Testing new Ranges for Sierpinski/Riesel

[rogue]

I fixed the problem, but I need to add the code that creates pl_trivial.txt. Right now my program is at least 4x faster than new-base.txt for that conjecture. It should be even faster after I add the trivial code.

[rogue]

Fixed. The outputs created by srbsieve now match the outputs from the new-base.txt script with one exception, pl_MOB.txt, pl_trivial.txt, and pl_prime.txt are not sorted in ascending k. After sorting the files and comparing, they do match.

In the test for R498, I only ran to n=10. srbsieve took about 5 minutes. prgw with new-base.txt took about 28 minutes. I strongly recommend that you use more phases if you take on that base for the simple reason that the size of the numbers grows quickly so PRP tests take much longer than for base 3. sieving is much more important than trial factoring as b and n increase.

[KEP]

I've just begun following tests of srbsieve version 2:

kMin: 20,000,000,001
kMax: 20,100,000,000
nMin: 1
nMax: 300
Phase 1: n=25 to n=80
Phase 2: n=81 to n=300
I'm using PFGW version 3.7.9

Also I've just begun using the starting base script following test:

kMin: 20,000,000,001
kMax: 20,100,000,000
nMin: 1
nMax: 300
I'm using PFGW version 3.7.9 with -f0

Regards

Kenneth

For version 2 of srbsieve, the percentage counter is showing accurately by moving from 0.0 to 100.0 I expect no problems to go through Phase 1 and Phase 2.

Sometime during next week, we will know, not only how much more efficient srbsieve is compared to using the starting bases script, but we will also know if version 2 of srbsieve goes through each phase without a problem and if srbsieve version 2 computes the same MOB, remain and primes as the starting bases script. I'll get back to all of you as the various tests has completed

Take care.

[lalera]

hi,
i tested the range R498; k=1-96306; n= to 5000
with srbsieve second version and the new bases script.
i compared the files pl_MOB.txt, pl_prime.txt and pl_remain.txt
and they do match.
the timings are:
R498; k=1-50000; n= to 5000
49:23:13 with srbsieve on a sbe3930k running win10 preview build 10162
R498; k=50000-96306; n= to 5000
48:17:15 with srbsieve on a sbe3930k running win10 preview build 10162
and
R498; k=1-50000; n= to 5000
~ 83 hours with new bases script on a xeon 2630 v3 running SL6.6 64bit with -f10
R498; k=50000-96306; n= to 5000
~ 81 hours with new bases script on a xeon 2630 v3 running SL6.6 64bit with -f10

i sorted the files with openoffice writer
and compared them with winmerge

[rogue]

Originally Posted by lalera

hi,
i tested the range R498; k=1-96306; n= to 5000
with srbsieve second version and the new bases script.
i compared the files pl_MOB.txt, pl_prime.txt and pl_remain.txt
and they do match.
the timings are:
R498; k=1-50000; n= to 5000
49:23:13 with srbsieve on a sbe3930k running win10 preview build 10162
R498; k=50000-96306; n= to 5000
48:17:15 with srbsieve on a sbe3930k running win10 preview build 10162
and
R498; k=1-50000; n= to 5000
~ 83 hours with new bases script on a xeon 2630 v3 running SL6.6 64bit with -f10
R498; k=50000-96306; n= to 5000
~ 81 hours with new bases script on a xeon 2630 v3 running SL6.6 64bit with -f10

i sorted the files with openoffice writer
and compared them with winmerge

What is a sbe3930k?

[lalera]

intel sandy bridge extreme 3930k
http://ark.intel.com/de/products/63697

[Siemelink]

Hi all,

how is srbsieve pronounced? For me it is shrub-sieve.

Cheers, Willem.

[pepi37]

I got this error when try to use one of yours ini files

Problem signature:
Problem Event Name: APPCRASH
Application Name: srbsieve.exe
Application Version: 0.0.0.0
Application Timestamp: 55a5c818
Fault Module Name: ntdll.dll
Fault Module Version: 6.1.7601.18247
Fault Module Timestamp: 521eaf24
Exception Code: c0000005
Exception Offset: 0000000000052f86
OS Version: 6.1.7601.2.1.0.256.1
Locale ID: 1050
Additional Information 1: 3b16
Additional Information 2: 3b1608b2c56e9dd74d377b1a632d715e
Additional Information 3: 70f7
Additional Information 4: 70f7b6b63abd6db44a1fc2ac8598596a

I look over Net and it is something related to security permissions ...

[KEP]

Originally Posted by pepi37

I got this error when try to use one of yours ini files

Problem signature:
Problem Event Name: APPCRASH
Application Name: srbsieve.exe
Application Version: 0.0.0.0
Application Timestamp: 55a5c818
Fault Module Name: ntdll.dll
Fault Module Version: 6.1.7601.18247
Fault Module Timestamp: 521eaf24
Exception Code: c0000005
Exception Offset: 0000000000052f86
OS Version: 6.1.7601.2.1.0.256.1
Locale ID: 1050
Additional Information 1: 3b16
Additional Information 2: 3b1608b2c56e9dd74d377b1a632d715e
Additional Information 3: 70f7
Additional Information 4: 70f7b6b63abd6db44a1fc2ac8598596a

I look over Net and it is something related to security permissions ...

Try checking your antivirus, it is most likely that it has denied srbsieve to execute any actions and to write anything to the disc. At first run, of srbsieve, I had to tell comodo to never run srbsieve in a sandbox and after that, srbsieve has run without a problem and srbsieve has written all files as should be written to the SSD

[KEP]

Originally Posted by KEP

kMin: 20,000,000,001
kMax: 20,100,000,000
nMin: 1
nMax: 300
Phase 1: n=25 to n=80
Phase 2: n=81 to n=300
I'm using PFGW version 3.7.9

This test, using version 2 of srbsieve has taken a total testing time for srbsieve: 26h 19m 22s
Sieving took 00h 39m 04s

Total testing time, using sieve and srbsieve (phase 1 and 2): 26h 58m 26s or about 6 to 7 times faster than using the starting bases script to n=300 with the -f0.

Even though I have no numbers to back up my following statement, then it appears that from completion of phase 2 through to completion of phase 6, the program srbsieve goes from being 6 to 7 times faster than starting bases script, to be only 2 to 3 times faster than starting bases script and sieve manually. I've no numbers to back up my statement, but maybe Axn if he sees this can come up with a universal formula to calculate the optimal sievedepth for the amount of k's and the size of the nMax-nMin in each phase. Any idea Axn? Anyone else?

I should mention that I have no idea why it appears that we are loosing this much efficiency through phases 3, 4, 5 and 6 (or if I plain has forgotten how long an n=1 to n=25K range takes for base 3), but if my observation is correct, then I think it has to have something to do with our sievesettings

Take care

[pepi37]

Originally Posted by KEP

Try checking your antivirus, it is most likely that it has denied srbsieve to execute any actions and to write anything to the disc. At first run, of srbsieve, I had to tell comodo to never run srbsieve in a sandbox and after that, srbsieve has run without a problem and srbsieve has written all files as should be written to the SSD

D:\SRBSIEVE>srbsieve
Status (00:00:00): Started with 50000 terms

Unhandled Exception: System.AccessViolationException: Attempted to read or write
protected memory. This is often an indication that other memory is corrupt.
at fprintf(_iobuf* , SByte* )
at removeTrivial()
at main(Int32 argc, SByte** argv)
at mainCRTStartup()

It looks like it is not Kaspersky in problem :( I disabled him and put it in trusted app.