callforsanity / opencurrent Goto Github PK

What steps will reproduce the problem?

Compiling Opencurrent out of the distribution I got a runtime
problem.  This is the output:


PGP100:/local/joergs/opencurrent-1.0.0/src/tests >utest
[INFO] Running on GPU 0
Running tests: RayleighTimingTest, RayleighNoSlipTest, RayleighTest,
LockExTest, LockExDoubleTest, NSTest, MultigridMixedTest,
MultigridDoubleTest, ProjectDoubleTimingTest, ProjectDoubleTest,
ProjectTest, MultigridTest, Advection3DDoubleSymmetryTest,
Advection3DDoubleTest, Advection3DDoubleSwirlTest,
Advection3DSymmetryTest, Advection3DTest, SamplingTest,
Grid1DReduceTest, Grid1DReduceDoubleTest, Reduce1DTimingTest,
Reduce1DDoubleTimingTest, Grid3DTest, Grid3DReduceTest,
Reduce3DTimingTest, Grid3DReduceDoubleTest, Reduce3DDoubleTimingTest,
NetCDFTest, GridNetCDFTest, Diffusion3DTest, Diffusion1DTest
running RayleighTimingTest
[ERROR] Sol_ProjectDivergence3DBase::convert_bc_to_poisson_eqn -
invalid boundary condition type 0
[ERROR] Sol_ProjectDivergence3DBase::convert_bc_to_poisson_eqn -
invalid boundary condition type 0
[ERROR] Sol_ProjectDivergence3DBase::convert_bc_to_poisson_eqn -
invalid boundary condition type 1065353216
[ERROR] Sol_ProjectDivergence3DBase::convert_bc_to_poisson_eqn -
invalid boundary condition type 1065353216
[ERROR] Sol_ProjectDivergence3DBase::convert_bc_to_poisson_eqn -
invalid boundary condition type 0
[ERROR] Sol_ProjectDivergence3DBase::convert_bc_to_poisson_eqn -
invalid boundary condition type 0
[ERROR] Sol_Divergence3DDevice::initialize_storage - invalid
dimensions for divergence
[ERROR] Sol_ProjectDivergence3DDevice::initialize_storage - failed to
initialize divergence_solver
[ERROR] Eqn_IncompressibleNS3D::set_parameters - failed on
_projection_solver initialization
[ASSERT] RayleighTimingTest::assert_true at
/local/joergs/opencurrent-1.0.0/src/tests/rayleightiming.cpp line 136
*** glibc detected *** utest: munmap_chunk(): invalid pointer: 0xbf991374 ***
======= Backtrace: =========


When  running the Rayleigh program in the apps folder a similar error occured.

I guess the whole problem must be related to a compiler problem.

I am running the program on a SUSE 10.1 Linux maschine with the default 
gcc compiler. The cuda version has been downloaded yesterday from nvidia 
and the demos are running fine. My graphic card is a GTX 285 with 2GB of 
memory. I had to change the initial size of the convection problem in 
order to fit into the 2GB but everything else is unchanged.

Just some extra infos:

PGP100:>uname -a                                                      
Linux PGP100 2.6.27.39-0.2-pae #1 SMP 2009-11-23 12:57:38 +0100 i686 i686
i386 GNU/Linux
PGP100:~ >
PGP100:~ >gcc -v
Using built-in specs.
Target: i586-suse-linux
Configured with: ../configure --prefix=/usr --infodir=/usr/share/info
--mandir=/usr/share/man --libdir=/usr/lib --libexecdir=/usr/lib
--enable-languages=c,c++,objc,fortran,obj-c++,java,ada
--enable-checking=release --with-gxx-include-dir=/usr/include/c++/4.3
--enable-ssp --disable-libssp --with-bugurl=http://bugs.opensuse.org/
--with-pkgversion='SUSE Linux' --disable-libgcj --disable-libmudflap
--with-slibdir=/lib --with-system-zlib --enable-__cxa_atexit
--enable-libstdcxx-allocator=new --disable-libstdcxx-pch
--enable-version-specific-runtime-libs --program-suffix=-4.3
--enable-linux-futex --without-system-libunwind --with-cpu=generic
--build=i586-suse-linux
Thread model: posix
gcc version 4.3.2 [gcc-4_3-branch revision 141291] (SUSE Linux) 
PGP100:~ >


PGP100:~ >nvcc -V
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2009 NVIDIA Corporation
Built on Thu_Jul__2_10:56:25_PDT_2009
Cuda compilation tools, release 2.3, V0.2.1221
PGP100:~ >


  Thanks a lot for any idea on how to advance !

