[Libomptarget] Only export needed definitions in the BC library

[jhuber6]

Summary

This patch adds the use of the -internalize-public-api-file option in
the internalization pass to internalize any definition that isn't
explicitly needed for the interface. This will allow us to perform more
optimizations on the file that normally would not have been possible
with functions internal to the library not being internal.

Depends on D130293

[jdoerfert]

can't we do __kmpc* and omp_* and llvm_* instead?

[jhuber6]

In D130298#3669752, @jdoerfert wrote:

can't we do __kmpc* and omp_* and llvm_* instead?

The internalize patch just takes full names. It would definitely be better to use wildcards. I could set up a patch to make it accept those as well, but it would take some extra time to get reviewed.

[jdoerfert]

In D130298#3669776, @jhuber6 wrote:

In D130298#3669752, @jdoerfert wrote:

can't we do __kmpc* and omp_* and llvm_* instead?

The internalize patch just takes full names. It would definitely be better to use wildcards. I could set up a patch to make it accept those as well, but it would take some extra time to get reviewed.

Add a TODO somewhere. LG Otherwise

[JonChesterfield]

Tempting to generate the exports file with nm | grep but probably not worth it. Thanks!

[jhuber6]

Changing to use globs after making a patch to update internalize to accept them.

openmp/libomptarget/DeviceRTL/CMakeLists.txt

set(src_files
   ${source_directory}/Synchronization.cpp
   ${source_directory}/Tasking.cpp
   ${source_directory}/Utils.cpp
   ${source_directory}/Workshare.cpp
 )
 
 set(clang_opt_flags -O3 -mllvm -openmp-opt-disable -DSHARED_SCRATCHPAD_SIZE=512)
 set(link_opt_flags  -O3        -openmp-opt-disable)
+set(link_export_flag -passes=internalize -internalize-public-api-file=${source_directory}/exports)
 
 # Prepend -I to each list element
 set (LIBOMPTARGET_LLVM_INCLUDE_DIRS_DEVICERTL "${LIBOMPTARGET_LLVM_INCLUDE_DIRS}")
 list(TRANSFORM LIBOMPTARGET_LLVM_INCLUDE_DIRS_DEVICERTL PREPEND "-I")
 
 # Set flags for LLVM Bitcode compilation.
 set(bc_flags -c -emit-llvm -std=c++17 -fvisibility=hidden
               ${clang_opt_flags} --offload-device-only
function(compileDeviceRTLLibrary target_cpu target_name)
   # Link to a bitcode library.
   add_custom_command(OUTPUT ${CMAKE_CURRENT_BINARY_DIR}/linked_${bclib_name}
       COMMAND ${LINK_TOOL}
         -o ${CMAKE_CURRENT_BINARY_DIR}/linked_${bclib_name} ${bc_files}
       DEPENDS ${bc_files}
       COMMENT "Linking LLVM bitcode ${bclib_name}"
   )
 
+  add_custom_command(OUTPUT ${CMAKE_CURRENT_BINARY_DIR}/internalized_${bclib_name}
+      COMMAND ${OPT_TOOL} ${link_export_flag} ${CMAKE_CURRENT_BINARY_DIR}/linked_${bclib_name}
+                      -o ${CMAKE_CURRENT_BINARY_DIR}/internalized_${bclib_name}
+      DEPENDS ${source_directory}/exports ${CMAKE_CURRENT_BINARY_DIR}/linked_${bclib_name}
+      COMMENT "Internalizing LLVM bitcode ${bclib_name}"
+  )
+
   add_custom_command(OUTPUT ${CMAKE_CURRENT_BINARY_DIR}/${bclib_name}
-      COMMAND ${OPT_TOOL} ${link_opt_flags} ${CMAKE_CURRENT_BINARY_DIR}/linked_${bclib_name}
+      COMMAND ${OPT_TOOL} ${link_opt_flags} ${CMAKE_CURRENT_BINARY_DIR}/internalized_${bclib_name}
                       -o ${CMAKE_CURRENT_BINARY_DIR}/${bclib_name}
-      DEPENDS ${CMAKE_CURRENT_BINARY_DIR}/linked_${bclib_name}
+      DEPENDS ${CMAKE_CURRENT_BINARY_DIR}/internalized_${bclib_name}
       COMMENT "Optimizing LLVM bitcode ${bclib_name}"
   )
 
   # Add a file-level dependency to ensure that llvm-link and opt are up-to-date.
   # By default, add_custom_command only builds the tool if the executable is missing
   if("${LINK_TOOL}" STREQUAL "$<TARGET_FILE:llvm-link>")
     add_custom_command(OUTPUT ${CMAKE_CURRENT_BINARY_DIR}/linked_${bclib_name}
       DEPENDS llvm-link

openmp/libomptarget/DeviceRTL/src/exports

+omp_*
+*llvm_*
+__kmpc_*