Configure makefile to use nvcc

OpenMP/linear-algebra/kernels/atax/.bench.sh

@@ -20,30 +20,30 @@ for dataset in MINI_DATASET SMALL_DATASET STANDARD_DATASET LARGE_DATASET EXTRALA
 do
     for c in $(seq 0 15)
     do
-        cflags="-D$dataset"
+        cxxflags="-D$dataset"
 
         if (( $c & 1 ))
         then
-            cflags="$cflags -DTOGGLE_INIT_ARRAY_1"
+            cxxflags="$cxxflags -DTOGGLE_INIT_ARRAY_1"
         fi 
 
         if (( $c & 2 ))
         then
-            cflags="$cflags -DTOGGLE_INIT_ARRAY_2"
+            cxxflags="$cxxflags -DTOGGLE_INIT_ARRAY_2"
         fi 
 
         if (( $c & 4 ))
         then
-            cflags="$cflags -DTOGGLE_KERNEL_ATAX_1"
+            cxxflags="$cxxflags -DTOGGLE_KERNEL_ATAX_1"
         fi 
 
         if (( $c & 8 ))
         then
-            cflags="$cflags -DTOGGLE_KERNEL_ATAX_2"
+            cxxflags="$cxxflags -DTOGGLE_KERNEL_ATAX_2"
         fi 
 
-        echo "Flags: $cflags"
-        make "EXTRA_CFLAGS=$cflags" clean all
+        echo "Flags: $cxxflags"
+        make "EXTRA_CXXFLAGS=$cxxflags" clean all
 
         run_benchmarks
     done

OpenMP/linear-algebra/kernels/atax/Makefile

@@ -1,21 +1,32 @@
 -include ../../../utilities/options.mk
 -include ../../../utilities/c2.mk
 
-# -DPOLYBENCH_TIME makes Polybench output the execution time of the program
-CFLAGS+= -DPOLYBENCH_TIME 
-# -Wall and -Wextra enable more warnings
-CFLAGS+= -Wall 
-CFLAGS+= -Wextra 
-# -O3 applies all compiler optimization, improving from 800ms to 300ms
-CFLAGS+= -O3 
-# -DTHREAD_COUNT allows us to alter the number of threads used in the whole file
-CFLAGS+= -DTHREAD_COUNT=4
-# Extend CFLAGS with command line parameters
-CFLAGS+= ${EXTRA_CFLAGS}
 
 # Disable make output
 MAKEFLAGS+= --silent
 
+# -DPOLYBENCH_TIME makes Polybench output the execution time of the program
+CXXFLAGS+= -DPOLYBENCH_TIME 
+# -Wall and -Wextra enable more warnings
+CXXFLAGS+= -Wall 
+CXXFLAGS+= -Wextra 
+# -O3 applies all compiler optimization, improving from 800ms to 300ms
+CXXFLAGS+= -O3 
+# Extend CFLAGS with command line parameters
+CXXFLAGS+= ${EXTRA_CXXFLAGS}
+
+# Select the location of the local CUDA install
+CUDA_HOME:=/opt/cuda
+# Specify the directory of the nvc compiler
+NVCC=$(CUDA_HOME)/bin/nvcc
+# Specify the flags for the nvc compiler
+NVCFLAGS:=$(CXXFLAGS) $(NVOPT)
+
+
+# Create an automake definition for .cu files
+$(BUILD_DIR)/%.cu.o: %.cu
+  $(MKDIR_P) $(dir $@)
+  $(NVCC) $(NVCFLAGS) -c $< -o $@
 
 .PHONY: bench
 

OpenMP/linear-algebra/kernels/atax/atax.cu

@@ -8,7 +8,7 @@
 
 /* Include benchmark-specific header. */
 /* Default data type is double, default size is 4000. */
-#include "atax.h"
+#include "atax.hu"
 
 // Workaround for the editor not finding M_PI
 // It is exclusive to the GNU C compiler