[Openmp-commits] [PATCH] D14031: [OpenMP] Initial implementation of OpenMP offloading library - libomptarget.

[Jonas Hahnfeld via Openmp-commits]

Hahnfeld added a comment.

Below some comments on the tests


================
Comment at: libomptarget/test/CMakeLists.txt:28-34
@@ +27,9 @@
+    "Path to llvm-lit")
+  find_program(LIT_EXECUTABLE NAMES llvm-lit ${LIBOMPTARGET_LLVM_LIT_EXECUTABLE})
+  if(NOT LIT_EXECUTABLE)
+    libomptarget_say("Cannot find llvm-lit.")
+    libomptarget_say("Please put llvm-lit in your PATH or set LIBOMPTARGET_LLVM_LIT_EXECUTABLE to its full path")
+    libomptarget_warning_say("The check-libomptarget target will not be available!")
+    return()
+  endif()
+  
----------------
Maybe we could later on think about sharing some testing setup between `libomp` and `libomptarget`...

================
Comment at: libomptarget/test/CMakeLists.txt:64-67
@@ +63,6 @@
+  
+  set(LIBOMPTARGET_OPENMP_HEADER_FOLDER "" CACHE STRING
+    "Path to folder containing omp.h")
+  set(LIBOMPTARGET_OPENMP_HOST_RTL_FOLDER "" CACHE STRING
+    "Path to folder containing libomp.h")
+else()
----------------
I think this can safely assume the relative location of `../runtime/src`

================
Comment at: libomptarget/test/offloading/offloading_success.c:9
@@ +8,3 @@
+int main(void) {
+  int isHost = 0;
+
----------------
Maybe make this `isHost = 1`. clang 3.7.1 seems to completely ignore the target code, so `omp_is_initial_device` is never called...

================
Comment at: libomptarget/test/offloading/offloading_success.cpp:9
@@ +8,3 @@
+int main(void) {
+  int isHost = 0;
+
----------------
likewise


http://reviews.llvm.org/D14031