My goal is to insert a cublas process in my cuda kernel so i looked at simpleDevLibCUBLAS example.
My compilation leads to that:
/usr/lib/nvidia-cuda-toolkit/bin/nvcc -ccbin g++ -m64 -gencode arch=compute_35,code=sm_35 -gencode arch=compute_37,code=sm_37 -gencode arch=compute_50,code=sm_50 -gencode arch=compute_52,code=sm_52 -gencode arch=compute_52,code=compute_52 -o simpleDevLibCUBLAS kernels.o simpleDevLibCUBLAS.o -lcublas -lcublas_device -lcudadevrt
nvlink error : Undefined reference to 'cublasCreate_v2' in
'kernels.o' (target: sm_35) nvlink error : Undefined reference to
'cublasSgemm_v2' in 'kernels.o' (target: sm_35) nvlink error :
Undefined reference to 'cublasDestroy_v2' in 'kernels.o' (target:
sm_35) Makefile:239 : la recette pour la cible « simpleDevLibCUBLAS »
a échouée make: *** [simpleDevLibCUBLAS] Erreur 255
I try to modify the Makefile but nothing works.
Can somehone give me a tips?
It looks like your linker does not know where to find the libraries specified by -lcublas -lcublas_device -lcudadevrt
You could tell it with the link option -L/path/to/cuda/libs
Related
I am trying to compile some software that contains a compile.sh shell script. When I run the script, the software fails to compile due to what it reports as a broken C compiler.
I have tried reinstalling C compilers, both within and outside of conda, to no avail.
Below, please find the output of running the compile shell script
(base) userA#server:~/SPAdes-3.13.0-dev$ ./spades_compile.sh
-- The C compiler identification is GNU 7.3.0
-- The CXX compiler identification is GNU 7.3.0
-- Check for working C compiler: /home/userA/anaconda3/bin/x86_64-conda_cos6-linux-gnu-cc
-- Check for working C compiler: /home/userA/anaconda3/bin/x86_64-conda_cos6-linux-gnu-cc -- broken
CMake Error at /home/userA/anaconda3/share/cmake-3.14/Modules/CMakeTestCCompiler.cmake:60 (message):
The C compiler
"/home/userA/anaconda3/bin/x86_64-conda_cos6-linux-gnu-cc"
is not able to compile a simple test program.
It fails with the following output:
Change Dir: /home/userA/SPAdes-3.13.0-dev/build_spades/CMakeFiles/CMakeTmp
Run Build Command(s):/usr/bin/make cmTC_708f4/fast
/usr/bin/make -f CMakeFiles/cmTC_708f4.dir/build.make CMakeFiles/cmTC_708f4.dir/build
make[1]: Entering directory '/home/userA/SPAdes-3.13.0-dev/build_spades/CMakeFiles/CMakeTmp'
Building C object CMakeFiles/cmTC_708f4.dir/testCCompiler.c.o
/home/userA/anaconda3/bin/x86_64-conda_cos6-linux-gnu-cc -march=nocona -mtune=haswell -ftree-vectorize -fPIC -fstack-protector-strong -fno-plt -O2 -ffunction-sections -pipe -o CMakeFiles/cmTC_708f4.dir/testCCompiler.c.o -c /home/userA/SPAdes-3.13.0-dev/build_spades/CMakeFiles/CMakeTmp/testCCompiler.c
Linking C executable cmTC_708f4
/home/userA/anaconda3/bin/cmake -E cmake_link_script CMakeFiles/cmTC_708f4.dir/link.txt --verbose=1
/home/userA/anaconda3/bin/x86_64-conda_cos6-linux-gnu-cc -march=nocona -mtune=haswell -ftree-vectorize -fPIC -fstack-protector-strong -fno-plt -O2 -ffunction-sections -pipe -Wl,-O2 -Wl,--sort-common -Wl,--as-needed -Wl,-z,relro -Wl,-z,now -Wl,--disable-new-dtags -Wl,--gc-sections -rdynamic CMakeFiles/cmTC_708f4.dir/testCCompiler.c.o -o cmTC_708f4
/home/userA/anaconda3/bin/x86_64-conda_cos6-linux-gnu-ld: cannot find crtbeginS.o: No such file or directory
/home/userA/anaconda3/bin/x86_64-conda_cos6-linux-gnu-ld: cannot find -lgcc
/home/userA/anaconda3/bin/x86_64-conda_cos6-linux-gnu-ld: cannot find -lgcc
collect2: error: ld returned 1 exit status
CMakeFiles/cmTC_708f4.dir/build.make:86: recipe for target 'cmTC_708f4' failed
make[1]: *** [cmTC_708f4] Error 1
make[1]: Leaving directory '/home/userA/SPAdes-3.13.0-dev/build_spades/CMakeFiles/CMakeTmp'
Makefile:121: recipe for target 'cmTC_708f4/fast' failed
make: *** [cmTC_708f4/fast] Error 2
CMake will not be able to correctly generate this project.
