Queue management for e_small and 15e queues

VBCurtis · 2020-09-03, 19:45

It's time for a new queue thread!
I grouped e_small and 15e together because the queue maintainers can more easily decide which 15e queue to send to, and requestors need only decide which siever is fastest.

Guidelines for e_small versus 15e (subject to change):

Jobs clearly faster on 14e can still go to the d queue; if there is any doubt whether 14e or 15e is faster, post them here for e_small.

Candidates for e_small should use lims no higher than 134M and Q-range no more than 170M (example: 30M-200M).

Candidates for 15e should use lims no higher than 200M, per Greg's request, to keep memory use low. Unequal lim's that average 200M should use similar memory and are OK.

If you post a job in this thread, it is assumed you test-sieved with 15e.

VBCurtis · 2020-09-03, 20:59

QUEUED AS 5_2_779m

Here's the first job for e_small: 5*2^779-1 (job name: 5_2_779m), an SNFS-236:

Code:

n:  6459594758238867125844847264039052931684724508838707484694754180272905420711047576478995415952837439779741853435639507801334614100750051563929718846753577973898536573574094180100967551289388791
type: snfs
skew: 0.963
deg: 6
c6: 5
c0: -2
Y1: -1
Y0: 1361129467683753853853498429727072845824
rlim: 67000000
alim: 40000000
lpbr: 31
lpba: 31
mfbr: 59
mfba: 58
rlambda: 2.7
alambda: 2.7

Sieving Q-range on the -a side (-r is about 10% slower) from 10M to 52M should yield 165M relations, enough for this job.
Test-sieve is 1kQ:

Code:

#10M 	4459 (0.082 sec)
#20M	4537 (0.090 sec)
#30M	4585 (0.097 sec)
#40M	2532 (0.112 sec)
#50M	3287 (0.124 sec)

I'd like to handle the matrix.

Test-sieving on 14e had sec/rel about 15% lower at the same Q, but having to sieve larger Q makes the overall time estimate longer.
Edit: memory use is under 200MB for this choice of lim. :)

RichD · 2020-09-11, 18:09

QUEUED AS 157__509_19m1

C215 from the OPN file t550.
This polynomial is subject to a higher dup rate.
Sieve on the algebraic side.
This is a 31/32 hybrid with average lim of 200M.
Ready for 15e (full).

Code:

n: 432730660529455347794704796012789152819886765487394609765005017743938589291416004376546543642656612650926905912341612833585509006657095243576134691260953586939652371227204668687273683549423070375850029287782241576971
# 15798461357509^19-1, difficulty: 250.77, skewness: 0.01, alpha: 0.00
# cost: 8.809e+18, est. time: 4194.76 GHz days (not accurate yet!)
skew: 0.006
lss: 0
c6: 15798461357509
c0: -1
Y1: -1
Y0: 3943159792077739562140129933566253623229
type: snfs
rlim: 268000000
alim: 134000000
lpbr: 32
lpba: 31
mfbr: 64
mfba: 90
rlambda: 2.7
alambda: 3.6

Trial sieving 5K blocks.

Code:

chris2be8 · 2020-09-14, 15:53

QUEUED AS f38_158p1

(38^158+1)/2275019624642515721256024045199705 from the Brent tables:

Code:

# Built Sun Sep 13 08:46:36 2020
# Estimated SNFS difficulty 250, GNFS equivalent 182, GNFS difficulty 217, degree 6
n: 1773467026397538884657719287790831691289090712887603204096026865371593098375852318606038829089958612369242895434255050470993200162820803145068057803997016837738483761883286630195767269736259949421100784434078434006761
type: snfs
c6: 1444
c0: 1
# Y0 = 38^26
Y0: 118678899924055521887798989790733175619584
# Y1 = 1^0
Y1: 1
# msieve rating: skew 0.30, size 2.237e-12, alpha -0.105, combined = 2.147e-13 rroots = 0
skew: 0.30
rlim: 120000000
alim: 120000000
lpbr: 31
lpba: 31
mfbr: 62
mfba: 62
rlambda: 2.6
alambda: 2.6

Sieve from 40M to 110M on the algebraic side (this should be suitable for 15e_small).

Test sieving 10 ranges:

Code:

I'll do the LA.

I'm hoping 15e will produce a smaller matrix than 14e with 3LP would.

Chris

RichD · 2020-09-16, 13:46

QUEUED AS 2212657_37m1

C219 from the MWRB file with OPN weight 375.
Ready for 15e_small.

Code:

n: 244157941837026457683416044663893801894719099565048856960414501248388335908064190453921494581019889822027895650099045830136481093754660010842073430491543816794613355436251849838857849430734172083022065036773939924822939
# 2212657^37-1, difficulty: 234.76, skewness: 0.09, alpha: 0.00
# cost: 2.53741e+18, est. time: 1208.29 GHz days (not accurate yet!)
skew: 0.088
c6: 2212657
c0: -1
Y1: -1
Y0: 117350400482093247116855868217109756449
type: snfs
rlim: 67000000
alim: 67000000
lpbr: 30
lpba: 30
mfbr: 60
mfba: 87
rlambda: 2.6
alambda: 3.5

Trial sieving 5K blocks.

