Since commit 2c71d051fa (Makefiles Generators: use compiler for
dependencies generation, 2020-10-18, v3.20.0-rc1~392^2) we invoke `nvcc`
for CUDA < 10.2 a second time in order to generate a depfile. When
`CMAKE_CUDA_HOST_COMPILER` is set, the second invocation is missing its
`-ccbin=` option, even after refactoring in commit 8981e3e7cc
(NVIDIA-CUDA: rely on new capabilities for deps generation, 2020-12-02,
v3.20.0-rc1~362^2).
Ideally we should move the `-ccbin=` flag into `Compiler/NVIDIA-CUDA`,
but that will add `CMAKE_CUDA_HOST_COMPILER` support on Windows in
command-line generators but not the Visual Studio generators.
For now, add the flag to the depfile command specifically.
Fixes: #22037
Cray 11.0 adds support for preprocessing with output written to a
specified file (instead of always next to the source). Use it to
enable Cray Fortran with the Ninja generator.
Patch-by: James Elliott
Fixes: #20731
The PGI ( and NVIDIA HPC ) compilers default C++ standard level
are based on the GCC system headers it is compiling against.
Therefore on newer platforms the default C++ level will be >= 11
and requesting C++98 compilation mode will fail as no explicit
flag will be set.
The NAG Fortran compiler's `-mdir` flag sets the module output
directory but does not add the directory to the search path for using
modules. This is inconsistent with other compilers like the GNU Fortran
compiler's `-J` flag that does both. In order to make these consistent,
add the module output directory with a `-I` flag on the NAG Fortran
compiler so that it will be searched when using modules too.
We already do this for the XL Fortran compiler since commit 210b0b99a9
(XL: Fix using Fortran modules from their output directory, 2020-02-28,
v3.18.0-rc1~640^2~1).
Tell the Fortran compiler to write preprocessor output directly to a
file, as we do for the GNU compiler. The previous "redirect stdout"
approach could break during ABI detection with some `mpif90` wrappers
that add version information to stdout when called with `-v`.
Issue: #21828
Tell the Fortran compiler to write preprocessor output directly to a
file, as we do for the GNU compiler. The previous "redirect stdout"
approach could break during ABI detection with some `mpif90` wrappers
that add version information to stdout when called with `-v`.
Fixes: #21828
`clang-cl` supports the `-imsvc` flag to tell the compiler an include
directory is intended for system paths. `icx` does not accept this
flag, even on MSVC platforms, so do not tell CMake that it exists.
Fixes: #21801
Signed-off-by: william.r.dieter <william.r.dieter@intel.com>
In commit bb61c2d024 (Clang: use -imsvc for system include dirs when
running on Windows, 2020-09-16, v3.19.0-rc1~162^2) we added `-imsvc`
for all Clang compilers targeting the MSVC ABI. However, the option
only exists for the MSVC-like front-end. The GNU-like front-ends
use `-isystem`.
Fixes: #21789
Using a single ID 'IntelLLVM' for the suite of Intel compilers based on
the LLVM backend. The 'IntelLLVM' ID are used for C, C++, and Fortran.
Data Parallel C++ will be handled in a separate commit.
The C and C++ definitions are based on the Clang definitions. The Intel
LLVM-based C and C++ compilers are based on the Clang front end, so
existing Clang options are more likely to be a good match than options
for the older Intel compilers.
Fortran is based on the older Fortran front end with the LLVM backend.
It has a similar interface to the older versions, though many options
are shared with the C and C++ compilers.
Fixes: #21561
Signed-off-by: William R. Dieter <william.r.dieter@intel.com>
Identify the compilers as `NVHPC` to distinguish it from the older PGI
compilers from which they evolved, and from other `NVIDIA` compilers.
Fixes: #20887
Revert commit 887f3a88a6 (Intel: Add Intel DPC++ compiler
identification, 2020-09-21, v3.19.0-rc1~124^2). The compiler has
already been released, and is more usable with CMake by pretending to be
upstream Clang than by identifying it as a compiler for which we have
not implemented support.
Fixes: #21551
Revert commit 5c3a93ab88 (Intel: Add Intel Clang compiler
identification, 2020-09-29, v3.19.0-rc1~68^2). The compiler has already
been released, and is more usable with CMake by pretending to be
upstream Clang than by identifying it as a compiler for which we have
not implemented support.
