merge release code into development

CMakeLists.txt

@@ -3,8 +3,22 @@
 # See https://llvm.org/LICENSE.txt for license information.
 # SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 #
+# 
+# Modifications Copyright (c) 2019 Advanced Micro Devices, Inc. All rights reserved.
+# Notified per clause 4(b) of the license.
+#
+# Last Modified: May 2020
+#
 
 cmake_minimum_required(VERSION 2.8)
+cmake_policy(SET CMP0057 NEW)
+
+if(${ENABLE_DEVEL_PACKAGE})
+  set(DEVEL_PACKAGE "devel/")
+endif()
+if(${ENABLE_RUN_PACKAGE})
+  set(RUN_PACKAGE "runtime/")
+endif()
 
 # In order to bootstrap the runtime library we need to skip
 # CMake's Fortran tests
@@ -53,7 +67,9 @@ set(CMAKE_HOST_SYSTEM_PROCESSOR ${TARGET_ARCHITECTURE})
 # standalone project, using LLVM as an external library:
 if( CMAKE_SOURCE_DIR STREQUAL CMAKE_CURRENT_SOURCE_DIR )
   project(Flang)
-
+  # Set default libdir to be "lib" for ROCm, distros will override this anyway:
+  set(CMAKE_INSTALL_LIBDIR "lib" CACHE STRING "Library install directory")
+  include(GNUInstallDirs)
   # Rely on llvm-config.
   set(CONFIG_OUTPUT)
 
@@ -68,8 +84,7 @@ if( CMAKE_SOURCE_DIR STREQUAL CMAKE_CURRENT_SOURCE_DIR )
       "--bindir"
       "--libdir"
       "--includedir"
-      "--prefix"
-      "--src-root")
+      "--prefix")
     execute_process(
       COMMAND ${CONFIG_COMMAND}
       RESULT_VARIABLE HAD_ERROR
@@ -133,6 +148,9 @@ if( CMAKE_SOURCE_DIR STREQUAL CMAKE_CURRENT_SOURCE_DIR )
   option(LLVM_INSTALL_TOOLCHAIN_ONLY
     "Only include toolchain files in the 'install' target." OFF)
 
+  option(LLVM_INSTALL_RUNTIME
+     "Build and install the flang runtime. Do this last" OFF)
+
   option(LLVM_FORCE_USE_OLD_HOST_TOOLCHAIN
     "Set to ON to force using an old, unsupported host toolchain." OFF)
 
@@ -172,6 +190,8 @@ Please install Python or specify the PYTHON_EXECUTABLE CMake variable.")
       set(LLVM_UTILS_PROVIDED ON)
     endif()
 
+   if (NOT LLVM_INSTALL_RUNTIME)
+
     if(EXISTS ${LLVM_MAIN_SRC_DIR}/utils/lit/lit.py)
       set(LLVM_LIT ${LLVM_MAIN_SRC_DIR}/utils/lit/lit.py)
       if(NOT LLVM_UTILS_PROVIDED)
@@ -181,6 +201,7 @@ Please install Python or specify the PYTHON_EXECUTABLE CMake variable.")
         set(LLVM_UTILS_PROVIDED ON)
         set(FLANG_TEST_DEPS FileCheck count not)
       endif()
+     endif()
       #set(UNITTEST_DIR ${LLVM_MAIN_SRC_DIR}/utils/unittest)
       #if(EXISTS ${UNITTEST_DIR}/googletest/include/gtest/gtest.h
       #    AND NOT EXISTS ${LLVM_LIBRARY_DIR}/${CMAKE_STATIC_LIBRARY_PREFIX}gtest${CMAKE_STATIC_LIBRARY_SUFFIX}
@@ -276,19 +297,23 @@ else()
   set(FLANG_HAS_VERSION_PATCHLEVEL 0)
 endif()
 
