convert fregrid_acc to fregrid_gpu

Makefile.am

@@ -56,8 +56,8 @@ if WITH_CHECK_PROGS
 SUBDIRS += t
 endif
 
-if ENABLE_ACC
-SUBDIRS += tools/libfrencutils_acc
-SUBDIRS += tools/fregrid_acc
-SUBDIRS += t_acc
+if ENABLE_GPU
+SUBDIRS += tools/libfrencutils_gpu
+SUBDIRS += tools/fregrid_gpu
+SUBDIRS += t_gpu
 endif

configure.ac

@@ -89,13 +89,13 @@ AM_CONDITIONAL([WITH_MPI_TESTS], [test x"$build_mpi" = x"yes" -a -z "$SKIP_MPI_T
 AC_ARG_ENABLE([acc],
             [AS_HELP_STRING([--enable-acc],
                             [Builds with OpenACC. This will result in a second executable for fregrid, fregrid_gpu.(default no)])])
-AS_IF([test ${enable_acc:-no} = yes],
-      [enable_acc=yes],
-      [enable_acc=no])
+AS_IF([test ${enable_gpu:-no} = yes],
+      [enable_gpu=yes],
+      [enable_gpu=no])
 # check compile flags
-AS_IF([test ${enable_acc} = yes],
+AS_IF([test ${enable_gpu} = yes],
       [GX_OPENACC_FLAGS()])
-AM_CONDITIONAL([ENABLE_ACC], [test "$enable_acc" = "yes"])
+AM_CONDITIONAL([ENABLE_GPU], [test "$enable_gpu" = "yes"])
 
 AC_CHECK_FUNCS(gettid, [], [])
 
@@ -135,7 +135,7 @@ else
 fi
 
 # Check if openacc.h exists
-if test "$enable_acc" = yes ; then
+if test "$enable_gpu" = yes ; then
    AC_CHECK_HEADERS([openacc.h], [], [AC_MSG_ERROR(["Cannot find OpenACC header file"])] )
    AC_MSG_WARN(Currently only NVIDIA compilers are supported to compile with OpenACC in FRE-NCTOOLS)
 fi
@@ -206,11 +206,11 @@ AC_CONFIG_FILES([Makefile
                  postprocessing/split_ncvars/Makefile
                  postprocessing/timavg/Makefile
                  tools/libfrencutils/Makefile
-                 tools/libfrencutils_acc/Makefile
+                 tools/libfrencutils_gpu/Makefile
                  tools/check_mask/Makefile
                  tools/cubic_utils/Makefile
                  tools/fregrid/Makefile
-                 tools/fregrid_acc/Makefile
+                 tools/fregrid_gpu/Makefile
                  tools/make_coupler_mosaic/Makefile
                  tools/make_hgrid/Makefile
                  tools/make_land_domain/Makefile
@@ -232,11 +232,11 @@ AC_CONFIG_FILES([Makefile
                  tools/simple_hydrog/rmvpr/Makefile
                  tools/simple_hydrog/libfmslite/Makefile
                  t/Makefile
-                 t_acc/Makefile
-                 t_acc/test_read_remap_file/Makefile
-                 t_acc/test_get_grid_cell_struct/Makefile
-                 t_acc/test_get_upbound_nxcells_2dx2d/Makefile
-                 t_acc/test_get_interp_order1/Makefile
+                 t_gpu/Makefile
+                 t_gpu/test_read_remap_file/Makefile
+                 t_gpu/test_get_grid_cell_struct/Makefile
+                 t_gpu/test_get_upbound_nxcells_2dx2d/Makefile
+                 t_gpu/test_get_interp_order1/Makefile
 
 ])
 

t_acc/test_get_grid_cell_struct/Makefile.am → t_gpu/test_get_grid_cell_struct/Makefile.am

@@ -20,14 +20,14 @@
 check_PROGRAMS = test_get_grid_cell_struct
 
 AM_CFLAGS =  $(NETCDF_CFLAGS) \
-            -I$(top_srcdir)/tools/fregrid_acc \
+            -I$(top_srcdir)/tools/fregrid_gpu \
             -I$(top_srcdir)/tools/libfrencutils \
-            -I$(top_srcdir)/tools/libfrencutils_acc -acc
+            -I$(top_srcdir)/tools/libfrencutils_gpu -acc
 