Issue: #21551
48aac247e9 Compile with explicit language flag when source LANGUAGE property is set
2e67a75acd Embarcadero: Simplify addition of -P flag for C++
Acked-by: Kitware Robot <kwrobot@kitware.com>
Merge-request: !5522
This change was originally made by commit 74b1c9fc8e (Explicitly specify
language flag when source LANGUAGE property is set, 2020-06-01,
v3.19.0-rc1~722^2), but it was reverted by commit 30aa715fac (Revert
"specify language flag when source LANGUAGE property is set",
2020-11-19) to restore compatibility with pre-3.19 behavior.
Implement the change again, but add policy CMP0119 to make this change
while preserving compatibility with existing projects.
Note that the `Compiler/{Clang,Intel,MSVC}-CXX` modules do not need to
specify `-TP` for their MSVC-like variants because we already use the
flag in `CMAKE_CXX_COMPILE_OBJECT`. Similarly for `Compiler/XL-CXX`
and `Platform/Windows-Embarcadero`.
Note also that this does not seem possible to implement for XL C.
Even with `-qsourcetype=c`, `xlc` complains about an unknown suffix:
`1501-218 (W) file /.../AltExtC.zzz contains an incorrect file suffix`.
It returns non-zero even with `-qsuppress=1501-218`.
Co-Author: Robert Maynard <robert.maynard@kitware.com>
Fixes: #14516, #20716
2c71d051fa Makefiles Generators: use compiler for dependencies generation
afd0f6785d Refactoring: Abstract Makefile line continuation format
b6068ce407 Refactoring: enhance include file filtering
3401403f69 Refactoring: Introduce place-holder for dependency target.
a97c41bf8b Refactoring: Makefiles Generators: Add support for various depends scanners
Acked-by: Kitware Robot <kwrobot@kitware.com>
Acked-by: Brad King <brad.king@kitware.com>
Merge-request: !5528
Each source compilation generates a dependencies file. These dependencies
files are consolidated in one file per target. This consolidation is done
as part of command 'cmake -E cmake_depends` launched before evaluation of
makefile dependency graph.
The consolidation uses the same approach as `CMake` dependencies management.
Fixes: #21321
These changes are in preparation of compiler generated dependencies support
for Makefiles generators
* compiler output and dependency target can be different for Makefiles generators
* resolve inconsistency naming for dependency file place-holder
Rename the `CheckPIESupported` helper functions so that they
don't clobber other internal functions. Also rename them to
document they can't be unified with `CheckCompilerFlag`.
Fixes: #21497
Revert commit 74b1c9fc8e (Explicitly specify language flag when source
LANGUAGE property is set, 2020-06-01, v3.19.0-rc1~722^2) and the lookup
tables from its two immediate ancestors. The purpose of that change was
to convert an explicit `LANGUAGE` source file property into an explicit
language specification compiler flag like `-x c`. This seems reasonable
since the property is documented as meaning "indicate what programming
language the source file is". It is also needed to help compilers deal
with non-standard source file extensions they don't recognize.
However, some projects have been setting `LANGUAGE C` on `.S` assembler
source files to mean "use the C compiler". Passing `-x c` for them
breaks the build because the `.S` sources are not written in C. These
projects should be updated to use `enable_language(ASM)`, for which
CMake often chooses the C compiler as the assembler when using
toolchains that support it (which would have to be the case for projects
using the approach).
Revert the change for now to preserve the old behavior for such projects.
We can re-introduce it with a policy in a future version of CMake.
Fixes: #21469
Issue: #14516, #20716
This fixes the following two issues with the CUDA support on QNX:
* cuda target name is not derived correctly (should be `aarch64-qnx`).
* linking `cudart` must not be linked against `rt`, `dl`, `pthread`.
This enables to use cmake's native cuda support on QNX.
Fixes: #21381
For NVCC the compiler takes care of device linking when passed the "-dlink"
flag.
Clang doesn't support such magic and requires the buildsystem to do the work
that NVCC does behind the scenes.
The implementation is based on Bazel's device linking documentation:
7cabcdf073/third_party/nccl/build_defs.bzl.tpl (L259)Closes: #20726
