[perf] Benchmark traces

Add a variant of cairo-perf that measures the time to replay traces.

7 files changed, 605 insertions, 3 deletions

diff --git a/boilerplate/cairo-boilerplate-system.c b/boilerplate/cairo-boilerplate-system.c
index 3855b73e..f13bd146 100644
--- a/boilerplate/cairo-boilerplate-system.c
+++ b/boilerplate/cairo-boilerplate-system.c
@@ -145,3 +145,18 @@ xunlink (const char *pathname)
 	exit (1);
     }
 }
+
+char *
+xstrdup (const char *str)
+{
+    if (str == NULL)
+	return NULL;
+
+    str = strdup (str);
+    if (str == NULL) {
+	fprintf (stderr, "Error: Out of memory. Exiting.\n");
+	exit (1);
+    }
+
+    return (char *) str;
+}
diff --git a/boilerplate/cairo-boilerplate-system.h b/boilerplate/cairo-boilerplate-system.h
index 51273a74..8c2dbe8a 100644
--- a/boilerplate/cairo-boilerplate-system.h
+++ b/boilerplate/cairo-boilerplate-system.h
@@ -48,4 +48,8 @@ xasprintf (char **strp, const char *fmt, ...) CAIRO_BOILERPLATE_PRINTF_FORMAT(2,
 void
 xunlink (const char *path);
 
+#define xstrdup cairo_boilerplate_xstrdup
+char *
+xstrdup (const char *str);
+
 #endif
diff --git a/perf/.gitignore b/perf/.gitignore
index a5d452f0..4716f5a6 100644
--- a/perf/.gitignore
+++ b/perf/.gitignore
@@ -1,6 +1,7 @@
 TAGS
 tags
 cairo-perf
+cairo-perf-trace
 cairo-perf-compare-backends
 cairo-perf-diff-files
 cairo-perf-graph-files
diff --git a/perf/Makefile.am b/perf/Makefile.am
index 653d93dd..bb2ce2eb 100644
--- a/perf/Makefile.am
+++ b/perf/Makefile.am
@@ -4,12 +4,14 @@ AM_CPPFLAGS =					\
 	-I$(srcdir)				\
 	-I$(top_srcdir)/boilerplate		\
 	-I$(top_srcdir)/src			\
+	-I$(top_srcdir)/util/cairo-script	\
 	-I$(top_builddir)/src			\
 	$(CAIRO_CFLAGS)
 
 AM_LDFLAGS = $(CAIRO_LDFLAGS)
 
 EXTRA_PROGRAMS += cairo-perf \
+		  cairo-perf-trace \
 		  cairo-perf-diff-files \
 		  cairo-perf-compare-backends \
 		  cairo-perf-graph-files
@@ -70,6 +72,21 @@ libcairoperf_la_SOURCES = \
 	cairo-stats.c		\
 	cairo-stats.h
 
+cairo_perf_trace_SOURCES =		\
+	cairo-perf-trace.c
+if CAIRO_HAS_WIN32_SURFACE
+cairo_perf_trace_SOURCES += cairo-perf-win32.c
+else
+if CAIRO_HAS_OS2_SURFACE
+cairo_perf_trace_SOURCES += cairo-perf-os2.c
+else
+cairo_perf_trace_SOURCES += cairo-perf-posix.c
+endif
+endif
+cairo_perf_trace_LDADD =		\
+	$(top_builddir)/util/cairo-script/libcairo-script-interpreter.la \
+	$(LDADD)
+
 cairo_perf_diff_files_SOURCES =	\
 	cairo-perf-diff-files.c
 
diff --git a/perf/README b/perf/README
index ca5f2a1c..ebc2daf0 100644
--- a/perf/README
+++ b/perf/README
@@ -179,6 +179,37 @@ added:
     above, three tests would be performed at sizes of 16x16, 32x32 and
     64x64.
 