 LDADD = $(NETCDF_LDFLAGS) $(NETCDF_LIBS) $(RPATH_FLAGS) \
-        $(top_builddir)/tools/fregrid_acc/interp_utils_acc.o \
+        $(top_builddir)/tools/fregrid_gpu/interp_utils_gpu.o \
         $(top_builddir)/tools/libfrencutils/libfrencutils.a \
-        $(top_builddir)/tools/libfrencutils_acc/libfrencutils_acc.a
+        $(top_builddir)/tools/libfrencutils_gpu/libfrencutils_gpu.a
 
 test_get_grid_cell_struct_SOURCES = test_get_grid_cell_struct.c
 

t_acc/test_get_grid_cell_struct/test_get_grid_cell_struct.c → t_gpu/test_get_grid_cell_struct/test_get_grid_cell_struct.c

@@ -18,18 +18,18 @@
  * <http://www.gnu.org/licenses/>.
  **********************************************************************/
 
-// This test tests the function test_get_grid_cell_struct used in fregrid_acc.
+// This test tests the function test_get_grid_cell_struct used in fregrid_gpu.
 // Properties of each grid cell in a smple made-up grid with no poles are computed
 // on the device.  This test ensures that the data transfer between the host and device
 // and computations have executed on the device as expected.
 
 #include <stdlib.h>
 #include <stdio.h>
 #include <math.h>
-#include "create_xgrid_utils_acc.h"
-#include "interp_utils_acc.h"
+#include "create_xgrid_utils_gpu.h"
+#include "interp_utils_gpu.h"
 #include "parameters.h"
-#include "globals_acc.h"
+#include "globals_gpu.h"
 
 #define NLON 36 // 36 cells in lon direction (36+1 grid points in the lon direction for each lat point)
 #define NLAT 4  // 4 cells in lat direction ( 4+1 grid points in the lat direction for each lon point)
@@ -79,10 +79,10 @@ int main(){
   }
 
   // copy grid to device
-  copy_grid_to_device_acc((NLON+1)*(NLAT+1), grid.latc, grid.lonc);
+  copy_grid_to_device_gpu((NLON+1)*(NLAT+1), grid.latc, grid.lonc);
 
   // get grid_cells
-  get_grid_cell_struct_acc( NLON, NLAT, &grid, &grid_cells);
+  get_grid_cell_struct_gpu( NLON, NLAT, &grid, &grid_cells);
 
   // get answers
   get_answers(grid.lonc, grid.latc, &answers);

t_acc/test_get_interp_order1/Makefile.am → t_gpu/test_get_interp_order1/Makefile.am

@@ -20,15 +20,15 @@
 check_PROGRAMS = test_get_interp_order1
 
 AM_CFLAGS =  $(NETCDF_CFLAGS) \
-            -I$(top_srcdir)/tools/fregrid_acc \
+            -I$(top_srcdir)/tools/fregrid_gpu \
             -I$(top_srcdir)/tools/libfrencutils \
-            -I$(top_srcdir)/tools/libfrencutils_acc -acc
+            -I$(top_srcdir)/tools/libfrencutils_gpu -acc
 
 LDADD = $(NETCDF_LDFLAGS) $(NETCDF_LIBS) $(RPATH_FLAGS) \
-        $(top_builddir)/tools/fregrid_acc/interp_utils_acc.o \
-        $(top_builddir)/tools/fregrid_acc/conserve_interp_acc.o \
+        $(top_builddir)/tools/fregrid_gpu/interp_utils_gpu.o \
+        $(top_builddir)/tools/fregrid_gpu/conserve_interp_gpu.o \
         $(top_builddir)/tools/libfrencutils/libfrencutils.a \
-        $(top_builddir)/tools/libfrencutils_acc/libfrencutils_acc.a
+        $(top_builddir)/tools/libfrencutils_gpu/libfrencutils_gpu.a
 
 test_get_interp_order1_SOURCES = test_get_interp_order1.c
 

t_acc/test_get_interp_order1/test_get_interp_order1.c → t_gpu/test_get_interp_order1/test_get_interp_order1.c

@@ -18,19 +18,19 @@
  * <http://www.gnu.org/licenses/>.
  **********************************************************************/
 
-// This test tests the function create_xgrid_2dx2d_order2_acc for a simple
+// This test tests the function create_xgrid_2dx2d_order2_gpu for a simple
 // case where the input grid is identical to the output grid.  The exchange
 // grid is identical to the input/output grid.  Only the parent indices corresponding
 // to each exchange cell are checked.
 
