Part I - This original post bounced off CMake mail list due to 40kb limit.
I am was attempting to build the nbody source in the CUDA sdk to both learn
CMake and CUDA build settings (this is a documentation of that experience -
heck maybe this will help someone else). After moving the sources into my
tree and creating the necessary CMakeList.txt file containing:
find( cuda )
MESSAGE( STATUS "Nbody - project is being generated" )
SET( NBODY_SRC
src/bodysystemcpu.cpp
src/bodysystemcuda.cpp
src/bodysystemcuda.cu
src/nbody.cpp
src/nbody_gold.cpp
src/nbody_kernel.cu
src/render_particles.cpp
)
MESSAGE( "NBODY_SRC = " ${NBODY_SRC} )
CUDA_ADD_EXECUTABLE( nbody ${NBODY_SRC} )
INSTALL_TARGETS( ${INSTALL_BINDIR} nbody)
1>------ Build started: Project: nbody, Configuration: Debug x64 ------
1>Building NVCC (Device) object cpp_source/nbody-new/Debug/
nbody_generated_nbody_kernel.cu.obj
1>nbody_kernel.cu
1>nbody_kernel.cu
1>tmpxft_00001ac4_00000000-3_nbody_kernel.cudafe1.gpu
1>tmpxft_00001ac4_00000000-8_nbody_kernel.cudafe2.gpu
1>tmpxft_00001ac4_00000000-3_nbody_kernel.cudafe1.cpp
1>tmpxft_00001ac4_00000000-13_nbody_kernel.ii
1>Building NVCC (Device) object
cpp_source/nbody-new/Debug/nbody_generated_bodysystemcuda.cu.obj
1>nbody_kernel.cu
1>nbody_kernel.cu
1>tmpxft_00001b5c_00000000-3_nbody_kernel.cudafe1.gpu
1>tmpxft_00001b5c_00000000-8_nbody_kernel.cudafe2.gpu
1>tmpxft_00001b5c_00000000-3_nbody_kernel.cudafe1.cpp
1>tmpxft_00001b5c_00000000-13_nbody_kernel.ii
1>Compiling...
1>render_particles.cpp
1>..\..\..\..\source\cpp\app\testing\cudaTests\nbody\src\render_particles.cpp(31)
: fatal error C1083: Cannot open include file: 'GL/glew.h': No such file or
directory
1>nbody_gold.cpp
1>nbody.cpp
1>..\..\..\..\source\cpp\app\testing\cudaTests\nbody\src\nbody.cpp(30) :
fatal error C1083: Cannot open include file: 'GL/glew.h': No such file or
directory
1>bodysystemcuda.cpp
1>..\..\..\..\source\cpp\app\testing\cudaTests\nbody\src\bodysystemcuda.cpp(32)
: fatal error C1083: Cannot open include file: 'cutil_inline.h': No such
file or directory
1>bodysystemcpu.cpp
1>..\..\..\..\source\cpp\app\testing\cudaTests\nbody\src\bodysystemcpu.cpp(37)
: fatal error C1083: Cannot open include file: 'cutil_inline.h': No such
file or directory
1>Generating Code...
1>Build log was saved at
"file://c:\projects\NIH2009\source\branches\brian\build\dvip4-Win64\cpp_source\nbody-new\nbody.dir\Debug\BuildLog.htm"
1>nbody - 4 error(s), 0 warning(s)
========== Build: 0 succeeded, 1 failed, 1 up-to-date, 0 skipped ==========
Ok and that's all fine and good except for the "Cannot open include file:
'GL/gluwl.h': No such file or directory.
So I compare above to the build of nbody supplied by NVIDIA
Compiling with CUDA Build Rule...
"C:\CUDA\bin\nvcc.exe" -arch sm_10 -ccbin "c:\Program Files
(x86)\Microsoft Visual Studio 9.0\VC\bin" -Xcompiler "/EHsc /W3 /nologo
/Od /Zi /MTd " -I"C:\CUDA\include" -I"./" -I"../../common/inc"
-maxrregcount=32 --compile -o "x64\Debug\bodysystemcuda.cu.obj" "c:\NVIDIA
Corporation\NVIDIA CUDA SDK\projects\nbody\bodysystemcuda.cu"
bodysystemcuda.cu
tmpxft_000011a0_00000000-3_bodysystemcuda.cudafe1.gpu
tmpxft_000011a0_00000000-8_bodysystemcuda.cudafe2.gpu
tmpxft_000011a0_00000000-3_bodysystemcuda.cudafe1.cpp
tmpxft_000011a0_00000000-13_bodysystemcuda.ii
Note: including lib: glut32.lib
Compiling...
cl : Command line warning D9035 : option 'Wp64' has been deprecated and will
be removed in a future release
bodysystemcpu.cpp
bodysystemcuda.cpp
nbody.cpp
Note: including lib: glut32.lib
nbody_gold.cpp
render_particles.cpp
Note: including lib: glut32.lib
param.cpp
paramgl.cpp
Note: including lib: glut32.lib
Generating Code...
