Adding upstream version 0.6.0.

Signed-off-by: Daniel Baumann <daniel@debian.org>

regressions/ck_hs/benchmark/parallel_bytestring.c

@@ -0,0 +1,602 @@
+/*
+ * Copyright 2012 Samy Al Bahra.
+ * 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 copyrighs
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyrighs
+ *    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.
+ */
+#include "../../common.h"
+#include <ck_hs.h>
+#include "../../../src/ck_ht_hash.h"
+#include <assert.h>
+#include <ck_epoch.h>
+#include <ck_malloc.h>
+#include <ck_pr.h>
+#include <ck_spinlock.h>
+
+#include <errno.h>
+#include <inttypes.h>
+#include <pthread.h>
+#include <signal.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <time.h>
+#include <unistd.h>
+
+static ck_hs_t hs CK_CC_CACHELINE;
+static char **keys;
+static size_t keys_length = 0;
+static size_t keys_capacity = 128;
+static ck_epoch_t epoch_hs;
+static ck_epoch_record_t epoch_wr;
+static int n_threads;
+static bool next_stage;
+
+enum state {
+	HS_STATE_STOP = 0,
+	HS_STATE_GET,
+	HS_STATE_STRICT_REPLACEMENT,
+	HS_STATE_DELETION,
+	HS_STATE_REPLACEMENT,
+	HS_STATE_COUNT
+};
+
+static ck_spinlock_t mtx = CK_SPINLOCK_INITIALIZER;
+static struct affinity affinerator = AFFINITY_INITIALIZER;
+static uint64_t accumulator[HS_STATE_COUNT];
+static int barrier[HS_STATE_COUNT];
+static int state;
+
+struct hs_epoch {
+	ck_epoch_entry_t epoch_entry;
+};
+
+COMMON_ALARM_DECLARE_GLOBAL(hs_alarm, alarm_event, next_stage)
+
+static void
+alarm_handler(int s)
+{
+
+	(void)s;
+	next_stage = true;
+	return;
+}
+
+static unsigned long
+hs_hash(const void *object, unsigned long seed)
+{
+	const char *c = object;
+	unsigned long h;
+
+	h = (unsigned long)MurmurHash64A(c, strlen(c), seed);
+	return h;
+}
+
+static bool
+hs_compare(const void *previous, const void *compare)
+{
+
+	return strcmp(previous, compare) == 0;
+}
+
+static void
+hs_destroy(ck_epoch_entry_t *e)
+{
+
+	free(e);
+	return;
+}
+
+static void *
+hs_malloc(size_t r)
+{
+	ck_epoch_entry_t *b;
+
+	b = malloc(sizeof(*b) + r);
+	return b + 1;
+}
+
+static void
+hs_free(void *p, size_t b, bool r)
+{
+	struct hs_epoch *e = p;
+
+	(void)b;
+
+	if (r == true) {
+		/* Destruction requires safe memory reclamation. */
+		ck_epoch_call(&epoch_wr, &(--e)->epoch_entry, hs_destroy);
+	} else {
+		free(--e);
+	}
+
+	return;
+}
+
+static struct ck_malloc my_allocator = {
+	.malloc = hs_malloc,
+	.free = hs_free
+};
+
+static void
+set_init(void)
+{
+	unsigned int mode = CK_HS_MODE_OBJECT | CK_HS_MODE_SPMC;
+
+#ifdef HS_DELETE
+	mode |= CK_HS_MODE_DELETE;
+#endif
+
+	ck_epoch_init(&epoch_hs);
+	ck_epoch_register(&epoch_hs, &epoch_wr);
+	common_srand48((long int)time(NULL));
+	if (ck_hs_init(&hs, mode, hs_hash, hs_compare, &my_allocator, 65536, common_lrand48()) == false) {
+		perror("ck_hs_init");
+		exit(EXIT_FAILURE);
+	}
+
+	return;
+}
+
+static bool
+set_remove(const char *value)
+{
+	unsigned long h;
+
+	h = CK_HS_HASH(&hs, hs_hash, value);
+	return (bool)ck_hs_remove(&hs, h, value);
+}
+
+static bool
+set_replace(const char *value)
+{
+	unsigned long h;
+	void *previous;
+
+	h = CK_HS_HASH(&hs, hs_hash, value);
+	return ck_hs_set(&hs, h, value, &previous);
+}
+
+static bool
+set_swap(const char *value)
+{
+	unsigned long h;
+	void *previous;
+
+	h = CK_HS_HASH(&hs, hs_hash, value);
