2 # This file is part of the GROMACS molecular simulation package.
4 # Copyright (c) 2017, by the GROMACS development team, led by
5 # Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
6 # and including many others, as listed in the AUTHORS file in the
7 # top-level source directory and at http://www.gromacs.org.
9 # GROMACS is free software; you can redistribute it and/or
10 # modify it under the terms of the GNU Lesser General Public License
11 # as published by the Free Software Foundation; either version 2.1
12 # of the License, or (at your option) any later version.
14 # GROMACS is distributed in the hope that it will be useful,
15 # but WITHOUT ANY WARRANTY; without even the implied warranty of
16 # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
17 # Lesser General Public License for more details.
19 # You should have received a copy of the GNU Lesser General Public
20 # License along with GROMACS; if not, see
21 # http://www.gnu.org/licenses, or write to the Free Software Foundation,
22 # Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
24 # If you want to redistribute modifications to GROMACS, please
25 # consider that scientific software is very special. Version
26 # control is crucial - bugs must be traceable. We will be happy to
27 # consider code for inclusion in the official distribution, but
28 # derived work must not be called official GROMACS. Details are found
29 # in the README & COPYING files - if they are missing, get the
30 # official version at http://www.gromacs.org.
32 # To help us fund GROMACS development, we humbly ask that you cite
33 # the research papers on the package. Check out http://www.gromacs.org.
39 # GROMACS developers README
40 # -------------------------
42 # This file is taken from CMake git tag v.3.4.3 Modules/FindCUDA.cmake
43 # and modified to use --include-path and --system-include for the
44 # include paths required for nvcc to function. Those are the supported
45 # flags for nvcc, so should continue to function. Several other
46 # supporting cmake files from Modules/FindCUDA/*cmake are also
47 # required, and these are also copied from the CMake repo, unmodified.
49 # The main file is modified further to make
50 # FindPackageHandleStandardArgs able to find those required supporting
53 # Once CMake 3.7.0 is required for GROMACS, then the fix for CMake
54 # issue #14201 contained therein may remove the need for GROMACS to
55 # import these files. Or CMake 3.8 supports CUDA natively (but 3.9
56 # required for MSVC support also) if that proves useful for
57 # GROMACS. However, the FindCUDA.cmake in CMake 3.9.0 still does not
58 # use include functionality for headers internal to CUDA that would
59 # permit the GROMACS use of -Wundef to lead to warning-free CUDA
62 # Standard FindCUDA documentation
63 # -------------------------------
65 # Tools for building CUDA C files: libraries and build dependencies.
67 # This script locates the NVIDIA CUDA C tools. It should work on linux,
68 # windows, and mac and should be reasonably up to date with CUDA C
71 # This script makes use of the standard find_package arguments of
72 # <VERSION>, REQUIRED and QUIET. CUDA_FOUND will report if an
73 # acceptable version of CUDA was found.
75 # The script will prompt the user to specify CUDA_TOOLKIT_ROOT_DIR if
76 # the prefix cannot be determined by the location of nvcc in the system
77 # path and REQUIRED is specified to find_package(). To use a different
78 # installed version of the toolkit set the environment variable
79 # CUDA_BIN_PATH before running cmake (e.g.
80 # CUDA_BIN_PATH=/usr/local/cuda1.0 instead of the default
81 # /usr/local/cuda) or set CUDA_TOOLKIT_ROOT_DIR after configuring. If
82 # you change the value of CUDA_TOOLKIT_ROOT_DIR, various components that
83 # depend on the path will be relocated.
85 # It might be necessary to set CUDA_TOOLKIT_ROOT_DIR manually on certain
86 # platforms, or to use a cuda runtime not installed in the default
87 # location. In newer versions of the toolkit the cuda library is
88 # included with the graphics driver- be sure that the driver version
89 # matches what is needed by the cuda runtime version.
91 # The following variables affect the behavior of the macros in the
92 # script (in alphebetical order). Note that any of these flags can be
93 # changed multiple times in the same directory before calling
94 # CUDA_ADD_EXECUTABLE, CUDA_ADD_LIBRARY, CUDA_COMPILE, CUDA_COMPILE_PTX,
95 # CUDA_COMPILE_FATBIN, CUDA_COMPILE_CUBIN or CUDA_WRAP_SRCS::
97 # CUDA_64_BIT_DEVICE_CODE (Default matches host bit size)
98 # -- Set to ON to compile for 64 bit device code, OFF for 32 bit device code.
99 # Note that making this different from the host code when generating object
100 # or C files from CUDA code just won't work, because size_t gets defined by
101 # nvcc in the generated source. If you compile to PTX and then load the
102 # file yourself, you can mix bit sizes between device and host.
104 # CUDA_ATTACH_VS_BUILD_RULE_TO_CUDA_FILE (Default ON)
105 # -- Set to ON if you want the custom build rule to be attached to the source
106 # file in Visual Studio. Turn OFF if you add the same cuda file to multiple
109 # This allows the user to build the target from the CUDA file; however, bad
110 # things can happen if the CUDA source file is added to multiple targets.
111 # When performing parallel builds it is possible for the custom build
112 # command to be run more than once and in parallel causing cryptic build
113 # errors. VS runs the rules for every source file in the target, and a
114 # source can have only one rule no matter how many projects it is added to.
115 # When the rule is run from multiple targets race conditions can occur on
116 # the generated file. Eventually everything will get built, but if the user
117 # is unaware of this behavior, there may be confusion. It would be nice if
118 # this script could detect the reuse of source files across multiple targets
119 # and turn the option off for the user, but no good solution could be found.
121 # CUDA_BUILD_CUBIN (Default OFF)
122 # -- Set to ON to enable and extra compilation pass with the -cubin option in
123 # Device mode. The output is parsed and register, shared memory usage is
124 # printed during build.
126 # CUDA_BUILD_EMULATION (Default OFF for device mode)
127 # -- Set to ON for Emulation mode. -D_DEVICEEMU is defined for CUDA C files
128 # when CUDA_BUILD_EMULATION is TRUE.
130 # CUDA_GENERATED_OUTPUT_DIR (Default CMAKE_CURRENT_BINARY_DIR)
131 # -- Set to the path you wish to have the generated files placed. If it is
132 # blank output files will be placed in CMAKE_CURRENT_BINARY_DIR.
133 # Intermediate files will always be placed in
134 # CMAKE_CURRENT_BINARY_DIR/CMakeFiles.
136 # CUDA_HOST_COMPILATION_CPP (Default ON)
137 # -- Set to OFF for C compilation of host code.
139 # CUDA_HOST_COMPILER (Default CMAKE_C_COMPILER, $(VCInstallDir)/bin for VS)
140 # -- Set the host compiler to be used by nvcc. Ignored if -ccbin or
141 # --compiler-bindir is already present in the CUDA_NVCC_FLAGS or
142 # CUDA_NVCC_FLAGS_<CONFIG> variables. For Visual Studio targets
143 # $(VCInstallDir)/bin is a special value that expands out to the path when
144 # the command is run from withing VS.
147 # CUDA_NVCC_FLAGS_<CONFIG>
148 # -- Additional NVCC command line arguments. NOTE: multiple arguments must be
149 # semi-colon delimited (e.g. --compiler-options;-Wall)
151 # CUDA_PROPAGATE_HOST_FLAGS (Default ON)
152 # -- Set to ON to propagate CMAKE_{C,CXX}_FLAGS and their configuration
153 # dependent counterparts (e.g. CMAKE_C_FLAGS_DEBUG) automatically to the
154 # host compiler through nvcc's -Xcompiler flag. This helps make the
155 # generated host code match the rest of the system better. Sometimes
156 # certain flags give nvcc problems, and this will help you turn the flag
157 # propagation off. This does not affect the flags supplied directly to nvcc
158 # via CUDA_NVCC_FLAGS or through the OPTION flags specified through
159 # CUDA_ADD_LIBRARY, CUDA_ADD_EXECUTABLE, or CUDA_WRAP_SRCS. Flags used for
160 # shared library compilation are not affected by this flag.
162 # CUDA_SEPARABLE_COMPILATION (Default OFF)
163 # -- If set this will enable separable compilation for all CUDA runtime object
164 # files. If used outside of CUDA_ADD_EXECUTABLE and CUDA_ADD_LIBRARY
165 # (e.g. calling CUDA_WRAP_SRCS directly),
166 # CUDA_COMPUTE_SEPARABLE_COMPILATION_OBJECT_FILE_NAME and
167 # CUDA_LINK_SEPARABLE_COMPILATION_OBJECTS should be called.
169 # CUDA_SOURCE_PROPERTY_FORMAT
170 # -- If this source file property is set, it can override the format specified
171 # to CUDA_WRAP_SRCS (OBJ, PTX, CUBIN, or FATBIN). If an input source file
172 # is not a .cu file, setting this file will cause it to be treated as a .cu
173 # file. See documentation for set_source_files_properties on how to set
176 # CUDA_USE_STATIC_CUDA_RUNTIME (Default ON)
177 # -- When enabled the static version of the CUDA runtime library will be used
178 # in CUDA_LIBRARIES. If the version of CUDA configured doesn't support
179 # this option, then it will be silently disabled.
181 # CUDA_VERBOSE_BUILD (Default OFF)
182 # -- Set to ON to see all the commands used when building the CUDA file. When
183 # using a Makefile generator the value defaults to VERBOSE (run make
184 # VERBOSE=1 to see output), although setting CUDA_VERBOSE_BUILD to ON will
185 # always print the output.
187 # The script creates the following macros (in alphebetical order)::
189 # CUDA_ADD_CUFFT_TO_TARGET( cuda_target )
190 # -- Adds the cufft library to the target (can be any target). Handles whether
191 # you are in emulation mode or not.
193 # CUDA_ADD_CUBLAS_TO_TARGET( cuda_target )
194 # -- Adds the cublas library to the target (can be any target). Handles
195 # whether you are in emulation mode or not.
197 # CUDA_ADD_EXECUTABLE( cuda_target file0 file1 ...
198 # [WIN32] [MACOSX_BUNDLE] [EXCLUDE_FROM_ALL] [OPTIONS ...] )
199 # -- Creates an executable "cuda_target" which is made up of the files
200 # specified. All of the non CUDA C files are compiled using the standard
201 # build rules specified by CMAKE and the cuda files are compiled to object
202 # files using nvcc and the host compiler. In addition CUDA_INCLUDE_DIRS is
203 # added automatically to include_directories(). Some standard CMake target
204 # calls can be used on the target after calling this macro
205 # (e.g. set_target_properties and target_link_libraries), but setting
206 # properties that adjust compilation flags will not affect code compiled by
207 # nvcc. Such flags should be modified before calling CUDA_ADD_EXECUTABLE,
208 # CUDA_ADD_LIBRARY or CUDA_WRAP_SRCS.
210 # CUDA_ADD_LIBRARY( cuda_target file0 file1 ...
211 # [STATIC | SHARED | MODULE] [EXCLUDE_FROM_ALL] [OPTIONS ...] )
212 # -- Same as CUDA_ADD_EXECUTABLE except that a library is created.