             Joerg

I can't run your example source code.
When I run incompress(in apps/incompress), it needs two grid file.

[ginide@GPUWorks apps]$ ./incompress 
usage: incompress <in.nc> <out.nc> [<tolerance>]

My machine has netCDF 3.6.3, but I can't make that files. 
I made some nc files, but that didn't work.
Please help me.

What steps will reproduce the problem?
1. when I ran ./utest or individually ran those tests
2.
3.

What is the expected output? What do you see instead?
I always met memory problem. However, my other cuda programs ran well.

What version of the product are you using? On what operating system?
opencurrent1.1.0, cuda 3.1, GTX480, Ubuntu8.04 64-bit

Please provide any additional information below.

error info:
=================================================
[INFO] Running on GPU 0
Running tests: RayleighTimingTest, RayleighNoSlipTest, RayleighTest, 
PCGDoubleTest, LockExTest, LockExDoubleTest, NSTest, MultigridMixedTest, 
MultigridDoubleTest, ProjectDoubleTimingTest, ProjectDoubleTest, ProjectTest, 
MultigridTest, Advection3DDoubleSymmetryTest, Advection3DDoubleTest, 
Advection3DDoubleSwirlTest, Advection3DSymmetryTest, Advection3DTest, 
SamplingTest, Grid1DReduceTest, Grid1DReduceDoubleTest, Reduce1DTimingTest, 
Reduce1DDoubleTimingTest, Grid3DTest, Grid3DReduceTest, Reduce3DTimingTest, 
Grid3DReduceDoubleTest, Reduce3DDoubleTimingTest, Diffusion3DTest, 
Diffusion1DTest
running RayleighTimingTest on thread 0
[ERROR] Grid3DDeviceF::init - cudaMalloc failed
[ERROR] Eqn_IncompressibleNS3D::set_parameters - failed on initializing w
[ASSERT] RayleighTimingTest::assert_true at 
/home/xueqianz/download/opencurrent-1.1.0/src/tests/rayleightiming.cpp line 136
[FAILED] RayleighTimingTest
running RayleighNoSlipTest on thread 0
run resolution 16
deltaT = 1674.000000
init min/max t = 52.312500 1621.687500
[ERROR] cudaMemcpy(HostToDevice) - CUDA error "out of memory"
[ERROR] Eqn_IncompressibleNS3D::set_parameters - failed copying to u
[ASSERT] RayleighNoSlipTest::assert_true at 
/home/xueqianz/download/opencurrent-1.1.0/src/tests/rayleighnoslip.cpp line 231
*** glibc detected *** ./utest: double free or corruption (out): 
0x00007fffea66bda0 ***
========================================

What steps will reproduce the problem?
1. build for sm_10 (using netcdf4 and a GeForce 8800 GTX)
2. make test

What is the expected output? 
All tests should be ok.

What do you see instead?
see attached unittest.txt

What version of the product are you using? On what operating system?
1.0.0 on Ubuntu 8.04 LTS

Please provide any additional information below.
I ran utest MultigridTest under a debugger and found that at line 388 of 
sol_mgpressure3ddev.cu the value of _num_level is not 6 as it should be 
but a very large value. Something is obviously overwritten. I also ran 
utest MultigridTest under valgrind and the output is in the attached file 
memcheck.txt.

What steps will reproduce the problem?
1. attempt to build on windows 7, 64bit
2. build in 64 bit mode, cause c++ compiler issue; build in 32 bit mode is fine 
but when running cause memory issues

What version of the product are you using? On what operating system?

opencurrent v1.1; cuda v3.2; windows 7 64 bit; trying to build on visual studio 
express 2009 and 2010 (both produce same issues after setting v10->v9)

