CUDALucas (a.k.a. MaclucasFFTW/CUDA 2.3/CUFFTW) - Page 13

kjaget · 2010-04-21, 16:51

Quote:

Originally Posted by jasonp

NVCC will not work windows systems unless you have MSVC (or the free MSVC Express) installed. Believe me, I wish it was otherwise.

This is my experience (and feeling) as well. Once you get past the simple porting issues you're going to see assorted strangeness because NVCC isn't meant to work with GCC under windows.

I'd be extremely happy to be proven wrong, though, so keep us up to date on your progress.

odin · 2010-04-23, 02:06

hello, can I first time LL test a 53M exponent with this program?

Thanks.

odin · 2010-04-23, 03:42

Also can this program utilize all the CUDA cores in a dual GPU card like the GTX 295? Or is it restricted to single GPU cards like the GTX 260?

msft · 2010-04-23, 05:56

Hi, odin

Quote:

Originally Posted by odin

hello, can I first time LL test a 53M exponent with this program?

Yes you can.

msft · 2010-04-23, 06:00

Quote:

Originally Posted by odin

Also can this program utilize all the CUDA cores in a dual GPU card like the GTX 295? Or is it restricted to single GPU cards like the GTX 260?

This program support single core GPU.
Unfortunately(good fortune to me :-) PCIx16 is too slow.

frmky · 2010-04-23, 06:08

Quote:

Originally Posted by odin

Also can this program utilize all the CUDA cores in a dual GPU card like the GTX 295? Or is it restricted to single GPU cards like the GTX 260?

With a dual GPU card, you can run one instance on each GPU. As is, the code runs on GPU 0. Change the cudaSetDevice() call from 0 to 1 and recompile to create a binary that runs on GPU 1.

frmky · 2010-04-23, 06:21

Quote:

Originally Posted by msft

PCIx16 is too slow.

Our system has a quad GPU setup (Tesla S1070 actually), which is connected as two GPUs per PCIe 1.1 x16 slot. Running it on all 4 GPUs simultaneously, one binary doing one test on each GPU, doesn't slow it at all.

odin · 2010-04-23, 16:29

Hi thanks for the replies.

Which of the following tests that mprime can run that this program can run:

Trial Factoring
P-1 Factoring
First Time LL tests for exponents between 50000000 - 60000000 range?
LL Double Tests.

If I want to run a first time LL test to find a prime on let's say exponent 500001000 what is the command?

Is it ./MacLucasFFTW 50001000 ?

Thanks.

TheJudger · 2010-04-23, 18:36

Hi,

Quote:

Originally Posted by odin

Which of the following tests that mprime can run that this program can run:

Trial Factoring
P-1 Factoring
First Time LL tests for exponents between 50000000 - 60000000 range?
LL Double Tests.

"First time LL" and "double checks" are the same, expect that double checks are usually in a lower range. Both can be done with msft's version of MacLucasFFTW on GPU (CUDA). Supported FFT sizes are 2M, 4M and 8M. A pessimistically guess: these FFT sizes should be able to process exponents up to 35M, 70M and 140M.

Oliver

TheJudger · 2010-04-23, 18:37

Quote:

Originally Posted by frmky

Our system has a quad GPU setup (Tesla S1070 actually), which is connected as two GPUs per PCIe 1.1 x16 slot. Running it on all 4 GPUs simultaneously, one binary doing one test on each GPU, doesn't slow it at all.

Maybe msft thought about one test distributed on multiple GPUs?

odin · 2010-04-23, 21:05

Quote:

Originally Posted by frmky

Our system has a quad GPU setup (Tesla S1070 actually), which is connected as two GPUs per PCIe 1.1 x16 slot. Running it on all 4 GPUs simultaneously, one binary doing one test on each GPU, doesn't slow it at all.

Hi, here is what I want to do. I'm currently building a system to do first time LL tests in search of a prime. The specs are:

Intel I5 750 2.66Ghz overclocked to 3.6Ghz
2GB 1333Mhz DDR3 RAM
GIGABYTE GA-H55M-S2H

If MacLucasFFTW can run on the GPU and use only 5% on the CPU then I would like to purchase something like a GTX 295 which is a Dual GPU card and run two instances of MacLucasFFTW on one CPU core and have each instance run on one GPU.

At the same time run 4 workers of mprime.

What I really want to know is how can I run MacLucasFFTW and give it an exponent to first time LL test which I can manually assign myself from the mprime website? Does this program support something like the worktodo.txt file in mprime to assign an exponent to test?

How do I make MacLucasFFTW create checkpoint file regularly like mprime in case my system restarts due to power loss?

If MacLucasFFTW finds a prime what would the result look like? Would it place the result in a text file so I could upload it to the mprime website?

Have you guys, msft and frmky, checked the power usage of your system? How many watts is it pulling now?

Does MacLucasFFTW support SLI?

Thanks.

2010-04-23, 02:06	#134
odin Apr 2010 3³ Posts	hello, can I first time LL test a 53M exponent with this program? Thanks. Last fiddled with by odin on 2010-04-23 at 02:13

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Don't DC/LL them with CudaLucas	LaurV	Data	131	2017-05-02 18:41
CUDALucas / cuFFT Performance on CUDA 7 / 7.5 / 8	Brain	GPU Computing	13	2016-02-19 15:53
CUDALucas: which binary to use?	Karl M Johnson	GPU Computing	15	2015-10-13 04:44
settings for cudaLucas	fairsky	GPU Computing	11	2013-11-03 02:08
Trying to run CUDALucas on Windows 8 CP	Rodrigo	GPU Computing	12	2012-03-07 23:20

2010-04-23, 03:42	#135
odin Apr 2010 3³ Posts	Also can this program utilize all the CUDA cores in a dual GPU card like the GTX 295? Or is it restricted to single GPU cards like the GTX 260?

2010-04-23, 16:29	#140
odin Apr 2010 3³ Posts	Hi thanks for the replies. Which of the following tests that mprime can run that this program can run: Trial Factoring P-1 Factoring First Time LL tests for exponents between 50000000 - 60000000 range? LL Double Tests. If I want to run a first time LL test to find a prime on let's say exponent 500001000 what is the command? Is it ./MacLucasFFTW 50001000 ? Thanks.