214 # CUDA_BUILD_CLEAN_TARGET()
215 # -- Creates a convience target that deletes all the dependency files
216 # generated. You should make clean after running this target to ensure the
217 # dependency files get regenerated.
219 # CUDA_COMPILE( generated_files file0 file1 ... [STATIC | SHARED | MODULE]
221 # -- Returns a list of generated files from the input source files to be used
222 # with ADD_LIBRARY or ADD_EXECUTABLE.
224 # CUDA_COMPILE_PTX( generated_files file0 file1 ... [OPTIONS ...] )
225 # -- Returns a list of PTX files generated from the input source files.
227 # CUDA_COMPILE_FATBIN( generated_files file0 file1 ... [OPTIONS ...] )
228 # -- Returns a list of FATBIN files generated from the input source files.
230 # CUDA_COMPILE_CUBIN( generated_files file0 file1 ... [OPTIONS ...] )
231 # -- Returns a list of CUBIN files generated from the input source files.
233 # CUDA_COMPUTE_SEPARABLE_COMPILATION_OBJECT_FILE_NAME( output_file_var
236 # -- Compute the name of the intermediate link file used for separable
237 # compilation. This file name is typically passed into
238 # CUDA_LINK_SEPARABLE_COMPILATION_OBJECTS. output_file_var is produced
239 # based on cuda_target the list of objects files that need separable
240 # compilation as specified by object_files. If the object_files list is
241 # empty, then output_file_var will be empty. This function is called
242 # automatically for CUDA_ADD_LIBRARY and CUDA_ADD_EXECUTABLE. Note that
243 # this is a function and not a macro.
245 # CUDA_INCLUDE_DIRECTORIES( path0 path1 ... )
246 # -- Sets the directories that should be passed to nvcc
247 # (e.g. nvcc -Ipath0 -Ipath1 ... ). These paths usually contain other .cu
252 # CUDA_LINK_SEPARABLE_COMPILATION_OBJECTS( output_file_var cuda_target
253 # nvcc_flags object_files)
255 # -- Generates the link object required by separable compilation from the given
256 # object files. This is called automatically for CUDA_ADD_EXECUTABLE and
257 # CUDA_ADD_LIBRARY, but can be called manually when using CUDA_WRAP_SRCS
258 # directly. When called from CUDA_ADD_LIBRARY or CUDA_ADD_EXECUTABLE the
259 # nvcc_flags passed in are the same as the flags passed in via the OPTIONS
260 # argument. The only nvcc flag added automatically is the bitness flag as
261 # specified by CUDA_64_BIT_DEVICE_CODE. Note that this is a function
262 # instead of a macro.
264 # CUDA_WRAP_SRCS ( cuda_target format generated_files file0 file1 ...
265 # [STATIC | SHARED | MODULE] [OPTIONS ...] )
266 # -- This is where all the magic happens. CUDA_ADD_EXECUTABLE,
267 # CUDA_ADD_LIBRARY, CUDA_COMPILE, and CUDA_COMPILE_PTX all call this
268 # function under the hood.
270 # Given the list of files (file0 file1 ... fileN) this macro generates
271 # custom commands that generate either PTX or linkable objects (use "PTX" or
272 # "OBJ" for the format argument to switch). Files that don't end with .cu
273 # or have the HEADER_FILE_ONLY property are ignored.
275 # The arguments passed in after OPTIONS are extra command line options to
276 # give to nvcc. You can also specify per configuration options by
277 # specifying the name of the configuration followed by the options. General
278 # options must preceed configuration specific options. Not all
279 # configurations need to be specified, only the ones provided will be used.
281 # OPTIONS -DFLAG=2 "-DFLAG_OTHER=space in flag"
283 # RELEASE --use_fast_math
284 # RELWITHDEBINFO --use_fast_math;-g
285 # MINSIZEREL --use_fast_math
287 # For certain configurations (namely VS generating object files with
288 # CUDA_ATTACH_VS_BUILD_RULE_TO_CUDA_FILE set to ON), no generated file will
289 # be produced for the given cuda file. This is because when you add the
290 # cuda file to Visual Studio it knows that this file produces an object file
291 # and will link in the resulting object file automatically.
293 # This script will also generate a separate cmake script that is used at
294 # build time to invoke nvcc. This is for several reasons.
296 # 1. nvcc can return negative numbers as return values which confuses
297 # Visual Studio into thinking that the command succeeded. The script now
298 # checks the error codes and produces errors when there was a problem.
300 # 2. nvcc has been known to not delete incomplete results when it
301 # encounters problems. This confuses build systems into thinking the
302 # target was generated when in fact an unusable file exists. The script
303 # now deletes the output files if there was an error.
305 # 3. By putting all the options that affect the build into a file and then
306 # make the build rule dependent on the file, the output files will be
307 # regenerated when the options change.
309 # This script also looks at optional arguments STATIC, SHARED, or MODULE to
310 # determine when to target the object compilation for a shared library.
311 # BUILD_SHARED_LIBS is ignored in CUDA_WRAP_SRCS, but it is respected in
312 # CUDA_ADD_LIBRARY. On some systems special flags are added for building
313 # objects intended for shared libraries. A preprocessor macro,
314 # <target_name>_EXPORTS is defined when a shared library compilation is
317 # Flags passed into add_definitions with -D or /D are passed along to nvcc.
321 # The script defines the following variables::
323 # CUDA_VERSION_MAJOR -- The major version of cuda as reported by nvcc.
324 # CUDA_VERSION_MINOR -- The minor version.
326 # CUDA_VERSION_STRING -- CUDA_VERSION_MAJOR.CUDA_VERSION_MINOR
328 # CUDA_TOOLKIT_ROOT_DIR -- Path to the CUDA Toolkit (defined if not set).
329 # CUDA_SDK_ROOT_DIR -- Path to the CUDA SDK. Use this to find files in the
330 # SDK. This script will not directly support finding
331 # specific libraries or headers, as that isn't
332 # supported by NVIDIA. If you want to change
333 # libraries when the path changes see the
334 # FindCUDA.cmake script for an example of how to clear
335 # these variables. There are also examples of how to
336 # use the CUDA_SDK_ROOT_DIR to locate headers or
337 # libraries, if you so choose (at your own risk).
338 # CUDA_INCLUDE_DIRS -- Include directory for cuda headers. Added automatically
339 # for CUDA_ADD_EXECUTABLE and CUDA_ADD_LIBRARY.
340 # CUDA_LIBRARIES -- Cuda RT library.
341 # CUDA_CUFFT_LIBRARIES -- Device or emulation library for the Cuda FFT
342 # implementation (alternative to:
343 # CUDA_ADD_CUFFT_TO_TARGET macro)
344 # CUDA_CUBLAS_LIBRARIES -- Device or emulation library for the Cuda BLAS
345 # implementation (alterative to:
346 # CUDA_ADD_CUBLAS_TO_TARGET macro).
347 # CUDA_cudart_static_LIBRARY -- Statically linkable cuda runtime library.
348 # Only available for CUDA version 5.5+
349 # CUDA_cupti_LIBRARY -- CUDA Profiling Tools Interface library.
350 # Only available for CUDA version 4.0+.
351 # CUDA_curand_LIBRARY -- CUDA Random Number Generation library.
352 # Only available for CUDA version 3.2+.
353 # CUDA_cusolver_LIBRARY -- CUDA Direct Solver library.
354 # Only available for CUDA version 7.0+.
355 # CUDA_cusparse_LIBRARY -- CUDA Sparse Matrix library.
356 # Only available for CUDA version 3.2+.
357 # CUDA_npp_LIBRARY -- NVIDIA Performance Primitives lib.
358 # Only available for CUDA version 4.0+.
359 # CUDA_nppc_LIBRARY -- NVIDIA Performance Primitives lib (core).
360 # Only available for CUDA version 5.5+.
361 # CUDA_nppi_LIBRARY -- NVIDIA Performance Primitives lib (image processing).
362 # Only available for CUDA version 5.5+.
363 # CUDA_npps_LIBRARY -- NVIDIA Performance Primitives lib (signal processing).
364 # Only available for CUDA version 5.5+.
365 # CUDA_nvcuvenc_LIBRARY -- CUDA Video Encoder library.
366 # Only available for CUDA version 3.2+.
368 # CUDA_nvcuvid_LIBRARY -- CUDA Video Decoder library.
369 # Only available for CUDA version 3.2+.
373 # James Bigler, NVIDIA Corp (nvidia.com - jbigler)
374 # Abe Stephens, SCI Institute -- http://www.sci.utah.edu/~abe/FindCuda.html
376 # Copyright (c) 2008 - 2009 NVIDIA Corporation. All rights reserved.
378 # Copyright (c) 2007-2009
379 # Scientific Computing and Imaging Institute, University of Utah
381 # This code is licensed under the MIT License. See the FindCUDA.cmake script
382 # for the text of the license.
386 # License for the specific language governing rights and limitations under
387 # Permission is hereby granted, free of charge, to any person obtaining a
388 # copy of this software and associated documentation files (the "Software"),
389 # to deal in the Software without restriction, including without limitation
390 # the rights to use, copy, modify, merge, publish, distribute, sublicense,
391 # and/or sell copies of the Software, and to permit persons to whom the
392 # Software is furnished to do so, subject to the following conditions:
394 # The above copyright notice and this permission notice shall be included
395 # in all copies or substantial portions of the Software.
397 # THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
398 # OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
399 # FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
400 # THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
401 # LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
402 # FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
403 # DEALINGS IN THE SOFTWARE.
405 ###############################################################################
409 # This macro helps us find the location of helper files we will need the full path to
410 macro(CUDA_FIND_HELPER_FILE _name _extension)
411 set(_full_name "${_name}.${_extension}")
412 # CMAKE_CURRENT_LIST_FILE contains the full path to the file currently being
413 # processed. Using this variable, we can pull out the current path, and
414 # provide a way to get access to the other files we need local to here.
415 get_filename_component(CMAKE_CURRENT_LIST_DIR "${CMAKE_CURRENT_LIST_FILE}" PATH)
416 set(CUDA_${_name} "${CMAKE_CURRENT_LIST_DIR}/FindCUDA/${_full_name}")
417 if(NOT EXISTS "${CUDA_${_name}}")
418 set(error_message "${_full_name} not found in ${CMAKE_CURRENT_LIST_DIR}/FindCUDA")
419 if(CUDA_FIND_REQUIRED)
420 message(FATAL_ERROR "${error_message}")
422 if(NOT CUDA_FIND_QUIETLY)
423 message(STATUS "${error_message}")
427 # Set this variable as internal, so the user isn't bugged with it.
428 set(CUDA_${_name} ${CUDA_${_name}} CACHE INTERNAL "Location of ${_full_name}" FORCE)
431 #####################################################################
432 ## CUDA_INCLUDE_NVCC_DEPENDENCIES
435 # So we want to try and include the dependency file if it exists. If
436 # it doesn't exist then we need to create an empty one, so we can
439 # If it does exist, then we need to check to see if all the files it
440 # depends on exist. If they don't then we should clear the dependency
441 # file and regenerate it later. This covers the case where a header
442 # file has disappeared or moved.
