Fixed another bug in code generation.

Author: isazi

kernel_tuner/utils/directives.py

@@ -535,6 +535,11 @@ def allocate_signature_memory(data: dict, preprocessor: list = None, user_dimens
 
     return args
 
+def add_new_line(line: str) -> str:
+    """Adds the new line character to the end of the line if not present"""
+    if line.rfind("\n") != len(line) - 1:
+        return line + "\n"
+    return line
 
 def add_present_openacc(
     code: str, langs: Code, data: dict, preprocessor: list = None, user_dimensions: dict = None
@@ -560,8 +565,7 @@ def add_present_openacc(
                         elif is_fortran(langs.language):
                             present_clause += add_present_openacc_fortran(name, size)
                 new_body += new_line + present_clause.rstrip() + "\n"
-    if new_body.rfind("\n") != len(new_body) - 1:
-        new_body += "\n"
+        new_body = add_new_line(new_body)
     return new_body
 
 

test/utils/test_directives.py

@@ -318,6 +318,6 @@ def test_add_present_openacc():
     assert add_present_openacc(code_f90, acc_f90, data, preprocessor, None) == expected_f90
     code_f90 = "!$acc parallel async num_workers(16) copy(array(:42))\n"
     assert add_present_openacc(code_f90, acc_f90, data, preprocessor, None) == code_f90
-    code_cxx = "#pragma acc parallel num_gangs(32)\n\t#pragma acc loop\n\t//for loop"
-    expected_cxx = "#pragma acc parallel num_gangs(32) present(array[:42])\n\t#pragma acc loop\n\t//for loop"
+    code_cxx = "#pragma acc parallel num_gangs(32)\n\t#pragma acc loop\n\t//for loop\n"
+    expected_cxx = "#pragma acc parallel num_gangs(32) present(array[:42])\n\t#pragma acc loop\n\t//for loop\n"
     assert add_present_openacc(code_cxx, acc_cxx, data, preprocessor, None) == expected_cxx