+	return ck_hs_fas(&hs, h, value, &previous);
+}
+
+static void *
+set_get(const char *value)
+{
+	unsigned long h;
+	void *v;
+
+	h = CK_HS_HASH(&hs, hs_hash, value);
+	v = ck_hs_get(&hs, h, value);
+	return v;
+}
+
+static bool
+set_insert(const char *value)
+{
+	unsigned long h;
+
+	h = CK_HS_HASH(&hs, hs_hash, value);
+	return ck_hs_put(&hs, h, value);
+}
+
+static size_t
+set_count(void)
+{
+
+	return ck_hs_count(&hs);
+}
+
+static bool
+set_reset(void)
+{
+
+	return ck_hs_reset(&hs);
+}
+
+static void *
+reader(void *unused)
+{
+	size_t i;
+	ck_epoch_record_t epoch_record;
+	int state_previous = HS_STATE_STOP;
+	int n_state = 0;
+	uint64_t s, j, a;
+
+	(void)unused;
+	if (aff_iterate(&affinerator) != 0)
+		perror("WARNING: Failed to affine thread");
+
+	s = j = a = 0;
+	ck_epoch_register(&epoch_hs, &epoch_record);
+	for (;;) {
+		j++;
+		ck_epoch_begin(&epoch_record, NULL);
+		s = rdtsc();
+		for (i = 0; i < keys_length; i++) {
+			char *r;
+
+			r = set_get(keys[i]);
+			if (r == NULL) {
+				if (n_state == HS_STATE_STRICT_REPLACEMENT) {
+					ck_error("ERROR: Did not find during replacement: %s\n", keys[i]);
+				}
+
+				continue;
+			}
+
+			if (strcmp(r, keys[i]) == 0)
+				continue;
+
+			ck_error("ERROR: Found invalid value: [%s] but expected [%s]\n", (char *)r, keys[i]);
+		}
+		a += rdtsc() - s;
+		ck_epoch_end(&epoch_record, NULL);
+
+		n_state = ck_pr_load_int(&state);
+		if (n_state != state_previous) {
+			ck_spinlock_lock(&mtx);
+			accumulator[state_previous] += a / (j * keys_length);
+			ck_spinlock_unlock(&mtx);
+
+			ck_pr_inc_int(&barrier[state_previous]);
+			while (ck_pr_load_int(&barrier[state_previous]) != n_threads + 1)
+				ck_pr_stall();
+
+			state_previous = n_state;
+			s = j = a = 0;
+		}
+	}
+
+	return NULL;
+}
+
+static uint64_t
+acc(size_t i)
+{
+	uint64_t r;
+
+	ck_spinlock_lock(&mtx);
+	r = accumulator[i];
+	ck_spinlock_unlock(&mtx);
+
+	return r;
+}
+
+int
+main(int argc, char *argv[])
+{
+	FILE *fp;
+	char buffer[512];
+	size_t i, j, r;
+	unsigned int d = 0;
+	uint64_t s, e, a, repeated;
+	char **t;
+	pthread_t *readers;
+	double p_r, p_d;
+
+	COMMON_ALARM_DECLARE_LOCAL(hs_alarm, alarm_event)
+
+	r = 20;
+	s = 8;
+	p_d = 0.5;
+	p_r = 0.5;
+	n_threads = CORES - 1;
+
+	if (argc < 2) {
+		ck_error("Usage: parallel <dictionary> [<interval length> <initial size> <readers>\n"
+		    " <probability of replacement> <probability of deletion> <epoch threshold>]\n");
+	}
+
+	if (argc >= 3)
+		r = atoi(argv[2]);
+
+	if (argc >= 4)
+		s = (uint64_t)atoi(argv[3]);
+
+	if (argc >= 5) {
+		n_threads = atoi(argv[4]);
+		if (n_threads < 1) {
+			ck_error("ERROR: Number of readers must be >= 1.\n");
+		}
+	}
+
+	if (argc >= 6) {
+		p_r = atof(argv[5]) / 100.00;
+		if (p_r < 0) {
+			ck_error("ERROR: Probability of replacement must be >= 0 and <= 100.\n");
+		}
+	}
+
+	if (argc >= 7) {
+		p_d = atof(argv[6]) / 100.00;
+		if (p_d < 0) {
+			ck_error("ERROR: Probability of deletion must be >= 0 and <= 100.\n");
+		}
+	}
+
+	COMMON_ALARM_INIT(hs_alarm, alarm_event, r)
+
+	affinerator.delta = 1;
+	readers = malloc(sizeof(pthread_t) * n_threads);
+	assert(readers != NULL);
+
+	keys = malloc(sizeof(char *) * keys_capacity);
+	assert(keys != NULL);
+
+	fp = fopen(argv[1], "r");
+	assert(fp != NULL);
+
+	while (fgets(buffer, sizeof(buffer), fp) != NULL) {