 #include <stdlib.h>
 #include <stdio.h>
 #include <openacc.h>
-#include "globals_acc.h"
-#include "interp_utils_acc.h"
-#include "conserve_interp_acc.h"
-#include "create_xgrid_utils_acc.h"
-#include "create_xgrid_acc.h"
+#include "globals_gpu.h"
+#include "interp_utils_gpu.h"
+#include "conserve_interp_gpu.h"
+#include "create_xgrid_utils_gpu.h"
+#include "create_xgrid_gpu.h"
 
 void generate_input_is_same_as_output_grids(Grid_config *input_grid, Grid_config *output_grid,
                                             Interp_per_input_tile *interp_answers);
@@ -152,30 +152,30 @@ int run_tests(Grid_config *input_grid, Grid_config *output_grid, Interp_per_inpu
   Grid_cells_struct_config output_grid_cells;
 
   //copy grid to device
-  copy_grid_to_device_acc(ngridpts_input, input_grid->latc, input_grid->lonc);
-  copy_grid_to_device_acc(ngridpts_output, output_grid->latc, output_grid->lonc);
+  copy_grid_to_device_gpu(ngridpts_input, input_grid->latc, input_grid->lonc);
+  copy_grid_to_device_gpu(ngridpts_output, output_grid->latc, output_grid->lonc);
 
   //get mask to skip input cells in creating interp
-  get_input_grid_mask_acc(ncells_input, &input_grid_mask);
+  get_input_grid_mask_gpu(ncells_input, &input_grid_mask);
 
   //get output grid cell info
-  get_grid_cell_struct_acc(nlon_output_cells, nlat_output_cells, output_grid, &output_grid_cells);
+  get_grid_cell_struct_gpu(nlon_output_cells, nlat_output_cells, output_grid, &output_grid_cells);
 
   //get bounding index
-  get_bounding_indices_acc(nlon_output_cells, nlat_output_cells, nlon_input_cells, nlat_input_cells,
+  get_bounding_indices_gpu(nlon_output_cells, nlat_output_cells, nlon_input_cells, nlat_input_cells,
                            output_grid->latc, input_grid->latc, &jlat_overlap_starts, &jlat_overlap_ends);
 
   //malloc and create arrays
-  create_upbound_nxcells_arrays_on_device_acc(ncells_input, &approx_nxcells_per_ij1, &ij2_start, &ij2_end);
+  create_upbound_nxcells_arrays_on_device_gpu(ncells_input, &approx_nxcells_per_ij1, &ij2_start, &ij2_end);
 
-  upbound_nxcells = get_upbound_nxcells_2dx2d_acc(input_grid->nxc, input_grid->nyc, output_grid->nxc, output_grid->nyc,
+  upbound_nxcells = get_upbound_nxcells_2dx2d_gpu(input_grid->nxc, input_grid->nyc, output_grid->nxc, output_grid->nyc,
                                                   jlat_overlap_starts, jlat_overlap_ends,
                                                   input_grid->lonc, input_grid->latc,
                                                   output_grid->lonc, output_grid->latc,
                                                   input_grid_mask, &output_grid_cells,
                                                   approx_nxcells_per_ij1, ij2_start, ij2_end);
 
-  nxcells = create_xgrid_2dx2d_order1_acc( nlon_input_cells, nlat_input_cells,
+  nxcells = create_xgrid_2dx2d_order1_gpu( nlon_input_cells, nlat_input_cells,
                                            nlon_output_cells, nlat_output_cells,
                                            jlat_overlap_starts, jlat_overlap_ends,
                                            input_grid->lonc, input_grid->latc,
@@ -184,9 +184,9 @@ int run_tests(Grid_config *input_grid, Grid_config *output_grid, Interp_per_inpu
                                            approx_nxcells_per_ij1, ij2_start, ij2_end,
                                            interp);
 
