mtsieve enhancements

[pepi37]

Originally Posted by rogue

I have committed some changes to gfndsieve/gfndsievecl to support ppsieve type sieving. Although I can only compare against an old version of OpenCL enabled ppsieve (version cl-0.2.3e), it is about 3x faster than it. I do not have any way to compare its speed against the speed of the CUDA enabled ppsieve. If someone has a current version of OpenCL enabled ppsieve, please share.

Is anyone interesting is doing some testing to compare the speed of either the CUDA or OpenCL ppsieve with gfndsievecl?

I posted a Windows build of gfndsievecl at sourceforge. Use the -r option to bypass building of the bitmap.

I have some more changes to boost the speed further, but I want to see where gnfdsievecl stacks up first.

You can download currently used app in http://www.primegrid.com/download/

[rogue]

Originally Posted by pepi37

You can download currently used app in http://www.primegrid.com/download/

Those builds are from 2011. I thought that they had more recent builds.

I've asked in the PrimeGrid developers-corner over at discord, but no response.

[rant]
This is one of the beefs I have with PrimeGrid. The are in a hurry to get people to build stuff for them, but they have a lot of difficulty with people that live outside of their world of influence, which includes me. I create much faster sievers than anything they have used and they basically ignore what I produce. They won't even discuss it. Very frustrating.
[/rant]

[MisterBitcoin]

Originally Posted by Happy5214

I'm not able to create new sieve files using srsieve2 (SVN rev 84):

$ ./srsieve2 -W 4 -n 5e4 -N 145e3 -P "1e9" -o 't17_b2.prp' -f B -s "34767*2^n-1"
srsieve2 v1.3, a program to find factors of k*b^n+c numbers for fixed b and variable k and n
Sieving with generic logic
Fatal Error:  Expected 95001 terms, but only set 0

Strangly enough i got this one on the latest version:

C:\Users\Administrator\Documents\cllr S649\r15\thread 1>srsieve2.exe -s remain.txt -n 25000 -N 100000 -P 15e9
srsieve2 v1.5.1, a program to find factors of k*b^n+c numbers for fixed b and va
riable k and n
Sieving with generic logic for p >= 3
Fatal Error:  Expected 825011 terms, but only set 0

I also got a bug with using multithreading on 1.4 (which was fixed); so i am back using v1.1 which works fine. Creating the sieve file on my main pc now. :D

[rogue]

Originally Posted by MisterBitcoin

Strangly enough i got this one on the latest version:

C:\Users\Administrator\Documents\cllr S649\r15\thread 1>srsieve2.exe -s remain.txt -n 25000 -N 100000 -P 15e9
srsieve2 v1.5.1, a program to find factors of k*b^n+c numbers for fixed b and va
riable k and n
Sieving with generic logic for p >= 3
Fatal Error:  Expected 825011 terms, but only set 0

I also got a bug with using multithreading on 1.4 (which was fixed); so i am back using v1.1 which works fine. Creating the sieve file on my main pc now. :D

I'll dig into it. I'm expecting it to be easy to track down and fix.

[rogue]

Originally Posted by MisterBitcoin

Strangly enough i got this one on the latest version:

C:\Users\Administrator\Documents\cllr S649\r15\thread 1>srsieve2.exe -s remain.txt -n 25000 -N 100000 -P 15e9
srsieve2 v1.5.1, a program to find factors of k*b^n+c numbers for fixed b and va
riable k and n
Sieving with generic logic for p >= 3
Fatal Error:  Expected 825011 terms, but only set 0

I also got a bug with using multithreading on 1.4 (which was fixed); so i am back using v1.1 which works fine. Creating the sieve file on my main pc now. :D

I have not been able to reproduce. Can you post the remain.txt that you used for input?

[KEP]

Originally Posted by rogue

I have not been able to reproduce. Can you post the remain.txt that you used for input?

