[libomptarget][nvptx] Include omp_data.cu in bitcode deviceRTL

[JonChesterfield]

Summary

[libomptarget][nvptx] Include omp_data.cu in bitcode deviceRTL

[tianshilei1992]

This makes steps for deprecating the static library in the near future.

[JonChesterfield]

Bug https://bugs.llvm.org/show_bug.cgi?id=49764 points at this change as increasing device memory use. Previously the bitcode and the archive were both linked, where this file was always in the archive. After this commit, no files will have been extracted from the archive (and shortly afterwards the archive was dropped entirely).

If no symbols from omp_data.cu were used in some test case, and the bitcode version is not being deadstripped, then data use would increase. Both of those requirements seem unlikely.

It would be good to check that the devicertl symbols are being internalized for nvptx, as otherwise unused variables with global visibility will tend to remain.

openmp/libomptarget/deviceRTLs/nvptx/CMakeLists.txt

if(LIBOMPTARGET_DEP_CUDA_FOUND)
   set(CUDA_PROPAGATE_HOST_FLAGS OFF)
 
   set(cuda_src_files
       ${devicertl_common_directory}/src/cancel.cu
       ${devicertl_common_directory}/src/critical.cu
       ${devicertl_common_directory}/src/data_sharing.cu
       ${devicertl_common_directory}/src/libcall.cu
       ${devicertl_common_directory}/src/loop.cu
+      ${devicertl_common_directory}/src/omp_data.cu
       ${devicertl_common_directory}/src/omptarget.cu
       ${devicertl_common_directory}/src/parallel.cu
       ${devicertl_common_directory}/src/reduction.cu
       ${devicertl_common_directory}/src/support.cu
       ${devicertl_common_directory}/src/sync.cu
       ${devicertl_common_directory}/src/task.cu
       src/target_impl.cu
   )
 
-  set(omp_data_objects ${devicertl_common_directory}/src/omp_data.cu)
-
   # Build library support for the highest compute capability the system supports
   # and always build support for sm_35 by default
   if (${LIBOMPTARGET_DEP_CUDA_ARCH} EQUAL 35)
     set(default_capabilities 35)
   else()
       set(default_capabilities "35,${LIBOMPTARGET_DEP_CUDA_ARCH}")
   endif()
 
if(LIBOMPTARGET_DEP_CUDA_FOUND)
   endif()
 
   # NVPTX runtime library has to be statically linked. Dynamic linking is not
   # yet supported by the CUDA toolchain on the device.
   set(BUILD_SHARED_LIBS OFF)
   set(CUDA_SEPARABLE_COMPILATION ON)
   list(APPEND CUDA_NVCC_FLAGS -I${devicertl_base_directory}
                               -I${devicertl_nvptx_directory}/src)
-  cuda_add_library(omptarget-nvptx STATIC ${cuda_src_files} ${omp_data_objects}
+  cuda_add_library(omptarget-nvptx STATIC ${cuda_src_files}
       OPTIONS ${CUDA_ARCH} ${CUDA_DEBUG} ${MAX_SM_DEFINITION})
 
   # Install device RTL under the lib destination folder.
   install(TARGETS omptarget-nvptx ARCHIVE DESTINATION "${OPENMP_INSTALL_LIBDIR}")
 
   target_link_libraries(omptarget-nvptx ${CUDA_LIBRARIES})