Linking...
Embedding manifest...
Performing Post-Build Event...
Note the ../../common/inc (yep - sdk include directory)
So then I look at my CMakeCache.txt and find:
//Semi-colon delimit multiple arguments.
CUDA_NVCC_FLAGS:STRING=
//Semi-colon delimit multiple arguments.
CUDA_NVCC_FLAGS_DEBUG:STRING=
//Semi-colon delimit multiple arguments.
CUDA_NVCC_FLAGS_MINSIZEREL:STRING=
//Semi-colon delimit multiple arguments.
CUDA_NVCC_FLAGS_RELEASE:STRING=
//Semi-colon delimit multiple arguments.
CUDA_NVCC_FLAGS_RELWITHDEBINFO:STRING=
...
...
//Path to a file.
CUDA_SDK_ROOT_DIR:PATH=C:/NVIDIA Corporation/NVIDIA CUDA SDK
//Path to a file.
CUDA_TOOLKIT_INCLUDE:PATH=C:/CUDA/include
I ask my self two questions:
1) Why is CUDA_NVCC_FLAGS not set to some default build setting that would
just work out of the box? - I get the answer to this question when I turn on
CUDA_VERBOSE_BUILD_MODE - It is being set under the hood.
2) There is a CUDA_TOOLKIT_INCLUDE, but no CUDA_SDK_INCLUDE (or _LIB)...
why? (this would prevent ${CUDA_SDK_ROOT_DIR}/common/inc - see below)
Maybe these guys could answer question 2:
James Bigler, NVIDIA Corp (nvidia.com - jbigler)
Abe Stephens, SCI Institute -- http://www.sci.utah.edu/~abe/FindCuda.html
A request here would be to seperate FindCUDA into :
FindCUDA
and
FindNVIDIACUDASDK
and add include and lib dirs environment vars.
I also turn on CUDA_VERBOSE_BUILD as I like spam (mmm spam) and seeing my
build parameters.
I then get:
blah... blah...blah (working stuff) then:
1>C:/CUDA/bin/nvcc.exe
C:/projects/NIH2009/source/branches/brian/source/cpp/app/testing/cudaTests/nbody/src/
nbody_kernel.cu -m64 -ccbin "c:/Program Files (x86)/Microsoft Visual Studio
9.0/VC/bin" -Xcompiler
,\"/DWIN32\",\"/D_WINDOWS\",\"/W3\",\"/Zm1000\",\"/EHsc\",\"/GR\",\"/D_DEBUG\",\"/MDd\",\"/Zi\",\"/Ob0\",\"/Od\",\"/RTC1\"
-DNVCC -c -o
C:/projects/NIH2009/source/branches/brian/build/dvip4-Win64/cpp_source/nbody-new/Debug/nbody_generated_bodysystemcuda.cu.obj
-IC:/CUDA/include -IC:/CUDA/include
1>nbody_kernel.cu
1>tmpxft_00000fac_00000000-3_nbody_kernel.cudafe1.gpu
1>tmpxft_00000fac_00000000-8_nbody_kernel.cudafe2.gpu
1>tmpxft_00000fac_00000000-3_nbody_kernel.cudafe1.cpp
1>tmpxft_00000fac_00000000-13_nbody_kernel.ii
1>Generated
C:/projects/NIH2009/source/branches/brian/build/dvip4-Win64/cpp_source/nbody-new/Debug/nbody_generated_bodysystemcuda.cu.obj
successfully.
1>Compiling...