I saw exactly same error on my machine. I in fact had to roll back to the version of srsieve2 that Reb has on his BoincConfederation website. This was when resuming and not when starting from scratch. It was for base S561. One big difference between 2.2.0 and between Rebs version was that 2.2.0 used 638000 KB while Rebs version used only 68000 KB. Maybe that can give you a lead

[rogue]

Originally Posted by KEP

I saw exactly same error on my machine. I in fact had to roll back to the version of srsieve2 that Reb has on his BoincConfederation website. This was when resuming and not when starting from scratch. It was for base S561. One big difference between 2.2.0 and between Rebs version was that 2.2.0 used 638000 KB while Rebs version used only 68000 KB. Maybe that can give you a lead

I suspect that there was a compile issue or linking issue as I cannot reproduce. I put a new build of srsieve2.7z over at sourceforge. Please try that. If that still fails on Windows, let me know. If using a file of sequences, please post that file.

[KEP]

Originally Posted by rogue

I suspect that there was a compile issue or linking issue as I cannot reproduce. I put a new build of srsieve2.7z over at sourceforge. Please try that. If that still fails on Windows, let me know. If using a file of sequences, please post that file.

It solved one problem, but there must be a problem with -W or -w as you see in this copy from the command prompt:

C:\20210101_S561_startup\srsieve2_n=3073-25K>srsieve2 -p 96467343541 -W 4 -w 10000 -i b561_n.abcd
srsieve2 v1.5.1, a program to find factors of k*b^n+c numbers for fixed b and variable k and n
Cannot use Legendre tables because square-free part of k is too large
Must use generic sieving logic because there is more than one sequence
Sieving with generic logic for p >= 96467343541
Split 5604 base 561 sequences into 5604 base 561^1 sequences.
Sieve started: 96467343541 < p < 2^62 with 7810916 terms (9001 < n < 25000, k*561^n+c)
Must use generic sieving logic because there is more than one sequence
Sieving with generic logic for p >= 96468358331
Split 5604 base 561 sequences into 5604 base 561^1 sequences.
Must use generic sieving logic because there is more than one sequence
Sieving with generic logic for p >= 96469371701
Split 5604 base 561 sequences into 5604 base 561^1 sequences.
Must use generic sieving logic because there is more than one sequence
Sieving with generic logic for p >= 96470381507
Split 5604 base 561 sequences into 5604 base 561^1 sequences.
Must use generic sieving logic because there is more than one sequence
Sieving with generic logic for p >= 96471394813
Split 5604 base 561 sequences into 5604 base 561^1 sequences.
Must use generic sieving logic because there is more than one sequence
Sieving with generic logic for p >= 96472407233
Split 5604 base 561 sequences into 5604 base 561^1 sequences.
p=0, nanM p/sec, no factors found
Must use generic sieving logic because there is more than one sequence
Sieving with generic logic for p >= 96473417713
Split 5604 base 561 sequences into 5604 base 561^1 sequences.
Must use generic sieving logic because there is more than one sequence
Sieving with generic logic for p >= 96474430733
Split 5604 base 561 sequences into 5604 base 561^1 sequences.
Must use generic sieving logic because there is more than one sequence
Sieving with generic logic for p >= 96475448009
Split 5604 base 561 sequences into 5604 base 561^1 sequences.

Without -W and -w everything checksout fine:

C:\20210101_S561_startup\srsieve2_n=3073-25K>srsieve2 -p 96467343541 -i b561_n.abcd
srsieve2 v1.5.1, a program to find factors of k*b^n+c numbers for fixed b and variable k and n
Cannot use Legendre tables because square-free part of k is too large
Must use generic sieving logic because there is more than one sequence
Sieving with generic logic for p >= 96467343541
Split 5604 base 561 sequences into 5604 base 561^1 sequences.
Sieve started: 96467343541 < p < 2^62 with 7810916 terms (9001 < n < 25000, k*561^n+c)
p=96469028651, 1.092K p/sec, 5 factors found at 17.30 sec per factor (last 1 min)

So somewthing got fixed while something got broken unfortunantly

Keep up the good work

[rogue]

I'm fairly certain I know the cause of this.

[rogue]

Apparently Apple dropped support of the OpenCL framework. It was never installed on my new Mac.

This means that to support the GPU on OS X that I will need to convert that code to Metal. I have no idea how hard that will be as I have never looked at the Metal framework.

[rogue]

Originally Posted by rogue

Apparently Apple dropped support of the OpenCL framework. It was never installed on my new Mac.

This means that to support the GPU on OS X that I will need to convert that code to Metal. I have no idea how hard that will be as I have never looked at the Metal framework.

Apparently I am incorrect on this. I needed to reinstall Xcode. The framework is there. Whew!