444 macro(CUDA_INCLUDE_NVCC_DEPENDENCIES dependency_file)
445 set(CUDA_NVCC_DEPEND)
446 set(CUDA_NVCC_DEPEND_REGENERATE FALSE)
449 # Include the dependency file. Create it first if it doesn't exist . The
450 # INCLUDE puts a dependency that will force CMake to rerun and bring in the
451 # new info when it changes. DO NOT REMOVE THIS (as I did and spent a few
452 # hours figuring out why it didn't work.
453 if(NOT EXISTS ${dependency_file})
454 file(WRITE ${dependency_file} "#FindCUDA.cmake generated file. Do not edit.\n")
456 # Always include this file to force CMake to run again next
457 # invocation and rebuild the dependencies.
458 #message("including dependency_file = ${dependency_file}")
459 include(${dependency_file})
461 # Now we need to verify the existence of all the included files
462 # here. If they aren't there we need to just blank this variable and
463 # make the file regenerate again.
464 # if(DEFINED CUDA_NVCC_DEPEND)
465 # message("CUDA_NVCC_DEPEND set")
467 # message("CUDA_NVCC_DEPEND NOT set")
470 #message("CUDA_NVCC_DEPEND found")
471 foreach(f ${CUDA_NVCC_DEPEND})
472 # message("searching for ${f}")
474 #message("file ${f} not found")
475 set(CUDA_NVCC_DEPEND_REGENERATE TRUE)
479 #message("CUDA_NVCC_DEPEND false")
480 # No dependencies, so regenerate the file.
481 set(CUDA_NVCC_DEPEND_REGENERATE TRUE)
484 #message("CUDA_NVCC_DEPEND_REGENERATE = ${CUDA_NVCC_DEPEND_REGENERATE}")
485 # No incoming dependencies, so we need to generate them. Make the
486 # output depend on the dependency file itself, which should cause the
488 if(CUDA_NVCC_DEPEND_REGENERATE)
489 set(CUDA_NVCC_DEPEND ${dependency_file})
490 #message("Generating an empty dependency_file: ${dependency_file}")
491 file(WRITE ${dependency_file} "#FindCUDA.cmake generated file. Do not edit.\n")
496 ###############################################################################
497 ###############################################################################
498 # Setup variables' defaults
499 ###############################################################################
500 ###############################################################################
502 # Allow the user to specify if the device code is supposed to be 32 or 64 bit.
503 if(CMAKE_SIZEOF_VOID_P EQUAL 8)
504 set(CUDA_64_BIT_DEVICE_CODE_DEFAULT ON)
506 set(CUDA_64_BIT_DEVICE_CODE_DEFAULT OFF)
508 option(CUDA_64_BIT_DEVICE_CODE "Compile device code in 64 bit mode" ${CUDA_64_BIT_DEVICE_CODE_DEFAULT})
510 # Attach the build rule to the source file in VS. This option
511 option(CUDA_ATTACH_VS_BUILD_RULE_TO_CUDA_FILE "Attach the build rule to the CUDA source file. Enable only when the CUDA source file is added to at most one target." ON)
513 # Prints out extra information about the cuda file during compilation
514 option(CUDA_BUILD_CUBIN "Generate and parse .cubin files in Device mode." OFF)
516 # Set whether we are using emulation or device mode.
517 option(CUDA_BUILD_EMULATION "Build in Emulation mode" OFF)
519 # Where to put the generated output.
520 set(CUDA_GENERATED_OUTPUT_DIR "" CACHE PATH "Directory to put all the output files. If blank it will default to the CMAKE_CURRENT_BINARY_DIR")
522 # Parse HOST_COMPILATION mode.
523 option(CUDA_HOST_COMPILATION_CPP "Generated file extension" ON)
525 # Extra user settable flags
526 set(CUDA_NVCC_FLAGS "" CACHE STRING "Semi-colon delimit multiple arguments.")
528 if(CMAKE_GENERATOR MATCHES "Visual Studio")
529 set(CUDA_HOST_COMPILER "$(VCInstallDir)bin" CACHE FILEPATH "Host side compiler used by NVCC")
532 AND "${CMAKE_C_COMPILER_ID}" MATCHES "Clang"
533 AND "${CMAKE_C_COMPILER}" MATCHES "/cc$")
534 # Using cc which is symlink to clang may let NVCC think it is GCC and issue
535 # unhandled -dumpspecs option to clang. Also in case neither
536 # CMAKE_C_COMPILER is defined (project does not use C language) nor
537 # CUDA_HOST_COMPILER is specified manually we should skip -ccbin and let
538 # nvcc use its own default C compiler.
539 # Only care about this on APPLE with clang to avoid
540 # following symlinks to things like ccache
541 if(DEFINED CMAKE_C_COMPILER AND NOT DEFINED CUDA_HOST_COMPILER)
542 get_filename_component(c_compiler_realpath "${CMAKE_C_COMPILER}" REALPATH)
543 # if the real path does not end up being clang then
544 # go back to using CMAKE_C_COMPILER
545 if(NOT "${c_compiler_realpath}" MATCHES "/clang$")
546 set(c_compiler_realpath "${CMAKE_C_COMPILER}")
549 set(c_compiler_realpath "")
551 set(CUDA_HOST_COMPILER "${c_compiler_realpath}" CACHE FILEPATH "Host side compiler used by NVCC")
553 set(CUDA_HOST_COMPILER "${CMAKE_C_COMPILER}"
554 CACHE FILEPATH "Host side compiler used by NVCC")
558 # Propagate the host flags to the host compiler via -Xcompiler
559 option(CUDA_PROPAGATE_HOST_FLAGS "Propage C/CXX_FLAGS and friends to the host compiler via -Xcompile" ON)
561 # Enable CUDA_SEPARABLE_COMPILATION
562 option(CUDA_SEPARABLE_COMPILATION "Compile CUDA objects with separable compilation enabled. Requires CUDA 5.0+" OFF)
564 # Specifies whether the commands used when compiling the .cu file will be printed out.
565 option(CUDA_VERBOSE_BUILD "Print out the commands run while compiling the CUDA source file. With the Makefile generator this defaults to VERBOSE variable specified on the command line, but can be forced on with this option." OFF)
568 CUDA_64_BIT_DEVICE_CODE
569 CUDA_ATTACH_VS_BUILD_RULE_TO_CUDA_FILE
570 CUDA_GENERATED_OUTPUT_DIR
571 CUDA_HOST_COMPILATION_CPP
573 CUDA_PROPAGATE_HOST_FLAGS
577 CUDA_SEPARABLE_COMPILATION
580 # Makefile and similar generators don't define CMAKE_CONFIGURATION_TYPES, so we
581 # need to add another entry for the CMAKE_BUILD_TYPE. We also need to add the
582 # standerd set of 4 build types (Debug, MinSizeRel, Release, and RelWithDebInfo)
583 # for completeness. We need run this loop in order to accomodate the addition
584 # of extra configuration types. Duplicate entries will be removed by
586 set(CUDA_configuration_types ${CMAKE_CONFIGURATION_TYPES} ${CMAKE_BUILD_TYPE} Debug MinSizeRel Release RelWithDebInfo)
587 list(REMOVE_DUPLICATES CUDA_configuration_types)
588 foreach(config ${CUDA_configuration_types})
589 string(TOUPPER ${config} config_upper)
590 set(CUDA_NVCC_FLAGS_${config_upper} "" CACHE STRING "Semi-colon delimit multiple arguments.")
591 mark_as_advanced(CUDA_NVCC_FLAGS_${config_upper})
594 ###############################################################################
595 ###############################################################################
596 # Locate CUDA, Set Build Type, etc.
597 ###############################################################################
598 ###############################################################################
600 macro(cuda_unset_include_and_libraries)
601 unset(CUDA_TOOLKIT_INCLUDE CACHE)
602 unset(CUDA_CUDART_LIBRARY CACHE)
603 unset(CUDA_CUDA_LIBRARY CACHE)
604 # Make sure you run this before you unset CUDA_VERSION.
605 if(CUDA_VERSION VERSION_EQUAL "3.0")
606 # This only existed in the 3.0 version of the CUDA toolkit
607 unset(CUDA_CUDARTEMU_LIBRARY CACHE)
609 unset(CUDA_cudart_static_LIBRARY CACHE)
610 unset(CUDA_cublas_LIBRARY CACHE)
611 unset(CUDA_cublasemu_LIBRARY CACHE)
612 unset(CUDA_cufft_LIBRARY CACHE)
613 unset(CUDA_cufftemu_LIBRARY CACHE)
614 unset(CUDA_cupti_LIBRARY CACHE)
615 unset(CUDA_curand_LIBRARY CACHE)
616 unset(CUDA_cusolver_LIBRARY CACHE)
617 unset(CUDA_cusparse_LIBRARY CACHE)
618 unset(CUDA_npp_LIBRARY CACHE)
619 unset(CUDA_nppc_LIBRARY CACHE)
620 unset(CUDA_nppi_LIBRARY CACHE)
621 unset(CUDA_npps_LIBRARY CACHE)
622 unset(CUDA_nvcuvenc_LIBRARY CACHE)
623 unset(CUDA_nvcuvid_LIBRARY CACHE)
625 unset(CUDA_USE_STATIC_CUDA_RUNTIME CACHE)
628 # Check to see if the CUDA_TOOLKIT_ROOT_DIR and CUDA_SDK_ROOT_DIR have changed,
629 # if they have then clear the cache variables, so that will be detected again.
630 if(NOT "${CUDA_TOOLKIT_ROOT_DIR}" STREQUAL "${CUDA_TOOLKIT_ROOT_DIR_INTERNAL}")
631 unset(CUDA_TOOLKIT_TARGET_DIR CACHE)
632 unset(CUDA_NVCC_EXECUTABLE CACHE)
633 cuda_unset_include_and_libraries()
634 unset(CUDA_VERSION CACHE)
637 if(NOT "${CUDA_TOOLKIT_TARGET_DIR}" STREQUAL "${CUDA_TOOLKIT_TARGET_DIR_INTERNAL}")
638 cuda_unset_include_and_libraries()
641 if(NOT "${CUDA_SDK_ROOT_DIR}" STREQUAL "${CUDA_SDK_ROOT_DIR_INTERNAL}")
642 # No specific variables to catch. Use this kind of code before calling
643 # find_package(CUDA) to clean up any variables that may depend on this path.
645 # unset(MY_SPECIAL_CUDA_SDK_INCLUDE_DIR CACHE)
646 # unset(MY_SPECIAL_CUDA_SDK_LIBRARY CACHE)
649 # Search for the cuda distribution.
650 if(NOT CUDA_TOOLKIT_ROOT_DIR)
652 # Search in the CUDA_BIN_PATH first.
