so some cutoff is useful for brevity. Default stage2norm produces *thousands* of candidates per hour, not helpful.

A few days of searching for the C198 should illuminate the worst stage2 size that produced a possibly-useful poly; we can set the cutoff for worthy candidates from -nps just above that number.[/QUOTE]

I gotcha. Do you normally combine -np1 and -nps or run them separately? Running separately would make more sense to me in terms of speed since you wouldn't be limited by CPU, although if the stage 2 norm is somewhat reduced, it may not be as big an issue.

I always run them together, but I'm surely willing to run them separately to note effects.

stage 2 norm should have no bearing whatsoever on -np1 or -nps time; it only controls which outputs from -nps efforts get written to file. Altering stage2norm has the same effect as leaving it default, sorting the .ms file by 10th column (sort -g -k 10), and truncating it.

Altering stage1norm changes what space the GPU looks at, AND what gets fed to the -nps size-optimization step.

Ah! I had thought stage 2 norm affected the actual operation of -nps (beyond what is output).

FYI, in running the C198, dividing the default stage 1 and stage 2 norms by 10 has still resulted in 4065 lines in the .ms file (in approximately 2 hours) when starting from the lowest A1.

For my initial go at the C198, I ran with stage 1 norm 1e29, found 270k stage-1 hits in 4.3 hours, ran -nps on all 270k hits in four minutes with 17 of them having stage-2 norm less than 5e26, and (in three and a half minutes) found the best Murphy_E score from one of those seventeen was 4.756e-15.

The msieve default stage-2 norms are much too high: I am rather hoping to update them based on the results of the searching I've done on various C18x setting up for nfs@home.

[QUOTE=jasonp;418830]It's the integral of the square of the algebraic polynomial over a rectangular sieve region; lower is better, this is basically the RMS size of one sieve value.[/QUOTE]

Is there an option to set the size of the region the poly's integrated over? Ideally we want to set it the same as the area the siever we plan to use sieves over.

Chris

Another comment, msieve runs -npr single threaded, so with my new GTX 970 I found I needed to split it up as follows: [code]
CPUS=4 # Number of CPUs to use

# Stage 1 can produce duplicate lines so use uniq to get rid of them if the whole line (not just the key) matches the previous line.
sort -g -k $KEY msieve.dat.ms | head -1000 | uniq | split -nr/$CPUS - $NAME.x

# Now run -npr in parallel over each set of entries.
for a in $NAME.x*
do
mv $a $a.msieve.dat.ms
# Now root optimise the best few results.
/home/chris/msieve-svn988.2/trunk/msieve -npr -v -s $a -l $a.log -p $N >$a.npr.sysout &
done

wait # Until they all finish
[/code]
WARNING, I've not tested that, it's updated so you can set the number of CPU in one place. And pulled out of the middle of a script.

Chris

Surely you want to run uniq before head!

My first 12 hr on the C198 yielded 31 hits of 5e26 or better, and 42 more between 5e26 and 7e26.
-npr on the first set yielded 4 polys of 4.6-15 or better, with best at 4.97.
-npr on the other 42 yielded 2 polys of 4.6e-15 or better, best 4.72.

I suggest those of you intending to give this a few days' effort should send/post polys of 4.7e-15 or better; each search should generate a couple dozen of these a week, though a lucky burst might produce 20 from a single hit.

Some data points from first slice of GPU search
 

7e6..8e6 with stage 1 norm 1e29, stage 2 norm 5e26 got me 276 stage-2 hits, which went through -npr with min_evalue=4e-15 to give 93 hits with e>5e-15 and best norm 5.877e-15.

The 5.877e-15 came from a .ms line ending 1.34e+26:
[code]
7718940 -18303914090282568 -4985239051957647559821781 1523101065645887047248423855483 176413646267219899511785503934614217048 -16029243451110913224986242168580653752345319 21364512235934882597 -113765533253841237794176648402461606488 -2.05 1.341987e+26
[/code]

The .ms line with smallest norm was

[code]
7414680 41633122213142779 1024731177395438183977054 -2255594558179915548725058504389 -23396362267492306545158206978439230744 15479160595648114695353768807764374689993469 25231289452558154117 -114684244297937959244409663229829581058 -1.74 3.517414e+25
[/code]

but -npr gave a best hit with 4.246e-15 from that ... are there parameters that can profitably be tweaked?

Second-best was
[code]
7526244 2710096378710132 -742064360986204763209815 -2500399870535819554262634778630 289551401493120657655213377118439460189 292188113410400452689806487694850345257673555 3093808275468295657 -114342209747469697220340926442807128378 -1.60 5.671853e+25
[/code]

from which -npr got a 5.166e-15

I have two hits in the 5e+25 range, and one poly better than 5e-15 at 5.33. I've had about 100 hits better than 5e+26 so far, using 2e29 for stage1 norm yields more hits per day than 1e29 on my 750ti.