merge aomp-dev into amd-staging

test/smoke-fort-dev/gpu_mem_alloc/Makefile

@@ -0,0 +1,19 @@
+NOOPT        = 1
+NOOMP        = 0
+SUPPORTED    = $(SUPPORTS_USM)
+OMP_FLAGS    = -fopenmp --offload-arch=$(AOMP_GPU)
+include ../../Makefile.defs
+
+TESTNAME     = test
+TESTSRC_MAIN = test.f90
+TESTSRC_AUX  =
+TESTSRC_ALL  = $(TESTSRC_MAIN) $(TESTSRC_AUX)
+
+FLANG        ?= flang
+OMP_BIN      = $(AOMP)/bin/$(FLANG)
+CFLAGS       = -ffast-amd-memory-allocator -lflang_rt.amd
+CC           = $(OMP_BIN) $(VERBOSE)
+#-ccc-print-phases
+#"-\#\#\#"
+
+include ../Makefile.rules

test/smoke-fort-dev/gpu_mem_alloc/test.f90

@@ -0,0 +1,49 @@
+module kernels
+    implicit none
+contains
+    subroutine kernel(array, n)
+        implicit none
+        integer, allocatable, dimension(:) :: array
+        integer :: n
+        integer :: i
+
+        !$omp target teams distribute parallel do has_device_addr(array)
+            do i = 1, n
+                array(i) = array(i) / 2
+            end do
+        !$omp end target teams distribute parallel do
+    end subroutine kernel
+
+end module kernels
+
+
+program test
+    use kernels
+    implicit none
+
+    integer, parameter :: n = 2*1024*1028
+
+    integer, allocatable, dimension(:) :: array
+    integer, dimension(n) :: chk
+    integer :: i
+
+    allocate(array(n))
+
+    array = 42
+    chk = 0
+
+    call kernel(array, n)
+
+    !$omp target teams distribute parallel do map(from:chk) has_device_addr(array)
+        do i = 1, n
+            chk(i) = array(i)
+        end do
+    !$omp end target teams distribute parallel do
+
+    if (.not.all(array.eq.21)) then
+        print '(A)', 'FAIL'
+        stop 1
+    else
+        print '(A)', 'PASS'
+    end if
+end program test