653 find_path(CUDA_TOOLKIT_ROOT_DIR
658 PATH_SUFFIXES bin bin64
659 DOC "Toolkit location."
662 # Now search default paths
663 find_path(CUDA_TOOLKIT_ROOT_DIR
667 DOC "Toolkit location."
670 if (CUDA_TOOLKIT_ROOT_DIR)
671 string(REGEX REPLACE "[/\\\\]?bin[64]*[/\\\\]?$" "" CUDA_TOOLKIT_ROOT_DIR ${CUDA_TOOLKIT_ROOT_DIR})
672 # We need to force this back into the cache.
673 set(CUDA_TOOLKIT_ROOT_DIR ${CUDA_TOOLKIT_ROOT_DIR} CACHE PATH "Toolkit location." FORCE)
675 if (NOT EXISTS ${CUDA_TOOLKIT_ROOT_DIR})
676 if(CUDA_FIND_REQUIRED)
677 message(FATAL_ERROR "Specify CUDA_TOOLKIT_ROOT_DIR")
678 elseif(NOT CUDA_FIND_QUIETLY)
679 message("CUDA_TOOLKIT_ROOT_DIR not found or specified")
684 # CUDA_NVCC_EXECUTABLE
685 find_program(CUDA_NVCC_EXECUTABLE
687 PATHS "${CUDA_TOOLKIT_ROOT_DIR}"
690 PATH_SUFFIXES bin bin64
693 # Search default search paths, after we search our own set of paths.
694 find_program(CUDA_NVCC_EXECUTABLE nvcc)
695 mark_as_advanced(CUDA_NVCC_EXECUTABLE)
697 if(CUDA_NVCC_EXECUTABLE AND NOT CUDA_VERSION)
698 # Compute the version.
699 execute_process (COMMAND ${CUDA_NVCC_EXECUTABLE} "--version" OUTPUT_VARIABLE NVCC_OUT)
700 string(REGEX REPLACE ".*release ([0-9]+)\\.([0-9]+).*" "\\1" CUDA_VERSION_MAJOR ${NVCC_OUT})
701 string(REGEX REPLACE ".*release ([0-9]+)\\.([0-9]+).*" "\\2" CUDA_VERSION_MINOR ${NVCC_OUT})
702 set(CUDA_VERSION "${CUDA_VERSION_MAJOR}.${CUDA_VERSION_MINOR}" CACHE STRING "Version of CUDA as computed from nvcc.")
703 mark_as_advanced(CUDA_VERSION)
705 # Need to set these based off of the cached value
706 string(REGEX REPLACE "([0-9]+)\\.([0-9]+).*" "\\1" CUDA_VERSION_MAJOR "${CUDA_VERSION}")
707 string(REGEX REPLACE "([0-9]+)\\.([0-9]+).*" "\\2" CUDA_VERSION_MINOR "${CUDA_VERSION}")
710 # Always set this convenience variable
711 set(CUDA_VERSION_STRING "${CUDA_VERSION}")
713 # Support for arm cross compilation with CUDA 5.5
714 if(CUDA_VERSION VERSION_GREATER "5.0" AND CMAKE_CROSSCOMPILING AND CMAKE_SYSTEM_PROCESSOR MATCHES "arm" AND EXISTS "${CUDA_TOOLKIT_ROOT_DIR}/targets/armv7-linux-gnueabihf")
715 set(CUDA_TOOLKIT_TARGET_DIR "${CUDA_TOOLKIT_ROOT_DIR}/targets/armv7-linux-gnueabihf" CACHE PATH "Toolkit target location.")
717 set(CUDA_TOOLKIT_TARGET_DIR "${CUDA_TOOLKIT_ROOT_DIR}" CACHE PATH "Toolkit target location.")
719 mark_as_advanced(CUDA_TOOLKIT_TARGET_DIR)
721 # Target CPU architecture
722 if(CUDA_VERSION VERSION_GREATER "5.0" AND CMAKE_CROSSCOMPILING AND CMAKE_SYSTEM_PROCESSOR MATCHES "arm")
723 set(_cuda_target_cpu_arch_initial "ARM")
725 set(_cuda_target_cpu_arch_initial "")
727 set(CUDA_TARGET_CPU_ARCH ${_cuda_target_cpu_arch_initial} CACHE STRING "Specify the name of the class of CPU architecture for which the input files must be compiled.")
728 mark_as_advanced(CUDA_TARGET_CPU_ARCH)
730 # CUDA_TOOLKIT_INCLUDE
731 find_path(CUDA_TOOLKIT_INCLUDE
732 device_functions.h # Header included in toolkit
733 PATHS "${CUDA_TOOLKIT_TARGET_DIR}" "${CUDA_TOOLKIT_ROOT_DIR}"
736 PATH_SUFFIXES include
739 # Search default search paths, after we search our own set of paths.
740 find_path(CUDA_TOOLKIT_INCLUDE device_functions.h)
741 mark_as_advanced(CUDA_TOOLKIT_INCLUDE)
743 # Set the user list of include dir to nothing to initialize it.
744 set (CUDA_NVCC_INCLUDE_ARGS_USER "")
745 set (CUDA_INCLUDE_DIRS ${CUDA_TOOLKIT_INCLUDE})
747 macro(cuda_find_library_local_first_with_path_ext _var _names _doc _path_ext )
748 if(CMAKE_SIZEOF_VOID_P EQUAL 8)
749 # CUDA 3.2+ on Windows moved the library directories, so we need the new
751 set(_cuda_64bit_lib_dir "${_path_ext}lib/x64" "${_path_ext}lib64" "${_path_ext}libx64" )
753 # CUDA 3.2+ on Windows moved the library directories, so we need to new
754 # (lib/Win32) and the old path (lib).
757 PATHS "${CUDA_TOOLKIT_TARGET_DIR}" "${CUDA_TOOLKIT_ROOT_DIR}"
760 PATH_SUFFIXES ${_cuda_64bit_lib_dir} "${_path_ext}lib/Win32" "${_path_ext}lib" "${_path_ext}libWin32"
764 # Search default search paths, after we search our own set of paths.
767 PATHS "/usr/lib/nvidia-current"
772 macro(cuda_find_library_local_first _var _names _doc)
773 cuda_find_library_local_first_with_path_ext( "${_var}" "${_names}" "${_doc}" "" )
776 macro(find_library_local_first _var _names _doc )
777 cuda_find_library_local_first( "${_var}" "${_names}" "${_doc}" "" )
782 cuda_find_library_local_first(CUDA_CUDART_LIBRARY cudart "\"cudart\" library")
783 if(CUDA_VERSION VERSION_EQUAL "3.0")
784 # The cudartemu library only existed for the 3.0 version of CUDA.
785 cuda_find_library_local_first(CUDA_CUDARTEMU_LIBRARY cudartemu "\"cudartemu\" library")
787 CUDA_CUDARTEMU_LIBRARY
790 if(NOT CUDA_VERSION VERSION_LESS "5.5")
791 cuda_find_library_local_first(CUDA_cudart_static_LIBRARY cudart_static "static CUDA runtime library")
792 mark_as_advanced(CUDA_cudart_static_LIBRARY)
794 if(CUDA_cudart_static_LIBRARY)
795 # Set whether to use the static cuda runtime.
796 option(CUDA_USE_STATIC_CUDA_RUNTIME "Use the static version of the CUDA runtime library if available" ON)
798 option(CUDA_USE_STATIC_CUDA_RUNTIME "Use the static version of the CUDA runtime library if available" OFF)
801 if(CUDA_USE_STATIC_CUDA_RUNTIME)
803 # Check for the dependent libraries. Here we look for pthreads.
804 if (DEFINED CMAKE_THREAD_PREFER_PTHREAD)
805 set(_cuda_cmake_thread_prefer_pthread ${CMAKE_THREAD_PREFER_PTHREAD})
807 set(CMAKE_THREAD_PREFER_PTHREAD 1)
809 # Many of the FindXYZ CMake comes with makes use of try_compile with int main(){return 0;}
810 # as the source file. Unfortunately this causes a warning with -Wstrict-prototypes and
811 # -Werror causes the try_compile to fail. We will just temporarily disable other flags
812 # when doing the find_package command here.
813 set(_cuda_cmake_c_flags ${CMAKE_C_FLAGS})
814 set(CMAKE_C_FLAGS "-fPIC")
815 find_package(Threads REQUIRED)
816 set(CMAKE_C_FLAGS ${_cuda_cmake_c_flags})
818 if (DEFINED _cuda_cmake_thread_prefer_pthread)
819 set(CMAKE_THREAD_PREFER_PTHREAD ${_cuda_cmake_thread_prefer_pthread})
820 unset(_cuda_cmake_thread_prefer_pthread)
822 unset(CMAKE_THREAD_PREFER_PTHREAD)
825 # Here is librt that has things such as, clock_gettime, shm_open, and shm_unlink.
826 find_library(CUDA_rt_LIBRARY rt)
827 find_library(CUDA_dl_LIBRARY dl)
828 if (NOT CUDA_rt_LIBRARY)
829 message(WARNING "Expecting to find librt for libcudart_static, but didn't find it.")
831 if (NOT CUDA_dl_LIBRARY)
832 message(WARNING "Expecting to find libdl for libcudart_static, but didn't find it.")
838 # CUPTI library showed up in cuda toolkit 4.0
839 if(NOT CUDA_VERSION VERSION_LESS "4.0")
840 cuda_find_library_local_first_with_path_ext(CUDA_cupti_LIBRARY cupti "\"cupti\" library" "extras/CUPTI/")
841 mark_as_advanced(CUDA_cupti_LIBRARY)
844 # Set the CUDA_LIBRARIES variable. This is the set of stuff to link against if you are
845 # using the CUDA runtime. For the dynamic version of the runtime, most of the
846 # dependencies are brough in, but for the static version there are additional libraries
847 # and linker commands needed.
848 # Initialize to empty
851 # If we are using emulation mode and we found the cudartemu library then use
852 # that one instead of cudart.
853 if(CUDA_BUILD_EMULATION AND CUDA_CUDARTEMU_LIBRARY)
854 list(APPEND CUDA_LIBRARIES ${CUDA_CUDARTEMU_LIBRARY})
855 elseif(CUDA_USE_STATIC_CUDA_RUNTIME AND CUDA_cudart_static_LIBRARY)
856 list(APPEND CUDA_LIBRARIES ${CUDA_cudart_static_LIBRARY} ${CMAKE_THREAD_LIBS_INIT})
858 list(APPEND CUDA_LIBRARIES ${CUDA_rt_LIBRARY})
861 list(APPEND CUDA_LIBRARIES ${CUDA_dl_LIBRARY})
864 # We need to add the default path to the driver (libcuda.dylib) as an rpath, so that
865 # the static cuda runtime can find it at runtime.
866 list(APPEND CUDA_LIBRARIES -Wl,-rpath,/usr/local/cuda/lib)
869 list(APPEND CUDA_LIBRARIES ${CUDA_CUDART_LIBRARY})
872 # 1.1 toolkit on linux doesn't appear to have a separate library on
874 cuda_find_library_local_first(CUDA_CUDA_LIBRARY cuda "\"cuda\" library (older versions only).")