-  free_grid_cell_struct_acc(ncells_output, &output_grid_cells);
-  free_upbound_nxcells_arrays_acc(ncells_input, &approx_nxcells_per_ij1, &ij2_start, &ij2_end);
-  free_input_grid_mask_acc(ncells_input, &input_grid_mask);
+  free_grid_cell_struct_gpu(ncells_output, &output_grid_cells);
+  free_upbound_nxcells_arrays_gpu(ncells_input, &approx_nxcells_per_ij1, &ij2_start, &ij2_end);
+  free_input_grid_mask_gpu(ncells_input, &input_grid_mask);
 
   return 0;
 

t_acc/test_get_upbound_nxcells_2dx2d/Makefile.am → t_gpu/test_get_upbound_nxcells_2dx2d/Makefile.am

@@ -20,15 +20,15 @@
 check_PROGRAMS = test_get_upbound_nxcells_2dx2d
 
 AM_CFLAGS =  $(NETCDF_CFLAGS) \
-            -I$(top_srcdir)/tools/fregrid_acc \
+            -I$(top_srcdir)/tools/fregrid_gpu \
             -I$(top_srcdir)/tools/libfrencutils \
-            -I$(top_srcdir)/tools/libfrencutils_acc -acc
+            -I$(top_srcdir)/tools/libfrencutils_gpu -acc
 
 LDADD = $(NETCDF_LDFLAGS) $(NETCDF_LIBS) $(RPATH_FLAGS) \
-        $(top_builddir)/tools/fregrid_acc/conserve_interp_acc.o \
-        $(top_builddir)/tools/fregrid_acc/interp_utils_acc.o \
+        $(top_builddir)/tools/fregrid_gpu/conserve_interp_gpu.o \
+        $(top_builddir)/tools/fregrid_gpu/interp_utils_gpu.o \
         $(top_builddir)/tools/libfrencutils/libfrencutils.a \
-        $(top_builddir)/tools/libfrencutils_acc/libfrencutils_acc.a
+        $(top_builddir)/tools/libfrencutils_gpu/libfrencutils_gpu.a
 
 test_get_upbound_nxcells_2dx2d_SOURCES = test_get_upbound_nxcells_2dx2d.c
 

t_acc/test_get_upbound_nxcells_2dx2d/test_get_upbound_nxcells_2dx2d.c → t_gpu/test_get_upbound_nxcells_2dx2d/test_get_upbound_nxcells_2dx2d.c

@@ -28,11 +28,11 @@
 #include <stdlib.h>
 #include <stdio.h>
 #include <openacc.h>
-#include "globals_acc.h"
-#include "conserve_interp_acc.h"
-#include "interp_utils_acc.h"
-#include "create_xgrid_utils_acc.h"
-#include "create_xgrid_acc.h"
+#include "globals_gpu.h"
+#include "conserve_interp_gpu.h"
+#include "interp_utils_gpu.h"
+#include "create_xgrid_utils_gpu.h"
+#include "create_xgrid_gpu.h"
 
 typedef struct {
   int ncells_input;
@@ -322,33 +322,33 @@ int run_tests(Grid_config *input_grid, Grid_config *output_grid, Grid_cells_stru
   double *input_grid_mask;
 
   //copy grid to device
-  copy_grid_to_device_acc(ngridpts_input, input_grid->latc, input_grid->lonc);
-  copy_grid_to_device_acc(ngridpts_output, output_grid->latc, output_grid->lonc);
+  copy_grid_to_device_gpu(ngridpts_input, input_grid->latc, input_grid->lonc);
+  copy_grid_to_device_gpu(ngridpts_output, output_grid->latc, output_grid->lonc);
 
   //get mask to skip input cells in creating xgrid
-  get_input_grid_mask_acc(ncells_input, &input_grid_mask);
+  get_input_grid_mask_gpu(ncells_input, &input_grid_mask);
   if( ! acc_is_present(input_grid_mask, ncells_input*sizeof(double)) ) {
     printf("INPUT_GRID_MASK IS NOT ON DEVICE!"); exit(1);
   }
 
   //get output grid cell info
-  get_grid_cell_struct_acc(nlon_output_cells, nlat_output_cells, output_grid, output_grid_cells);
+  get_grid_cell_struct_gpu(nlon_output_cells, nlat_output_cells, output_grid, output_grid_cells);
 
