core: replace STARPU_ENABLE_STATS with an envvar

This allows to enable some statistics without re-building.

configure.ac

@@ -1613,17 +1613,6 @@ if  test x$enable_model_debug = xyes; then
 	AC_DEFINE(STARPU_MODEL_DEBUG, [1], [enable performance model debug])
 fi
 
-AC_MSG_CHECKING(whether statistics should be generated)
-AC_ARG_ENABLE(stats, [AS_HELP_STRING([--enable-stats],
-			[enable statistics])],
-			enable_stats=$enableval, enable_stats=no)
-AC_MSG_RESULT($enable_stats)
-AC_SUBST(STATS, $enable_stats)
-AC_SUBST(STARPU_ENABLE_STATS, $enable_stats)
-if test x$enable_stats = xyes; then
-        AC_DEFINE(STARPU_ENABLE_STATS, [1], [enable statistics])
-fi
-
 AC_MSG_CHECKING(whether memory stats should be displayed)
 AC_ARG_ENABLE(memory-stats, [AS_HELP_STRING([--enable-memory-stats],
 			     [enable memory stats])],

doc/doxygen/chapters/13offline_performance_tools.doxy

@@ -556,8 +556,8 @@ Node #3
 \section DataStatistics Data Statistics
 
 Different data statistics can be displayed at the end of the execution
-of the application. To enable them, you need to pass the option
-\ref enable-stats "--enable-stats" when calling <c>configure</c>. When calling
+of the application. To enable them, you need to define the environment
+variable \ref STARPU_ENABLE_STATS. When calling
 starpu_shutdown() various statistics will be displayed,
 execution, MSI cache statistics, allocation cache statistics, and data
 transfer statistics. The display can be disabled by setting the

doc/doxygen/chapters/40environment_variables.doxy

@@ -705,6 +705,13 @@ When set to <c>1</c>, this variable indicates that StarPU should automatically
 generate a Paje trace when starpu_shutdown() is called.
 </dd>
 
+<dt>STARPU_ENABLE_STATS</dt>
+<dd>
+\anchor STARPU_ENABLE_STATS
+\addindex __env__STARPU_ENABLE_STATS
+When defined, enable gathering various data statistics (\ref DataStatistics).
+</dd>
+
 <dt>STARPU_MEMORY_STATS</dt>
 <dd>
 \anchor STARPU_MEMORY_STATS

doc/doxygen/chapters/41configure_options.doxy

@@ -410,14 +410,6 @@ version of ViTE (at least r1430).
 Enable additional trace events which describes locks behaviour.
 </dd>
 
-<dt>--enable-stats</dt>
-<dd>
-\anchor enable-stats
-\addindex __configure__--enable-stats
-(see ../../src/datawizard/datastats.c)
-Enable gathering of various data statistics (\ref DataStatistics).
-</dd>
-
 <dt>--enable-maxbuffers</dt>
 <dd>
 \anchor enable-maxbuffers

src/core/workers.c

@@ -1154,11 +1154,13 @@ int starpu_initialize(struct starpu_conf *user_conf, int *argc, char ***argv)
 #ifdef STARPU_MODEL_DEBUG
 	_STARPU_DISP("Warning: StarPU was configured with --enable-model-debug, which slows down a bit\n");
 #endif
-#ifdef STARPU_ENABLE_STATS
-	_STARPU_DISP("Warning: StarPU was configured with --enable-stats, which slows down a bit\n");
-#endif
 #endif
 
+	if (starpu_getenv("STARPU_ENABLE_STATS")) {
+		_STARPU_DISP("Warning: STARPU_ENABLE_STATS is enabled, "
+			     "which slows down a bit\n");
+	}
+
 #if defined(_WIN32) && !defined(__CYGWIN__)
 	WSADATA wsadata;
 	WSAStartup(MAKEWORD(1,0), &wsadata);

src/datawizard/datastats.c

@@ -20,31 +20,38 @@
 #include <datawizard/coherency.h>
 #include <common/config.h>
 