881 #######################
882 # Look for some of the toolkit helper libraries
883 macro(FIND_CUDA_HELPER_LIBS _name)
884 cuda_find_library_local_first(CUDA_${_name}_LIBRARY ${_name} "\"${_name}\" library")
885 mark_as_advanced(CUDA_${_name}_LIBRARY)
888 #######################
889 # Disable emulation for v3.1 onward
890 if(CUDA_VERSION VERSION_GREATER "3.0")
891 if(CUDA_BUILD_EMULATION)
892 message(FATAL_ERROR "CUDA_BUILD_EMULATION is not supported in version 3.1 and onwards. You must disable it to proceed. You have version ${CUDA_VERSION}.")
896 # Search for additional CUDA toolkit libraries.
897 if(CUDA_VERSION VERSION_LESS "3.1")
898 # Emulation libraries aren't available in version 3.1 onward.
899 find_cuda_helper_libs(cufftemu)
900 find_cuda_helper_libs(cublasemu)
902 find_cuda_helper_libs(cufft)
903 find_cuda_helper_libs(cublas)
904 if(NOT CUDA_VERSION VERSION_LESS "3.2")
905 # cusparse showed up in version 3.2
906 find_cuda_helper_libs(cusparse)
907 find_cuda_helper_libs(curand)
909 find_cuda_helper_libs(nvcuvenc)
910 find_cuda_helper_libs(nvcuvid)
913 if(CUDA_VERSION VERSION_GREATER "5.0")
914 # In CUDA 5.5 NPP was splitted onto 3 separate libraries.
915 find_cuda_helper_libs(nppc)
916 find_cuda_helper_libs(nppi)
917 find_cuda_helper_libs(npps)
918 set(CUDA_npp_LIBRARY "${CUDA_nppc_LIBRARY};${CUDA_nppi_LIBRARY};${CUDA_npps_LIBRARY}")
919 elseif(NOT CUDA_VERSION VERSION_LESS "4.0")
920 find_cuda_helper_libs(npp)
922 if(NOT CUDA_VERSION VERSION_LESS "7.0")
923 # cusolver showed up in version 7.0
924 find_cuda_helper_libs(cusolver)
927 if (CUDA_BUILD_EMULATION)
928 set(CUDA_CUFFT_LIBRARIES ${CUDA_cufftemu_LIBRARY})
929 set(CUDA_CUBLAS_LIBRARIES ${CUDA_cublasemu_LIBRARY})
931 set(CUDA_CUFFT_LIBRARIES ${CUDA_cufft_LIBRARY})
932 set(CUDA_CUBLAS_LIBRARIES ${CUDA_cublas_LIBRARY})
935 ########################
936 # Look for the SDK stuff. As of CUDA 3.0 NVSDKCUDA_ROOT has been replaced with
937 # NVSDKCOMPUTE_ROOT with the old CUDA C contents moved into the C subdirectory
938 find_path(CUDA_SDK_ROOT_DIR common/inc/cutil.h
940 "$ENV{NVSDKCOMPUTE_ROOT}/C"
942 "[HKEY_LOCAL_MACHINE\\SOFTWARE\\NVIDIA Corporation\\Installed Products\\NVIDIA SDK 10\\Compute;InstallDir]"
944 "/Developer/GPU\ Computing/C"
947 # Keep the CUDA_SDK_ROOT_DIR first in order to be able to override the
948 # environment variables.
949 set(CUDA_SDK_SEARCH_PATH
950 "${CUDA_SDK_ROOT_DIR}"
951 "${CUDA_TOOLKIT_ROOT_DIR}/local/NVSDK0.2"
952 "${CUDA_TOOLKIT_ROOT_DIR}/NVSDK0.2"
953 "${CUDA_TOOLKIT_ROOT_DIR}/NV_CUDA_SDK"
954 "$ENV{HOME}/NVIDIA_CUDA_SDK"
955 "$ENV{HOME}/NVIDIA_CUDA_SDK_MACOSX"
959 # Example of how to find an include file from the CUDA_SDK_ROOT_DIR
961 # find_path(CUDA_CUT_INCLUDE_DIR
963 # PATHS ${CUDA_SDK_SEARCH_PATH}
964 # PATH_SUFFIXES "common/inc"
965 # DOC "Location of cutil.h"
968 # # Now search system paths
969 # find_path(CUDA_CUT_INCLUDE_DIR cutil.h DOC "Location of cutil.h")
971 # mark_as_advanced(CUDA_CUT_INCLUDE_DIR)
974 # Example of how to find a library in the CUDA_SDK_ROOT_DIR
976 # # cutil library is called cutil64 for 64 bit builds on windows. We don't want
977 # # to get these confused, so we are setting the name based on the word size of
980 # if(CMAKE_SIZEOF_VOID_P EQUAL 8)
981 # set(cuda_cutil_name cutil64)
983 # set(cuda_cutil_name cutil32)
986 # find_library(CUDA_CUT_LIBRARY
987 # NAMES cutil ${cuda_cutil_name}
988 # PATHS ${CUDA_SDK_SEARCH_PATH}
989 # # The new version of the sdk shows up in common/lib, but the old one is in lib
990 # PATH_SUFFIXES "common/lib" "lib"
991 # DOC "Location of cutil library"
994 # # Now search system paths
995 # find_library(CUDA_CUT_LIBRARY NAMES cutil ${cuda_cutil_name} DOC "Location of cutil library")
996 # mark_as_advanced(CUDA_CUT_LIBRARY)
997 # set(CUDA_CUT_LIBRARIES ${CUDA_CUT_LIBRARY})
1001 #############################
1002 # Check for required components
1003 set(CUDA_FOUND TRUE)
1005 set(CUDA_TOOLKIT_ROOT_DIR_INTERNAL "${CUDA_TOOLKIT_ROOT_DIR}" CACHE INTERNAL
1006 "This is the value of the last time CUDA_TOOLKIT_ROOT_DIR was set successfully." FORCE)
1007 set(CUDA_TOOLKIT_TARGET_DIR_INTERNAL "${CUDA_TOOLKIT_TARGET_DIR}" CACHE INTERNAL
1008 "This is the value of the last time CUDA_TOOLKIT_TARGET_DIR was set successfully." FORCE)
1009 set(CUDA_SDK_ROOT_DIR_INTERNAL "${CUDA_SDK_ROOT_DIR}" CACHE INTERNAL
1010 "This is the value of the last time CUDA_SDK_ROOT_DIR was set successfully." FORCE)
1012 include(FindPackageHandleStandardArgs)
1013 find_package_handle_standard_args(CUDA
1015 CUDA_TOOLKIT_ROOT_DIR
1016 CUDA_NVCC_EXECUTABLE
1025 ###############################################################################
1026 ###############################################################################
1028 ###############################################################################
1029 ###############################################################################
1031 ###############################################################################
1032 # Add include directories to pass to the nvcc command.
1033 macro(CUDA_INCLUDE_DIRECTORIES)
1034 foreach(dir ${ARGN})
1035 list(APPEND CUDA_NVCC_INCLUDE_ARGS_USER -I${dir})
1040 ##############################################################################
1041 cuda_find_helper_file(parse_cubin cmake)
1042 cuda_find_helper_file(make2cmake cmake)
1043 cuda_find_helper_file(run_nvcc cmake)
1045 ##############################################################################
1046 # Separate the OPTIONS out from the sources
1048 macro(CUDA_GET_SOURCES_AND_OPTIONS _sources _cmake_options _options)
1050 set( ${_cmake_options} )
1052 set( _found_options FALSE )
1053 foreach(arg ${ARGN})
1054 if("x${arg}" STREQUAL "xOPTIONS")
1055 set( _found_options TRUE )
1057 "x${arg}" STREQUAL "xWIN32" OR
1058 "x${arg}" STREQUAL "xMACOSX_BUNDLE" OR
1059 "x${arg}" STREQUAL "xEXCLUDE_FROM_ALL" OR
1060 "x${arg}" STREQUAL "xSTATIC" OR
1061 "x${arg}" STREQUAL "xSHARED" OR
1062 "x${arg}" STREQUAL "xMODULE"
1064 list(APPEND ${_cmake_options} ${arg})
1066 if ( _found_options )
1067 list(APPEND ${_options} ${arg})
1069 # Assume this is a file
1070 list(APPEND ${_sources} ${arg})
1076 ##############################################################################
1077 # Parse the OPTIONS from ARGN and set the variables prefixed by _option_prefix
1079 macro(CUDA_PARSE_NVCC_OPTIONS _option_prefix)
1080 set( _found_config )
1081 foreach(arg ${ARGN})
1082 # Determine if we are dealing with a perconfiguration flag
1083 foreach(config ${CUDA_configuration_types})
1084 string(TOUPPER ${config} config_upper)
1085 if (arg STREQUAL "${config_upper}")
1086 set( _found_config _${arg})
1087 # Set arg to nothing to keep it from being processed further
1093 list(APPEND ${_option_prefix}${_found_config} "${arg}")
1098 ##############################################################################
1099 # Helper to add the include directory for CUDA only once
1100 function(CUDA_ADD_CUDA_INCLUDE_ONCE)
1101 get_directory_property(_include_directories INCLUDE_DIRECTORIES)
1103 if(_include_directories)
1104 foreach(dir ${_include_directories})
1105 if("${dir}" STREQUAL "${CUDA_INCLUDE_DIRS}")
1111 include_directories(${CUDA_INCLUDE_DIRS})
1115 function(CUDA_BUILD_SHARED_LIBRARY shared_flag)
1116 set(cmake_args ${ARGN})
1117 # If SHARED, MODULE, or STATIC aren't already in the list of arguments, then
1118 # add SHARED or STATIC based on the value of BUILD_SHARED_LIBS.
1119 list(FIND cmake_args SHARED _cuda_found_SHARED)
1120 list(FIND cmake_args MODULE _cuda_found_MODULE)
1121 list(FIND cmake_args STATIC _cuda_found_STATIC)
1122 if( _cuda_found_SHARED GREATER -1 OR
1123 _cuda_found_MODULE GREATER -1 OR
1124 _cuda_found_STATIC GREATER -1)
1125 set(_cuda_build_shared_libs)
1127 if (BUILD_SHARED_LIBS)
1128 set(_cuda_build_shared_libs SHARED)
1130 set(_cuda_build_shared_libs STATIC)
1133 set(${shared_flag} ${_cuda_build_shared_libs} PARENT_SCOPE)
1136 ##############################################################################
1137 # Helper to avoid clashes of files with the same basename but different paths.
1138 # This doesn't attempt to do exactly what CMake internals do, which is to only
1139 # add this path when there is a conflict, since by the time a second collision
1140 # in names is detected it's already too late to fix the first one. For
1141 # consistency sake the relative path will be added to all files.
1142 function(CUDA_COMPUTE_BUILD_PATH path build_path)
1143 #message("CUDA_COMPUTE_BUILD_PATH([${path}] ${build_path})")
1144 # Only deal with CMake style paths from here on out
1145 file(TO_CMAKE_PATH "${path}" bpath)
1146 if (IS_ABSOLUTE "${bpath}")
1147 # Absolute paths are generally unnessary, especially if something like
1148 # file(GLOB_RECURSE) is used to pick up the files.