1>render_particles.cpp
1>..\..\..\..\source\cpp\app\testing\cudaTests\nbody\src\render_particles.cpp(31)
: fatal error C1083: Cannot open include file: 'GL/glew.h': No such file or
directory
1>nbody_gold.cpp
1>nbody.cpp
1>..\..\..\..\source\cpp\app\testing\cudaTests\nbody\src\nbody.cpp(30) :
fatal error C1083: Cannot open include file: 'GL/glew.h': No such file or
directory
1>bodysystemcuda.cpp
1>..\..\..\..\source\cpp\app\testing\cudaTests\nbody\src\bodysystemcuda.cpp(32)
: fatal error C1083: Cannot open include file: 'cutil_inline.h': No such
file or directory
1>bodysystemcpu.cpp
1>..\..\..\..\source\cpp\app\testing\cudaTests\nbody\src\bodysystemcpu.cpp(37)
: fatal error C1083: Cannot open include file: 'cutil_inline.h': No such
file or directory
1>Generating Code...
1>Build log was saved at
"file://c:\projects\NIH2009\source\branches\brian\build\dvip4-Win64\cpp_source\nbody-new\nbody.dir\Debug\BuildLog.htm"
1>nbody - 4 error(s), 0 warning(s)
========== Build: 0 succeeded, 1 failed, 1 up-to-date, 0 skipped ==========
The I notice that CMAKE is doublely uber certain to add C:/CUDA/include to
the include path (NOTE: -IC:/CUDA/include -IC:/CUDA/include above) , but not
C:\NVIDIA Corporation\NVIDIA CUDA SDK\common\inc.
It looks as though a cpp file cannot find glew.h and cutil_inline.h in the
NVIDIA CUDA SDK directory
So I then add to my CMakeLists.txt file (I think in a vain attempt to make
it work):
INCLUDE_DIRECTORIES( ${CUDA_SDK_ROOT_DIR}/common/inc )
"C:/Program Files (x86)/CMake 2.8/bin/cmake.exe" -E remove
C:/projects/NIH2009/source/branches/brian/build/dvip4-Win64/cpp_source/nbody-new/CMakeFiles/nbody_generated_nbody_kernel.cu.obj.depend.tmp
C:/projects/NIH2009/source/branches/brian/build/dvip4-Win64/cpp_source/nbody-new/CMakeFiles/nbody_generated_nbody_kernel.cu.obj.NVCC-depend
-- Generating
C:/projects/NIH2009/source/branches/brian/build/dvip4-Win64/cpp_source/nbody-new/Debug/nbody_generated_bodysystemcuda.cu.obj
C:/CUDA/bin/nvcc.exe
C:/projects/NIH2009/source/branches/brian/source/cpp/app/testing/cudaTests/nbody/src/
nbody_kernel.cu -m64 -ccbin "c:/Program Files (x86)/Microsoft Visual Studio
9.0/VC/bin" -Xcompiler
,\"/DWIN32\",\"/D_WINDOWS\",\"/W3\",\"/Zm1000\",\"/EHsc\",\"/GR\",\"/D_DEBUG\",\"/MDd\",\"/Zi\",\"/Ob0\",\"/Od\",\"/RTC1\"
-DNVCC -c -o
C:/projects/NIH2009/source/branches/brian/build/dvip4-Win64/cpp_source/nbody-new/Debug/nbody_generated_bodysystemcuda.cu.obj
-IC:/CUDA/include "-IC:/NVIDIA Corporation/NVIDIA CUDA SDK/common/inc"
-IC:/CUDA/include
nbody_kernel.cutmpxft_0000167c_00000000-3_nbody_kernel.cudafe1.gpu
tmpxft_0000167c_00000000-8_nbody_kernel.cudafe2.gpu
tmpxft_0000167c_00000000-3_nbody_kernel.cudafe1.cpp
tmpxft_0000167c_00000000-13_nbody_kernel.ii Generated
C:/projects/NIH2009/source/branches/brian/build/dvip4-Win64/cpp_source/nbody-new/Debug/nbody_generated_bodysystemcuda.cu.obj
successfully. Compiling... render_particles.cpp Note: including lib:
glut32.lib nbody.cpp Note: including lib: glut32.lib bodysystemcuda.cpp
bodysystemcpu.cpp Generating Code... Compiling manifest to resources...