+		buffer[strlen(buffer) - 1] = '\0';
+		keys[keys_length++] = strdup(buffer);
+		assert(keys[keys_length - 1] != NULL);
+
+		if (keys_length == keys_capacity) {
+			t = realloc(keys, sizeof(char *) * (keys_capacity *= 2));
+			assert(t != NULL);
+			keys = t;
+		}
+	}
+
+	t = realloc(keys, sizeof(char *) * keys_length);
+	assert(t != NULL);
+	keys = t;
+
+	set_init();
+
+	for (i = 0; i < (size_t)n_threads; i++) {
+		if (pthread_create(&readers[i], NULL, reader, NULL) != 0) {
+			ck_error("ERROR: Failed to create thread %zu.\n", i);
+		}
+	}
+
+	for (i = 0; i < keys_length; i++)
+		d += set_insert(keys[i]) == false;
+
+	fprintf(stderr, " [S] %d readers, 1 writer.\n", n_threads);
+	fprintf(stderr, " [S] %zu entries stored and %u duplicates.\n\n",
+	    set_count(), d);
+
+	fprintf(stderr, " ,- BASIC TEST\n");
+	fprintf(stderr, " | Executing SMR test...");
+	a = 0;
+	for (j = 0; j < r; j++) {
+		if (set_reset() == false) {
+			ck_error("ERROR: Failed to reset hash table.\n");
+		}
+
+		s = rdtsc();
+		for (i = 0; i < keys_length; i++)
+			d += set_insert(keys[i]) == false;
+		e = rdtsc();
+		a += e - s;
+	}
+	fprintf(stderr, "done (%" PRIu64 " ticks)\n", a / (r * keys_length));
+
+	fprintf(stderr, " | Executing replacement test...");
+	a = 0;
+	for (j = 0; j < r; j++) {
+		s = rdtsc();
+		for (i = 0; i < keys_length; i++)
+			set_replace(keys[i]);
+		e = rdtsc();
+		a += e - s;
+	}
+	fprintf(stderr, "done (%" PRIu64 " ticks)\n", a / (r * keys_length));
+
+	fprintf(stderr, " | Executing get test...");
+	a = 0;
+	for (j = 0; j < r; j++) {
+		s = rdtsc();
+		for (i = 0; i < keys_length; i++) {
+			if (set_get(keys[i]) == NULL) {
+				ck_error("ERROR: Unexpected NULL value.\n");
+			}
+		}
+		e = rdtsc();
+		a += e - s;
+	}
+	fprintf(stderr, "done (%" PRIu64 " ticks)\n", a / (r * keys_length));
+
+	a = 0;
+	fprintf(stderr, " | Executing removal test...");
+	for (j = 0; j < r; j++) {
+		s = rdtsc();
+		for (i = 0; i < keys_length; i++)
+			set_remove(keys[i]);
+		e = rdtsc();
+		a += e - s;
+
+		for (i = 0; i < keys_length; i++)
+			set_insert(keys[i]);
+	}
+	fprintf(stderr, "done (%" PRIu64 " ticks)\n", a / (r * keys_length));
+
+	fprintf(stderr, " | Executing negative look-up test...");
+	a = 0;
+	for (j = 0; j < r; j++) {
+		s = rdtsc();
+		for (i = 0; i < keys_length; i++) {
+			set_get("\x50\x03\x04\x05\x06\x10");
+		}
+		e = rdtsc();
+		a += e - s;
+	}
+	fprintf(stderr, "done (%" PRIu64 " ticks)\n", a / (r * keys_length));
+
+	ck_epoch_record_t epoch_temporary = epoch_wr;
+	ck_epoch_synchronize(&epoch_wr);
+
+	fprintf(stderr, " '- Summary: %u pending, %u peak, %lu reclamations -> "
+	    "%u pending, %u peak, %lu reclamations\n\n",
+	    epoch_temporary.n_pending, epoch_temporary.n_peak, epoch_temporary.n_dispatch,
+	    epoch_wr.n_pending, epoch_wr.n_peak, epoch_wr.n_dispatch);
+
+	fprintf(stderr, " ,- READER CONCURRENCY\n");
+	fprintf(stderr, " | Executing reader test...");
+
+	ck_pr_store_int(&state, HS_STATE_GET);
+	while (ck_pr_load_int(&barrier[HS_STATE_STOP]) != n_threads)
+		ck_pr_stall();
+	ck_pr_inc_int(&barrier[HS_STATE_STOP]);
+	common_sleep(r);
+	ck_pr_store_int(&state, HS_STATE_STRICT_REPLACEMENT);
+	while (ck_pr_load_int(&barrier[HS_STATE_GET]) != n_threads)
+		ck_pr_stall();
+
+	fprintf(stderr, "done (reader = %" PRIu64 " ticks)\n",