+set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -fcommon -Wno-implicit-function-declaration -Wno-implicit-int -Wno-int-conversion -Wno-enum-constexpr-conversion -Wno-incompatible-function-pointer-types -w")
+
 # Add appropriate flags for GCC
 if (LLVM_COMPILER_IS_GCC_COMPATIBLE)
-  set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++11 -fno-common -Woverloaded-virtual -Wcast-qual -fno-strict-aliasing -pedantic -Wno-long-long -Wall -W -Wno-unused-parameter -Wwrite-strings")
+	set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++11 -fno-common -Woverloaded-virtual -Wcast-qual -fno-strict-aliasing -pedantic -Wno-long-long -Wall -W -Wno-unused-parameter -Wwrite-strings -g -DOMP_OFFLOAD_LLVM -Wno-implicit-function-declaration -Wno-implicit-int -Wno-enum-constexpr-conversion -Wno-incompatible-function-pointer-types -w")
   option(WITH_WERROR "Compile with '-Werror' C compiler flag" ON)
   if (WITH_WERROR)
-     set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -Werror")
+	  # set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -Werror")
   endif ()
 endif ()
 
 if (APPLE)
   set(CMAKE_MODULE_LINKER_FLAGS "-Wl,-flat_namespace -Wl,-undefined -Wl,suppress")
 endif ()
 
+find_library(LIBQUADMATH_LOC quadmath NAMES libquadmath.so libquadmath.so.0 REQUIRED)
+
 macro(add_flang_library name)
   llvm_process_sources(srcs ${ARGN})
   if (MODULE)
@@ -305,13 +330,13 @@ macro(add_flang_library name)
   endif( LLVM_COMMON_DEPENDS )
 
   llvm_config( ${name} ${LLVM_LINK_COMPONENTS} )
-  target_link_libraries( ${name} ${LLVM_COMMON_LIBS} )
+  target_link_libraries( ${name} ${LLVM_COMMON_LIBS} ${LIBQUADMATH_LOC})
 #  link_system_libs( ${name} )  # getd of cmake warning messages
 
   install(TARGETS ${name}
-    LIBRARY DESTINATION lib${LLVM_LIBDIR_SUFFIX}
-    ARCHIVE DESTINATION lib${LLVM_LIBDIR_SUFFIX}
-    RUNTIME DESTINATION bin)
+    LIBRARY DESTINATION ${RUN_PACKAGE}${CMAKE_INSTALL_LIBDIR}
+    ARCHIVE DESTINATION ${RUN_PACKAGE}${CMAKE_INSTALL_LIBDIR}
+    RUNTIME DESTINATION ${DEVEL_PACKAGE}${CMAKE_INSTALL_BINDIR})
   set_target_properties(${name} PROPERTIES FOLDER "Flang libraries")
 endmacro(add_flang_library)
 
@@ -328,16 +353,18 @@ include_directories(BEFORE
 # Direct module files to build include directory
 set(CMAKE_Fortran_MODULE_DIRECTORY ${CMAKE_CURRENT_BINARY_DIR}/include)
 
+if (FLANG_BUILD_RUNTIME)
 # Install Fortran module files
 install(DIRECTORY ${CMAKE_Fortran_MODULE_DIRECTORY}/
-  DESTINATION include
+  DESTINATION ${DEVEL_PACKAGE}${CMAKE_INSTALL_INCLUDEDIR}
   )
+endif()
 
 # Install Fortran OpenMP include file
 # Copy omp_lib.h file, not the symlink
 get_filename_component(OMP_LIB_H_PATH ${CMAKE_CURRENT_SOURCE_DIR}/include/omp_lib.h REALPATH)
 install(FILES ${OMP_LIB_H_PATH}
-  DESTINATION include
+  DESTINATION ${DEVEL_PACKAGE}${CMAKE_INSTALL_INCLUDEDIR}
   )
 
 add_definitions( -D_GNU_SOURCE )
@@ -353,6 +380,22 @@ mark_as_advanced(FLANG_EXECUTABLE_VERSION LIBFLANG_LIBRARY_VERSION)
 
 option(FLANG_LLVM_EXTENSIONS "enable the Flang LLVM extensions" OFF)
 
