genefer/CUDA

[Ken_g6]

Hey, all,

I've just made a Windows build of CUDAGenefer 0.94. It's built with the 2.3 SDK instead of the 3.2 SDK, so it will probably be S-L-O-W. But if this works I'll try to install the 3.2 SDK in a VM and build with that.

[rogue]

Originally Posted by Ken_g6

Hey, all,

I've just made a Windows build of CUDAGenefer 0.94. It's built with the 2.3 SDK instead of the 3.2 SDK, so it will probably be S-L-O-W. But if this works I'll try to install the 3.2 SDK in a VM and build with that.

Did you build your code from msft's latest code, i.e. the code that works with PRPNet?

[Ken_g6]

Originally Posted by rogue

Did you build your code from msft's latest code, i.e. the code that works with PRPNet?

I built from the code in his post of December 14, which says it supports genefer 2.2.0.

It looks like it may not be working with CUDA 3.2, so I'm trying to build with that next.

[msft]

Hi ,ken_g6
This version was add your advice.

[Ken_g6]

Thanks! The build is now updated to 0.95, and built with CUDA 3.2.

[msft]

Hi ,
Linux 64bit exec file.

[IBethune]

I've built MacIntel 32 and 64 executables from the 0.95 code - http://www.pyramid-productions.net/d...el_0.95.tar.gz.

I haven't been able to test these for real since I don't have a compute capability 1.3 card, but I see no reason why it shouldn't work since I didn't change anything from the Linux version.

If you need any more versions of this compiled on the Mac please send me an email as I don't read this forum very regularly!

(iain<at>pyramid<dash>productions<dot>net)

Cheers

- Iain

[msft]

Hi ,IBethune
Thank you very much for your help.

[IBethune]

Cross-posted from PrimeGrid, but hope no-one minds!

I've updated the build at http://www.pyramid-productions.net/d...el_0.95.tar.gz to run on Mac OS X 10.5 and above.

However, I also just checked the current options available through the Apple store, and there is no DP-capable card available - the MacBook Pros have Geforce 320M or GT 330M, and the Mac Pros and iMacs all have ATI only, so this is a bit academic!

Msft, I added a check near the start of main() to test if the device is cc 1.3 or above, otherwise people who tried the code would get error messages back from the first cufft call. The modified source is in the tar file if you want to use it in the main code.

Cheers

- Iain

[msft]

$ ./GeneferCUDA -r
GeneferCUDA 0.98 (CUDA3.2) based on Genefer v2.2.0
Copyright (C) 2001-2003, Yves Gallot (v1.3)
Copyright (C) 2009, Mark Rodenkirch, David Underbakke (v2.2.0)
Copyright (C) 2010, Shoichiro Yamada (CUDA)
A program for finding large probable generalized Fermat primes.

230234^8192+1 is composite. (RES=5972a302c255e081)
  (43927 digits) (err = 0.0024) (time = 0:03:00) 00:09:25
151902^16384+1 is composite. (RES=0a1a98e69b17f1e9)
  (84895 digits) (err = 0.0017) (time = 0:05:51) 00:15:16
177444^32768+1 is a probable composite. (RES=f69cc3e2334a43a8)
  (172002 digits) (err = 0.0034) (time = 0:13:44) 00:29:00
157476^65536+1 is a probable composite. (RES=9f64b3f0d545615c)
  (340605 digits) (err = 0.0042) (time = 0:34:36) 01:03:36
52186^131072+1 is composite. (RES=1b196d6c0e4d778f)
  (618340 digits) (err = 0.0007) (time = 1:22:27) 02:26:03
70000^262144+1 is a probable composite. (RES=fa15b4b858fd2ff0)
  (1270114 digits) (err = 0.0020) (time = 2:10:04) 04:36:07

$ ./GeneferCUDA -b
GeneferCUDA 0.98 (CUDA3.2) based on Genefer v2.2.0
Copyright (C) 2001-2003, Yves Gallot (v1.3)
Copyright (C) 2009, Mark Rodenkirch, David Underbakke (v2.2.0)
Copyright (C) 2010, Shoichiro Yamada (CUDA)
A program for finding large probable generalized Fermat primes.

Generalized Fermat Number Bench
2009574^8192+1  Time: 428 us/mul.       Err: 3.82e-01   51636 digits
1632282^16384+1 Time: 453 us/mul.       Err: 2.53e-01   101791 digits
1325824^32768+1 Time: 490 us/mul.       Err: 2.19e-01   200622 digits
1076904^65536+1 Time: 636 us/mul.       Err: 1.72e-01   395325 digits
874718^131072+1 Time: 815 us/mul.       Err: 3.47e-01   778813 digits
710492^262144+1 Time: 1.15 ms/mul.      Err: 4.21e-01   1533952 digits
577098^524288+1 Time: 2.1 ms/mul.       Err: 2.01e-01   3020555 digits
468750^1048576+1        Time: 3.96 ms/mul.      Err: 1.64e-01   5946413 digits
380742^2097152+1        Time: 8.14 ms/mul.      Err: 3.63e-01   11703432 digits

Delete cudaSetDeviceFlags(cudaDeviceBlockingSync);

[msft]

Ver.0.99