Expected result: properly compiled software
Actual result: report of a broken C compiler
uranus#uranus:~/lib/SDL2-2.0.5$ make
/bin/bash build-scripts/updaterev.sh
LTLINK build/libSDL2.la
build/.libs/SDL_mirvideo.o: In function 'MIR_InitDisplayFromOutput':
/home/uranus/lib/SDL2-2.0.5/src/video/mir/SDL_mirvideo.c:258: undefined reference to 'mir_output_get_current_mode'
collect2: error: ld returned 1 exit status
Makefile:147: recipe for target 'build/libSDL2.la' failed
make: *** [build/libSDL2.la] Error 1
SDL2 Configure Summary:
Building Shared Libraries
Building Static Libraries
Enabled modules : atomic audio video render events joystick haptic power filesystem threads timers file loadso cpuinfo assembly
Assembly Math : mmx 3dnow sse sse2
Audio drivers : disk dummy oss alsa(dynamic) pulse(dynamic) sndio
Video drivers : dummy x11(dynamic) opengl opengl_es2 mir(dynamic)
X11 libraries : xcursor xdbe xinerama xinput2 xinput2_multitouch xrandr xscrnsaver xshape xvidmode
Input drivers : linuxev linuxkd
Using libudev : YES
Using dbus : YES
Using ime : YES
Using ibus : NO
Using fcitx : NO
The linker needs the name of the library where is this function.
I fixed it, adding in my Makefile file "-lmirclient".
# diff Makefile Makefile-ORG
26c26
< EXTRA_LDFLAGS = -Wl,--no-undefined -lm -ldl -lasound -lm -ldl -lpthread -lsndio -lX11 -lXext -lXcursor -lXinerama -lXi -lXrandr -lXss -lXxf86vm -lpthread -lrt -lmirclient
---
> EXTRA_LDFLAGS = -Wl,--no-undefined -lm -ldl -lasound -lm -ldl -lpthread -lsndio -lX11 -lXext -lXcursor -lXinerama -lXi -lXrandr -lXss -lXxf86vm -lpthread -lrt
#
A quick fix is to set static linking to the mir library when running configure:
./configure --enable-mir-shared=no
This causes the EXTRA_LDFLAGS variable to have the -lmirclient option added to it, as per Juan's answer.
As for why there is a problem, see bug 3539, which was resolved in a patch shortly after.
Looking at the patch, it appears there was a missing definition needed to allow dynamic linking to the 'mir_output_get_current_mode' function mentioned in the error message.
i had the same problem and i couldn't solve
but i used SDL2.0.3 and i didn't face any problems and it worked and Compiled fine with me
You can install it from here
https://sourceforge.net/projects/libsdl/files/SDL/2.0.3/
I am trying to install FFTW on my mac with OpenMP enabled. I initially had trouble configuring, but that issue was solved here.
Now after configuring I type "make" and I get the following error:
libtool: compile: mpicc -DHAVE_CONFIG_H -I. -I.. -I../kernel -I../dft -I../rdft -I../api -I../tests -I../libbench2 -O3 -fomit-frame-pointer -mtune=native -malign-double -fstrict-aliasing -fno-schedule-insns -ffast-math -Wa,-q -Wl,-no_compact_unwind -MT any-true.lo -MD -MP -MF .deps/any-true.Tpo -c any-true.c -o any-true.o
clang: error: unknown argument: '-malign-double'
clang: warning: -Wl,-no_compact_unwind: 'linker' input unused
clang: error: unsupported argument '-q' to option 'Wa,'
clang: warning: optimization flag '-fno-schedule-insns' is not supported
make[3]: *** [any-true.lo] Error 1
make[2]: *** [all] Error 2
make[1]: *** [all-recursive] Error 1
make: *** [all] Error 2
What is the issue here and how can I solve it?