-#ifdef STARPU_ENABLE_STATS
+static inline int starpu_enable_stats(void)
+{
+	return !!starpu_getenv("STARPU_ENABLE_STATS");
+}
+
 /* measure the cache hit ratio for each node */
 static unsigned hit_cnt[STARPU_MAXNODES];
 static unsigned miss_cnt[STARPU_MAXNODES];
-#endif
 
 void _starpu_msi_cache_hit(unsigned node STARPU_ATTRIBUTE_UNUSED)
 {
-#ifdef STARPU_ENABLE_STATS
+	if (!starpu_enable_stats())
+		return;
+
 	STARPU_HG_DISABLE_CHECKING(hit_cnt[node]);
 	hit_cnt[node]++;
-#endif
 }
 
 void _starpu_msi_cache_miss(unsigned node STARPU_ATTRIBUTE_UNUSED)
 {
-#ifdef STARPU_ENABLE_STATS
+	if (!starpu_enable_stats())
+		return;
+
 	STARPU_HG_DISABLE_CHECKING(miss_cnt[node]);
 	miss_cnt[node]++;
-#endif
 }
 
 void _starpu_display_msi_stats(void)
 {
-#ifdef STARPU_ENABLE_STATS
+	if (!starpu_enable_stats())
+		return;
+
 	unsigned node;
 	unsigned total_hit_cnt = 0;
 	unsigned total_miss_cnt = 0;
@@ -70,35 +77,35 @@ void _starpu_display_msi_stats(void)
 		}
 	}
 	fprintf(stderr, "#---------------------\n");
-#endif
 }
 
 /* measure the efficiency of our allocation cache */
-
-#ifdef STARPU_ENABLE_STATS
 static unsigned alloc_cnt[STARPU_MAXNODES];
 static unsigned alloc_cache_hit_cnt[STARPU_MAXNODES];
-#endif
 
 void _starpu_allocation_cache_hit(unsigned node STARPU_ATTRIBUTE_UNUSED)
 {
-#ifdef STARPU_ENABLE_STATS
+	if (!starpu_enable_stats())
+		return;
+
 	STARPU_HG_DISABLE_CHECKING(alloc_cache_hit_cnt[node]);
 	alloc_cache_hit_cnt[node]++;
-#endif
 }
 
 void _starpu_data_allocation_inc_stats(unsigned node STARPU_ATTRIBUTE_UNUSED)
 {
-#ifdef STARPU_ENABLE_STATS
+	if (!starpu_enable_stats())
+		return;
+
 	STARPU_HG_DISABLE_CHECKING(alloc_cnt[node]);
 	alloc_cnt[node]++;
-#endif
 }
 
 void _starpu_display_alloc_cache_stats(void)
 {
-#ifdef STARPU_ENABLE_STATS
+	if (!starpu_enable_stats())
+		return;
+
 	fprintf(stderr, "\n#---------------------\n");
 	fprintf(stderr, "Allocation cache stats:\n");
 	unsigned node;
@@ -115,20 +122,18 @@ void _starpu_display_alloc_cache_stats(void)
 			fprintf(stderr, "No allocation on node %d\n", node);
 	}
 	fprintf(stderr, "#---------------------\n");
-#endif
 }
 
 /* measure the amount of data transfers between each pair of nodes */
-#ifdef STARPU_ENABLE_STATS
 static size_t comm_amount[STARPU_MAXNODES][STARPU_MAXNODES];
-#endif /* STARPU_ENABLE_STATS */
 
 void _starpu_comm_amounts_inc(unsigned src  STARPU_ATTRIBUTE_UNUSED, unsigned dst  STARPU_ATTRIBUTE_UNUSED, size_t size  STARPU_ATTRIBUTE_UNUSED)
 {
-#ifdef STARPU_ENABLE_STATS
+	if (!starpu_enable_stats())
+		return;
+
 	STARPU_HG_DISABLE_CHECKING(comm_amount[src][dst]);
 	comm_amount[src][dst] += size;
-#endif /* STARPU_ENABLE_STATS */
 }
 
 void _starpu_display_comm_amounts(void)
@@ -137,7 +142,9 @@ void _starpu_display_comm_amounts(void)
 #  warning TODO. The information displayed here seems to be similar to the one displayed by starpu_profiling_bus_helper_display_summary()
 #endif
 
-#ifdef STARPU_ENABLE_STATS
+	if (!starpu_enable_stats())
+		return;
+
 	unsigned src, dst;
 	size_t sum = 0;
 
@@ -163,6 +170,5 @@ void _starpu_display_comm_amounts(void)
 					dst, src, ((float)comm_amount[dst][src])/(1024*1024));
 		}
 	fprintf(stderr, "#---------------------\n");
-#endif
 }