+	    acc(HS_STATE_GET) / n_threads);
+
+	fprintf(stderr, " | Executing strict replacement test...");
+
+	a = repeated = 0;
+	common_alarm(alarm_handler, &alarm_event, r);
+
+	ck_pr_inc_int(&barrier[HS_STATE_GET]);
+	for (;;) {
+		repeated++;
+		s = rdtsc();
+		for (i = 0; i < keys_length; i++) {
+			if (i & 1) {
+				set_replace(keys[i]);
+			} else {
+				set_swap(keys[i]);
+			}
+		}
+		e = rdtsc();
+		a += e - s;
+
+		if (next_stage == true) {
+			next_stage = false;
+			break;
+		}
+	}
+
+	ck_pr_store_int(&state, HS_STATE_DELETION);
+	while (ck_pr_load_int(&barrier[HS_STATE_STRICT_REPLACEMENT]) != n_threads)
+		ck_pr_stall();
+	set_reset();
+	ck_epoch_synchronize(&epoch_wr);
+	fprintf(stderr, "done (writer = %" PRIu64 " ticks, reader = %" PRIu64 " ticks)\n",
+	    a / (repeated * keys_length), acc(HS_STATE_STRICT_REPLACEMENT) / n_threads);
+
+	common_alarm(alarm_handler, &alarm_event, r);
+
+	fprintf(stderr, " | Executing deletion test (%.2f)...", p_d * 100);
+	a = repeated = 0;
+	ck_pr_inc_int(&barrier[HS_STATE_STRICT_REPLACEMENT]);
+	for (;;) {
+		double delete;
+
+		repeated++;
+		s = rdtsc();
+		for (i = 0; i < keys_length; i++) {
+			set_insert(keys[i]);
+			if (p_d != 0.0) {
+				delete = common_drand48();
+				if (delete <= p_d)
+					set_remove(keys[i]);
+			}
+		}
+		e = rdtsc();
+		a += e - s;
+
+		if (next_stage == true) {
+			next_stage = false;
+			break;
+		}
+	}
+	ck_pr_store_int(&state, HS_STATE_REPLACEMENT);
+	while (ck_pr_load_int(&barrier[HS_STATE_DELETION]) != n_threads)
+		ck_pr_stall();
+
+	set_reset();
+	ck_epoch_synchronize(&epoch_wr);
+	fprintf(stderr, "done (writer = %" PRIu64 " ticks, reader = %" PRIu64 " ticks)\n",
+	    a / (repeated * keys_length), acc(HS_STATE_DELETION) / n_threads);
+
+	common_alarm(alarm_handler, &alarm_event, r);
+
+	fprintf(stderr, " | Executing replacement test (%.2f)...", p_r * 100);
+	a = repeated = 0;
+	ck_pr_inc_int(&barrier[HS_STATE_DELETION]);
+	for (;;) {
+		double delete, replace;
+
+		repeated++;
+		s = rdtsc();
+		for (i = 0; i < keys_length; i++) {
+			set_insert(keys[i]);
+			if (p_d != 0.0) {
+				delete = common_drand48();
+				if (delete <= p_d)
+					set_remove(keys[i]);
+			} else {
+				delete = 0.0;
+			}
+
+			if (p_r != 0.0) {
+				replace = common_drand48();
+				if (replace <= p_r) {
+					if ((i & 1) || (delete <= p_d)) {
+						set_replace(keys[i]);
+					} else {
+						set_swap(keys[i]);
+					}
+				}
+			}
+		}
+		e = rdtsc();
+		a += e - s;
+
+		if (next_stage == true) {
+			next_stage = false;
+			break;
+		}
+	}
+	ck_pr_store_int(&state, HS_STATE_STOP);
+	while (ck_pr_load_int(&barrier[HS_STATE_REPLACEMENT]) != n_threads)
+		ck_pr_stall();
+	set_reset();
+	ck_epoch_synchronize(&epoch_wr);
+	fprintf(stderr, "done (writer = %" PRIu64 " ticks, reader = %" PRIu64 " ticks)\n",
+	    a / (repeated * keys_length), acc(HS_STATE_REPLACEMENT) / n_threads);
+
+	ck_pr_inc_int(&barrier[HS_STATE_REPLACEMENT]);
+	epoch_temporary = epoch_wr;
+	ck_epoch_synchronize(&epoch_wr);
+
+	fprintf(stderr, " '- Summary: %u pending, %u peak, %lu reclamations -> "
+	    "%u pending, %u peak, %lu reclamations\n\n",
+	    epoch_temporary.n_pending, epoch_temporary.n_peak, epoch_temporary.n_dispatch,
+	    epoch_wr.n_pending, epoch_wr.n_peak, epoch_wr.n_dispatch);
+	return 0;
+}
+