+
+How to benchmark traces
+-----------------------
+Using cairo-trace you can record the exact sequence of graphic operations
+made by an application and replay them later. These traces can then be
+used to benchmark the various backends and patches.
+
+To record a trace:
+$ cairo-trace --no-mark-dirty --no-callers $APPLICATION [$ARGV]
+
+--no-mark-dirty is useful for applications that are paranoid about
+surfaces being modified by external plugins outside of their control, the
+prime example here is firefox.
+--no-callers disables the symbolic caller lookup and so speeds tracing
+(dramatically for large c++ programs) and similarly speeds up the replay
+as the files are much smaller.
+
+The output file will be called $APPLICATION.$PID.trace, the actual path
+written to will be displayed on the terminal.
+
+Then to use cairo-perf-trace:
+$ ./cairo-perf-trace $APPLICATION.$PID.trace
+
+Alternatively you can put the trace into perf/traces, or set
+CAIRO_TRACE_DIR to point to your trace directory, and the trace will be
+included in the performance tests.
+
+If you record an interesting trace, please consider sharing it by compressing
+it, LZMA preferred, and posting a link to cairo@cairographics.org.
+
+
 How to run cairo-perf-diff on WINDOWS
 -------------------------------------
 This section explains the specifics of running cairo-perf-diff under
diff --git a/perf/cairo-perf-compare-backends.c b/perf/cairo-perf-compare-backends.c
index 3f280ce9..5c8d2fc4 100644
--- a/perf/cairo-perf-compare-backends.c
+++ b/perf/cairo-perf-compare-backends.c
@@ -122,9 +122,13 @@ test_diff_print (test_diff_t			*diff,
     double test_time;
     double change;
 
-    printf ("(%s, size: %d)\n",
-	    diff->tests[0]->name,
-	    diff->tests[0]->size);
+    if (diff->tests[0]->size != 0) {
+	printf ("(%s, size: %d)\n",
+		diff->tests[0]->name,
+		diff->tests[0]->size);
+    } else {
+	printf ("(%s\n", diff->tests[0]->name);
+    }
 
