PyCUDA February 2016

pycuda@tiker.net

7 participants
7 discussions

pycuda 2015.1.3 test_driver.py fails once

by Dorin Niculescu

Hi all, I have a fresh installed ubuntu 14.04.3 distribution, with cuda 7.0 and pycuda 2015.1.3. I'm using an NVIDIA GTX 960 card and the latest driver 361.28. All the installation went well but when i run test_driver.py i get : ============================= test session starts ============================== platform linux2 -- Python 2.7.6, pytest-2.8.7, py-1.4.31, pluggy-0.3.1 rootdir: /opt/pycuda-2015.1.3, inifile: collected 23 items test_driver.py ................x...... ===================== 22 passed, 1 xfailed in 5.91 seconds ===================== Can you please help me to understand why the test fails once? Thank you very much, Dorin

5 years, 4 months

Re: [PyCUDA] linear interpolation / trapezoidal rule integration

by Lev Givon

Received from Walter White on Thu, Feb 25, 2016 at 08:48:59AM EST: > Hello, > > are there functions for linear interpolation and trapezoidal rule > integration in CUDA? > Something like interp1d or trapz in in Python and Matlab. > Or how would you do it? So far I am iterating trough a list manually. > > Kind regards, > Joe There are 1D and 2D trapezoidal integration methods available in scikit-cuda, although their performance probably could be improved: http://scikit-cuda.readthedocs.org/en/latest/reference_integrate.html -- Lev Givon Bionet Group | Neurokernel Project http://lebedov.github.io/ http://neurokernel.github.io/

5 years, 5 months

linear interpolation / trapezoidal rule integration

by Walter White

Hello, are there functions for linear interpolation and trapezoidal rule integration in CUDA? Something like interp1d or trapz in in Python and Matlab. Or how would you do it? So far I am iterating trough a list manually. Kind regards, Joe

5 years, 5 months

Re: [PyCUDA] Newbie Question concerning data transfer and kernel compilation

by Andreas Kloeckner

Yuan Chen <chenyuan920911(a)gmail.com> writes: > Hi, > > I just start to use pycuda to do some gpu computing. > > However, I found that transfering numpy arrays to gpu costs a lot of time > and so does compiling the source. > > I am using the SourceModule now and as far as I know, for example, I have a > file called try.py and a kernel function called searching(float *arr), the > question is > > 1) Everytime I run the try.py, the searching function is compiled once, > and cached later until the codes end. So I am wondering if I can > perminantly save that function and load the saved function so that I don't > have to compile it when I run the script. PyCUDA caches the binaries for your source code as much as possible. So once you compile the same code a second time, SourceModule construction should be quite fast. Are you finding otherwise? > 2) Is there a way that make transfering data faster? I read the documents, > is the managed memory gonna help with this? Read about page-locked host memory. Those transfers are a fair bit faster than non-page-locked ones, since the hardware can do them on its own. Andreas

5 years, 5 months

Newbie Question concerning data transfer and kernel compilation

by Yuan Chen

Hi, I just start to use pycuda to do some gpu computing. However, I found that transfering numpy arrays to gpu costs a lot of time and so does compiling the source. I am using the SourceModule now and as far as I know, for example, I have a file called try.py and a kernel function called searching(float *arr), the question is 1) Everytime I run the try.py, the searching function is compiled once, and cached later until the codes end. So I am wondering if I can perminantly save that function and load the saved function so that I don't have to compile it when I run the script. 2) Is there a way that make transfering data faster? I read the documents, is the managed memory gonna help with this? Thanks a lot for help. Best Regards, Yuan Chen