+if ( FLANG_EXECUTABLE_VERSION )
+  add_definitions( -DFLANG_VERSION="${FLANG_EXECUTABLE_VERSION} " )
+  add_definitions( -DFLANG_VERSION_MAJOR="${FLANG_VERSION_MAJOR} " )
+  add_definitions( -DFLANG_VERSION_MINOR="${FLANG_VERSION_MINOR} " )
+endif()
+execute_process(COMMAND git log -1 --format=format:%H WORKING_DIRECTORY ${CMAKE_CURRENT_SOURCE_DIR} OUTPUT_VARIABLE FLANG_SHA RESULT_VARIABLE rc)
+
+if(NOT rc EQUAL "0")
+  add_definitions( -DFLANG_SHA="Flang Sha NULL ")
+  message("Warning: Unable to get Flang Sha. Return Code: ${rc} Flang Sha: NULL!!!")
+else()
+  string(STRIP ${FLANG_SHA} FLANG_SHA)
+  message(STATUS "MY_VAR=${FLANG_SHA}")
+  add_definitions( -DFLANG_SHA="Flang Sha ${FLANG_SHA} ")
+endif()
+
 option(FLANG_INCLUDE_TESTS
        "Generate build targets for the Flang unit tests."
        ${LLVM_INCLUDE_TESTS})
@@ -367,11 +410,14 @@ set(FLANG_LIB_DIR ${CMAKE_CURRENT_BINARY_DIR}/lib)
 set(FLANG_RTE_LIB_DIR ${CMAKE_CURRENT_BINARY_DIR}/lib)
 add_definitions( -DPGFLANG )
 
-add_subdirectory(lib)
-add_subdirectory(runtime)
-add_subdirectory(utils)
-add_subdirectory(tools)
-#add_subdirectory(test)
+if (LLVM_INSTALL_RUNTIME)
+   add_subdirectory(runtime)
+else()
+   add_subdirectory(lib)
+   add_subdirectory(utils)
+   add_subdirectory(tools)
+   #add_subdirectory(test)
+endif()
 
 #option(FLANG_BUILD_EXAMPLES "Build FLANG example programs by default." OFF)
 #if (FLANG_BUILD_EXAMPLES)
@@ -386,6 +432,17 @@ if (FLANG_OPENMP_GPU_NVIDIA)
   add_definitions("-DOMP_OFFLOAD_LLVM")
 endif()
 
+# AOCC Begin
+option(FLANG_OPENMP_GPU_AMD "Enable OpenMP AMD Accelerator Offload." OFF)
+if (FLANG_OPENMP_GPU_AMD)
+  add_definitions("-DOMP_OFFLOAD_AMD")
+endif()
+
+if (FLANG_OPENMP_GPU_NVIDIA OR FLANG_OPENMP_GPU_AMD)
+  set(OMP_OFFLOADING_BUILD 1)
+endif()
+# AOCC End
+
 if( FLANG_INCLUDE_TESTS )
 #  if(EXISTS ${LLVM_MAIN_SRC_DIR}/utils/unittest/googletest/include/gtest/gtest.h)
 #    add_subdirectory(unittests)

docs/CMakeLists.txt

@@ -93,7 +93,7 @@ if (LLVM_ENABLE_DOXYGEN)
 
   if (NOT LLVM_INSTALL_TOOLCHAIN_ONLY)
     install(DIRECTORY ${CMAKE_CURRENT_BINARY_DIR}/web/html
-      DESTINATION docs/html)
+      DESTINATION ${DEVEL_PACKAGE}docs/html)
   endif()
 endif()
 endif()

include/flang/Error/errmsg-in.n

@@ -4,6 +4,12 @@
 .\" * SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 .\" *
 .\" */
+.\"/* 
+.\" * Modifications Copyright (c) 2019 Advanced Micro Devices, Inc. All rights reserved.
+.\" * Notified per clause 4(b) of the license.
+.\" *
+.\" * Last Modified: May 2020
+.\" */
 .NS 23 "Error Messages" "Appendix II - "
 .af EN 001    \" define format for EN
 .de MS
@@ -711,7 +717,7 @@ Messages 280-300 RESERVED for directive handling
 .MS W 281 "Directive ignored - $ $"
 .MS F 282 "#error $"
 This message appears when preprocessing a file and a #error line is reached.
-.MS F 283 "#warning $"
+.MS S 283 "#warning $"
 This message appears when preprocessing a file and a #warning line is reached.
 .MS W 284 "A statement with an HPF keyword must begin with $ - $"
 HPF keyword statements (like DISTRIBUTE, ALIGN) must appear on lines
@@ -1333,6 +1339,8 @@ LAUNCH_BOUNDS is now allowed on ATTRIBUTES(DEVICE) or host subprograms.
 .MS E 552 "LAUNCH_BOUNDS() values must be positive"
 The LAUNCH_BOUNDS maximum number of threads and minimum grid size must
 be positive integer values.
+.MS W 603 "Unsupported clause specified for the vector directive. Only the always/never clauses are supported."
+.MS W 604 "Unsupported clause specified for the omp simd directive. The directive will be ignored."
 .MS S 901 "#elif after #else"
 A preprocessor #elif directive was found after a #else directive; only
 #endif is allowed in this context.
@@ -1500,6 +1508,7 @@ A DO CONCURRENT or FORALL construct or statement may not specify an index name m
 .MS S 1061 "The definition of function return type of $ does not match its declaration type"
 .MS S 1062 "LOCAL_INIT variable does not have an outside variable of the same name - $"
 A DO CONCURRENT variable with LOCAL_INIT locality must have a host variable of the same name.
+.MS S 1063 "Real number has a ‘q’ exponent and an explicit kind"
 Starting from 1100, Reserved for OpenMP GPU
 .MS S 1198 "OpenMP GPU - The feature is not implemented yet for the target device"
 .MS W 1199 "OpenMP GPU - \"$\" is ignored for the target device"
@@ -1523,3 +1532,4 @@ Starting from 1100, Reserved for OpenMP GPU
 .MS S 1217 "Left hand side of polymorphic assignment must be allocatable - $"
 .MS S 1218 "$ statement may not appear in a BLOCK construct."
 .MS S 1219 "Unimplemented feature: $."
+.MS W 1220 "$ is uninitialized at $"

