CUDALucas (a.k.a. MaclucasFFTW/CUDA 2.3/CUFFTW)

[flashjh]

All, been out of the loop for a while. We just about settled in FL. I would like to get Carl's, et al, updates incorporated and working. Sourceforge has 2.04b, but I don't see 2.05. Can anyone catch me up on current developments, etc.?

Thanks

[msft]

Originally Posted by kracker

If you need testers for anything I have two 7770's.

I appreciate your offer.

[msft]

Originally Posted by owftheevil

Do you plan on make a LL test out of this?

Some day.

[msft]

New Version.
HD7750:

$ sh -x ./run.sh
+ rm *.o a.out
+ g++ -c main.cpp -I /opt/AMDAPP/include/ -I /opt/clAmdFft-1.10.321/include/
+ g++ -c clFFTPlans.cpp -I /opt/AMDAPP/include/ -I /opt/clAmdFft-1.10.321/include/
+ g++ main.o clFFTPlans.o /opt/clAmdFft-1.10.321/lib64/libclAmdFft.Runtime.so -lOpenCL -lfftw3
+ export LD_LIBRARY_PATH=:/opt/clAmdFft-1.10.321/lib64/
+ time ./a.out
Using device: Capeverde
AmdFFT size= 1048576 time= 4.200000 msec
AmdFFT size= 2097152 time= 7.900000 msec
AmdFFT size= 3145728 time= 27.000000 msec
AmdFFT size= 4194304 time= 15.900000 msec
AmdFFT size= 5242880 time= 80.200000 msec
AmdFFT size= 6291456 time= 137.800000 msec
23.93user 71.93system 45:40.64elapsed 3%CPU (0avgtext+0avgdata 970032maxresident)k
0inputs+18136outputs (0major+120912minor)pagefaults 0swaps
$

[msft]

OpenCLucas ?
HD7750:

$ time ./a.out 216091
Using device: Capeverde
1 16384
---
215001 32768
216001 32768
M( 216091 )P, n = 32768, MacLucasFFTW v8.1 Ballester

real 12m17.124s
user 5m10.547s
sys 3m3.279s

2048k FFT 94 msec

[kracker]

Or CLLucas

I might install linux just for tinkering with it...

[Karl M Johnson]

OCLucas

[dbaugh]

I am having trouble with exponent 58715819. I have run it several times using p95v279win64 on different processors and got the following results.
M58715819 is not prime. Res64: A25E1DD191A340D0. We4: 36299458,27352921,00000000, AID: C19BB90856CE3B67E66D58C6682991CB
M58715819 is not prime. Res64: A25E1DD191A340D0. We4: 3BB22B6F,53977644,01000200, AID: C19BB90856CE3B67E66D58C6682991CB
M58715819 is not prime. Res64: A25E1DD191A340D0. We4: 349B894F,51331660,00000000, AID: C19BB90856CE3B67E66D58C6682991CB

One gave errors which is why the multiple runs but, they all gave the same Res64.

I ran CUDALucas 2.01 on a 590 (each half) and a 560Ti and got the following results.

M( 58715819 )C, 0x0ac290f866461586, n = 3670016, CUDALucas v2.01
M( 58715819 )C, 0x0ac290f866461586, n = 3670016, CUDALucas v2.01
M( 58715819 )C, 0x0ac290f866461586, n = 3670016, CUDALucas v2.01

I thought I had found the solution by moving to CUDALucas 2.03 on the 590 and got a Res64 that did not match anything previously.

M( 58715819 )C, 0x0be4ee92a264fc3b, n = 3670016, CUDALucas v2.03

What is going on? Here are the last few screen prints of the final run in case it helps. Very low err.

Iteration 58600000 M( 58715819 )C, 0xabbbf85c1135a70a, n = 3670016, CUDALucas v2.03 err = 0.0166 (12:55 real, 7.7443 ms/iter, ETA 12:54)
Iteration 58700000 M( 58715819 )C, 0x5b8436b841f162ef, n = 3670016, CUDALucas v2.03 err = 0.0166 (13:03 real, 7.8299 ms/iter, ETA 0:00)
M( 58715819 )C, 0x0be4ee92a264fc3b, n = 3670016, CUDALucas v2.03

Best regards,

David

P.S. My Res64's have always matched before and after on other similar exponents between p95 and CUDALucas.

[Karl M Johnson]

Due to this thread, can someone please compile a 64 bit CUDA 5.5 binary for sm_35 arch?

[Robert_JD]

Originally Posted by Karl M Johnson

Due to this thread, can someone please compile a 64 bit CUDA 5.5 binary for sm_35 arch?

Sure thing

Presently using Win8 x64 compiled on MSVS 12 update 3, using CuLu 2.03 source.

[Karl M Johnson]

2.55 ms/iter vs 2.94 ms/iter(M47).
I'm using the latest and greatest WHQL of 320.49.
It must be the toolkit, not the drivers.
Neat, thanks!

Notice the difference, btw:

26.09.2012  00:46        26,093,928 cufft64_50_35.dll
11.07.2013  14:06        74,730,784 cufft64_55.dll