     for (i = 0; i < diff->num_tests; i++) {
 	test_time = diff->tests[i]->stats.min_ticks;
diff --git a/perf/cairo-perf-trace.c b/perf/cairo-perf-trace.c
new file mode 100644
index 00000000..b7441f9e
--- /dev/null
+++ b/perf/cairo-perf-trace.c
@@ -0,0 +1,530 @@
+/* -*- Mode: c; c-basic-offset: 4; indent-tabs-mode: t; tab-width: 8; -*- */
+/*
+ * Copyright © 2006 Mozilla Corporation
+ * Copyright © 2006 Red Hat, Inc.
+ * Copyright © 2009 Chris Wilson
+ *
+ * Permission to use, copy, modify, distribute, and sell this software
+ * and its documentation for any purpose is hereby granted without
+ * fee, provided that the above copyright notice appear in all copies
+ * and that both that copyright notice and this permission notice
+ * appear in supporting documentation, and that the name of
+ * the authors not be used in advertising or publicity pertaining to
+ * distribution of the software without specific, written prior
+ * permission. The authors make no representations about the
+ * suitability of this software for any purpose.  It is provided "as
+ * is" without express or implied warranty.
+ *
+ * THE AUTHORS DISCLAIM ALL WARRANTIES WITH REGARD TO THIS
+ * SOFTWARE, INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND
+ * FITNESS, IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR ANY SPECIAL,
+ * INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER
+ * RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION
+ * OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR
+ * IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+ *
+ * Authors: Vladimir Vukicevic <vladimir@pobox.com>
+ *          Carl Worth <cworth@cworth.org>
+ *          Chris Wilson <chris@chris-wilson.co.uk>
+ */
+
+#define _GNU_SOURCE 1	/* for sched_getaffinity() */
+
+#include "cairo-perf.h"
+#include "cairo-stats.h"
+
+#include "cairo-boilerplate-getopt.h"
+#include <cairo-script-interpreter.h>
+
+/* For basename */
+#ifdef HAVE_LIBGEN_H
+#include <libgen.h>
+#endif
+
+#include <sys/types.h>
+#include <dirent.h>
+
+#if HAVE_FCFINI
+#include <fontconfig/fontconfig.h>
+#endif
+
+#ifdef HAVE_SCHED_H
+#include <sched.h>
+#endif
+
+#define CAIRO_PERF_ITERATIONS_DEFAULT	10
+#define CAIRO_PERF_LOW_STD_DEV		0.3
+#define CAIRO_PERF_STABLE_STD_DEV_COUNT	3
+
+/* Some targets just aren't that interesting for performance testing,
+ * (not least because many of these surface types use a meta-surface
+ * and as such defer the "real" rendering to later, so our timing
+ * loops wouldn't count the real work, just the recording by the
+ * meta-surface. */
+static cairo_bool_t
+target_is_measurable (cairo_boilerplate_target_t *target)
+{
+    if (target->content != CAIRO_CONTENT_COLOR_ALPHA)
+	return FALSE;
+
+    switch (target->expected_type) {
+    case CAIRO_SURFACE_TYPE_IMAGE:
+	if (strcmp (target->name, "pdf") == 0 ||
+	    strcmp (target->name, "ps") == 0)
+	{
+	    return FALSE;
+	}
+	else
+	{
+	    return TRUE;
+	}
+    case CAIRO_SURFACE_TYPE_XLIB:
+	if (strcmp (target->name, "xlib-fallback") == 0)
+	{
+	    return FALSE;
+	}
+	else
+	{
+	    return TRUE;
+	}
+    case CAIRO_SURFACE_TYPE_XCB:
+    case CAIRO_SURFACE_TYPE_GLITZ:
+    case CAIRO_SURFACE_TYPE_QUARTZ:
+    case CAIRO_SURFACE_TYPE_WIN32:
+    case CAIRO_SURFACE_TYPE_BEOS:
+    case CAIRO_SURFACE_TYPE_DIRECTFB:
+#if CAIRO_VERSION_MAJOR > 1 || (CAIRO_VERSION_MAJOR == 1 && CAIRO_VERSION_MINOR > 2)
+    case CAIRO_SURFACE_TYPE_OS2:
+#endif
+	return TRUE;
