Trial division with CUDA (mmff) -- used, but runs like new!

[ET_]

Thank you George, we at FermatSearch were waiting some GPU code for our ranges!

I hope I will be able to finish my tweakings to the Double Mersennes site tonight.

Just a thought: I successfully compiled mmff on my GTX 275, that has cc 1.3.
Then I set up a mmff.ini file and a worktodo.txt file.

The program run happily, and (obviously) gave a wrong result:

luigi@luigi-ubuntu:~/luigi/CUDA/mmff/mmff-0.20$ ./mmff.exe 
mmff v0.20mmff (64bit built)

Compiletime options
  THREADS_PER_BLOCK         256
  MORE_CLASSES              enabled

Runtime options
  GPU Sieving               enabled
  GPUSievePrimes            82486
  GPUSieveSize              16M bits
WARNING: Cannot read GPUSieveProcessSize from mmff.ini, using default value (8)
  GPUSieveProcessSize       8K bits
  WorkFile                  worktodo.txt
  Checkpoints               enabled
  CheckpointDelay           300s
  Stages                    enabled
  StopAfterFactor           disabled
  PrintMode                 compact
  V5UserID                  (none)
  ComputerID                (none)
WARNING, no ProgressFormat specified in mmff.ini, using default
  TimeStampInResults        no

CUDA version info
  binary compiled for CUDA  4.10
  CUDA runtime version      4.10
  CUDA driver version       4.20

CUDA device info
  name                      GeForce GTX 275
  compute capability        1.3
  maximum threads per block 512
  number of multiprocessors 30 (240 shader cores)
  clock rate                1404MHz

got assignment: F31 bit_min=75 bit_max=76
Starting trial factoring F31 from 2^75 to 2^76 (k range: 5460000000000 to 5470000000000)
 k_min = 5459999996220
 k_max = 5470000000000
Using GPU kernel "mfaktc_barrett89_F30_61gs"
ERROR: Class problems.  Factor divisible by 2, 3, 5, 7, or 11

I guess the program may check the compute capability of the GPU before starting, as well as CUDA runtime and version.

Luigi

[c10ck3r]

Originally Posted by flashjh

Windows 64 executable: http://www.sendspace.com/file/zy4t27

Compiled with CUDA 4.2, sm_20 and sm_30

Let me know how it works.

@Prime95: I had to change line 243 of tf_common_gs.cu to:
#ifdefined (TF_BARRETT_76BIT_GS) || defined (TF_BARRETT_77BIT_GS) || defined (TF_BARRETT_79BIT_GS)

Could I beg and plead for a Win32 exe? :)

[flashjh]

Originally Posted by c10ck3r

Could I beg and plead for a Win32 exe? :)

Windows 32 executable: http://www.sendspace.com/file/rx2d4u

Compiled Cuda 4.2 | sm_20 & sm_30

Let me know if it works.

[c10ck3r]

I should know in ~4 hours. Thank you!

[ET_]

Originally Posted by c10ck3r

I should know in ~4 hours. Thank you!

Let me know how the program works, I will add it to the site.

Luigi

[firejuggler]

win32 version work on my win 64, with cudart32_42_9.dll
oh and even if it was already ran

no factor for F26 from 2^67 to 2^68 (k range: 1000000000000 to 10000000000000) [mmff 0.20mmff mfaktc_barrett89_F0_29gs]
no factor for F26 from 2^68 to 2^69 (k range: 1000000000000 to 10000000000000) [mmff 0.20mmff mfaktc_barrett89_F0_29gs]
no factor for F26 from 2^69 to 2^70 (k range: 1000000000000 to 10000000000000) [mmff 0.20mmff mfaktc_barrett89_F0_29gs]
no factor for F26 from 2^70 to 2^71 (k range: 1000000000000 to 10000000000000) [mmff 0.20mmff mfaktc_barrett89_F0_29gs]
no factor for F26 from 2^71 to 2^72 (k range: 1000000000000 to 10000000000000) [mmff 0.20mmff mfaktc_barrett89_F0_29gs]

[ET_]

Originally Posted by firejuggler

win32 version work on my win 64, with cudart32_42_9.dll
oh and even if it was already ran

no factor for F26 from 2^67 to 2^68 (k range: 1000000000000 to 10000000000000) [mmff 0.20mmff mfaktc_barrett89_F0_29gs]
no factor for F26 from 2^68 to 2^69 (k range: 1000000000000 to 10000000000000) [mmff 0.20mmff mfaktc_barrett89_F0_29gs]
no factor for F26 from 2^69 to 2^70 (k range: 1000000000000 to 10000000000000) [mmff 0.20mmff mfaktc_barrett89_F0_29gs]
no factor for F26 from 2^70 to 2^71 (k range: 1000000000000 to 10000000000000) [mmff 0.20mmff mfaktc_barrett89_F0_29gs]
no factor for F26 from 2^71 to 2^72 (k range: 1000000000000 to 10000000000000) [mmff 0.20mmff mfaktc_barrett89_F0_29gs]

May I add you to FermatSearch as Firejuggler, or o you prefer your real name/country?



Luigi

[firejuggler]

sure, add me as firejuggler

[firejuggler]

For FermatSearch, you use 'k' value, not bit value.. for example : FermatFactor=26,6.9e10,4e12

no factor for F26 from 2^64 to 2^68 (k range: 69000000000 to 4000000000000) [mmff 0.20mmff mfaktc_barrett89_F0_29gs]

[NormanRKN]

grrrr. i´m to stupid to get a range for fermat and double mersenne numbers.
i´ve to wait for a userfriendly reservation interface.

[ET_]

Originally Posted by firejuggler

For FermatSearch, you use 'k' value, not bit value.. for example : FermatFactor=26,6.9e10,4e12

no factor for F26 from 2^64 to 2^68 (k range: 69000000000 to 4000000000000) [mmff 0.20mmff mfaktc_barrett89_F0_29gs]

I already credited your name for 2⁶⁷ to 2⁷² i.e. k=549,755,809,680 to 17,592,186,044,416.

In fact, , since R = k*2^exponent+2)+1, 2⁶⁷/2⁽²⁶⁺²⁾ = 549,755,809,680, while 2⁷²/2⁽²⁶⁺²⁾ = 17,592,186,044,416

BTW, that's why 2⁷² as k_end is higher than 2⁷² as k_start...

Luigi