Trying to make opencurrent-1.1.0

on 31% it says
[  31%] Building CXX object 
ocustorage/CMakeFiles/ocustorage.dir/gridnetcdf.cpp.o
/home/bravo/opencurrent-1.1.0/src/ocustorage/gridnetcdf.cpp: In member function 
‘bool ocu::NetCDFGrid3DWriter::open(const char*, int, int, int, float, float, 
float)’:
/home/bravo/opencurrent-1.1.0/src/ocustorage/gridnetcdf.cpp:204: error: 
‘NC_NETCDF4’ was not declared in this scope
/home/bravo/opencurrent-1.1.0/src/ocustorage/gridnetcdf.cpp: In member function 
‘bool ocu::NetCDFGrid3DWriter::define_variable(const char*, nc_type, 
ocu::GridStaggering, float, float, float)’:
/home/bravo/opencurrent-1.1.0/src/ocustorage/gridnetcdf.cpp:281: error: 
‘nc_def_var_deflate’ was not declared in this scope
/home/bravo/opencurrent-1.1.0/src/ocustorage/gridnetcdf.cpp: In member function 
‘bool ocu::NetCDFGrid3DReader::open(const char*)’:
/home/bravo/opencurrent-1.1.0/src/ocustorage/gridnetcdf.cpp:492: error: 
‘nc_inq_varids’ was not declared in this scope

Maybe it's because of new CUDA 3.2? or i'm just linux noob
my system and card:
ubuntu 10.04
GF104 geforce 460, sm_21
netcdf-4.1.1 compiled by myself

What steps will reproduce the problem?
1. Build OpenCurrent with CMake
2. Open RUN_TESTS project in VS2012
3. Build solution
4. Build failure due to unresolved external symbols

What version of the product are you using? On what operating system?
OpenCurrent 1.1.0
Windows 7 64-bits
Visual Studio 2012
CUDA 5.5

Please provide any additional information below.
I have also tried to create a new empty project, add OpenCurrent files into it, 
fix includes then build the project but still some unresolved external symbols 
errors are thrown.

NSTest is failing (the multigrid solver isn't converging for some reason, 
though it appears to be spurious, as the L2 error is zero!)

Furthermore, due to a missing ASSERT, this failure is going unreported.

What steps will reproduce the problem?
1.make test
2.
3.

What is the expected output? What do you see instead?

0% tests passed, 17 tests failed out of 17

What version of the product are you using? On what operating system?

opencurrent 1.1.0,Linux 2.6.18-238.19.1.el5, x86_64 GNU/Linux

Please provide any additional information below.

LastTest.log

"Diffusion1DTest" start time: Oct 04 17:23 EDT
Output:
----------------------------------------------------------
[INFO] Running on GPU 0
Running tests: Diffusion1DTest
running Diffusion1DTest on thread 0
[ERROR] Sol_LaplacianCentered1DDevice::apply_boundary_conditions - CUDA error 
"invalid device function "
[ERROR] Sol_LaplacianCentered1DDevice::solve - CUDA error "invalid device 
function "
[ERROR] Grid1DDeviceF::linear_combination - CUDA error "invalid device function 
"
[WARNING] Failure in file 
"~/opencurrent-1.1.0/src/ocuequation/eqn_diffusion1d.cpp" at line 91
[ERROR] Equation::advance - failed to advance equation

CUDA_INSTALL_PREFIX was obsoleted with the new FindCUDA.  The new variable
is called CUDA_TOOLKIT_ROOT_DIR.

The template System file should be updated to reflect this.

I have finish generating the codes using CMake but stuck at using make to
compile the code:
Here is the errors I met.
[cch2000@chimerasub01 ~]$ cd cmake-2.6.4-Linux-i386/bin/
[cch2000@chimerasub01 bin]$ make
[  7%] Built target ocuutil
[ 30%] Built target ocustorage
[ 63%] Built target ocuequation
make[2]: *** No rule to make target `/vnodehomes/cch2000/lib/libnetcdf.so',
needed by `apps/cubicrayleigh'.  Stop.
make[1]: *** [apps/CMakeFiles/cubicrayleigh.dir/all] Error 2
make: *** [all] Error 2
I download the NetCDF library but no "libnetcdf.so" is there. I wonder if
anybody to help me find where the problem is. Thanks and have a good day.

running incompress didn't work. 
When running 'make test', opencurrent pass all of test.

[root@GPUWorks apps]# make incompress
[ 48%] Built target ocuequation
[ 60%] Built target ocuutil
[ 98%] Built target ocustorage
[100%] Built target incompress
[root@GPUWorks apps]# ./incompress 
This program can only be compiled with double precision and NetCDF support
[root@GPUWorks apps]# 

My machine has a Tesla c2050. 
I think c2050 support double precision and I installed opencurrent with sm20. I 
also install netCDF4.4.1.