Microsoft (R) Windows (R) Resource Compiler Version 6.0.5724.0 Copyright (C)
Microsoft Corporation. All rights reserved. Linking... LINK : fatal error
LNK1104: cannot open file 'glut32.lib'
And I would sure hope so as this is a 64bit system even though both
glut32.lib and glut64.lib can be found in (either way the original NVIDIA
nbody compiles for x64 and still links to glut32.lib):
C:\NVIDIA Corporation\NVIDIA CUDA SDK\common\lib
So I add:
link_directories( ${CUDA_SDK_ROOT_DIR}/common/lib )
And I get:
Linking...
bodysystemcuda.obj : error LNK2019: unresolved external symbol
allocateNBodyArrays referenced in function "protected: virtual void __cdecl
BodySystemCUDA::_initialize(int)" (?_initial...@bodysystemcuda@@mea...@z)
bodysystemcuda.obj : error LNK2019: unresolved external symbol
registerGLBufferObject referenced in function "protected: virtual void
__cdecl BodySystemCUDA::_initialize(int)" (?_initial...@bodysystemcuda
@@mea...@z)
bodysystemcuda.obj : error LNK2001: unresolved external symbol
__imp___glewGetBufferParameteriv
bodysystemcuda.obj : error LNK2001: unresolved external symbol
__imp___glewBufferData
...
There is lots more of this wherever this came from
...
render_particles.obj : error LNK2001: unresolved external symbol
__imp___glewAttachShader
render_particles.obj : error LNK2001: unresolved external symbol
__imp___glewCreateProgram
render_particles.obj : error LNK2001: unresolved external symbol
__imp___glewCompileShader
render_particles.obj : error LNK2001: unresolved external symbol
__imp___glewShaderSource
render_particles.obj : error LNK2001: unresolved external symbol
__imp___glewCreateShader
C:\projects\NIH2009\source\branches\brian\build\Windows-6.1\ouput\bin\Debug\nbody.exe
: fatal error LNK1120: 60 unresolved externals
So then I add:
add_library( cudart STATIC IMPORTED )
add_library( cutil64D STATIC IMPORTED )
add_library( glew64 STATIC IMPORTED )
add_library( glut64 STATIC IMPORTED )
#add_library( paramgl64 STATIC IMPORTED ) <- An attempt to build param and
paramgl into a lib
# Gotta love the syntax below... Is there anything less verbose than this in
CMAKE?
set_property(TARGET cudart PROPERTY IMPORTED_LOCATION
${CUDA_TOOLKIT_ROOT_DIR}/lib/cudart.lib )
set_property(TARGET cutil64D PROPERTY IMPORTED_LOCATION
${CUDA_SDK_ROOT_DIR}/common/lib/cutil64D.lib )
set_property(TARGET glew64 PROPERTY IMPORTED_LOCATION
${CUDA_SDK_ROOT_DIR}/common/lib/glew64.lib )
set_property(TARGET glut64 PROPERTY IMPORTED_LOCATION
${CUDA_SDK_ROOT_DIR}/common/lib/glut64.lib )
#set_property(TARGET paramgl64 PROPERTY IMPORTED_LOCATION
${CUDA_SDK_ROOT_DIR}/common/lib/paramgl64.lib )
#CUDA_ADD_LIBRARY( paramgl64 ${CUDA_SDK_ROOT_DIR}/common/src/param.cpp
${CUDA_SDK_ROOT_DIR}/common/src/paramgl.cpp STATIC )
#SET(NBODY_LIBS cudart cutil64D glew64 glut64 paramgl64 )
SET(NBODY_LIBS cudart cutil64D glew64 glut64 )
CUDA_ADD_EXECUTABLE( nbody ${NBODY_SRC} )
target_link_libraries (nbody ${NBODY_LIBS} )
And I get among other things:
1>nbody.obj : error LNK2019: unresolved external symbol "public: __cdecl
ParamListGL::ParamListGL(char const *)" (??0ParamListGL@@q...@pebd@Z)
referenced in function "void __cdecl initParameters(void)" (?initParameters@
@YAXXZ)
1>nbody.obj : error LNK2019: unresolved external symbol "public: void
__cdecl ParamListGL::Render(int,int,bool)" (?ren...@paramlistgl@@qeaaxh...@z)
referenced in function "void __cdecl display(void)" (?display@@YAXXZ)
1>nbody.obj : error LNK2019: unresolved external symbol "public: bool
__cdecl ParamListGL::Mouse(int,int,int,int)" (?mo...@paramlistgl
@@qeaa_nh...@z) referenced in function "void __cdecl mouse(int,int,int,int)"
(?mouse@@yaxh...@z)
1>nbody.obj : error LNK2019: unresolved external symbol "public: bool
__cdecl ParamListGL::Motion(int,int)" (?mot...@paramlistgl@@qeaa_...@z)
referenced in function "void __cdecl motion(int,int)" (?motion@@ya...@z)
1>nbody.obj : error LNK2019: unresolved external symbol "public: void
__cdecl ParamListGL::Special(int,int,int)" (?spec...@paramlistgl@@qeaax...@z)
referenced in function "void __cdecl special(int,int,int)" (?special@
@yax...@z)
So I change:
SET( NBODY_SRC
src/bodysystem.h
src/render_particles.h
src/bodysystemcpu.h
src/bodysystemcuda.h
src/bodysystemcuda.cu
src/nbody_kernel.cu
src/bodysystemcpu.cpp
src/bodysystemcuda.cpp
src/nbody.cpp
src/nbody_gold.cpp
src/render_particles.cpp
${CUDA_SDK_ROOT_DIR}/common/src/param.cpp
${CUDA_SDK_ROOT_DIR}/common/src/paramgl.cpp
)
To add param and pramgl
And I get:
1>------ Build started: Project: nbody, Configuration: Debug x64 ------
1>Linking...