Code:

RichD · 2020-09-18, 02:22

QUEUED AS 24571_53m1

C198 from the MWRB file with OPN weight 204.
Ready for 15e_small.

Code:

n: 120022126957597320333159139744749404373249697036952220381427694377455743982586488845453342973632927519924916431246372481337729424921899953831789484868842237453739032873223269277598731643098113494097
# 24571^53-1, difficulty: 237.08, skewness: 5.39, alpha: 0.00
# cost: 3.0485e+18, est. time: 1451.66 GHz days (not accurate yet!)
skew: 5.392
c6: 1
c0: -24571
Y1: -1
Y0: 3264415921067916092024692109591027749531
type: snfs
rlim: 60000000
alim: 60000000
lpbr: 30
lpba: 30
mfbr: 59
mfba: 59
rlambda: 2.6
alambda: 2.6

Trial sieving 5K blocks.

Code:

  Q  Yield
 20M 14093
 50M 11005
 78M  9247

RichD · 2020-09-18, 23:37

QUEUED AS 266__199_23m1

C219 from the OPN t550 file.
This polynomial will likely have a higher dup rate.
Sieve on the algebraic side. Ready for 15e.

Code:

n: 137127115955321142987909585337685535490299341309524961160449232993744480103667801681580148620298422515804715416311944176589442786040363878636684356409586040159225969734216908176479899874367102838723642222767610245888737
# 26677735199^23-1, difficulty: 250.23, skewness: 54.66, alpha: 0.00
# cost: 8.44962e+18, est. time: 4023.63 GHz days (not accurate yet!)
lss: 0
skew: 54.663
c6: 1
c0: -26677735199
Y1: -1
Y0: 506519103884710452257843504879469971299201
type: snfs
rlim: 200000000
alim: 200000000
lpbr: 32
lpba: 32
mfbr: 64
mfba: 93
rlambda: 2.7
alambda: 3.6

Trial sieving 5K blocks.

Code:

It appears the yield at 240M is a bit of an abnormally. I increased the range to 260M but that is too much. Extrapolating across 200-260M it appears 240M is reality the appropriate bound. Suggest sieving 20-240M.

VBCurtis · 2020-09-21, 20:54

QUEUED AS 5_2_782m1 aim for 220M rels

5*2^782-1 is ready for e_small:

Code:

n:  74672565937734849184549317893144798774103084875897937938305774051716011233953050940576863926650464924879249366877321888900487290584727475906915294417487936156017751807937443332383635777529100936547703467595965038711879813929551
type: snfs
skew: 0.62
deg: 6
c6: 20
c0: -1
Y1: -1
Y0: 1361129467683753853853498429727072845824
rlim: 67000000
alim: 50000000
lpbr: 32
lpba: 31
mfbr: 59
mfba: 58
rlambda: 2.7
alambda: 2.7

I'm already bending the e_small rules with 31/32LP, but the small mfb settings keep relations needed in line with a "normal" 31LP job.

Suggest Q from 8-50M for 220M relations. EDIT: test sieve is Q=1k each.

Code:

#10M	7458 (0.064 sec)
#20M	6674 (0.070 sec)
#25M	6018 (0.076 sec)
#30M	4121 (0.081 sec)
#35M	6490 (0.080 sec)
#40M	3369 (0.082 sec)
#45M	6351 (0.082 sec)
#50M	5744 (0.091 sec)

RichD · 2020-09-22, 11:55

QUEUED AS 63601_53m1

C224 from the OPN file t550.
Ready for 15e_small.

Code:

n: 67749594138192999484481337779609791089946519974483100876417916932898674510964055585599587197573271837346434660151524272739811371485950696879866369715862373783851271259851239197022982927478110854188273922799549074130882065239
# 63601^53-1, difficulty: 259.39, skewness: 6.32, alpha: 0.00
# cost: 1.68721e+19, est. time: 8034.32 GHz days (not accurate yet!)
skew: 6.318
c6: 1
c0: -63601
Y1: -1
Y0: 17028464846217236119938196992072301523132401
type: snfs
rlim: 134000000
alim: 134000000
lpbr: 31
lpba: 31
mfbr: 62
mfba: 90
rlambda: 2.7
alambda: 3.6

Trial sieving 5K blocks.