+    case CAIRO_SURFACE_TYPE_PDF:
+    case CAIRO_SURFACE_TYPE_PS:
+    case CAIRO_SURFACE_TYPE_SVG:
+    default:
+	return FALSE;
+    }
+}
+
+cairo_bool_t
+cairo_perf_can_run (cairo_perf_t	*perf,
+		    const char		*name)
+{
+    unsigned int i;
+    char *copy, *dot;
+
+    if (perf->num_names == 0)
+	return TRUE;
+
+    copy = xstrdup (name);
+    dot = strchr (copy, '.');
+    if (dot != NULL)
+	*dot = '\0';
+
+    for (i = 0; i < perf->num_names; i++)
+	if (strstr (copy, perf->names[i]))
+	    break;
+
+    free (copy);
+
+    return i != perf->num_names;
+}
+
+static void
+clear_surface (cairo_surface_t *surface)
+{
+    cairo_t *cr = cairo_create (surface);
+    cairo_set_operator (cr, CAIRO_OPERATOR_CLEAR);
+    cairo_paint (cr);
+    cairo_destroy (cr);
+}
+
+static cairo_surface_t *
+_similar_surface_create (void *closure,
+			 cairo_content_t content,
+			 double width, double height)
+{
+    return cairo_surface_create_similar (closure, content, width, height);
+}
+
+static void
+execute (cairo_perf_t		 *perf,
+	 cairo_script_interpreter_t *csi,
+	 cairo_surface_t	 *target,
+	 const char		 *trace)
+{
+    static cairo_bool_t first_run = TRUE;
+    unsigned int i;
+    cairo_perf_ticks_t *times;
+    cairo_stats_t stats = {0.0, 0.0};
+    int low_std_dev_count;
+    char *trace_cpy, *name, *dot;
+    const cairo_script_interpreter_hooks_t hooks = {
+	.closure = target,
+	.surface_create = _similar_surface_create
+    };
+
+    trace_cpy = xstrdup (trace);
+    name = basename (trace_cpy);
+    dot = strchr (name, '.');
+    if (dot)
+	*dot = '\0';
+
+    if (perf->list_only) {
+	printf ("%s\n", name);
+	free (trace_cpy);
+	return;
+    }
+
+    if (first_run) {
+	if (perf->raw) {
+	    printf ("[ # ] %s.%-s %s %s %s ...\n",
+		    "backend", "content", "test-size", "ticks-per-ms", "time(ticks)");
+	}
+
+	if (perf->summary) {
+	    fprintf (perf->summary,
+		     "[ # ] %8s %28s %8s %8s %5s %5s %s\n",
+		     "backend", "test", "min(ticks)", "min(ms)", "median(ms)",
+		     "stddev.", "iterations");
+	}
+	first_run = FALSE;
+    }
+
+    times = perf->times;
+
+    if (perf->summary) {
+	fprintf (perf->summary,
+		 "[%3d] %8s %26s ",
+		 perf->test_number,
+		 perf->target->name,
+		 name);
+	fflush (perf->summary);
+    }
+
+    cairo_script_interpreter_install_hooks (csi, &hooks);
+
+    low_std_dev_count = 0;
+    for (i =0; i < perf->iterations; i++) {
+	cairo_perf_yield ();
+	cairo_perf_timer_start ();
+
+	cairo_script_interpreter_run (csi, trace);
+	clear_surface (target); /* queue a write to the sync'ed surface */
+
+	cairo_perf_timer_stop ();
+	times[i] = cairo_perf_timer_elapsed ();
+
+	if (perf->raw) {
+	    if (i == 0)
+		printf ("[*] %s.%s %s.%d %g",
+			perf->target->name,
+			"rgba",
+			name,
+			0,
+			cairo_perf_ticks_per_second () / 1000.0);
+	    printf (" %lld", (long long) times[i]);
+	} else if (! perf->exact_iterations) {
+	    if (i > 0) {
+		_cairo_stats_compute (&stats, times, i+1);
+
+		if (stats.std_dev <= CAIRO_PERF_LOW_STD_DEV) {
+		    low_std_dev_count++;
+		    if (low_std_dev_count >= CAIRO_PERF_STABLE_STD_DEV_COUNT)
+			break;
+		} else {
+		    low_std_dev_count = 0;
+		}
+	    }
+	}
+    }
+
+    if (perf->raw)
+	printf ("\n");
+
+    if (perf->summary) {
+	_cairo_stats_compute (&stats, times, i);
+	fprintf (perf->summary,
+		 "%10lld %#8.3f %#8.3f %#5.2f%% %3d\n",
+		 (long long) stats.min_ticks,
+		 (stats.min_ticks * 1000.0) / cairo_perf_ticks_per_second (),