1>bodysystemcuda.obj : error LNK2019: unresolved external symbol
allocateNBodyArrays referenced in function "protected: virtual void __cdecl
BodySystemCUDA::_initialize(int)" (?_initial...@bodysystemcuda@@mea...@z)
1>bodysystemcuda.obj : error LNK2019: unresolved external symbol
registerGLBufferObject referenced in function "protected: virtual void
__cdecl BodySystemCUDA::_initialize(int)" (?_initial...@bodysystemcuda
@@mea...@z)
1>bodysystemcuda.obj : error LNK2019: unresolved external symbol
unregisterGLBufferObject referenced in function "protected: virtual void
__cdecl BodySystemCUDA::_finalize(void)" (?_final...@bodysystemcuda
@@MEAAXXZ)
1>bodysystemcuda.obj : error LNK2019: unresolved external symbol
deleteNBodyArrays referenced in function "protected: virtual void __cdecl
BodySystemCUDA::_finalize(void)" (?_final...@bodysystemcuda@@MEAAXXZ)
1>bodysystemcuda.obj : error LNK2019: unresolved external symbol
setDeviceSoftening referenced in function "public: virtual void __cdecl
BodySystemCUDA::setSoftening(float)" (?setsoften...@bodysystemcuda@@uea...@z
)
1>bodysystemcuda.obj : error LNK2019: unresolved external symbol
integrateNbodySystem referenced in function "public: virtual void __cdecl
BodySystemCUDA::update(float)" (?upd...@bodysystemcuda@@uea...@z)
1>bodysystemcuda.obj : error LNK2019: unresolved external symbol
copyArrayFromDevice referenced in function "public: virtual float * __cdecl
BodySystemCUDA::getArray(enum BodySystem::BodyArray)"
(?getar...@bodysystemcuda@@ueaapeamw4bodyar...@bodysystem@@@Z)
1>bodysystemcuda.obj : error LNK2019: unresolved external symbol
copyArrayToDevice referenced in function "public: virtual void __cdecl
BodySystemCUDA::setArray(enum BodySystem::BodyArray,float const *)"
(?setar...@bodysystemcuda@@ueaaxw4bodyar...@bodysystem@@p...@z)
1>bodysystemcuda.obj : error LNK2019: unresolved external symbol threadSync
referenced in function "public: virtual void __cdecl
BodySystemCUDA::synchronizeThreads(void)const "
(?synchronizethre...@bodysystemcuda@@UEBAXXZ)
1>C:\projects\NIH2009\source\branches\brian\build\Windows-6.1\ouput\bin\Debug\nbody.exe
: fatal error LNK1120: 9 unresolved externals
1>Build log was saved at
"file://c:\projects\NIH2009\source\branches\brian\build\dvip4-Win64\cpp_source\nbody-new\nbody.dir\Debug\BuildLog.htm"
1>nbody - 10 error(s), 0 warning(s)
========== Build: 0 succeeded, 1 failed, 1 up-to-date, 0 skipped ==========
... to be continued
--
Brian J. Davis
_______________________________________________
Powered by www.kitware.com
Visit other Kitware open-source projects at
http://www.kitware.com/opensource/opensource.html
Please keep messages on-topic and check the CMake FAQ at:
http://www.cmake.org/Wiki/CMake_FAQ
Follow this link to subscribe/unsubscribe:
http://www.cmake.org/mailman/listinfo/cmake
