- add a (pretty naive) test to measure the latency between 2 GPUs

tests/Makefile.am

@@ -100,6 +100,7 @@ check_PROGRAMS += 				\
 	microbenchs/tasks-overhead		\
 	microbenchs/prefetch_data_on_node 	\
 	microbenchs/redundant_buffer		\
+	microbenchs/local_pingpong		\
 	overlap/overlap
 
 core_restart_SOURCES =				\
@@ -192,6 +193,9 @@ datawizard_sync_and_notify_data_SOURCES =	\
 microbenchs_redundant_buffer_SOURCES =		\
 	microbenchs/redundant_buffer.c
 
+microbenchs_local_pingpong_SOURCES =		\
+	microbenchs/local_pingpong.c
+
 if USE_CUDA
 datawizard_sync_and_notify_data_SOURCES +=	\
 	datawizard/sync_and_notify_data_kernels.cu

tests/microbenchs/local_pingpong.c

@@ -0,0 +1,111 @@
+/*
+ * StarPU
+ * Copyright (C) INRIA 2008-2009 (see AUTHORS file)
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU Lesser General Public License as published by
+ * the Free Software Foundation; either version 2.1 of the License, or (at
+ * your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * See the GNU Lesser General Public License in COPYING.LGPL for more details.
+ */
+
+#include <stdio.h>
+#include <unistd.h>
+#include <errno.h>
+#include <starpu.h>
+#include <stdlib.h>
+
+static size_t vector_size = 1;
+
+static niter = 10000;
+static unsigned cnt;
+
+static unsigned finished = 0;
+
+starpu_data_handle v_handle;
+static unsigned *v;
+
+static char worker_0_name[128];
+static char worker_1_name[128];
+static uint32_t memory_node_0;
+static uint32_t memory_node_1;
+
+struct timeval start;
+struct timeval end;
+
+int main(int argc, char **argv)
+{
+	starpu_init(NULL);
+
+	/* Create a piece of data */
+	starpu_malloc_pinned_if_possible((void **)&v, vector_size);
+	starpu_register_vector_data(&v_handle, 0, (uintptr_t)v, vector_size, 1);
+
+	/* Find a pair of memory nodes */
+	if (starpu_get_cuda_worker_count() > 1)
+	{
+		/* Take the two devices that come first */
+		int nworkers = (int)starpu_get_worker_count();
+
+		unsigned found_node_0 = 0;
+
+		int w;
+		for (w = 0; w < nworkers; w++)
+		{
+			if (starpu_get_worker_type(w) == STARPU_CUDA_WORKER)
+			{
+				if (!found_node_0)
+				{
+					memory_node_0 = starpu_get_worker_memory_node(w);
+					starpu_get_worker_name(w, worker_0_name, 128);
+					found_node_0 = 1;
+				}
+				else {
+					memory_node_1 = starpu_get_worker_memory_node(w);
+					starpu_get_worker_name(w, worker_1_name, 128);
+					break;
+				}
+			}
+		}
+
+		fprintf(stderr, "Ping-pong will be done between %s (node %d) and %s (node %d)\n",
+					worker_0_name, memory_node_0, worker_1_name, memory_node_1);
+	}
+
+	unsigned iter;
+
+	/* warm up */
+	unsigned nwarmupiter = 128;
+	for (iter = 0; iter < nwarmupiter; iter++)
+	{
+		 _starpu_prefetch_data_on_node_with_mode(v_handle, memory_node_0, 0, STARPU_RW);
+		 _starpu_prefetch_data_on_node_with_mode(v_handle, memory_node_1, 0, STARPU_RW);
+	}
+
+	gettimeofday(&start, NULL);
+
+	for (iter = 0; iter < niter; iter++)
+	{
+		starpu_trace_user_event(0x42);
+		 _starpu_prefetch_data_on_node_with_mode(v_handle, memory_node_0, 0, STARPU_RW);
+		starpu_trace_user_event(0x43);
+		 _starpu_prefetch_data_on_node_with_mode(v_handle, memory_node_1, 0, STARPU_RW);
+	}
+
+	gettimeofday(&end, NULL);
+
+	double timing = (double)((end.tv_sec - start.tv_sec)*1000000 +
+					(end.tv_usec - start.tv_usec));
+
+	fprintf(stderr, "Took %f ms\n", timing/1000);
+	fprintf(stderr, "Avg. transfer time : %f us\n", timing/(2*niter));
+
+	starpu_shutdown();
+
+	return 0;
+}