include/fp-folding.h

@@ -23,6 +23,18 @@
  *  Note: The types have names like "float32_t" to distinguish them
  *  from integers, while the routines use names like "real32" to
  *  distinguish them from complex operations that might be added.
+ *
+ */
+/* 
+ * Modifications Copyright (c) 2019 Advanced Micro Devices, Inc. All rights reserved.
+ * Notified per clause 4(b) of the license.
+ *
+ *
+ * Added support for quad precision
+ * Last modified: Feb 2020
+ *
+ * Support for nearest intrinsic
+ * Last modified: 01 March 2020
  */
 
 #ifndef FP_FOLDING_H_
@@ -72,7 +84,6 @@ enum fold_relation {
 
 enum fold_relation fold_real32_compare(const float32_t *x, const float32_t *y);
 enum fold_relation fold_real64_compare(const float64_t *x, const float64_t *y);
-enum fold_relation fold_real128_compare(const float128_t *x, const float128_t *y);
 
 /*
  *  Operations.  These all return a status code.
@@ -86,6 +97,8 @@ enum fold_status {
   FOLD_INEXACT = -4,
 };
 
+#ifdef FOLD_LDBL_128BIT
+enum fold_relation fold_real128_compare(const float128_t *x, const float128_t *y);
 enum fold_status fold_int32_from_real32(int32_t *res, const float32_t *arg);
 enum fold_status fold_int64_from_real32(int64_t *res, const float32_t *arg);
 enum fold_status fold_uint64_from_real32(uint64_t *res, const float32_t *arg);
@@ -104,6 +117,7 @@ enum fold_status fold_real32_divide(float32_t *res, const float32_t *x, const fl
 enum fold_status fold_real32_pow(float32_t *res, const float32_t *x, const float32_t *y);
 enum fold_status fold_real32_sin(float32_t *res, const float32_t *arg);
 enum fold_status fold_real32_cos(float32_t *res, const float32_t *arg);
+enum fold_status fold_real32_cotan(float32_t *res, const float32_t *arg);
 enum fold_status fold_real32_tan(float32_t *res, const float32_t *arg);
 enum fold_status fold_real32_asin(float32_t *res, const float32_t *arg);
 enum fold_status fold_real32_acos(float32_t *res, const float32_t *arg);
@@ -130,11 +144,17 @@ enum fold_status fold_real64_divide(float64_t *res, const float64_t *x, const fl
 enum fold_status fold_real64_pow(float64_t *res, const float64_t *x, const float64_t *y);
 enum fold_status fold_real64_sin(float64_t *res, const float64_t *arg);
 enum fold_status fold_real64_cos(float64_t *res, const float64_t *arg);
+enum fold_status fold_real64_cotan(float64_t *res, const float64_t *arg);
 enum fold_status fold_real64_tan(float64_t *res, const float64_t *arg);
 enum fold_status fold_real64_asin(float64_t *res, const float64_t *arg);
 enum fold_status fold_real64_acos(float64_t *res, const float64_t *arg);
 enum fold_status fold_real64_atan(float64_t *res, const float64_t *arg);
 enum fold_status fold_real64_atan2(float64_t *res, const float64_t *x, const float64_t *y);
+//AOCC Begin
+enum fold_status fold_real32_nearest(float32_t *res, const float32_t *x, const float32_t *y);
+enum fold_status fold_real64_nearest(float64_t *res, const float64_t *x, const float64_t *y);
+enum fold_status fold_real128_nearest(float128_t *res, const float128_t *x, const float128_t *y);
+//AOCC End
 enum fold_status fold_real64_exp(float64_t *res, const float64_t *arg);
 enum fold_status fold_real64_log(float64_t *res, const float64_t *arg);
 enum fold_status fold_real64_log10(float64_t *res, const float64_t *arg);
@@ -156,6 +176,7 @@ enum fold_status fold_real128_divide(float128_t *res, const float128_t *x, const