You could try passing the location of the gcc compiler its absolute path rather than the binary name, just as
CC=/usr/local/gcc-6.1.0/bin/gcc-6.1.0 ./configure --enable-mpi --enable-threads --enable-openmp
EDIT
After having a brief chat with OP, it looks like he omitted part of the error and the error came from the mpicc compiler used (due to --enable-mpi in the configure stage). According to the user, they were using OpenMPI and according to this FAQ the user can tune the C compiler by using the OMPI_CC environtment variable. So my suggestion was to build the FFTW library as
OMPI_CC=/usr/local/gcc-6.1.0/bin/gcc-6.1.0 make
I am trying to compile rJava using on OSX 10.7.5
install.packages("rJava",type="source")
The motive is this. I am using code compiled with jdk1.8.0_05 and call it using rJava. When I do this there is a mismatch between the class file version of code compiled with jdk1.8.0_05 and the class files rJava recognizes.
llvm-gcc-4.2 -arch x86_64 -std=gnu99 -c -o rjava.o rjava.c -g -Iinclude -DRIF_HAS_CSTACK
-DRIF_HAS_RSIGHAND -mtune=core2 -g -O2 -
I/Library/Java/JavaVirtualMachines/jdk1.8.0_05.jdk/Contents/Home/jre/../include -
I/Library/Java/JavaVirtualMachines/jdk1.8.0_05.jdk/Contents/Home/jre/../include/darwin -
fno-common -
I/Library/Java/JavaVirtualMachines/jdk1.8.0_05.jdk/Contents/Home/jre/../include -
I/Library/Java/JavaVirtualMachines/jdk1.8.0_05.jdk/Contents/Home/jre/../include/darwin
llvm-gcc-4.2 -arch x86_64 -std=gnu99 -o libjri.jnilib Rengine.o jri.o Rcallbacks.o Rinit.o
globals.o rjava.o -dynamiclib -framework JavaVM -F/Library/Frameworks/R.framework/.. -
framework R -llzma -licucore -lm -liconv
ld: library not found for -llzma
collect2: ld returned 1 exit status
make[2]: *** [libjri.jnilib] Error 1
make[1]: *** [src/JRI.jar] Error 2
make: *** [jri] Error 2
ERROR: compilation failed for package ‘rJava’
I have installed xz using homebrew but that didn't help.
Can I not use rJava to call code compiled with jdk1.8.0_05 ?
rJava binaries are installed and working properly. It is the source compilation that causes this.
I have executed R CMD javareconf.
Java interpreter : /usr/bin/java
Java version : 1.8.0_05
Java home path : /Library/Java/JavaVirtualMachines/jdk1.8.0_05.jdk/Contents/Home/jre
Java compiler : /usr/bin/javac
Java headers gen.: /usr/bin/javah
Java archive tool: /usr/bin/jar
Non-system Java on OS X
trying to compile and link a JNI progam
detected JNI cpp flags : -I$(JAVA_HOME)/../include -I$(JAVA_HOME)/../include/darwin
detected JNI linker flags : -L$(JAVA_HOME)/lib/server -ljvm
llvm-gcc-4.2 -arch x86_64 -std=gnu99 -I/Library/Frameworks/R.framework/Resources/include -
DNDEBUG -I/Library/Java/JavaVirtualMachines/jdk1.8.0_05.jdk/Contents/Home/jre/../include -
I/Library/Java/JavaVirtualMachines/jdk1.8.0_05.jdk/Contents/Home/jre/../include/darwin -
I/usr/local/include -fPIC -mtune=core2 -g -O2 -c conftest.c -o conftest.o
llvm-gcc-4.2 -arch x86_64 -std=gnu99 -dynamiclib -Wl,-headerpad_max_install_names -undefined dynamic_lookup -single_module -multiply_defined suppress -L/usr/local/lib -L/usr/local/lib -o conftest.so conftest.o -L/Library/Java/JavaVirtualMachines/jdk1.8.0_05.jdk/Contents/Home/jre/lib/server -ljvm -F/Library/Frameworks/R.framework/.. -framework R -Wl,-framework -Wl,CoreFoundation
JAVA_HOME : /Library/Java/JavaVirtualMachines/jdk1.8.0_05.jdk/Contents/Home/jre
Java library path: $(JAVA_HOME)/lib/server
JNI cpp flags : -I$(JAVA_HOME)/../include -I$(JAVA_HOME)/../include/darwin
JNI linker flags : -L$(JAVA_HOME)/lib/server -ljvm
Updating Java configuration in /Library/Frameworks/R.framework/Resources
Done.
