This test visits a geometry on the GPU, and writes down data such as PlacedVolume
IDs, daughters, logical volume IDs and transformations.
This data is copied to the host, and compared with the host geometry.

Stephan Hageboeck authored Jun 16, 2021 and Sandro Christian Wenzel committed Jul 05, 2021

When the flag -G is passed, nvcc generates un-optimised cuda code. In order
to obtain faster code and have line numbers, this is replaced by "-lineinfo".

Stephan Hageboeck authored Jun 18, 2021 and Sandro Christian Wenzel committed Jul 05, 2021

Add a regular expression to .clang-format that protects doxygen syntax.
The previously existing regex seems to be a relic from include what you use,
but no such comments are left in the project.

Stephan Hageboeck authored Jun 17, 2021 and Sandro Christian Wenzel committed Jul 05, 2021

In order to allow for multiple build configurations (e.g. debug and release),
git now ignores all build*/ directories.

The same condition was returning different results in CPU vs. GPU.
As a result, a CMS polycone was getting an extra 0-length ZSection
on the GPU, which was later causing discrepancies in the CPU vs. GPU
tracking.

Adding a tolerance parameter to the FP comparison fixed the divergence.

Stephan Hageboeck authored May 25, 2021 and Gabriele Cosmo committed Jun 08, 2021

Also add a check that paths configured for doxygen exist.

This adds a test that compares steps for different navigators
againt the simplest navigator, NewSimpleNavigator. If a step mismatch
is found, it is reported and the test fails.

Andrei Gheata authored May 03, 2021 and Gabriele Cosmo committed May 20, 2021

Three tests were added to ctest. Thanks to Aswath Srikanthan for contributing to TestTube, TestTet and TestParaboloid.

Benjamin Morgan authored May 17, 2021 and Gabriele Cosmo committed May 20, 2021

Reported by Guilherme Lima: If a VecGeom install is present in a
path that's added to include_directories, this will appear first
in the set of -I paths passed to nvcc. These will be found/used in
place of the project local ones, causing compile errors.

Triaged to FindCUDA not respecting the CMake variable
CMAKE_INCLUDE_DIRECTORIES_PROJECT_BEFORE when creating the overall
include paths.

Call cuda_include_directories with project source and binary directories
before cuda_add_library to ensure these are at the front on the include
search path for nvcc. Minor hack as cuda_include_directories is usually
intended for .cu/.cuh files. Temporary until migration to CMake native
CUDA support.

Benjamin Morgan authored Apr 30, 2021 and Andrei Gheata committed May 20, 2021

Need to specify use of libc++ on macOS has been obsolete since 10.9
(Mavericks) which is no longer supported.

Benjamin Morgan authored Mar 12, 2021 and Andrei Gheata committed May 20, 2021

Only place in vecgeom library that uses dlfcn.h is in implementation
of GeoManager, so link can be PRIVATE scope.

Link all tests explicitly to CMAKE_DL_LIBS as dlfcn.h also used in
globalbenchmarks/NavigationKernelBenchmarker.cpp. Linking non-users
is very low cost.

Benjamin Morgan authored Mar 12, 2021 and Andrei Gheata committed May 20, 2021

VECGEOM_CXX_FLAGS is used to hold "private" flags needed only at
VecGeom compile time. VECGEOM_COMPILATION_FLAGS is used to hold
"public" flags needed to build VecGeom *and* that a client of the
library must use when compileing/linking against VecGeom. Both
variables are appended in public-private order to CMAKE_CXX_FLAGS
given correct behaviour.

Remove flags set in both VECGEOM_CXX_FLAGS and
VECGEOM_COMPILATION_FLAGS, only setting them in the latter.

Benjamin Morgan authored Mar 12, 2021 and Andrei Gheata committed May 20, 2021

No change in behaviour, emphasizes commonality.

Benjamin Morgan authored Jan 22, 2021 and Andrei Gheata committed May 20, 2021

CMake prior to 3.10 did not fully know C++ compile features for
the Intel compiler. VecGeom currently supports 3.8 and newer.

Add a module to set known compile features for newer Intel compilers
when CMake 3.10 or earlier is used. remove custom setting of -std
flag in VECGEOM_CXX_FLAGS.

Benjamin Morgan authored Mar 12, 2021 and Andrei Gheata committed May 20, 2021

GNU/Intel use the same flag, Clang has the same concept just a different
name

             normal. Also contains a bug fix happening for the cone when rmin==rmax

             Cone generally calls all the functions of ConeStruct.
             But for Normal Calculation, there is a redundant definition of
             Normal in UnplacedCone and in ConeStruct, which is actually not
             required and cone can safely call Normal of ConeStruct.

             Now Cone calls the Normal of ConeStruct.

             NOTE : CONE STILL DOESN'T HAVE A VECTORIZED DEFINITION OF NORMAL