1150 string(FIND "${bpath}" "${CMAKE_CURRENT_BINARY_DIR}" _binary_dir_pos)
1151 if (_binary_dir_pos EQUAL 0)
1152 file(RELATIVE_PATH bpath "${CMAKE_CURRENT_BINARY_DIR}" "${bpath}")
1154 file(RELATIVE_PATH bpath "${CMAKE_CURRENT_SOURCE_DIR}" "${bpath}")
1158 # This recipe is from cmLocalGenerator::CreateSafeUniqueObjectFileName in the
1162 string(REGEX REPLACE "^[/]+" "" bpath "${bpath}")
1163 # Avoid absolute paths by removing ':'
1164 string(REPLACE ":" "_" bpath "${bpath}")
1165 # Avoid relative paths that go up the tree
1166 string(REPLACE "../" "__/" bpath "${bpath}")
1168 string(REPLACE " " "_" bpath "${bpath}")
1170 # Strip off the filename. I wait until here to do it, since removin the
1171 # basename can make a path that looked like path/../basename turn into
1172 # path/.. (notice the trailing slash).
1173 get_filename_component(bpath "${bpath}" PATH)
1175 set(${build_path} "${bpath}" PARENT_SCOPE)
1176 #message("${build_path} = ${bpath}")
1179 ##############################################################################
1180 # This helper macro populates the following variables and setups up custom
1181 # commands and targets to invoke the nvcc compiler to generate C or PTX source
1182 # dependent upon the format parameter. The compiler is invoked once with -M
1183 # to generate a dependency file and a second time with -cuda or -ptx to generate
1184 # a .cpp or .ptx file.
1186 # cuda_target - Target name
1187 # format - PTX, CUBIN, FATBIN or OBJ
1188 # FILE1 .. FILEN - The remaining arguments are the sources to be wrapped.
1189 # OPTIONS - Extra options to NVCC
1191 # generated_files - List of generated files
1192 ##############################################################################
1193 ##############################################################################
1195 macro(CUDA_WRAP_SRCS cuda_target format generated_files)
1197 # If CMake doesn't support separable compilation, complain
1198 if(CUDA_SEPARABLE_COMPILATION AND CMAKE_VERSION VERSION_LESS "2.8.10.1")
1199 message(SEND_ERROR "CUDA_SEPARABLE_COMPILATION isn't supported for CMake versions less than 2.8.10.1")
1202 # Set up all the command line flags here, so that they can be overridden on a per target basis.
1206 # Emulation if the card isn't present.
1207 if (CUDA_BUILD_EMULATION)
1209 set(nvcc_flags ${nvcc_flags} --device-emulation -D_DEVICEEMU -g)
1211 # Device mode. No flags necessary.
1214 if(CUDA_HOST_COMPILATION_CPP)
1215 set(CUDA_C_OR_CXX CXX)
1217 if(CUDA_VERSION VERSION_LESS "3.0")
1218 set(nvcc_flags ${nvcc_flags} --host-compilation C)
1220 message(WARNING "--host-compilation flag is deprecated in CUDA version >= 3.0. Removing --host-compilation C flag" )
1222 set(CUDA_C_OR_CXX C)
1225 set(generated_extension ${CMAKE_${CUDA_C_OR_CXX}_OUTPUT_EXTENSION})
1227 if(CUDA_64_BIT_DEVICE_CODE)
1228 set(nvcc_flags ${nvcc_flags} -m64)
1230 set(nvcc_flags ${nvcc_flags} -m32)
1233 if(CUDA_TARGET_CPU_ARCH)
1234 set(nvcc_flags ${nvcc_flags} "--target-cpu-architecture=${CUDA_TARGET_CPU_ARCH}")
1237 # This needs to be passed in at this stage, because VS needs to fill out the
1238 # value of VCInstallDir from within VS. Note that CCBIN is only used if
1239 # -ccbin or --compiler-bindir isn't used and CUDA_HOST_COMPILER matches
1240 # $(VCInstallDir)/bin.
1241 if(CMAKE_GENERATOR MATCHES "Visual Studio")
1242 set(ccbin_flags -D "\"CCBIN:PATH=$(VCInstallDir)bin\"" )
1247 # Figure out which configure we will use and pass that in as an argument to
1248 # the script. We need to defer the decision until compilation time, because
1249 # for VS projects we won't know if we are making a debug or release build
1251 if(CMAKE_GENERATOR MATCHES "Visual Studio")
1252 set( CUDA_build_configuration "$(ConfigurationName)" )
1254 set( CUDA_build_configuration "${CMAKE_BUILD_TYPE}")
1257 # Initialize our list of includes with the user ones followed by the CUDA system ones.
1258 set(CUDA_NVCC_INCLUDE_ARGS ${CUDA_NVCC_INCLUDE_ARGS_USER} --system-include;${CUDA_INCLUDE_DIRS})
1259 # Get the include directories for this directory and use them for our nvcc command.
1260 # Remove duplicate entries which may be present since include_directories
1261 # in CMake >= 2.8.8 does not remove them.
1262 get_directory_property(CUDA_NVCC_INCLUDE_DIRECTORIES INCLUDE_DIRECTORIES)
1263 list(REMOVE_DUPLICATES CUDA_NVCC_INCLUDE_DIRECTORIES)
1264 if(CUDA_NVCC_INCLUDE_DIRECTORIES)
1265 foreach(dir ${CUDA_NVCC_INCLUDE_DIRECTORIES})
1266 list(APPEND CUDA_NVCC_INCLUDE_ARGS --include-path;${dir})
1270 # Reset these variables
1271 set(CUDA_WRAP_OPTION_NVCC_FLAGS)
1272 foreach(config ${CUDA_configuration_types})
1273 string(TOUPPER ${config} config_upper)
1274 set(CUDA_WRAP_OPTION_NVCC_FLAGS_${config_upper})
1277 CUDA_GET_SOURCES_AND_OPTIONS(_cuda_wrap_sources _cuda_wrap_cmake_options _cuda_wrap_options ${ARGN})
1278 CUDA_PARSE_NVCC_OPTIONS(CUDA_WRAP_OPTION_NVCC_FLAGS ${_cuda_wrap_options})
1280 # Figure out if we are building a shared library. BUILD_SHARED_LIBS is
1281 # respected in CUDA_ADD_LIBRARY.
1282 set(_cuda_build_shared_libs FALSE)
1284 list(FIND _cuda_wrap_cmake_options SHARED _cuda_found_SHARED)
1285 list(FIND _cuda_wrap_cmake_options MODULE _cuda_found_MODULE)
1286 if(_cuda_found_SHARED GREATER -1 OR _cuda_found_MODULE GREATER -1)
1287 set(_cuda_build_shared_libs TRUE)
1290 list(FIND _cuda_wrap_cmake_options STATIC _cuda_found_STATIC)
1291 if(_cuda_found_STATIC GREATER -1)
1292 set(_cuda_build_shared_libs FALSE)
1296 if(_cuda_build_shared_libs)
1297 # If we are setting up code for a shared library, then we need to add extra flags for
1298 # compiling objects for shared libraries.
1299 set(CUDA_HOST_SHARED_FLAGS ${CMAKE_SHARED_LIBRARY_${CUDA_C_OR_CXX}_FLAGS})
1301 set(CUDA_HOST_SHARED_FLAGS)
1303 # Only add the CMAKE_{C,CXX}_FLAGS if we are propagating host flags. We
1304 # always need to set the SHARED_FLAGS, though.
1305 if(CUDA_PROPAGATE_HOST_FLAGS)
1306 set(_cuda_host_flags "set(CMAKE_HOST_FLAGS ${CMAKE_${CUDA_C_OR_CXX}_FLAGS} ${CUDA_HOST_SHARED_FLAGS})")
1308 set(_cuda_host_flags "set(CMAKE_HOST_FLAGS ${CUDA_HOST_SHARED_FLAGS})")
1311 set(_cuda_nvcc_flags_config "# Build specific configuration flags")
1312 # Loop over all the configuration types to generate appropriate flags for run_nvcc.cmake
1313 foreach(config ${CUDA_configuration_types})
1314 string(TOUPPER ${config} config_upper)
1315 # CMAKE_FLAGS are strings and not lists. By not putting quotes around CMAKE_FLAGS
1316 # we convert the strings to lists (like we want).
1318 if(CUDA_PROPAGATE_HOST_FLAGS)
1319 # nvcc chokes on -g3 in versions previous to 3.0, so replace it with -g
1320 set(_cuda_fix_g3 FALSE)
1322 if(CMAKE_COMPILER_IS_GNUCC)
1323 if (CUDA_VERSION VERSION_LESS "3.0" OR
1324 CUDA_VERSION VERSION_EQUAL "4.1" OR
1325 CUDA_VERSION VERSION_EQUAL "4.2"
1327 set(_cuda_fix_g3 TRUE)
1331 string(REPLACE "-g3" "-g" _cuda_C_FLAGS "${CMAKE_${CUDA_C_OR_CXX}_FLAGS_${config_upper}}")
1333 set(_cuda_C_FLAGS "${CMAKE_${CUDA_C_OR_CXX}_FLAGS_${config_upper}}")
1336 set(_cuda_host_flags "${_cuda_host_flags}\nset(CMAKE_HOST_FLAGS_${config_upper} ${_cuda_C_FLAGS})")
1339 # Note that if we ever want CUDA_NVCC_FLAGS_<CONFIG> to be string (instead of a list
1340 # like it is currently), we can remove the quotes around the
1341 # ${CUDA_NVCC_FLAGS_${config_upper}} variable like the CMAKE_HOST_FLAGS_<CONFIG> variable.
1342 set(_cuda_nvcc_flags_config "${_cuda_nvcc_flags_config}\nset(CUDA_NVCC_FLAGS_${config_upper} ${CUDA_NVCC_FLAGS_${config_upper}} ;; ${CUDA_WRAP_OPTION_NVCC_FLAGS_${config_upper}})")
1345 # Process the C++11 flag. If the host sets the flag, we need to add it to nvcc and
1346 # remove it from the host. This is because -Xcompile -std=c++ will choke nvcc (it uses
1347 # the C preprocessor). In order to get this to work correctly, we need to use nvcc's
1348 # specific c++11 flag.
1349 if( "${_cuda_host_flags}" MATCHES "-std=c\\+\\+11")
1350 # Add the c++11 flag to nvcc if it isn't already present. Note that we only look at
1351 # the main flag instead of the configuration specific flags.
1352 if( NOT "${CUDA_NVCC_FLAGS}" MATCHES "-std;c\\+\\+11" )
1353 list(APPEND nvcc_flags --std c++11)
1355 string(REGEX REPLACE "[-]+std=c\\+\\+11" "" _cuda_host_flags "${_cuda_host_flags}")
1358 # Get the list of definitions from the directory property
1359 get_directory_property(CUDA_NVCC_DEFINITIONS COMPILE_DEFINITIONS)
1360 if(CUDA_NVCC_DEFINITIONS)
1361 foreach(_definition ${CUDA_NVCC_DEFINITIONS})
1362 list(APPEND nvcc_flags "-D${_definition}")
1366 if(_cuda_build_shared_libs)
1367 list(APPEND nvcc_flags "-D${cuda_target}_EXPORTS")
1370 # Reset the output variable
1371 set(_cuda_wrap_generated_files "")
1373 # Iterate over the macro arguments and create custom
1374 # commands for all the .cu files.
