Merge branch 'amd-stg-openmp' of https://github.com/ROCm-Developer-To…

…ols/aomp into amd-stg-openmp

test/smoke/Makefile

@@ -26,6 +26,7 @@ TESTS_DIR = \
     flang_isystem_prob \
     flang_math \
     flang_omp_map \
+    flang_omp_red_3d \
     flang_real16_prob \
     flang_tgt_alloc_ptr \
     function \

test/smoke/check_smoke.sh

@@ -39,7 +39,7 @@ echo "                   A non-zero exit code means a failure occured." >> check
 echo "Tests that need to be visually inspected: devices, pfspecify, pfspecify_str, stream" >> check-smoke.txt
 echo "***********************************************************************************" >> check-smoke.txt
 
-known_fails="targ_static target_teams_reduction tasks omp_get_initial flang_gen_sptr_prob"
+known_fails="targ_static target_teams_reduction tasks omp_get_initial flang_gen_sptr_prob flang_omp_red_3d"
 
 if [ "$SKIP_FAILURES" == 1 ] ; then
   skip_tests=$known_fails

test/smoke/flang_omp_red_3d/Makefile

@@ -0,0 +1,16 @@
+include ../Makefile.defs
+
+TESTNAME     = flang_omp_red_3d
+TESTSRC_MAIN = flang_omp_red_3d.f90
+TESTSRC_AUX  =
+TESTSRC_ALL  = $(TESTSRC_MAIN) $(TESTSRC_AUX)
+
+TARGET       = -fopenmp -O3 -w -i8 -fopenmp-targets=amdgcn-amd-amdhsa -Xopenmp-target=amdgcn-amd-amdhsa -march=$(AOMP_GPU)
+
+FLANG        = flang
+OMP_BIN      = $(AOMP)/bin/$(FLANG)
+CC           = $(OMP_BIN) $(VERBOSE)
+#-ccc-print-phases
+#"-\#\#\#"
+
+include ../Makefile.rules

test/smoke/flang_omp_red_3d/flang_omp_red_3d.f90

@@ -0,0 +1,88 @@
+program TargetReduction_3D
+
+    use OMP_LIB
+    implicit none
+
+    integer :: &
+      iV, jV, kV, &
+      StopCode
+    integer, dimension ( 3 ) :: &
+      lV, uV, oV
+    real, dimension ( :, :, : ), allocatable :: &
+      FEP_1, FEP_2, FEP_3, &
+      FEM_1, FEM_2, FEM_3 
+    real :: &
+      MaxSpeed
+    real, parameter :: &
+      DEFINED_MAX_SPEED = 10.0
+
+    StopCode = 0
+
+    allocate ( FEP_1 ( 32, 16, 8 ) )
+    allocate ( FEP_2 ( 32, 16, 8 ) )
+    allocate ( FEP_3 ( 32, 16, 8 ) )
+    allocate ( FEM_1 ( 32, 16, 8 ) )
+    allocate ( FEM_2 ( 32, 16, 8 ) )
+    allocate ( FEM_3 ( 32, 16, 8 ) )
+
+    call random_number ( FEP_1 )
+    call random_number ( FEP_2 )
+    call random_number ( FEP_3 )
+    call random_number ( FEM_1 )
+    call random_number ( FEM_2 )
+    call random_number ( FEM_3 )
+
+    lV = lbound ( FEP_1 )
+    uV = ubound ( FEP_1 )
+
+    FEM_1 = -1.0 * FEM_1
+    FEM_2 = -1.0 * FEM_2
+    FEM_3 = -1.0 * FEM_3
+
+    FEP_1 ( 15, 15, 4 ) = DEFINED_MAX_SPEED
+
+    !$OMP  target enter data &
+    !$OMP  map ( to: FEP_1, FEP_2, FEP_3, FEM_1, FEM_2, FEM_3 )
+
+    MaxSpeed = - huge ( 1.0 )
+
+    print*, 'MaxSpeed init',  MaxSpeed 
+    print*, 'Num devices: ', omp_get_num_devices()
+    print*, 'Expected MaxSpeed: ', DEFINED_MAX_SPEED
+
+    !$OMP  target teams distribute parallel do simd collapse ( 3 ) &
+    !$OMP  schedule ( static, 1 ) private ( iV, jV, kV ) &
+    !$OMP  reduction ( max : MaxSpeed ) 
+    do kV = lV ( 3 ), uV ( 3 )   
+      do jV = lV ( 2 ), uV ( 2 )  
+        do iV = lV ( 1 ), uV ( 1 )
+          MaxSpeed &
+            = max (  FEP_1 ( iV, jV, kV ),  FEP_2 ( iV, jV, kV ), & 
+                     FEP_3 ( iV, jV, kV ), -FEM_1 ( iV, jV, kV ), & 
+                    -FEM_2 ( iV, jV, kV ), -FEM_3 ( iV, jV, kV ), MaxSpeed  )
+        end do
+      end do
+    end do
+    !$OMP  end target teams distribute parallel do simd
+
+    print*, 'MaxSpeed reduced on device ', MaxSpeed
+    if ( MaxSpeed /= DEFINED_MAX_SPEED ) then
+      print*, 'Reduction on device: FAILED'
+      StopCode = 1
+    end if
+
+    MaxSpeed = - huge ( 1.0 )
+    MaxSpeed = max ( maxval (  FEP_1 ), maxval (  FEP_2 ), &
+                     maxval (  FEP_3 ), maxval ( -FEM_1 ), &           
+                     maxval ( -FEM_2 ), maxval ( -FEM_3 ), MaxSpeed  )
+
+    print*, 'MaxSpeed reduced on host ', MaxSpeed
+    if ( MaxSpeed /= DEFINED_MAX_SPEED ) then
+      print*, 'Reduction on host: FAILED'
+      StopCode = 1 
+    end if
+
+    if ( StopCode /= 0 ) stop StopCode
+
+end program TargetReduction_3D
+