+		 (stats.median_ticks * 1000.0) / cairo_perf_ticks_per_second (),
+		 stats.std_dev * 100.0, stats.iterations);
+	fflush (perf->summary);
+    }
+
+    perf->test_number++;
+    free (trace_cpy);
+}
+
+static void
+usage (const char *argv0)
+{
+    fprintf (stderr,
+"Usage: %s [-l] [-r] [-v] [-i iterations] [test-names ... | traces ...]\n"
+"       %s -l\n"
+"\n"
+"Run the cairo performance test suite over the given tests (all by default)\n"
+"The command-line arguments are interpreted as follows:\n"
+"\n"
+"  -r	raw; display each time measurement instead of summary statistics\n"
+"  -v	verbose; in raw mode also show the summaries\n"
+"  -i	iterations; specify the number of iterations per test case\n"
+"  -l	list only; just list selected test case names without executing\n"
+"\n"
+"If test names are given they are used as sub-string matches so a command\n"
+"such as \"cairo-perf-trace firefox\" can be used to run all firefox traces.\n"
+"Alternatively, you can specify a list of filenames to execute.\n",
+	     argv0, argv0);
+}
+
+static void
+parse_options (cairo_perf_t *perf, int argc, char *argv[])
+{
+    int c;
+    const char *iters;
+    char *end;
+    int verbose = 0;
+
+    if ((iters = getenv ("CAIRO_PERF_ITERATIONS")) && *iters)
+	perf->iterations = strtol (iters, NULL, 0);
+    else
+	perf->iterations = CAIRO_PERF_ITERATIONS_DEFAULT;
+    perf->exact_iterations = 0;
+
+    perf->raw = FALSE;
+    perf->list_only = FALSE;
+    perf->names = NULL;
+    perf->num_names = 0;
+    perf->summary = stdout;
+
+    while (1) {
+	c = _cairo_getopt (argc, argv, "i:lrv");
+	if (c == -1)
+	    break;
+
+	switch (c) {
+	case 'i':
+	    perf->exact_iterations = TRUE;
+	    perf->iterations = strtoul (optarg, &end, 10);
+	    if (*end != '\0') {
+		fprintf (stderr, "Invalid argument for -i (not an integer): %s\n",
+			 optarg);
+		exit (1);
+	    }
+	    break;
+	case 'l':
+	    perf->list_only = TRUE;
+	    break;
+	case 'r':
+	    perf->raw = TRUE;
+	    perf->summary = NULL;
+	    break;
+	case 'v':
+	    verbose = 1;
+	    break;
+	default:
+	    fprintf (stderr, "Internal error: unhandled option: %c\n", c);
+	    /* fall-through */
+	case '?':
+	    usage (argv[0]);
+	    exit (1);
+	}
+    }
+
+    if (verbose && perf->summary == NULL)
+	perf->summary = stderr;
+
+    if (optind < argc) {
+	perf->names = &argv[optind];
+	perf->num_names = argc - optind;
+    }
+}
+
+static int
+check_cpu_affinity (void)
+{
+#ifdef HAVE_SCHED_GETAFFINITY
+    cpu_set_t affinity;
+    int i, cpu_count;
+
+    if (sched_getaffinity (0, sizeof (affinity), &affinity)) {
+        perror ("sched_getaffinity");
+        return -1;
+    }
+
+    for (i = 0, cpu_count = 0; i < CPU_SETSIZE; ++i) {
+        if (CPU_ISSET (i, &affinity))
+            ++cpu_count;
+    }
+
+    if (cpu_count > 1) {
+	fputs ("WARNING: cairo-perf has not been bound to a single CPU.\n",
+	       stderr);
+        return -1;
+    }
+
+    return 0;
+#else
+    fputs ("WARNING: Cannot check CPU affinity for this platform.\n",
+	   stderr);
+    return -1;
+#endif
+}
+
+static void
+cairo_perf_fini (cairo_perf_t *perf)
+{
+    cairo_boilerplate_free_targets (perf->targets);
+    free (perf->times);
+#if 0 /* XXX */
+    cairo_debug_reset_static_data ();
+#if HAVE_FCFINI
+    FcFini ();
+#endif
+#endif
+}
+
+static cairo_bool_t
+have_trace_filenames (cairo_perf_t *perf)
+{
+    unsigned int i;
+
+    if (perf->num_names == 0)
+	return FALSE;
+