1375 foreach(file ${ARGN})
1376 # Ignore any file marked as a HEADER_FILE_ONLY
1377 get_source_file_property(_is_header ${file} HEADER_FILE_ONLY)
1378 # Allow per source file overrides of the format. Also allows compiling non-.cu files.
1379 get_source_file_property(_cuda_source_format ${file} CUDA_SOURCE_PROPERTY_FORMAT)
1380 if((${file} MATCHES "\\.cu$" OR _cuda_source_format) AND NOT _is_header)
1382 if(NOT _cuda_source_format)
1383 set(_cuda_source_format ${format})
1385 # If file isn't a .cu file, we need to tell nvcc to treat it as such.
1386 if(NOT ${file} MATCHES "\\.cu$")
1387 set(cuda_language_flag -x=cu)
1389 set(cuda_language_flag)
1392 if( ${_cuda_source_format} MATCHES "OBJ")
1393 set( cuda_compile_to_external_module OFF )
1395 set( cuda_compile_to_external_module ON )
1396 if( ${_cuda_source_format} MATCHES "PTX" )
1397 set( cuda_compile_to_external_module_type "ptx" )
1398 elseif( ${_cuda_source_format} MATCHES "CUBIN")
1399 set( cuda_compile_to_external_module_type "cubin" )
1400 elseif( ${_cuda_source_format} MATCHES "FATBIN")
1401 set( cuda_compile_to_external_module_type "fatbin" )
1403 message( FATAL_ERROR "Invalid format flag passed to CUDA_WRAP_SRCS or set with CUDA_SOURCE_PROPERTY_FORMAT file property for file '${file}': '${_cuda_source_format}'. Use OBJ, PTX, CUBIN or FATBIN.")
1407 if(cuda_compile_to_external_module)
1408 # Don't use any of the host compilation flags for PTX targets.
1409 set(CUDA_HOST_FLAGS)
1410 set(CUDA_NVCC_FLAGS_CONFIG)
1412 set(CUDA_HOST_FLAGS ${_cuda_host_flags})
1413 set(CUDA_NVCC_FLAGS_CONFIG ${_cuda_nvcc_flags_config})
1416 # Determine output directory
1417 cuda_compute_build_path("${file}" cuda_build_path)
1418 set(cuda_compile_intermediate_directory "${CMAKE_CURRENT_BINARY_DIR}/CMakeFiles/${cuda_target}.dir/${cuda_build_path}")
1419 if(CUDA_GENERATED_OUTPUT_DIR)
1420 set(cuda_compile_output_dir "${CUDA_GENERATED_OUTPUT_DIR}")
1422 if ( cuda_compile_to_external_module )
1423 set(cuda_compile_output_dir "${CMAKE_CURRENT_BINARY_DIR}")
1425 set(cuda_compile_output_dir "${cuda_compile_intermediate_directory}")
1429 # Add a custom target to generate a c or ptx file. ######################
1431 get_filename_component( basename ${file} NAME )
1432 if( cuda_compile_to_external_module )
1433 set(generated_file_path "${cuda_compile_output_dir}")
1434 set(generated_file_basename "${cuda_target}_generated_${basename}.${cuda_compile_to_external_module_type}")
1435 set(format_flag "-${cuda_compile_to_external_module_type}")
1436 file(MAKE_DIRECTORY "${cuda_compile_output_dir}")
1438 set(generated_file_path "${cuda_compile_output_dir}/${CMAKE_CFG_INTDIR}")
1439 set(generated_file_basename "${cuda_target}_generated_${basename}${generated_extension}")
1440 if(CUDA_SEPARABLE_COMPILATION)
1441 set(format_flag "-dc")
1443 set(format_flag "-c")
1447 # Set all of our file names. Make sure that whatever filenames that have
1448 # generated_file_path in them get passed in through as a command line
1449 # argument, so that the ${CMAKE_CFG_INTDIR} gets expanded at run time
1450 # instead of configure time.
1451 set(generated_file "${generated_file_path}/${generated_file_basename}")
1452 set(cmake_dependency_file "${cuda_compile_intermediate_directory}/${generated_file_basename}.depend")
1453 set(NVCC_generated_dependency_file "${cuda_compile_intermediate_directory}/${generated_file_basename}.NVCC-depend")
1454 set(generated_cubin_file "${generated_file_path}/${generated_file_basename}.cubin.txt")
1455 set(custom_target_script "${cuda_compile_intermediate_directory}/${generated_file_basename}.cmake")
1457 # Setup properties for obj files:
1458 if( NOT cuda_compile_to_external_module )
1459 set_source_files_properties("${generated_file}"
1461 EXTERNAL_OBJECT true # This is an object file not to be compiled, but only be linked.
1465 # Don't add CMAKE_CURRENT_SOURCE_DIR if the path is already an absolute path.
1466 get_filename_component(file_path "${file}" PATH)
1467 if(IS_ABSOLUTE "${file_path}")
1468 set(source_file "${file}")
1470 set(source_file "${CMAKE_CURRENT_SOURCE_DIR}/${file}")
1473 if( NOT cuda_compile_to_external_module AND CUDA_SEPARABLE_COMPILATION)
1474 list(APPEND ${cuda_target}_SEPARABLE_COMPILATION_OBJECTS "${generated_file}")
1477 # Bring in the dependencies. Creates a variable CUDA_NVCC_DEPEND #######
1478 cuda_include_nvcc_dependencies(${cmake_dependency_file})
1480 # Convience string for output ###########################################
1481 if(CUDA_BUILD_EMULATION)
1482 set(cuda_build_type "Emulation")
1484 set(cuda_build_type "Device")
1487 # Build the NVCC made dependency file ###################################
1488 set(build_cubin OFF)
1489 if ( NOT CUDA_BUILD_EMULATION AND CUDA_BUILD_CUBIN )
1490 if ( NOT cuda_compile_to_external_module )
1491 set ( build_cubin ON )
1495 # Configure the build script
1496 configure_file("${CUDA_run_nvcc}" "${custom_target_script}" @ONLY)
1498 # So if a user specifies the same cuda file as input more than once, you
1499 # can have bad things happen with dependencies. Here we check an option
1500 # to see if this is the behavior they want.
1501 if(CUDA_ATTACH_VS_BUILD_RULE_TO_CUDA_FILE)
1502 set(main_dep MAIN_DEPENDENCY ${source_file})
1504 set(main_dep DEPENDS ${source_file})
1507 if(CUDA_VERBOSE_BUILD)
1508 set(verbose_output ON)
1509 elseif(CMAKE_GENERATOR MATCHES "Makefiles")
1510 set(verbose_output "$(VERBOSE)")
1512 set(verbose_output OFF)
1515 # Create up the comment string
1516 file(RELATIVE_PATH generated_file_relative_path "${CMAKE_BINARY_DIR}" "${generated_file}")
1517 if(cuda_compile_to_external_module)
1518 set(cuda_build_comment_string "Building NVCC ${cuda_compile_to_external_module_type} file ${generated_file_relative_path}")
1520 set(cuda_build_comment_string "Building NVCC (${cuda_build_type}) object ${generated_file_relative_path}")
1523 # Build the generated file and dependency file ##########################
1525 OUTPUT ${generated_file}
1526 # These output files depend on the source_file and the contents of cmake_dependency_file
1528 DEPENDS ${CUDA_NVCC_DEPEND}
1529 DEPENDS ${custom_target_script}
1530 # Make sure the output directory exists before trying to write to it.
1531 COMMAND ${CMAKE_COMMAND} -E make_directory "${generated_file_path}"
1532 COMMAND ${CMAKE_COMMAND} ARGS
1533 -D verbose:BOOL=${verbose_output}
1535 -D build_configuration:STRING=${CUDA_build_configuration}
1536 -D "generated_file:STRING=${generated_file}"
1537 -D "generated_cubin_file:STRING=${generated_cubin_file}"
1538 -P "${custom_target_script}"
1539 WORKING_DIRECTORY "${cuda_compile_intermediate_directory}"
1540 COMMENT "${cuda_build_comment_string}"
1543 # Make sure the build system knows the file is generated.
1544 set_source_files_properties(${generated_file} PROPERTIES GENERATED TRUE)
1546 list(APPEND _cuda_wrap_generated_files ${generated_file})
1548 # Add the other files that we want cmake to clean on a cleanup ##########
1549 list(APPEND CUDA_ADDITIONAL_CLEAN_FILES "${cmake_dependency_file}")
1550 list(REMOVE_DUPLICATES CUDA_ADDITIONAL_CLEAN_FILES)
1551 set(CUDA_ADDITIONAL_CLEAN_FILES ${CUDA_ADDITIONAL_CLEAN_FILES} CACHE INTERNAL "List of intermediate files that are part of the cuda dependency scanning.")
1556 # Set the return parameter
1557 set(${generated_files} ${_cuda_wrap_generated_files})
1560 function(_cuda_get_important_host_flags important_flags flag_string)
1561 if(CMAKE_GENERATOR MATCHES "Visual Studio")
1562 string(REGEX MATCHALL "/M[DT][d]?" flags "${flag_string}")
1563 list(APPEND ${important_flags} ${flags})
1565 string(REGEX MATCHALL "-fPIC" flags "${flag_string}")
1566 list(APPEND ${important_flags} ${flags})
1568 set(${important_flags} ${${important_flags}} PARENT_SCOPE)
1571 ###############################################################################
1572 ###############################################################################
1573 # Separable Compilation Link
1574 ###############################################################################
1575 ###############################################################################
1577 # Compute the filename to be used by CUDA_LINK_SEPARABLE_COMPILATION_OBJECTS
1578 function(CUDA_COMPUTE_SEPARABLE_COMPILATION_OBJECT_FILE_NAME output_file_var cuda_target object_files)
1580 set(generated_extension ${CMAKE_${CUDA_C_OR_CXX}_OUTPUT_EXTENSION})
1581 set(output_file "${CMAKE_CURRENT_BINARY_DIR}/CMakeFiles/${cuda_target}.dir/${CMAKE_CFG_INTDIR}/${cuda_target}_intermediate_link${generated_extension}")
1586 set(${output_file_var} "${output_file}" PARENT_SCOPE)
1589 # Setup the build rule for the separable compilation intermediate link file.
1590 function(CUDA_LINK_SEPARABLE_COMPILATION_OBJECTS output_file cuda_target options object_files)
1593 set_source_files_properties("${output_file}"
1595 EXTERNAL_OBJECT TRUE # This is an object file not to be compiled, but only
1597 GENERATED TRUE # This file is generated during the build
1600 # For now we are ignoring all the configuration specific flags.