You need a library: "library not found for -llzma"
Try download and install XZ 5.0.7: http://macpkg.sourceforge.net/
I am trying to compile the examples of libhand, but I get the following output:
/usr/bin/ld: /usr/local/lib/libOgreMainStatic.a(OgreDynLib.cpp.o): undefined reference to symbol 'dlopen##GLIBC_2.1'
/usr/bin/ld: note: 'dlopen##GLIBC_2.1' is defined in DSO /usr/lib/gcc/i686-linux-gnu/4.6/../../../i386-linux-gnu/libdl.so so try adding it to the linker command line
/usr/lib/gcc/i686-linux-gnu/4.6/../../../i386-linux-gnu/libdl.so: could not read symbols: Invalid operation
collect2: ld returned 1 exit status
make[2]: *** [bend_one_finger] Error 1
make[1]: *** [CMakeFiles/bend_one_finger.dir/all] Error 2
make: *** [all] Error 2
The same thing happens when I try to compile my own example using the direction from libhand example. I believe it has something to do with the configuration file of libhand (LibHandConfig.cmake).
The same example compiles fine when I build it as a part of libhand, the problem occurs when I try to use libhand as a package (using cmake's FIND_PACKAGE() )
Not a complete answer, but just a step toward:
try to issue the following command after you got the make error:
/usr/bin/c++ CMakeFiles/example.dir/example.cc.o -o example -rdynamic -ldl -lXt /path/libs/libHand/hand_cpp/dist/libhand_hog.a /path/libs/libHand/hand_cpp/dist/libhand_renderer.a /path/libs/libHand/hand_cpp/dist/libhand_utils.a -lopencv_core -lopencv_flann -lopencv_imgproc -lopencv_highgui -lopencv_features2d -lopencv_calib3d -lopencv_cudaarithm -lopencv_cudawarping -lopencv_ml -lopencv_objdetect -lopencv_cuda -lopencv_cudafilters -lopencv_cudaimgproc -lopencv_video -lopencv_legacy -lopencv_cudaoptflow -lopencv_photo -lopencv_videostab -lopencv_ts -lopencv_cudacodec -lopencv_ocl -lopencv_superres -lopencv_cudafeatures2d -lopencv_nonfree -lopencv_stitching -lopencv_softcascade -lopencv_shape -lopencv_optim -lopencv_cudastereo -lopencv_cudabgsegm -lopencv_contrib -lopencv_bioinspired /path/libs/libHand/hand_cpp/dist/libdot_sceneloader.a /path/libs/libHand/hand_cpp/dist/libtinyxml.a /usr/local/lib/libOgreMainStatic.a -lzzip -lz -lfreeimage -lfreetype -lSM -lICE -lX11 -lXext -lXaw -lXrandr -ltbb /usr/local/lib/OGRE/libRenderSystem_GLStatic.a -lGLU -lGL /usr/local/lib/libOgreMainStatic.a -lzzip -lz -lfreeimage -lfreetype -lSM -lICE -lX11 -lXext -lXaw -lXrandr -ltbb /usr/local/lib/OGRE/libRenderSystem_GLStatic.a -lGLU -lGL /usr/local/lib/OGRE/libPlugin_OctreeSceneManagerStatic.a -ldl
If it succeeds than you have to manually add -ldl to the end of your libraries list ( target_link_libraries(example ${SOME_LIBS} ${OTHER_LIBS} -ldl) ). It seems that the configuration part of LibOGRE is slightly incorrect (it doesn't include a library libdl necessary to carelessly link application with it.