+    for (i = 0; i < perf->num_names; i++)
+	if (access (perf->names[i], R_OK) == 0)
+	    return TRUE;
+
+    return FALSE;
+}
+
+static void
+cairo_perf_trace (cairo_perf_t *perf,
+		  cairo_boilerplate_target_t *target,
+		  cairo_script_interpreter_t *csi,
+		  const char *trace)
+{
+    cairo_surface_t *surface;
+    void *closure;
+
+    surface = (target->create_surface) (NULL,
+					CAIRO_CONTENT_COLOR_ALPHA,
+					1, 1,
+					1, 1,
+					CAIRO_BOILERPLATE_MODE_PERF,
+					0,
+					&closure);
+    if (surface == NULL) {
+	fprintf (stderr,
+		 "Error: Failed to create target surface: %s\n",
+		 target->name);
+	return;
+    }
+
+    cairo_perf_timer_set_synchronize (target->synchronize, closure);
+
+    execute (perf, csi, surface, trace);
+
+    cairo_surface_destroy (surface);
+
+    if (target->cleanup)
+	target->cleanup (closure);
+}
+
+int
+main (int argc, char *argv[])
+{
+    cairo_perf_t perf;
+    cairo_script_interpreter_t *csi;
+    const char *trace_dir = "traces";
+    cairo_bool_t names_are_traces;
+    unsigned int n;
+    int i;
+
+    parse_options (&perf, argc, argv);
+
+    if (! perf.list_only && check_cpu_affinity ()) {
+        fputs ("NOTICE: cairo-perf and the X server should be bound to CPUs (either the same\n"
+	       "or separate) on SMP systems. Not doing so causes random results when the X\n"
+	       "server is moved to or from cairo-perf's CPU during the benchmarks:\n"
+	       "\n"
+	       "    $ sudo taskset -cp 0 $(pidof X)\n"
+	       "    $ taskset -cp 1 $$\n"
+	       "\n"
+	       "See taskset(1) for information about changing CPU affinity.\n\n",
+	       stderr);
+    }
+
+    if (getenv ("CAIRO_TRACE_DIR") != NULL)
+	trace_dir = getenv ("CAIRO_TRACE_DIR");
+
+    perf.targets = cairo_boilerplate_get_targets (&perf.num_targets, NULL);
+    perf.times = xmalloc (perf.iterations * sizeof (cairo_perf_ticks_t));
+
+    csi = cairo_script_interpreter_create ();
+
+    /* do we have a list of filenames? */
+    names_are_traces = have_trace_filenames (&perf);
+
+    for (i = 0; i < perf.num_targets; i++) {
+        cairo_boilerplate_target_t *target = perf.targets[i];
+
+	if (! perf.list_only && ! target_is_measurable (target))
+	    continue;
+
+	perf.target = target;
+	perf.test_number = 0;
+
+	if (names_are_traces) {
+	    for (n = 0; n < perf.num_names; n++) {
+		if (access (perf.names[n], R_OK) == 0)
+		    cairo_perf_trace (&perf, target, csi, perf.names[n]);
+	    }
+	} else {
+	    DIR *dir;
+	    struct dirent *de;
+
+	    dir = opendir (trace_dir);
+	    if (dir == NULL) {
+		fprintf (stderr,
+			 "Error: Failed to open trace directory '%s'.\n"
+			 "Have you cloned the trace repository?\n"
+			 "  git clone git://anongit.freedesktop.org/~ickle/traces\n"
+			 "  cd traces && make\n"
+			 "Or set the env.var CAIRO_TRACE_DIR to point to your traces?\n",
+			 trace_dir);
+
+		return 1;
+	    }
+
+	    while ((de = readdir (dir)) != NULL) {
+		char *trace;
+		const char *dot;
+
+		dot = strrchr (de->d_name, '.');
+		if (dot == NULL)
+		    continue;
+		if (strcmp (dot, ".trace"))
+		    continue;
+
+		if (! cairo_perf_can_run (&perf, de->d_name))
+		    continue;
+
+		xasprintf (&trace, "%s/%s", trace_dir, de->d_name);
+		cairo_perf_trace (&perf, target, csi, trace);
+		free (trace);
+
+	    }
+	    closedir (dir);
+	}
+
+	if (perf.list_only)
+	    break;
+    }
+
+    cairo_script_interpreter_destroy (csi);
+    cairo_perf_fini (&perf);
+
+    return 0;
+}