1602 CUDA_PARSE_NVCC_OPTIONS(nvcc_flags ${options})
1603 if(CUDA_64_BIT_DEVICE_CODE)
1604 list(APPEND nvcc_flags -m64)
1606 list(APPEND nvcc_flags -m32)
1608 # If -ccbin, --compiler-bindir has been specified, don't do anything. Otherwise add it here.
1609 list( FIND nvcc_flags "-ccbin" ccbin_found0 )
1610 list( FIND nvcc_flags "--compiler-bindir" ccbin_found1 )
1611 if( ccbin_found0 LESS 0 AND ccbin_found1 LESS 0 AND CUDA_HOST_COMPILER )
1612 list(APPEND nvcc_flags -ccbin "\"${CUDA_HOST_COMPILER}\"")
1615 # Create a list of flags specified by CUDA_NVCC_FLAGS_${CONFIG} and CMAKE_${CUDA_C_OR_CXX}_FLAGS*
1616 set(config_specific_flags)
1618 foreach(config ${CUDA_configuration_types})
1619 string(TOUPPER ${config} config_upper)
1620 # Add config specific flags
1621 foreach(f ${CUDA_NVCC_FLAGS_${config_upper}})
1622 list(APPEND config_specific_flags $<$<CONFIG:${config}>:${f}>)
1624 set(important_host_flags)
1625 _cuda_get_important_host_flags(important_host_flags "${CMAKE_${CUDA_C_OR_CXX}_FLAGS_${config_upper}}")
1626 foreach(f ${important_host_flags})
1627 list(APPEND flags $<$<CONFIG:${config}>:-Xcompiler> $<$<CONFIG:${config}>:${f}>)
1630 # Add CMAKE_${CUDA_C_OR_CXX}_FLAGS
1631 set(important_host_flags)
1632 _cuda_get_important_host_flags(important_host_flags "${CMAKE_${CUDA_C_OR_CXX}_FLAGS}")
1633 foreach(f ${important_host_flags})
1634 list(APPEND flags -Xcompiler ${f})
1637 # Add our general CUDA_NVCC_FLAGS with the configuration specifig flags
1638 set(nvcc_flags ${CUDA_NVCC_FLAGS} ${config_specific_flags} ${nvcc_flags})
1640 file(RELATIVE_PATH output_file_relative_path "${CMAKE_BINARY_DIR}" "${output_file}")
1642 # Some generators don't handle the multiple levels of custom command
1643 # dependencies correctly (obj1 depends on file1, obj2 depends on obj1), so
1644 # we work around that issue by compiling the intermediate link object as a
1645 # pre-link custom command in that situation.
1646 set(do_obj_build_rule TRUE)
1647 if (MSVC_VERSION GREATER 1599 AND MSVC_VERSION LESS 1800)
1648 # VS 2010 and 2012 have this problem.
1649 set(do_obj_build_rule FALSE)
1652 if (do_obj_build_rule)
1654 OUTPUT ${output_file}
1655 DEPENDS ${object_files}
1656 COMMAND ${CUDA_NVCC_EXECUTABLE} ${nvcc_flags} -dlink ${object_files} -o ${output_file}
1658 COMMENT "Building NVCC intermediate link file ${output_file_relative_path}"
1661 get_filename_component(output_file_dir "${output_file}" DIRECTORY)
1663 TARGET ${cuda_target}
1665 COMMAND ${CMAKE_COMMAND} -E echo "Building NVCC intermediate link file ${output_file_relative_path}"
1666 COMMAND ${CMAKE_COMMAND} -E make_directory "${output_file_dir}"
1667 COMMAND ${CUDA_NVCC_EXECUTABLE} ${nvcc_flags} ${flags} -dlink ${object_files} -o "${output_file}"
1673 ###############################################################################
1674 ###############################################################################
1676 ###############################################################################
1677 ###############################################################################
1678 macro(CUDA_ADD_LIBRARY cuda_target)
1680 CUDA_ADD_CUDA_INCLUDE_ONCE()
1682 # Separate the sources from the options
1683 CUDA_GET_SOURCES_AND_OPTIONS(_sources _cmake_options _options ${ARGN})
1684 CUDA_BUILD_SHARED_LIBRARY(_cuda_shared_flag ${ARGN})
1685 # Create custom commands and targets for each file.
1686 CUDA_WRAP_SRCS( ${cuda_target} OBJ _generated_files ${_sources}
1687 ${_cmake_options} ${_cuda_shared_flag}
1688 OPTIONS ${_options} )
1690 # Compute the file name of the intermedate link file used for separable
1692 CUDA_COMPUTE_SEPARABLE_COMPILATION_OBJECT_FILE_NAME(link_file ${cuda_target} "${${cuda_target}_SEPARABLE_COMPILATION_OBJECTS}")
1695 add_library(${cuda_target} ${_cmake_options}
1701 # Add a link phase for the separable compilation if it has been enabled. If
1702 # it has been enabled then the ${cuda_target}_SEPARABLE_COMPILATION_OBJECTS
1703 # variable will have been defined.
1704 CUDA_LINK_SEPARABLE_COMPILATION_OBJECTS("${link_file}" ${cuda_target} "${_options}" "${${cuda_target}_SEPARABLE_COMPILATION_OBJECTS}")
1706 target_link_libraries(${cuda_target}
1710 # We need to set the linker language based on what the expected generated file
1711 # would be. CUDA_C_OR_CXX is computed based on CUDA_HOST_COMPILATION_CPP.
1712 set_target_properties(${cuda_target}
1714 LINKER_LANGUAGE ${CUDA_C_OR_CXX}
1720 ###############################################################################
1721 ###############################################################################
1723 ###############################################################################
1724 ###############################################################################
1725 macro(CUDA_ADD_EXECUTABLE cuda_target)
1727 CUDA_ADD_CUDA_INCLUDE_ONCE()
1729 # Separate the sources from the options
1730 CUDA_GET_SOURCES_AND_OPTIONS(_sources _cmake_options _options ${ARGN})
1731 # Create custom commands and targets for each file.
1732 CUDA_WRAP_SRCS( ${cuda_target} OBJ _generated_files ${_sources} OPTIONS ${_options} )
1734 # Compute the file name of the intermedate link file used for separable
1736 CUDA_COMPUTE_SEPARABLE_COMPILATION_OBJECT_FILE_NAME(link_file ${cuda_target} "${${cuda_target}_SEPARABLE_COMPILATION_OBJECTS}")
1739 add_executable(${cuda_target} ${_cmake_options}
1745 # Add a link phase for the separable compilation if it has been enabled. If
1746 # it has been enabled then the ${cuda_target}_SEPARABLE_COMPILATION_OBJECTS
1747 # variable will have been defined.
1748 CUDA_LINK_SEPARABLE_COMPILATION_OBJECTS("${link_file}" ${cuda_target} "${_options}" "${${cuda_target}_SEPARABLE_COMPILATION_OBJECTS}")
1750 target_link_libraries(${cuda_target}
1754 # We need to set the linker language based on what the expected generated file
1755 # would be. CUDA_C_OR_CXX is computed based on CUDA_HOST_COMPILATION_CPP.
1756 set_target_properties(${cuda_target}
1758 LINKER_LANGUAGE ${CUDA_C_OR_CXX}
1764 ###############################################################################
1765 ###############################################################################
1766 # (Internal) helper for manually added cuda source files with specific targets
1767 ###############################################################################
1768 ###############################################################################
1769 macro(cuda_compile_base cuda_target format generated_files)
1771 # Separate the sources from the options
1772 CUDA_GET_SOURCES_AND_OPTIONS(_sources _cmake_options _options ${ARGN})
1773 # Create custom commands and targets for each file.
1774 CUDA_WRAP_SRCS( ${cuda_target} ${format} _generated_files ${_sources} ${_cmake_options}
1775 OPTIONS ${_options} )
1777 set( ${generated_files} ${_generated_files})
1781 ###############################################################################
1782 ###############################################################################
1784 ###############################################################################
1785 ###############################################################################
1786 macro(CUDA_COMPILE generated_files)
1787 cuda_compile_base(cuda_compile OBJ ${generated_files} ${ARGN})
1790 ###############################################################################
1791 ###############################################################################
1793 ###############################################################################
1794 ###############################################################################
1795 macro(CUDA_COMPILE_PTX generated_files)
1796 cuda_compile_base(cuda_compile_ptx PTX ${generated_files} ${ARGN})
1799 ###############################################################################
1800 ###############################################################################
1801 # CUDA COMPILE FATBIN
1802 ###############################################################################
1803 ###############################################################################
1804 macro(CUDA_COMPILE_FATBIN generated_files)
1805 cuda_compile_base(cuda_compile_fatbin FATBIN ${generated_files} ${ARGN})
1808 ###############################################################################
1809 ###############################################################################
1810 # CUDA COMPILE CUBIN
1811 ###############################################################################
1812 ###############################################################################
1813 macro(CUDA_COMPILE_CUBIN generated_files)
1814 cuda_compile_base(cuda_compile_cubin CUBIN ${generated_files} ${ARGN})
1818 ###############################################################################
1819 ###############################################################################
1820 # CUDA ADD CUFFT TO TARGET
1821 ###############################################################################
1822 ###############################################################################
1823 macro(CUDA_ADD_CUFFT_TO_TARGET target)
1824 if (CUDA_BUILD_EMULATION)
1825 target_link_libraries(${target} ${CUDA_cufftemu_LIBRARY})
1827 target_link_libraries(${target} ${CUDA_cufft_LIBRARY})
1831 ###############################################################################
1832 ###############################################################################
1833 # CUDA ADD CUBLAS TO TARGET
1834 ###############################################################################
1835 ###############################################################################
1836 macro(CUDA_ADD_CUBLAS_TO_TARGET target)
1837 if (CUDA_BUILD_EMULATION)
1838 target_link_libraries(${target} ${CUDA_cublasemu_LIBRARY})
1840 target_link_libraries(${target} ${CUDA_cublas_LIBRARY})
1844 ###############################################################################
1845 ###############################################################################
1846 # CUDA BUILD CLEAN TARGET
1847 ###############################################################################
1848 ###############################################################################
1849 macro(CUDA_BUILD_CLEAN_TARGET)
1850 # Call this after you add all your CUDA targets, and you will get a convience
1851 # target. You should also make clean after running this target to get the
1852 # build system to generate all the code again.
1854 set(cuda_clean_target_name clean_cuda_depends)
1855 if (CMAKE_GENERATOR MATCHES "Visual Studio")
1856 string(TOUPPER ${cuda_clean_target_name} cuda_clean_target_name)
1858 add_custom_target(${cuda_clean_target_name}
1859 COMMAND ${CMAKE_COMMAND} -E remove ${CUDA_ADDITIONAL_CLEAN_FILES})
1861 # Clear out the variable, so the next time we configure it will be empty.
1862 # This is useful so that the files won't persist in the list after targets
1863 # have been removed.
1864 set(CUDA_ADDITIONAL_CLEAN_FILES "" CACHE INTERNAL "List of intermediate files that are part of the cuda dependency scanning.")