   //get bounding indices
-  get_bounding_indices_acc(nlon_output_cells, nlat_output_cells, nlon_input_cells, nlat_input_cells,
+  get_bounding_indices_gpu(nlon_output_cells, nlat_output_cells, nlon_input_cells, nlat_input_cells,
                            output_grid->latc, input_grid->latc, &jlat_overlap_starts, &jlat_overlap_ends);
   if(jlat_overlap_starts != 0) printf("SMETHING IS WRONG WITH JLAT_OVERLAP_STARTS %d\n", jlat_overlap_starts);
   if(jlat_overlap_ends != nlat_input_cells-1) // this is what was in the original.
     printf("SOMETHING IS WRONG WITH JLAT_OVERLAP_ENDS %d %d\n", jlat_overlap_ends, nlat_input_cells-1);
 
   //malloc and create arrays
-  create_upbound_nxcells_arrays_on_device_acc(ncells_input, &p_approx_nxcells_per_ij1, &p_ij2_start, &p_ij2_end);
+  create_upbound_nxcells_arrays_on_device_gpu(ncells_input, &p_approx_nxcells_per_ij1, &p_ij2_start, &p_ij2_end);
 
   // check to ensure all data have been transferred/created to device
   check_data_on_device(input_grid, output_grid, p_approx_nxcells_per_ij1, p_ij2_start, p_ij2_end,
                        output_grid_cells);
 
-  *upbound_nxcells = get_upbound_nxcells_2dx2d_acc(input_grid->nxc, input_grid->nyc, output_grid->nxc, output_grid->nyc,
+  *upbound_nxcells = get_upbound_nxcells_2dx2d_gpu(input_grid->nxc, input_grid->nyc, output_grid->nxc, output_grid->nyc,
                                                    jlat_overlap_starts, jlat_overlap_ends,
                                                    input_grid->lonc, input_grid->latc, output_grid->lonc, output_grid->latc,
                                                    input_grid_mask, output_grid_cells,
@@ -358,7 +358,7 @@ int run_tests(Grid_config *input_grid, Grid_config *output_grid, Grid_cells_stru
   *ij2_start = p_ij2_start;
   *ij2_end = p_ij2_end;
 
-  free_input_grid_mask_acc(ncells_input, &input_grid_mask);
+  free_input_grid_mask_gpu(ncells_input, &input_grid_mask);
 
   return 0;
 
@@ -372,8 +372,8 @@ void cleanup_test(Answers *answers, Grid_config *input_grid, Grid_config *output
   int ncells_output = output_grid->nxc * output_grid->nyc;
   int ncells_input  = input_grid->nxc * input_grid->nyc;
 
-  free_grid_cell_struct_acc(ncells_output, output_grid_cells);
-  free_upbound_nxcells_arrays_acc(ncells_input, &approx_nxcells_per_ij1, &ij2_start, &ij2_end);
+  free_grid_cell_struct_gpu(ncells_output, output_grid_cells);
+  free_upbound_nxcells_arrays_gpu(ncells_input, &approx_nxcells_per_ij1, &ij2_start, &ij2_end);
 
 }
 

t_acc/test_read_remap_file/Makefile.am → t_gpu/test_read_remap_file/Makefile.am

@@ -21,14 +21,14 @@ check_PROGRAMS = test_read_remap_file
 
 AM_CFLAGS = $(NETCDF_CFLAGS) \
             -I$(top_srcdir)/tools/libfrencutils  \
-            -I$(top_srcdir)/tools/libfrencutils_acc \
-            -I$(top_srcdir)/tools/fregrid_acc -acc
+            -I$(top_srcdir)/tools/libfrencutils_gpu \
+            -I$(top_srcdir)/tools/fregrid_gpu -acc
 
 LDADD = $(NETCDF_LDFLAGS) $(NETCDF_LIBS) $(RPATH_FLAGS) \
-        $(top_builddir)/tools/fregrid_acc/conserve_interp_acc.o \
-        $(top_builddir)/tools/fregrid_acc/interp_utils_acc.o \
+        $(top_builddir)/tools/fregrid_gpu/conserve_interp_gpu.o \
+        $(top_builddir)/tools/fregrid_gpu/interp_utils_gpu.o \
         $(top_builddir)/tools/libfrencutils/libfrencutils.a \
-        $(top_builddir)/tools/libfrencutils_acc/libfrencutils_acc.a
+        $(top_builddir)/tools/libfrencutils_gpu/libfrencutils_gpu.a
 
 test_read_remap_file_SOURCES = test_read_remap_file.c