What version of the product are you using? On what operating system?
CentOS 5.5 - 64bit. CUDA 3.2, Tesla C2050

Please ignore

What steps will reproduce the problem?
1. Followed instructions in System.cmake.src to cmake. 
   Had to add the following lines:
    SET(NetCDF_INCLUDE_DIR "/usr/include/")
    SET(NetCDF_LIBRARY "/usr/lib/libnetcdf.so")

2. Ran make. Got to 8% and were then interrupted by: 
<path>/opencurrent-1.1.0/src/ocustorage/transfer_routines.cu(144): error: 
identifier "printf" is undefined
<^^Repeated 10 more times for different lines in the same file^^>
11 errors detected in the compilation of 
"/tmp/tmpxft_00002674_00000000-4_transfer_routines.cpp1.ii".


What version of the product are you using? On what operating system?
OpenCurrent v1.1.0 release (tar.gz file)
GeForce GTX 580 (compute capability 2.0, 1536 Mb memory).
Ubuntu 11.04 x86_64

Please provide any additional information below.

I am using VS2005+ Vista32. I think there are some settings of my compiler
that forbid using "std" in other name spaces to pass... I wonder where I
could make these changes to compile the code correctly. Thanks and have a
good day.

I'm working with a 1Gb card (Quadro FX 3800, if it matters) and the test 
suite fails (cuda_malloc, unsurprisingly) when I attempt to run it as it's 
currently checked in on HEAD.  RayleighTiming, at least, fails, as it 
defaults to trying a 384^2 x 192 grid, which I can't allocate.  Easily fixed 
by modifying the test, but I think that the test should probably run 
unmodified on any reasonable card.

One simple change I can see would be querying the card for available memory, 
and then running each test at a resolution that fills most of the available 
memory (up to some maximum.)  If you are amenable to this change, I'm happy 
to make it for you, in fact.

What steps will reproduce the problem?
1. run pcg program
2. run cubicrayleigh with Rayleigh_run() 
3. run cubicrayleigh with Nusselt_run()

What is the expected output? What do you see instead?
None of them converged, and the while loop is dead
Message similar to:
[INFO] Sol_MultigridPressure3DBase::do_fmg - error after 3 iterations: L2 = 
0.000000 (221243.654747x), Linf = 0.000000 (-nanx)
There is a "nan", which likely lead to segmentation faults or dead loop 

What version of the product are you using? On what operating system?
Ubuntu 10.04LTS 64bit, CUDA 3.2, GTX460 Fermi

Please provide any additional information below.
I have noticed that two of the 18 unit tests failed due to segmentation faults 
and it should be related to 3D pressure solver. 
The unit test diagnostic info is as below.
******************************************************************
The following tests FAILED:
     12 - MultigridMixedTest (SEGFAULT)
     17 - PCGTest (SEGFAULT)
Errors while running CTest
make: *** [test] Error 8
******************************************************************
Thanks!

What steps will reproduce the problem?
1.make
2.
3.

What is the expected output? What do you see instead?
[ 56%] Building CXX object 
ocuequation/CMakeFiles/ocuequation.dir/eqn_diffusion3d.cpp.o

opencurrent-1.1.0/src/ocuequation/eqn_diffusion3d.cpp(235): error: identifier 
"check_ok" is undefined
    check_ok(CoArrayManager::barrier_exchange(this->_negx_handle));


What version of the product are you using? On what operating system?
opencurrent-1.1.0;
2.6.32-gentoo-r24 x86_64 GNU/Linux

Please provide any additional information below.

When running the RayleighTest unit test I encounter a consistent
"unspecified launch failure".  This error is observed for both release and
debug versions.

What steps will reproduce the problem?
1. Run the RayleighTest unit test ( ./utest -gpu 0 RayleighTest ) Failure
observed for the "run resolution 32" case.  See rt_output.txt for sample
output.

What version of the product are you using? On what operating system?
Running current release version available via hg clone.

Running on CentOS 5.4 x86-64 system (uname -a):

Linux manzano 2.6.18-164.el5 #1 SMP Thu Sep 3 03:28:30 EDT 2009 x86_64
x86_64 x86_64 GNU/Linux

Using GeForce GTX 260 MaxCore device (216 cores / 27 multiprocessors).  See
dq_output.txt for more details.