 enum fold_status fold_real128_pow(float128_t *res, const float128_t *x, const float128_t *y);
 enum fold_status fold_real128_sin(float128_t *res, const float128_t *arg);
 enum fold_status fold_real128_cos(float128_t *res, const float128_t *arg);
+enum fold_status fold_real128_cotan(float128_t *res, const float128_t *arg);
 enum fold_status fold_real128_tan(float128_t *res, const float128_t *arg);
 enum fold_status fold_real128_asin(float128_t *res, const float128_t *arg);
 enum fold_status fold_real128_acos(float128_t *res, const float128_t *arg);
@@ -164,6 +185,40 @@ enum fold_status fold_real128_atan2(float128_t *res, const float128_t *x, const
 enum fold_status fold_real128_exp(float128_t *res, const float128_t *arg);
 enum fold_status fold_real128_log(float128_t *res, const float128_t *arg);
 enum fold_status fold_real128_log10(float128_t *res, const float128_t *arg);
+// AOCC begin
+// To support quad precision REAL128 type
+#else
+enum fold_relation fold_real128_compare(const __float128 *x, const __float128 *y);
+enum fold_status fold_int32_from_real128(int32_t *res, const __float128 *arg);
+enum fold_status fold_int64_from_real128(int64_t *res, const __float128 *arg);
+enum fold_status fold_uint32_from_real128(uint32_t *res, const __float128 *arg);
+enum fold_status fold_uint64_from_real128(uint64_t *res, const __float128 *arg);
+enum fold_status fold_real128_from_int64(__float128 *res, const int64_t *arg);
+enum fold_status fold_real128_from_uint64(__float128 *res, const uint64_t *arg);
+enum fold_status fold_real128_from_real32(__float128 *res, const float32_t *arg);
+enum fold_status fold_real128_from_real64(__float128 *res, const float64_t *arg);
+enum fold_status fold_real128_negate(__float128 *res, const __float128 *arg);
+enum fold_status fold_real128_abs(__float128 *res, const __float128 *arg);
+enum fold_status fold_real128_sqrt(__float128 *res, const __float128 *arg);
+enum fold_status fold_real128_add(__float128 *res, const __float128 *x, const __float128 *y);
+enum fold_status fold_real128_subtract(__float128 *res, const __float128 *x, const __float128 *y);
+enum fold_status fold_real128_multiply(__float128 *res, const __float128 *x, const __float128 *y);
+enum fold_status fold_real128_divide(__float128 *res, const __float128 *x, const __float128 *y);
+enum fold_status fold_real128_pow(__float128 *res, const __float128 *x, const __float128 *y);
+enum fold_status fold_real128_sin(__float128 *res, const __float128 *arg);
+enum fold_status fold_real128_cos(__float128 *res, const __float128 *arg);
+enum fold_status fold_real128_cotan(__float128 *res, const __float128 *arg);
+enum fold_status fold_real128_tan(__float128 *res, const __float128 *arg);
+enum fold_status fold_real128_asin(__float128 *res, const __float128 *arg);
+enum fold_status fold_real128_acos(__float128 *res, const __float128 *arg);
+enum fold_status fold_real128_atan(__float128 *res, const __float128 *arg);
+enum fold_status fold_real128_atan2(__float128 *res, const __float128 *x, const __float128 *y);
+enum fold_status fold_real128_nearest(__float128 *res, const __float128 *x, const __float128 *y);
+enum fold_status fold_real128_exp(__float128 *res, const __float128 *arg);
+enum fold_status fold_real128_log(__float128 *res, const __float128 *arg);
+enum fold_status fold_real128_log10(__float128 *res, const __float128 *arg);
+#endif
+// AOCC end
 
 #ifdef __cplusplus
 }