5 years, 5 months

Installing pycuda on Win 10 notebook

by Arnold Tunick

Hi, 1. I'm having trouble installing pycuda on a Win 10 notebook.2. I installed Cuda 5.53. I downloaded pycuda pycuda-2015.1.3 from git4. I downloaded boost_1_59_0.5. I ran configure.py. 6. My siteconf.py file looks like this:BOOST_INC_DIR = ['C:\\local\\boost_1_59_0\\boost'] BOOST_LIB_DIR = ['C:\\local\\boost_1_59_0\\lib64-msvc-12.0'] BOOST_COMPILER = 'gcc43' USE_SHIPPED_BOOST = True BOOST_PYTHON_LIBNAME = ['boost_python-vc120-mt-gd-1_59'] BOOST_THREAD_LIBNAME = ['boost_thread-vc120-mt-gd-1_59'] CUDA_TRACE = False CUDA_ROOT = 'C:\\Program Files\\NVIDIA GPU Computing Toolkit\\CUDA\\v5.5' CUDA_ENABLE_GL = False CUDA_ENABLE_CURAND = True CUDADRV_LIB_DIR = ['${CUDA_ROOT}\\lib\\x64'] CUDADRV_LIBNAME = ['cuda'] CUDART_LIB_DIR = ['${CUDA_ROOT}\\lib\\x64'] CUDART_LIBNAME = ['cudart'] CURAND_LIB_DIR = ['${CUDA_ROOT}\\lib\\x64'] CURAND_LIBNAME = ['curand']. 7. My errors when I run "python setup.py install" are as follows:C:/SciSoft/WinPython-64bit-2.7.9.4/tools/mingw32/bin/../lib/gcc/x86_64-w64-mingw32/4.9.2/../../../../x86_64-w64-mingw32/lib/../lib/libmingw32.a(lib64_libmingw32_a-atonexit.o):atonexit.c:(.text+0xb0):multiple definition of `atexit'C:\SciSoft\WinPython-64bit-2.7.9.4\scripts\..\python-2.7.9.amd64\libs/libmsvcr90.a(dodnibs01081.o):(.text+0x0):first defined hereC:/SciSoft/WinPython-64bit-2.7.9.4/tools/mingw32/bin/../lib/gcc/x86_64-w64-mingw32/4.9.2/../../../../x86_64-w64-mingw32/lib/../lib/libmingw32.a(lib64_libmingw32_a-mingw_helpers.o):mingw_helpers.c:(.text+0x0):multiple definition of `_decode_pointer'C:\SciSoft\WinPython-64bit-2.7.9.4\scripts\..\python-2.7.9.amd64\libs/libmsvcr90.a(dodnibs00231.o):(.text+0x0):first defined hereC:/SciSoft/WinPython-64bit-2.7.9.4/tools/mingw32/bin/../lib/gcc/x86_64-w64-mingw32/4.9.2/../../../../x86_64-w64-mingw32/lib/../lib/libmingw32.a(lib64_libmingw32_a-mingw_helpers.o):mingw_helpers.c:(.text+0x10):multiple definition of `_encode_pointer'C:\SciSoft\WinPython-64bit-2.7.9.4\scripts\..\python-2.7.9.amd64\libs/libmsvcr90.a(dodnibs00241.o):(.text+0x0):first defined hereC:\SciSoft\WinPython-64bit-2.7.9.4\scripts\..\python-2.7.9.amd64\libs/libmsvcr90.a(dodnibs00288.o):(.text+0x0):multiple definition of `_fpreset'C:/SciSoft/WinPython-64bit-2.7.9.4/tools/mingw32/bin/../lib/gcc/x86_64-w64-mingw32/4.9.2/../../../../x86_64-w64-mingw32/lib/../lib/libmingw32.a(lib64_libmingw32_a-CRT_fp8.o):CRT_fp8.c:(.text+0x0):first defined herebuild\temp.win-amd64-2.7\Release\bpl-subset\bpl_subset\libs\thread\src\win32\thread.o:thread.cpp:(.text+0x1f4):undefined reference to `InterlockedIncrement'build\temp.win-amd64-2.7\Release\bpl-subset\bpl_subset\libs\thread\src\win32\thread.o:thread.cpp:(.text+0x217):undefined reference to `InterlockedDecrement'build\temp.win-amd64-2.7\Release\bpl-subset\bpl_subset\libs\thread\src\win32\thread.o:thread.cpp:(.text+0x2b6):undefined reference to `InterlockedIncrement'build\temp.win-amd64-2.7\Release\bpl-subset\bpl_subset\libs\thread\src\win32\thread.o:thread.cpp:(.text+0x2f2):undefined reference to `InterlockedDecrement'build\temp.win-amd64-2.7\Release\bpl-subset\bpl_subset\libs\thread\src\win32\thread.o:thread.cpp:(.text+0x3ba):undefined reference to `InterlockedDecrement'build\temp.win-amd64-2.7\Release\bpl-subset\bpl_subset\libs\thread\src\win32\thread.o:thread.cpp:(.text+0x41b):undefined reference to `InterlockedDecrement'build\temp.win-amd64-2.7\Release\bpl-subset\bpl_subset\libs\thread\src\win32\thread.o:thread.cpp:(.text+0x4d5):undefined reference to `InterlockedDecrement'build\temp.win-amd64-2.7\Release\bpl-subset\bpl_subset\libs\thread\src\win32\thread.o:thread.cpp:(.text+0x5b3):undefined reference to `InterlockedDecrement'build\temp.win-amd64-2.7\Release\bpl-subset\bpl_subset\libs\thread\src\win32\thread.o:thread.cpp:(.text+0x856):undefined reference to `InterlockedIncrement'build\temp.win-amd64-2.7\Release\bpl-subset\bpl_subset\libs\thread\src\win32\thread.o:thread.cpp:(.text+0xb0d):undefined reference to `InterlockedDecrement'build\temp.win-amd64-2.7\Release\bpl-subset\bpl_subset\libs\thread\src\win32\thread.o:thread.cpp:(.text+0xccb):undefined reference to `InterlockedIncrement'build\temp.win-amd64-2.7\Release\bpl-subset\bpl_subset\libs\thread\src\win32\thread.o:thread.cpp:(.text+0x2f28):undefined reference to `InterlockedDecrement'build\temp.win-amd64-2.7\Release\bpl-subset\bpl_subset\libs\thread\src\win32\thread.o:thread.cpp:(.text+0x3acf):undefined reference to `InterlockedDecrement'C:/SciSoft/WinPython-64bit-2.7.9.4/tools/mingw32/bin/../lib/gcc/x86_64-w64-mingw32/4.9.2/../../../../x86_64-w64-mingw32/bin/ld.exe:build\temp.win-amd64-2.7\Release\bpl-subset\bpl_subset\libs\thread\src\win32\thread.o:bad reloc address 0x8 in section `.data'collect2.exe: error: ld returned 1 exit statuserror: command'C:\\SciSoft\\WinPython-64bit-2.7.9.4\\scripts\\..\\python-2.7.9.amd64\\..\\tools\\mingw32\\bin\\g++.exe'failed with exit status 1.8. I look on the archive and found patches to mitigate the InterlockedDecrement and InterlockedIncrement errors, but they appeared to be already in place. (unless I was looking at the wrong files in \boost). . 9. Any comments that you can provide would be helpful.. 10. FYI, running "pip install pycuda" also fails (I think due to the following automated lib and lib64 path assignment error, i.e., mixed forward/backward slash and lib64 should be .\v5.5\lib\x64.. .... -LC:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v5.5/lib""-LC:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v5.5/lib64""-LC:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v5.5/lib""-LC:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v5.5/lib64"-Lc:\scisoft\winpython-64bit-2.7.9.4\python-2.7.9.amd64\libs-Lc:\scisoft\winpython-64bit-2.7.9.4\python-2.7.9.amd64\PCbuild\amd64 -lcuda-lcurand -lpython27 -lmsvcr90 -o build\lib.win-amd64-2.7\pycuda\_driver.pyd C:/SciSoft/WinPython-64bit-2.7.9.4/tools/mingw32/bin/../lib/gcc/x86_64-w64-mingw32/4.9.2/../../../../x86_64-w64-mingw32/bin/ld.exe:cannot find -lcuda C:/SciSoft/WinPython-64bit-2.7.9.4/tools/mingw32/bin/../lib/gcc/x86_64-w64-mingw32/4.9.2/../../../../x86_64-w64-mingw32/bin/ld.exe:cannot find -lcurand. Best Regards,Arnold TunickUS Army Research Laboratory

