Newer milestone thread - Page 87

kracker · 2013-12-03, 18:42

Now... Can I get a list of areas worked/ing on now so I don't step on anyone?

lycorn · 2013-12-03, 19:20

Me, myself and I: 2.52M,62,63.
I noticed David Campeau has been actively working on 7M,64,65.
If everybody cares to reserve the expos and to only take unreserved, no stepping will occur.

lycorn · 2013-12-03, 19:28

Quote:

Originally Posted by kracker

a list of areas worked

For that purpose, GIMPS Visualization will be your friend. You´ll easily spot the number of exponents done to each level at any range, and drilling down the report you´ll end up in the "Factoring Limits" page of GIMPS. Once there, adjust the parameters to suit your needs and you´re set. Getting the reports in text format is great.

kracker · 2013-12-03, 19:42

Quote:

Originally Posted by lycorn

For that purpose, GIMPS Visualization will be your friend. You´ll easily spot the number of exponents done to each level at any range, and drilling down the report you´ll end up in the "Factoring Limits" page of GIMPS. Once there, adjust the parameters to suit your needs and you´re set. Getting the reports in text format is great.

Just for fun, started 3(.0)M 63 to 64. On mfakto, strangely it is faster than exponents in the LL range...

(170 GHz when it stabilized)

Jayder · 2013-12-03, 20:08

I'm working (slowly) on 4M to 64.

bloodIce · 2013-12-03, 20:20

I think that it is more realistic and more even to have everything done to 64bits. We are talking about 150K expos, and about 20K of them are really hard to go to 64bits. Anyhow, it will be achievable goal if all expos>1,000,000 go to 64bits (~130K). One difficulty I see is that not everyone has access to the special LowEx mfaktc edition. It is a dream of mine to have it, but there are only chosen few to use it.

P.S.: I took some above 6,666,000 and plan (slowly) to go to 6,700,000 from 63 to 64.

flashjh · 2013-12-03, 20:32

Quote:

Originally Posted by bloodIce

I think that it is more realistic and more even to have everything done to 64bits. We are talking about 150K expos, and about 20K of them are really hard to go to 64bits. Anyhow, it will be achievable goal if all expos>1,000,000 go to 64bits (~130K). One difficulty I see is that not everyone has access to the special LowEx mfaktc edition. It is a dream of mine to have it, but there are only chosen few to use it.

It has been tested a lot and produces reliable results, but GPU sieve only works from 2^64 and up, right now :) I compiled a version that specifies my username in the results line, that way the work can be redone if something is missed. With TF it's not so critical as LL/DC though.

ET_ · 2013-12-03, 21:40

Quote:

Originally Posted by bloodIce

I think that it is more realistic and more even to have everything done to 64bits. We are talking about 150K expos, and about 20K of them are really hard to go to 64bits. Anyhow, it will be achievable goal if all expos>1,000,000 go to 64bits (~130K). One difficulty I see is that not everyone has access to the special LowEx mfaktc edition. It is a dream of mine to have it, but there are only chosen few to use it.

P.S.: I took some above 6,666,000 and plan (slowly) to go to 6,700,000 from 63 to 64.

If you work on Linux, "I may be of assistance" ((C) Starlost) with source code and compile guide.
You don't need the "LowEx version" to run expos > 2M.

Luigi

bloodIce · 2013-12-03, 21:49

Yes, I am under Linux. I know that I do not need the LowEx for M>1,000,000, but I want to do some work in the 800,000 range. That is why I am very curious what happens with mfatkc bellow 1,000,000 and why the limit is not 100,000 or even 10,000.

ET_ · 2013-12-03, 21:58

Quote:

Originally Posted by bloodIce

Yes, I am under Linux. I know that I do not need the LowEx for M>1,000,000, but I want to do some work in the 800,000 range. That is why I am very curious what happens with mfatkc bellow 1,000,000 and why the limit is not 100,000 or even 10,000.

TheJudger may help here. From what I recall, the overhead needed to load all the multiprocessors was too high for a very opttimized program.

You could still have a "tweaked" version that is not as aggressive (i.e. work less efficiently), but till now it was not a high priority need.

If you had 16 cores, you could still test my Factor5 (it was a little more efficent in that configuration than Prime95 when I tested it on a Prescott). And maybe Ernst has his parallel version of Factor as well.

HTH

Luigi

ET_ · 2013-12-04, 09:06

My second range of 100 completed, and, well...

Code:

M2340673 has a factor: 6466054245513040351 [TF:62:63:mfaktc 0.19 75bit_mul32]
found 1 factor for M2340673 from 2^62 to 2^63 [mfaktc 0.19 75bit_mul32]

Luigi

2013-12-03, 19:20	#948
lycorn "GIMFS" Sep 2002 Oeiras, Portugal 2×11×67 Posts	Me, myself and I: 2.52M,62,63. I noticed David Campeau has been actively working on 7M,64,65. If everybody cares to reserve the expos and to only take unreserved, no stepping will occur. Last fiddled with by lycorn on 2013-12-03 at 19:24

2013-12-03, 20:20	#952
bloodIce Feb 2010 Sweden 173 Posts	2^64 I think that it is more realistic and more even to have everything done to 64bits. We are talking about 150K expos, and about 20K of them are really hard to go to 64bits. Anyhow, it will be achievable goal if all expos>1,000,000 go to 64bits (~130K). One difficulty I see is that not everyone has access to the special LowEx mfaktc edition. It is a dream of mine to have it, but there are only chosen few to use it. P.S.: I took some above 6,666,000 and plan (slowly) to go to 6,700,000 from 63 to 64. Last fiddled with by bloodIce on 2013-12-03 at 20:29

2013-12-04, 09:06	#957
ET_ Banned "Luigi" Aug 2002 Team Italia 12D3₁₆ Posts	My second range of 100 completed, and, well... Code: M2340673 has a factor: 6466054245513040351 [TF:62:63:mfaktc 0.19 75bit_mul32] found 1 factor for M2340673 from 2^62 to 2^63 [mfaktc 0.19 75bit_mul32] Luigi

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Newer X64 build needed	Googulator	Msieve	73	2020-08-30 07:47
Performance of cuda-ecm on newer hardware?	fivemack	GMP-ECM	14	2015-02-12 20:10
Cause this don't belong in the milestone thread	bcp19	Data	30	2012-09-08 15:09
Newer msieves are slow on Core i7	mklasson	Msieve	9	2009-02-18 12:58
Use of large memory pages possible with newer linux kernels	Dresdenboy	Software	3	2003-12-08 14:47

2013-12-03, 18:42	#947
kracker "Mr. Meeseeks" Jan 2012 California, USA 2168₁₀ Posts	Now... Can I get a list of areas worked/ing on now so I don't step on anyone?

2013-12-03, 20:08	#951
Jayder Dec 2012 2×139 Posts	I'm working (slowly) on 4M to 64.

2013-12-03, 21:49	#955
bloodIce Feb 2010 Sweden 173 Posts	Yes, I am under Linux. I know that I do not need the LowEx for M>1,000,000, but I want to do some work in the 800,000 range. That is why I am very curious what happens with mfatkc bellow 1,000,000 and why the limit is not 100,000 or even 10,000.