I forgot: How do I reserve a 100M digit Mersenne number?

[JuanTutors]

Originally Posted by ATH

You do not really need to do anything else, but since the test will probably take a very long time, you should check how far it has been factored, so you do not spend many months PRP testing it, while a factor could have been found in hours or days.

If you want I can trial factor it a bit on the GPU, you can PM me the exponent, it is up to you.

I thought that was a feature that Prime95 took care of before going the final test. Was it removed as a feature? I don't mind TFing myself when I get home, and will check how far it's been TFed if the worktodo file tells me. I put Prime95 on my desktop too and it is doing P-1 factoring on its mersenne before PRP testing. My laptop is currently doing a DC so I don't know what it will do.

[dcheuk]

Originally Posted by dcheuk

Yes sorry I didn't know what I was typing lol.

So I ran cudapm1 with this line (in worktodo.txt):

PFactor=1,2,333898333,-1,81,2

It just finished running stage 1 after a day but the program auto exits (or maybe it crashes) when stage 2 begins, immediately after the message printed on the screen:

Using up to 6560M GPU memory.
Selected B1=3130000, B2=71207500, 3.82% chance of finding a factor 
Using B1 = 3130000 from savefile. 
Continuing stage 2 from a partial result of M333898333 fft length = 20480K 
Starting stage 2. 
Using b1 = 3130000, b2 = 71207500, d = 2310, e = 12, nrp = 21
(program auto exists after this line ...)

Any help would be appreciated ...

@dominicanpapi82 Sorry for hijacking your thread here, I should have opened a new thread. Too late now.

For 333898333 I completed P-1 Stage 1 but cannot get my cudaPm1 to continue running stage 2. So I'm gonna skip it and just run PRP.

If anyone interested to run P-1 Stage 2, I have uploaded my stage 1 saves and worktodo.txt for cudaPm1 in this (Google Drive) link below. I will remove the files upon completion of PRP or P-1.

https://drive.google.com/open?id=1Ph...Trkb-uhTc5iLhx

[JuanTutors]

Originally Posted by ATH

You do not really need to do anything else, but since the test will probably take a very long time, you should check how far it has been factored, so you do not spend many months PRP testing it, while a factor could have been found in hours or days.

If you want I can trial factor it a bit on the GPU, you can PM me the exponent, it is up to you.

ATH my two worktodo files on my two computers have the same thing, minus the exponent and the hex:

PRP=HEX,1,2,EXPONENT,-1,77,2

Should I do anything to TF further?

@dcheuk Just open up a new thread. Really not that hard.

[ATH]

Originally Posted by dominicanpapi82

I thought that was a feature that Prime95 took care of before going the final test. Was it removed as a feature? I don't mind TFing myself when I get home, and will check how far it's been TFed if the worktodo file tells me. I put Prime95 on my desktop too and it is doing P-1 factoring on its mersenne before PRP testing. My laptop is currently doing a DC so I don't know what it will do.

Prime95 was never designed to do all this automatically on a 100M digit exponent only back at the current wavefront, and Prime95 is not used for trial factoring at all now when GPUs are hundreds of times faster at it.

PRP=HEX,1,2,EXPONENT,-1,77,2

means that the exponent has been trial factored to 2^77, which is not enough for a 332M exponent, it should be at least 2^80 or higher.

[JuanTutors]

Originally Posted by ATH

Prime95 was never designed to do all this automatically on a 100M digit exponent only back at the current wavefront, and Prime95 is not used for trial factoring at all now when GPUs are hundreds of times faster at it.

PRP=HEX,1,2,EXPONENT,-1,77,2

means that the exponent has been trial factored to 2^77, which is not enough for a 332M exponent, it should be at least 2^80 or higher.

So what should I do to TF the exponents? And how high specifically should I TF it? I think all my computers have an onboard gpu, some better than others. Does it hurt throughput at all if I do some simultaneous TF and PRP? My goal is really to do 100M digit PRP tests but if I can contribute without damaging my gpu that would be a nice bonus.

[ATH]

Onboard gpus are pretty slow compared to dedicated gpus and I'm not sure if they run mfaktc using CUDA, but you can try:

https://www.mersenneforum.org/mfaktc/mfaktc-0.21/

I would try "mfaktc-0.21.win.cuda80.zip" or "mfaktc-0.21.win.cuda65.zip" and make a worktodo.txt with:

Factor=N/A,<exponent>,77,78

and see if it works and how long it takes.

[JuanTutors]

Originally Posted by ATH

Onboard gpus are pretty slow compared to dedicated gpus and I'm not sure if they run mfaktc using CUDA, but you can try:

https://www.mersenneforum.org/mfaktc/mfaktc-0.21/

I would try "mfaktc-0.21.win.cuda80.zip" or "mfaktc-0.21.win.cuda65.zip" and make a worktodo.txt with:

Factor=N/A,<exponent>,77,78

and see if it works and how long it takes.

I tried "mfaktc-0.21.win.cuda65.zip" on my laptop. It asked me for permission to run it, and I said yes. It says 1d05h with the directions followed exactly as you said it.

I also tried both "mfaktc-0.21.win.cuda80.zip" and "mfaktc-0.21.win.cuda65.zip" on my desktop. With cuda80, a window popped up and went away in about 1/60 second. With cuda65, it asked me if I want to run the program, and I said yes. When I did that, a window popped up and went away the same as it did with cuda80. Clicking on the program again now just makes that same window pop up and away before I can even see what the window is.