5 years, 5 months

BOOST dependency needed or not?

by radek＠blue.alter.pl

Hi, I've read on https://wiki.tiker.net/PyCuda/Installation/Linux that "Boost C++ libraries are no longer a dependency of PyCUDA." Also on Mac installation wiki it's written that you do not need boost for pycuda. Yet I'm unable to install pycuda WITHOUT boost on my system, is it possible? How? Or is it a bad wording, and boost on linux is needed to properly compile and install pycuda. I'm on ubuntu 15.10 with fresh MANUAL installation of cuda-7.5 which works ok. Here is error message during pycuda attempt: radek@black:~/pycuda-2015.1.3$ python setup.py build running build running build_py running build_ext building '_driver' extension x86_64-linux-gnu-gcc -pthread -fwrapv -Wall -O3 -DNDEBUG -fno-strict-aliasing -D_FORTIFY_SOURCE=2 -g -fstack-protector-strong -Wformat -Werror=format-security -fPIC -DPYGPU_PYCUDA=1 -DPYGPU_PACKAGE=pycuda -DHAVE_CURAND=1 -Isrc/cpp -I/usr/local/cuda-7.5/include -I/usr/lib/python2.7/dist-packages/numpy/core/include -I/usr/include/python2.7 -c src/cpp/cuda.cpp -o build/temp.linux-x86_64-2.7/src/cpp/cuda.o In file included from src/cpp/cuda.cpp:1:0: src/cpp/cuda.hpp:30:32: fatal error: boost/shared_ptr.hpp: No such file or directory compilation terminated. error: command 'x86_64-linux-gnu-gcc' failed with exit status 1 -- r

5 years, 5 months

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

PyCUDA February 2016