LL with OpenCL

[pinhodecarlos]

Can you guys test 5*2^3264650-1? Then post your GPU model and timings. Thank you in advance. Carlos

[Bdot]

Originally Posted by kracker

@Bdot: So I finally got Ubuntu x64 working.. do you think it is easier to compile for windows on windows or on linux(gcc-mingw)?

Well, I always used native compilation and never tried cross-compiling (except that 32-bit mfakto was built on 64-bit too). I may not be the expert you seek.

[msft]

Originally Posted by pinhodecarlos

Can you guys test 5*2^3264650-1? Then post your GPU model and timings. Thank you in advance. Carlos

This program not support 5*2^3264650-1.

[msft]

New Version.
Code clean up.

$ pwd
/opt/AMDAPP/samples/opencl/cl/0.32
$ tar -xvf 0.32.tar.bz2
0.32/
0.32/clFFTPlans.cpp
0.32/Makefile
0.32/MatrixMulDouble_Kernels.cl
0.32/run.sh
0.32/clFFTPlans.h
0.32/CUDALucas.c
0.32/MatrixMulDouble.hpp
$ cd 0.32/
0.32$ sh -x ./run.sh
+ make
g++ -Wpointer-arith -Wfloat-equal -O3 -I /opt/AMDAPP/include/ -I /opt/clAmdFft-1.10.321/include/ -g3 -ffor-scope -I ../../../../../samples/opencl/SDKUti
l/include -I ../../../../../samples/bolt/BoltUtil/include -I ../../../../../samples/C++Amp/AmpUtil/include -I "/opt/AMDAPP/include" -I ../../../../../inc
lude -o build/debug/x86_64//CUDALucas.o -c CUDALucas.c
g++ -Wpointer-arith -Wfloat-equal -O3 -I /opt/AMDAPP/include/ -I /opt/clAmdFft-1.10.321/include/ -g3 -ffor-scope -I ../../../../../samples/opencl/SDKUti
l/include -I ../../../../../samples/bolt/BoltUtil/include -I ../../../../../samples/C++Amp/AmpUtil/include -I "/opt/AMDAPP/include" -I ../../../../../inc
lude -o build/debug/x86_64//clFFTPlans.o -c clFFTPlans.cpp
g++ -o build/debug/x86_64/CUDALucas build/debug/x86_64//CUDALucas.o build/debug/x86_64//clFFTPlans.o -lpthread -ldl -L/usr/X11R6/lib /opt/clAmdFft-1.10.321/li
b64/libclAmdFft.Runtime.so -lOpenCL -lSDKUtil -lOpenCL -L../../../../../lib/x86_64 -L../../../../../TempSDKUtil/lib/x86_64 -L"/opt/AMDAPP/lib/x86_64"
install -D build/debug/x86_64/CUDALucas ../../../../../samples/opencl/bin/x86_64/CUDALucas
+ mv build/debug/x86_64/CUDALucas ./
+ export LD_LIBRARY_PATH=:/opt/clAmdFft-1.10.321/lib64/:/opt/clAmdFft-1.10.321/lib64/
+ time ./CUDALucas 1398269

start M1398269 fft length = 73728

Iteration 1380000 M( 1398269 )C, 0x6650b95f00196d18, n = 73728, CUDALucas v1.66 err = 0.1108 (0:23 real, 2.3527 ms/iter, ETA 0:23)
Iteration 1390000 M( 1398269 )C, 0x554ae339bfea8fae, n = 73728, CUDALucas v1.66 err = 0.1108 (0:23 real, 2.2749 ms/iter, ETA 0:00)
M( 1398269 )P, n = 73728, CUDALucas v1.66

[kracker]

I get the error while compiling CUDALucas.c
"error: _mkdir is not declared"

Probably something I'm doing wrong or missing..
http://imagebin.org/264640

[msft]

Originally Posted by kracker

I get the error while compiling CUDALucas.c
"error: _mkdir is not declared"

Probably something I'm doing wrong or missing..
http://imagebin.org/264640

http://cygwin.com/cygwin-api/compati...html#std-susv4
cygwin supprt mkdir.

Imagebin hold image only 14 days.

[msft]

New version.
Fix bug.

[msft]

CUDALucs.c:

#define kErrLimit (0.35)
change to
#define kErrLimit (1000)

You can check performance.

HD7750:

Iteration 20000 M( 38000000 )C, 0x8f824ba3f8ea30fd, n = 2097152, CUDALucas v1.66 err = 0.5 (2:49 real, 16.9246 ms/iter, ETA 178:30:28)

Iteration 20000 M( 78000000 )C, 0x4f41e11de1d0dbf7, n = 4194304, CUDALucas v1.66 err = 0.5 (5:56 real, 35.6646 ms/iter, ETA 772:26:05)

[kracker]

Ok, this is weird.

line 1422 S

	{
#ifdef linux
	  mode_t mode = S_IRWXU | S_IRGRP | S_IXGRP | S_IROTH | S_IXOTH;
	  if (mkdir (folder, mode) != 0)
	    printf
	      ("mkdir: cannot create directory `%s': File exists\n", folder);
#else
	  if (_mkdir (folder) != 0)
	    printf
	      ("mkdir: cannot create directory `%s': File exists\n", folder);
#endif
	}

that doesn't work but cutting it to only the linux-ment code to this

	{
	  mode_t mode = S_IRWXU | S_IRGRP | S_IXGRP | S_IROTH | S_IXOTH;
	  if (mkdir (folder, mode) != 0)
	    printf
	      ("mkdir: cannot create directory `%s': File exists\n", folder);
	}

works/compiles. Hmm...

[msft]

Hi,
Start up guide.

1) install ubuntu Desktop 64bit 12.04 LTS
2) install amd-driver-installer-catalyst-13-4-linux-x86.x86_64.zip
3) install AMD-APP-SDK-v2.8-lnx64.tgz
4) install clAmdFft-1.10.321.tar.gz
5) install g++
6) install libglu1-mesa-dev

[LaurV]

let it as in the original, but use a #define linux in front of it and see what it does. Maybe your problem is that you don't have defined it (or definition is not accessible or not included in that file or local environment?)

(edit: obviously I was replying to kracker, but crossposted with msft, btw good work msft!)