Add syscall_timing, a simple timing micro-benchmark for some

characteristic system calls.  I've been sending this to people for
a while, and figured it would be more efficient to just put it in
CVS.

tools/tools/syscall_timing/Makefile

@@ -0,0 +1,9 @@
+#
+# $FreeBSD$
+#
+
+PROG=	syscall_timing
+CFLAGS+=	-static -O
+NOMAN=	yes
+
+.include <bsd.prog.mk>

tools/tools/syscall_timing/syscall_timing.c

@@ -0,0 +1,186 @@
+/*-
+ * Copyright (c) 2003-2004 Robert N. M. Watson
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ *
+ * $FreeBSD$
+ */
+
+#include <sys/types.h>
+#include <sys/socket.h>
+#include <sys/time.h>
+
+#include <assert.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <unistd.h>
+
+#define timespecsub(vvp, uvp)                                           \
+        do {                                                            \
+                (vvp)->tv_sec -= (uvp)->tv_sec;                         \
+                (vvp)->tv_nsec -= (uvp)->tv_nsec;                       \
+                if ((vvp)->tv_nsec < 0) {                               \
+                        (vvp)->tv_sec--;                                \
+                        (vvp)->tv_nsec += 1000000000;                   \
+                }                                                       \
+        } while (0)
+
+inline void
+test_getuid(int num)
+{
+	int i;
+
+	/*
+	 * Thread-local data should require no locking if system
+	 * call is MPSAFE.
+	 */
+	for (i = 0; i < num; i++)
+		getuid();
+}
+
+inline void
+test_getppid(int num)
+{
+	int i;
+
+	/*
+	 * This is process-local, but can change, so will require a
+	 * lock.
+	 */
+	for (i = 0; i < num; i++)
+		getppid();
+}
+
+inline void
+test_clock_gettime(int num)
+{
+	struct timespec ts;
+	int i;
+
+	for (i = 0; i < num; i++) {
+		if (clock_gettime(CLOCK_REALTIME, &ts) == -1) {
+			perror("clock_gettime");
+			exit(-1);
+		}
+	}
+}
+
+inline void
+test_pipe(int num)
+{
+	int i;
+
+	/*
+	 * pipe creation is expensive, as it will allocate a new file
+	 * descriptor, allocate a new pipe, hook it all up, and return.
+	 * Destroying is also expensive, as we now have to free up
+	 * the file descriptors and return the pipe.
+	 */
+	for (i = 0; i < num; i++) {
+		int fd[2];
+		if (pipe(fd) == -1) {
+			perror("pipe");
+			exit(-1);
+		}
+
+		close(fd[0]);
+		close(fd[1]);
+	}
+}
+
+inline void
+test_socket(int num)
+{
+	int i, so;
+
+	/*
+	 * Sockets are also expensive, but unlike pipes, currently
+	 * require Giant.
+	 */
+	for (i = 0; i < num; i++) {
+		so = socket(PF_LOCAL, SOCK_STREAM, 0);
+		if (so == -1) {
+			perror("socket");
+			exit(-1);
+		}
+		close(so);
+	}
+}
+
+int
+main(int argc, char *argv[])
+{
+	struct timespec ts_start, ts_end, ts_res;
+	int count;
+
+	if (argc != 3) {
+		fprintf(stderr, "syscall_timing [iterations] [test]\n");
+		exit(-1);
+	}
+	count = atoi(argv[1]);
+
+	assert(clock_getres(CLOCK_REALTIME, &ts_res) == 0);
+	printf("Clock resolution: %d.%09lu\n", ts_res.tv_sec, ts_res.tv_nsec);
+
+	if (strcmp(argv[2], "getuid") == 0) {
+		assert(clock_gettime(CLOCK_REALTIME, &ts_start) == 0);
+		test_getuid(count);
+		assert(clock_gettime(CLOCK_REALTIME, &ts_end) == 0);
+	} else if (strcmp(argv[2], "getppid") == 0) {
+		assert(clock_gettime(CLOCK_REALTIME, &ts_start) == 0);
+		test_getppid(count);
+		assert(clock_gettime(CLOCK_REALTIME, &ts_end) == 0);
+	} else if (strcmp(argv[2], "clock_gettime") == 0) {
+		assert(clock_gettime(CLOCK_REALTIME, &ts_start) == 0);
+		test_clock_gettime(count);
+		assert(clock_gettime(CLOCK_REALTIME, &ts_end) == 0);
+	} else if (strcmp(argv[2], "pipe") == 0) {
+		assert(clock_gettime(CLOCK_REALTIME, &ts_start) == 0);
+		test_pipe(count);
+		assert(clock_gettime(CLOCK_REALTIME, &ts_end) == 0);
+	} else if (strcmp(argv[2], "socket") == 0) {
+		assert(clock_gettime(CLOCK_REALTIME, &ts_start) == 0);
+		test_socket(count);
+		assert(clock_gettime(CLOCK_REALTIME, &ts_end) == 0);
+	 } else {
+		fprintf(stderr, "syscal [iterations] [test]\n");
+		exit(-1);
+	}
+
+	timespecsub(&ts_end, &ts_start);
+
+	printf("%d.%09lu for %d iterations\n", ts_end.tv_sec,
+	    ts_end.tv_nsec, count);
+
+	/*
+	 * Note.  This assumes that each iteration takes less than
+	 * a second, and that our total nanoseconds doesn't exceed
+	 * the room in our arithmetic unit.  Fine for system calls,
+	 * but not for long things.
+	 */
+	ts_end.tv_sec *= 1000000000 / count;
+	printf("0.%09lu per/iteration\n", 
+	    ts_end.tv_sec + ts_end.tv_nsec / count);
+	return (0);
+}