What steps will reproduce the problem?
1. I entered ..../src/, then typed cmake ocuequation/
2.
3.

What is the expected output? What do you see instead?

CMake Error at CMakeLists.txt:15 (CUDA_ADD_LIBRARY):
  Unknown CMake command "CUDA_ADD_LIBRARY".


What version of the product are you using? On what operating system?
1.0.1, ubuntu 8.04 64-bit

Please provide any additional information below.

What version of the product are you using? On what operating system?

Opencurrent 1.1.0 on Ubuntu 10.10

Please provide any additional information below.

I am trying to put timing statements into the code to compare the performance 
of a single GPU and a 2GPU solution for the same case. I can't seem to get 
anything to print out. In which function is the choice made whether to solve on 
1 or 2 GPUs? I don't mind which test case is used in the first instance.

Is there by any chance an existing test script for this?

Thanks
Matt

http://code.google.com/p/opencurrent/wiki/OpenCurrent
states;

"To build OpenCurrent, you must first copy the file in the top-level
directory System.cmake.src to System.src, "

however the comments in System.cmake.src and the information further down
the page both state to copy it to System.cmake and not System.src   
Attempting to compile also seems to indicate that System.cmake is the file
the built system is looking for.


Further down the page the built instructions say

> mkdir sm13-rel  (or whatever you want to call it)
> cd sm13-rel
> cmake ../src
> cd sm13-rel
> make"

however there is no sm13-rel dir inside the sm13-rel dir so I think the
second cd command shouldn't be there.



There may also an issue with the instructions saying to set in System.cmake
"CUDA_INSTALL_PREFIX - where CUDA is installed. The defaults will normally
be correct, but you may have done something funky. "

for me on Linux (kubuntu 7.04) atleast that doesn't seem to work.    I
tried setting it to my cuda install location and got the error;

> cmake ../src/
CUDA_TOOLKIT_ROOT_DIR not found or specified
-- Could NOT find CUDA  (missing:  CUDA_TOOLKIT_ROOT_DIR
CUDA_NVCC_EXECUTABLE CUDA_INCLUDE_DIRS CUDA_CUDART_LIBRARY)
-- Found NETCDF: /vol/apps/netcdf-4.0.1_64/lib/libnetcdf.so
-- Found NETCDF: /vol/apps/netcdf-4.0.1_64/lib/libnetcdf.so
CMake Error: The following variables are used in this project, but they are
set to NOTFOUND.
Please set them or make sure they are set and tested correctly in the CMake
files:
CUDA_CUDART_LIBRARY (ADVANCED)
    linked by target "ocuutil" in directory
/src/opencurrent/opencurrent-1.0.0/src/ocuutil
    linked by target "ocustorage" in directory
/src/opencurrent/opencurrent-1.0.0/src/ocustorage
    linked by target "ocuequation" in directory
/src/opencurrent/opencurrent-1.0.0/src/ocuequation
    linked by target "cubicrayleigh" in directory
/src/opencurrent/opencurrent-1.0.0/src/apps
    linked by target "utest" in directory
/src/opencurrent/opencurrent-1.0.0/src/tests
CUDA_TOOLKIT_INCLUDE (ADVANCED)
   used as include directory in directory
/src/opencurrent/opencurrent-1.0.0/src/ocuutil
   used as include directory in directory
/src/opencurrent/opencurrent-1.0.0/src/ocustorage
   used as include directory in directory
/src/opencurrent/opencurrent-1.0.0/src/ocuequation
   used as include directory in directory
/src/opencurrent/opencurrent-1.0.0/src/apps
   used as include directory in directory
/src/opencurrent/opencurrent-1.0.0/src/tests

-- Configuring incomplete, errors occurred!



however if I added in 
SET(CUDA_TOOLKIT_ROOT_DIR "/mycuda-installdir/")

cmake then seems to find the cuda install;

> cmake ../src/
-- Found NETCDF: /vol/apps/netcdf-4.0.1_64/lib/libnetcdf.so
-- Found NETCDF: /vol/apps/netcdf-4.0.1_64/lib/libnetcdf.so
-- Configuring done
-- Generating done
-- Build files have been written to:
/src/opencurrent/opencurrent-1.0.0/sm13-rel













What steps will reproduce the problem?
1.
2.
3.

What is the expected output? What do you see instead?


What version of the product are you using? On what operating system?


Please provide any additional information below.