1009e1d7 Merge branch 'backport-clang-format-cuda' into clang-format-cuda
33a65941 Tests: Run clang-format on CUDA code
fa5188d0 Merge branch 'backport-clang-format-cuda' into clang-format-cuda
c4a61350 Tests: Run clang-format on CUDA code
0b2f8ae3 Modules: Run clang-format on CUDA code
592e3d2f clang-format.bash: Format CUDA source files too
Add `.clang-format` configuration files for Cuda test directories that
use `Standard: Cpp11`. Otherwise clang-format splits the triple angle
brackets used for CUDA kernels.
e0f39312 cmCommands: add commands directly to cmState
72dcbbe5 split Default commands into Scripting and Project
7fb14775 cmDisallowedCommand: extract policy checking from cmCommand
615e2a17 cmCommand: Don't prefix error message with command name
45b49099 cmCommand: make noncopyable
5396bc92 cmCommand: remove member Helper
Since commit v3.8.0-rc1~232^2 (execute_process: Add ENCODING option for
Windows child process output, 2016-11-23) we decode child process output
using the console's active codepage by default. This differs from
previous versions of CMake and is therefore incompatible. Changing this
default will require a policy, so for now revert the default behavior
back to performing no decoding.
Reported-by: Nils Gladitz <nilsgladitz@gmail.com>
166b3393 Tests/Cuda: Select a CUDA device supporting compute 3.0
cbe4d595 Tests/Cuda: Return a non-zero code if errors occurred
8731701c Tests/Cuda: Use memory allocated on the GPU in the kernels
0ae5386a Tests/Cuda: Add missing separable compilation property
ce19607f Tests/Cuda: Fix missing CUDA static library at runtime on macOS
008ed80d Tests/Cuda: Output error messages to std::cerr instead of std::cout
c0d7bb83 Tests/Cuda: Print asynchronous error messages, if any
21a125cd Tests/Cuda: Print error message if mixed_kernel failed
eebb2be8 Tests/Cuda: Add identifiers to error messages
84f3c87b Tests/Cuda: Print error message if an error occurred
As kernel launches are asynchronous, a `cudaGetLastError()` right after
the kernel launch might be executed while the kernel is still running.
Synchronizing the device will ensure that all the work is completed
before progressing further on, and allows to catch errors that were
previously missed.
The `cudaGetLastError()` after the `cudaDeviceSynchronize()` is there
to reset the error variable to `cudaSuccess`.
Since upstream curl commit curl-7_52_0~131 (URL-parser: for
file://[host]/ URLs, the [host] must be localhost, 2016-11-11) we can no
longer use URLs of the form `file://c:/...` on Windows. These worked
only accidentally before. Use `file:///c:/...` instead.
30abf145 Help: Fix cmake code block warnings produced by Sphinx 1.4
97917900 Utilities/Sphinx: Port qthelp css workaround to Sphinx 1.4
971384c2 Utilities/Sphinx: Port cmake extension to Sphinx 1.4
Brad King