Question about twinsieve
 

Hey,


I am experimenting a bit with twinsieve.exe.
I tried the following command: twinsieve.exe -k 1 -K 100000000 -b 2 -n 268768 -P 1e14 -W 12
After a while I got this this error message.

[Quote]
p=6057931104593, 53.01M p/sec, 99902590 factors found at 3.15 sec per factor, 6.1% done. ETC 2020-04-10 04:47
Fatal Error: Something is wrong. Counted terms (96257) != expected terms (97304)[/Quote]Can someone please explain to me what I did wrong ? Sieving for -k 1 -K 10000000 -P 1e12 for the same base resulted in no error.


Thanks in advance.

[QUOTE=hunson;542195]I am experimenting a bit with twinsieve.exe.
I tried the following command: twinsieve.exe -k 1 -K 100000000 -b 2 -n 268768 -P 1e14 -W 12
After a while I got this this error message.

Can someone please explain to me what I did wrong ? Sieving for -k 1 -K 10000000 -P 1e12 for the same base resulted in no error.[/QUOTE]

Which version and build date? I will have to take a look.

Hey,


twinsieve is from the mtsieve-framework version 1.9.5, version 1.2.
Sorry can't find the build date.


My machine:
Win10-64bit
AMD Ryzen 3600 (6 cores, 12 threads)

16 GB RAM




Thanks for helping out.

[QUOTE=hunson;542200]Hey,


twinsieve is from the mtsieve-framework version 1.9.5, version 1.2.
Sorry can't find the build date.


My machine:
Win10-64bit
AMD Ryzen 3600 (6 cores, 12 threads)

16 GB RAM




Thanks for helping out.[/QUOTE]

Old bug you must use start point that is not even

What start point do you mean ? In my case k is not even and works for smaller ranges (1-10.000.000).

[QUOTE=hunson;542200]twinsieve is from the mtsieve-framework version 1.9.5, version 1.2.[/QUOTE]

Try 1.9.6 (assuming I posted it). I fixed a patch when using multiple threads. I haven't had time to try to reproduce this.

Note that for twinsieve, it might be difficult to get maximum utilization from multiple threads due to how fast it goes thru a chunk of work.

I ran a test with 1.9.6 framework. No issues. This means that either the bug has been fixed or I don't have the right conditions for reproducing it.

[QUOTE=rogue;542465]I ran a test with 1.9.6 framework. No issues. This means that either the bug has been fixed or I don't have the right conditions for reproducing it.[/QUOTE]


I cannot find 1.9.6


e:\MTSIEVE\mtsieve1957\twin>twinsieve -P10000000000 -W6 -k 6 -K 990000000 -n 50000 -b10 -fN -r -o twinb10.txt
twinsieve v1.2, a program to find factors of k*b^n+1/-1 numbers for fixed b and n and variable k
Sieve started: 1 < p < 1e10 with 890999995 terms (6 < k < 990000000, k*10^50000) (expecting 864178223 factors)
Sieve completed at p=10000000019.
Processor time: 104.38 sec. (3.52 sieving) (1.93 cores)
Fatal Error: Something is wrong. Counted terms (2331345) != expected terms (2349970)

I posted 1.9.6 executables on the mtsieve home page. Hopefully that resolves the multi-threading issue.

[QUOTE=rogue;542478]I posted 1.9.6 executables on the mtsieve home page. Hopefully that resolves the multi-threading issue.[/QUOTE]
Speed is same...as CPU usage.

[QUOTE=pepi37;542479]Speed is same...as CPU usage.[/QUOTE]

I would not expect any speed gains. The best way to gain speed with this sieve is to do one of two things:

1) Run multiple copies then use the factor files to eliminate terms.
2) When running one copy, use -w to increase the number of primes per chunk of work.

In any case I'm hoping that the Fatal Error is gone.

Thanks for the quick reply. I will try it out and will post the result.


regards.

Hey,


I tested the new version, it works without any problems. Thanks for the quick fix !
My parameters: -k 1 -K 500000000 -w 1e7 -P 1e14 -W 12

With -w 1e7 it seems that I get around 10M more primes/s (42 vs. ~52M).


kind regards.


hunson

I tried to run twinsieve v1.4, but it crashed after a few seconds.
On Linux, it reported as a segfault, and on Windows it reported as either an ntdll error or a ucrtbase error.

[QUOTE=lordlouckster;625223]I tried to run twinsieve v1.4, but it crashed after a few seconds.
On Linux, it reported as a segfault, and on Windows it reported as either an ntdll error or a ucrtbase error.[/QUOTE]

What are your command line arguments?