Code:

chris2be8 · 2020-09-22, 15:55

(45^151-1)/63078885472494611673536293105353925054750194716086374535711136045644 from the Brent tables:

Code:

# Built Mon Sep 21 08:22:11 2020
# Estimated SNFS difficulty 250, GNFS equivalent 180, GNFS difficulty 182, degree 6
n: 68423544642699068666660722658578585739778701040501239366410260033128453974246282171203196958199390763793606531696477868701497644829767439070008488861290359304528140553156651329442171
type: snfs
c6: 45
c0: -1
# Y0 = 45^25
Y0: 213950272706703123083784878253936767578125
# Y1 = 1^0
Y1: 1
# msieve rating: skew 0.53, size 2.770e-12, alpha 0.936, combined = 2.464e-13 rroots = 2
skew: 0.53
rlim: 120000000
alim: 120000000
lpbr: 31
lpba: 31
mfbr: 62
mfba: 62
rlambda: 2.6
alambda: 2.6

Sieve on 15e_small from 40M to 140M on the algebraic side (this may be a bit too much but I'm still getting to know 15e).

Test sieving 10k ranges:

Code:

Chris

swellman · 2020-09-22, 16:24

Quote:

Originally Posted by chris2be8

(45^151-1)/63078885472494611673536293105353925054750194716086374535711136045644 from the Brent tables:

Code:

# Built Mon Sep 21 08:22:11 2020
# Estimated SNFS difficulty 250, GNFS equivalent 180, GNFS difficulty 182, degree 6
n: 68423544642699068666660722658578585739778701040501239366410260033128453974246282171203196958199390763793606531696477868701497644829767439070008488861290359304528140553156651329442171
type: snfs
c6: 45
c0: -1
# Y0 = 45^25
Y0: 213950272706703123083784878253936767578125
# Y1 = 1^0
Y1: 1
# msieve rating: skew 0.53, size 2.770e-12, alpha 0.936, combined = 2.464e-13 rroots = 2
skew: 0.53
rlim: 120000000
alim: 120000000
lpbr: 31
lpba: 31
mfbr: 62
mfba: 62
rlambda: 2.6
alambda: 2.6

Sieve on 15e_small from 40M to 140M on the algebraic side (this may be a bit too much but I'm still getting to know 15e).

Test sieving 10k ranges:

Code:

Chris

I queued this job as f45_151m1, though I hedged a bit on the sieving range and went with 40-130M. Can add more Q if needed but sieving on the algebraic side can be noisy - sometimes the yield exceeds estimates.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
System management notes	kriesel	kriesel	10	2021-09-01 15:56
Run down the queue on MPRIME without quitting GIMPS	Rodrigo	Software	7	2018-05-25 13:26
Improving the queue management.	debrouxl	NFS@Home	10	2018-05-06 21:05
split a prime95 queue & client installation	joblack	Information & Answers	1	2009-01-06 08:45
Power Management settings	PrimeCroat	Hardware	3	2004-02-17 19:11

2020-09-03, 19:45	#1
VBCurtis "Curtis" Feb 2005 Riverside, CA 53·113 Posts	Queue management for e_small and 15e queues It's time for a new queue thread! I grouped e_small and 15e together because the queue maintainers can more easily decide which 15e queue to send to, and requestors need only decide which siever is fastest. Guidelines for e_small versus 15e (subject to change): Jobs clearly faster on 14e can still go to the d queue; if there is any doubt whether 14e or 15e is faster, post them here for e_small. Candidates for e_small should use lims no higher than 134M and Q-range no more than 170M (example: 30M-200M). Candidates for 15e should use lims no higher than 200M, per Greg's request, to keep memory use low. Unequal lim's that average 200M should use similar memory and are OK. If you post a job in this thread, it is assumed you test-sieved with 15e. Last fiddled with by VBCurtis on 2021-01-31 at 17:55 Reason: Jan 2021: Edited Q-range guidance for e-small, removed "30 or 31 LP"

2020-09-03, 20:59	#2
VBCurtis "Curtis" Feb 2005 Riverside, CA 53×113 Posts	QUEUED AS 5_2_779m Here's the first job for e_small: 52^779-1 (job name: 5_2_779m), an SNFS-236: Code: n: 6459594758238867125844847264039052931684724508838707484694754180272905420711047576478995415952837439779741853435639507801334614100750051563929718846753577973898536573574094180100967551289388791 type: snfs skew: 0.963 deg: 6 c6: 5 c0: -2 Y1: -1 Y0: 1361129467683753853853498429727072845824 rlim: 67000000 alim: 40000000 lpbr: 31 lpba: 31 mfbr: 59 mfba: 58 rlambda: 2.7 alambda: 2.7 Sieving Q-range on the -a side (-r is about 10% slower) from 10M to 52M should yield 165M relations, enough for this job. Test-sieve is 1kQ: Code: #10M 4459 (0.082 sec) #20M 4537 (0.090 sec) #30M 4585 (0.097 sec) #40M 2532 (0.112 sec) #50M 3287 (0.124 sec) I'd like to handle the matrix. Test-sieving on 14e had sec/rel about 15% lower at the same Q, but having to sieve larger Q makes the overall time estimate longer. Edit: memory use is under 200MB for this choice of lim. :) Last fiddled with by swellman on 2020-09-04 at 00:52*