Updated bench

elementary/benchmark/main.c

@@ -30,7 +30,9 @@ void init_chunk_elems(chunk_info_t *ci) {
 }
 
 double compute_next_chunk_elems(struct dispatch_context *dc, struct futhark_context *fc, chunk_info_t *ci) {
+    double start0 = MPI_Wtime();
     struct futhark_i8_1d *fut_chunk_with_envelope = get_chunk_with_envelope(dc, fc, 1, futhark_new_i8_1d);
+    double finish0 = MPI_Wtime();
 
     // Warmup
     struct futhark_i8_1d *fut_next_chunk_elems;
@@ -38,7 +40,7 @@ double compute_next_chunk_elems(struct dispatch_context *dc, struct futhark_cont
     futhark_context_sync(fc);
     futhark_free_i8_1d(fc, fut_next_chunk_elems);
 
-    double start = MPI_Wtime();
+    double start1 = MPI_Wtime();
     for (int i = 0; i < N_ITERATIONS; ++i) {
         futhark_entry_next_chunk_elems(fc, &fut_next_chunk_elems, fut_chunk_with_envelope);
         futhark_context_sync(fc);
@@ -46,10 +48,10 @@ double compute_next_chunk_elems(struct dispatch_context *dc, struct futhark_cont
     }
     futhark_values_i8_1d(fc, fut_next_chunk_elems, ci->data);
     futhark_context_sync(fc);
-    double finish = MPI_Wtime();
+    double finish1 = MPI_Wtime();
 
     futhark_free_i8_1d(fc, fut_chunk_with_envelope);
-    return finish - start;
+    return (finish0 - start0) + (finish1 - start1);
 }
 
 int main(int argc, char *argv[]) {

game_of_life/benchmark/main.c

@@ -34,14 +34,16 @@ void init_chunk_board(chunk_info_t *ci) {
 }
 
 double compute_next_chunk_board(struct dispatch_context *dc, struct futhark_context *fc, chunk_info_t *ci) {
+    double start0 = MPI_Wtime();
     struct futhark_i8_2d *fut_chunk_with_envelope = get_chunk_with_envelope(dc, fc, 1, futhark_new_i8_2d);
+    double finish0 = MPI_Wtime();
 
     struct futhark_i8_2d *fut_next_chunk_board;
     futhark_entry_next_chunk_board(fc, &fut_next_chunk_board, fut_chunk_with_envelope);
     futhark_context_sync(fc);
     futhark_free_i8_2d(fc, fut_next_chunk_board);
 
-    double start = MPI_Wtime();
+    double start1 = MPI_Wtime();
     for (int i = 0; i < N_ITERATIONS; ++i) {
         futhark_entry_next_chunk_board(fc, &fut_next_chunk_board, fut_chunk_with_envelope);
         futhark_context_sync(fc);
@@ -49,10 +51,10 @@ double compute_next_chunk_board(struct dispatch_context *dc, struct futhark_cont
     }
     futhark_values_i8_2d(fc, fut_next_chunk_board, ci->data);
     futhark_context_sync(fc);
-    double finish = MPI_Wtime();
+    double finish1 = MPI_Wtime();
 
     futhark_free_i8_2d(fc, fut_chunk_with_envelope);
-    return finish - start;
+    return (finish0 - start0) + (finish1 - start1);
 }
 
 int main(int argc, char *argv[]) {

lattice_boltzmann/main.c

@@ -42,7 +42,9 @@ convert_chunk_with_envelope(struct futhark_context *fc, const void *data, int64_
 }
 
 double compute_next_lbm(struct dispatch_context *dc, struct futhark_context *fc, chunk_info_t *ci) {
+    double start0 = MPI_Wtime();
     struct futhark_f32_4d *fut_chunk_with_envelope = get_chunk_with_envelope(dc, fc, 1, convert_chunk_with_envelope);
+    double finish0 = MPI_Wtime();
 
     struct futhark_f32_4d *fut_next_chunk_lbm;
     // Warmup
@@ -50,7 +52,7 @@ double compute_next_lbm(struct dispatch_context *dc, struct futhark_context *fc,
     futhark_context_sync(fc);
     futhark_free_f32_4d(fc, fut_next_chunk_lbm);
 
-    double start = MPI_Wtime();
+    double start1 = MPI_Wtime();
     for (int i = 0; i < N_ITERATIONS; ++i) {
         futhark_entry_next_chunk_lbm(fc, &fut_next_chunk_lbm, fut_chunk_with_envelope);
         futhark_context_sync(fc);
@@ -58,10 +60,10 @@ double compute_next_lbm(struct dispatch_context *dc, struct futhark_context *fc,
     }
     futhark_values_f32_4d(fc, fut_next_chunk_lbm, ci->data);
     futhark_context_sync(fc);
-    double finish = MPI_Wtime();
+    double finish1 = MPI_Wtime();
 
     futhark_free_f32_4d(fc, fut_chunk_with_envelope);
-    return finish - start;
+    return (finish0 - start0) + (finish1 - start1);
 }
 
 int main(int argc, char *argv[]) {