[kriesel]

Originally Posted by ATH

Onboard gpus are pretty slow compared to dedicated gpus and I'm not sure if they run mfaktc using CUDA,

I'm rather sure they don't run mfaktc. Intel integrated graphics processors I've encountered take mfakto and an Intel OpenCl driver. Typical is of order 20. GhzD/day TF and there's a noticeable related reduction of the prime95 PRP throughput, since the cpu and igp share the combined TDP of one package, and may also be limited by sharing system memory bandwidth, unlike a discrete gpu which mostly uses its own memory and has its own thermal budget.

try mfakto 0.15-pre6. Something like the following to leave some traces behind.
cmd /k mfakto >>mfakto-logfile.txt

[JuanTutors]

Originally Posted by kriesel

I'm rather sure they don't run mfaktc. Intel integrated graphics processors I've encountered take mfakto and an Intel OpenCl driver. Typical is of order 20. GhzD/day TF and there's a noticeable related reduction of the prime95 PRP throughput, since the cpu and igp share the combined TDP of one package, and may also be limited by sharing system memory bandwidth, unlike a discrete gpu which mostly uses its own memory and has its own thermal budget.



dcheuk, try mfakto 0.15-pre6. Something like the following to leave some traces behind.
cmd /k mfakto >>mfakto-logfile.txt

I actually tried installing mfakto but I didn't understand the instructions. Got stuck at step 1.2

[kriesel]

Originally Posted by dominicanpapi82

I actually tried installing mfakto but I didn't understand the instructions. Got stuck at step 1.2

Verify separately with an opencl test utility that you have a working opencl driver for the igp or gpu. https://www.mersenneforum.org/showpo...74&postcount=6
If you don't have a working opencl driver for it, there's no point in trying mfakto until you fix that.

Skip 1.2 (compilation of mfakto)
Use an executable from the mersenne.ca mirror or other source indicated in https://www.mersenneforum.org/showpo...91&postcount=2

Make sure you have the right opencl device number. On my i7-8750h laptop, with igp and discrete gpu, the cpu, igp, and gpu all show up in the opencl device listing. I run mfakto there on the UHD630 igp with a batch file containing

mfakto -d 11 >>mfakto.txt

along with prime95 on the cpu cores and mfaktc on the GTX1050Ti gpu.

Some other things to watch out for https://www.mersenneforum.org/showpo...60&postcount=3

Once it was working, and the ini file configured, the beginning of an mfakto selftest looked like this:

mfakto 0.15pre6-Win (64bit build)

Runtime options
  Inifile                   mfakto.ini
  Verbosity                 1
  SieveOnGPU                yes
  MoreClasses               yes
  GPUSievePrimes            81157
  GPUSieveProcessSize       24Ki bits
  GPUSieveSize              96Mi bits
  FlushInterval             0
  WorkFile                  worktodo.txt
  ResultsFile               results.txt
  Checkpoints               enabled
  CheckpointDelay           300s
  Stages                    enabled
  StopAfterFactor           class
  PrintMode                 full
  V5UserID                  kriesel
  ComputerID                peregine-uhd630
  TimeStampInResults        yes
  VectorSize                2
  GPUType                   AUTO
  SmallExp                  no
  UseBinfile                mfakto_Kernels.elf
Compiletime options

Select device - Get device info:
WARNING: Unknown GPU name, assuming GCN. Please post the device name "Intel(R) UHD Graphics 630 (Intel(R) Corporation)" to http://www.mersenneforum.org/showthread.php?t=15646 to have it added to mfakto. Set GPUType in mfakto.ini to select a GPU type yourself to avoid this warning.

OpenCL device info
  name                      Intel(R) UHD Graphics 630 (Intel(R) Corporation)
  device (driver) version   OpenCL 2.1 NEO  (23.20.16.4973)
  maximum threads per block 256
  maximum threads per grid  16777216
  number of multiprocessors 24 (1536 compute elements)
  clock rate                1100MHz

Automatic parameters
  threads per grid          0
  optimizing kernels for    GCN

Loading binary kernel file mfakto_Kernels.elf
Compiling kernels.
  GPUSievePrimes (adjusted) 81206
  GPUsieve minimum exponent 1037054
######### testcase 1/34071 (M67094119[81-82]) #########
Starting trial factoring M67094119 from 2^81 to 2^82 (7299.20GHz-days)
Date    Time | class   Pct |   time     ETA | GHz-d/day    Sieve     Wait
Dec 16 17:04 | 1272   0.1% |  0.755    n.a. |      n.a.    81206    0.00%
 M67094119 has a factor: 2623030050785206718821057 (81.117509 bits, 857.719011 GHz-d)

If you get something like that, things are looking promising.

If you don't get it to work, no worries, an igp is not many net GhzDays/day anyway. On an i7-7500U laptop, the HD620 produces about 20GhzD/day TF, but cuts the dual-cpu (PRP or LL) prime95 throughput in half when mfakto is running. It's a net gain in total throughput, but a very small one, mostly done for entertainment.

There are even slower discrete gpus than that; I benchmarked an old NVS295 in mfaktc at around 2.5 GhzD/day as I recall; not worth the electricity if I'm buying it, and I was.

A little tuning on a decently fast midrange gpu can produce increases in throughput that dwarf the total throughput of an igp.

[JuanTutors]

this is what I get (see attached)