numam-dpdk/app/test/test_hash_scaling.c
Pablo de Lara 5915699153 hash: fix scaling by reducing contention
If using multiple cores on a system with hardware transactional
memory support, thread scaling does not work, as there was a single
point in the hash library which is a bottleneck for all threads,
which is the "free_slots" ring, which stores all the indices of
the free slots in the table.

This patch fixes the problem, by creating a local cache per logical core,
which stores locally indices of free slots,
so most times, writer threads will not interfere each other.

Fixes: 48a3991196 ("hash: replace with cuckoo hash implementation")

Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
2015-11-04 01:04:10 +01:00

224 lines
6.3 KiB
C

/*-
* BSD LICENSE
*
* Copyright(c) 2015 Intel Corporation. All rights reserved.
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
*
* * Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* * 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.
* * Neither the name of Intel Corporation nor the names of its
* contributors may be used to endorse or promote products derived
* from this software without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS 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 COPYRIGHT
* OWNER 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 <rte_cycles.h>
#include <rte_hash.h>
#include <rte_hash_crc.h>
#include <rte_spinlock.h>
#include <rte_launch.h>
#include "test.h"
/*
* Check condition and return an error if true. Assumes that "handle" is the
* name of the hash structure pointer to be freed.
*/
#define RETURN_IF_ERROR(cond, str, ...) do { \
if (cond) { \
printf("ERROR line %d: " str "\n", __LINE__, \
##__VA_ARGS__); \
if (handle) \
rte_hash_free(handle); \
return -1; \
} \
} while (0)
enum locking_mode_t {
NORMAL_LOCK,
LOCK_ELISION,
NULL_LOCK
};
struct {
uint32_t num_iterations;
struct rte_hash *h;
rte_spinlock_t *lock;
int locking_mode;
} tbl_scaling_test_params;
static rte_atomic64_t gcycles;
static int test_hash_scaling_worker(__attribute__((unused)) void *arg)
{
uint64_t i, key;
uint32_t thr_id = rte_sys_gettid();
uint64_t begin, cycles = 0;
switch (tbl_scaling_test_params.locking_mode) {
case NORMAL_LOCK:
for (i = 0; i < tbl_scaling_test_params.num_iterations; i++) {
/* different threads get different keys because
we use the thread-id in the key computation
*/
key = rte_hash_crc(&i, sizeof(i), thr_id);
begin = rte_rdtsc_precise();
rte_spinlock_lock(tbl_scaling_test_params.lock);
rte_hash_add_key(tbl_scaling_test_params.h, &key);
rte_spinlock_unlock(tbl_scaling_test_params.lock);
cycles += rte_rdtsc_precise() - begin;
}
break;
case LOCK_ELISION:
for (i = 0; i < tbl_scaling_test_params.num_iterations; i++) {
key = rte_hash_crc(&i, sizeof(i), thr_id);
begin = rte_rdtsc_precise();
rte_spinlock_lock_tm(tbl_scaling_test_params.lock);
rte_hash_add_key(tbl_scaling_test_params.h, &key);
rte_spinlock_unlock_tm(tbl_scaling_test_params.lock);
cycles += rte_rdtsc_precise() - begin;
}
break;
default:
for (i = 0; i < tbl_scaling_test_params.num_iterations; i++) {
key = rte_hash_crc(&i, sizeof(i), thr_id);
begin = rte_rdtsc_precise();
rte_hash_add_key(tbl_scaling_test_params.h, &key);
cycles += rte_rdtsc_precise() - begin;
}
}
rte_atomic64_add(&gcycles, cycles);
return 0;
}
/*
* Do scalability perf tests.
*/
static int
test_hash_scaling(int locking_mode)
{
static unsigned calledCount = 1;
uint32_t num_iterations = 1024*1024;
uint64_t i, key;
struct rte_hash_parameters hash_params = {
.entries = num_iterations*2,
.key_len = sizeof(key),
.hash_func = rte_hash_crc,
.hash_func_init_val = 0,
.socket_id = rte_socket_id(),
.extra_flag = RTE_HASH_EXTRA_FLAGS_TRANS_MEM_SUPPORT
};
struct rte_hash *handle;
char name[RTE_HASH_NAMESIZE];
rte_spinlock_t lock;
rte_spinlock_init(&lock);
snprintf(name, 32, "test%u", calledCount++);
hash_params.name = name;
handle = rte_hash_create(&hash_params);
RETURN_IF_ERROR(handle == NULL, "hash creation failed");
tbl_scaling_test_params.num_iterations =
num_iterations/rte_lcore_count();
tbl_scaling_test_params.h = handle;
tbl_scaling_test_params.lock = &lock;
tbl_scaling_test_params.locking_mode = locking_mode;
rte_atomic64_init(&gcycles);
rte_atomic64_clear(&gcycles);
/* fill up to initial size */
for (i = 0; i < num_iterations; i++) {
key = rte_hash_crc(&i, sizeof(i), 0xabcdabcd);
rte_hash_add_key(tbl_scaling_test_params.h, &key);
}
rte_eal_mp_remote_launch(test_hash_scaling_worker, NULL, CALL_MASTER);
rte_eal_mp_wait_lcore();
unsigned long long int cycles_per_operation =
rte_atomic64_read(&gcycles)/
(tbl_scaling_test_params.num_iterations*rte_lcore_count());
const char *lock_name;
switch (locking_mode) {
case NORMAL_LOCK:
lock_name = "normal spinlock";
break;
case LOCK_ELISION:
lock_name = "lock elision";
break;
default:
lock_name = "null lock";
}
printf("--------------------------------------------------------\n");
printf("Cores: %d; %s mode -> cycles per operation: %llu\n",
rte_lcore_count(), lock_name, cycles_per_operation);
printf("--------------------------------------------------------\n");
/* CSV output */
printf(">>>%d,%s,%llu\n", rte_lcore_count(), lock_name,
cycles_per_operation);
rte_hash_free(handle);
return 0;
}
static int
test_hash_scaling_main(void)
{
int r = 0;
if (rte_lcore_count() == 1)
r = test_hash_scaling(NULL_LOCK);
if (r == 0)
r = test_hash_scaling(NORMAL_LOCK);
if (!rte_tm_supported()) {
printf("Hardware transactional memory (lock elision) is NOT supported\n");
return r;
}
printf("Hardware transactional memory (lock elision) is supported\n");
if (r == 0)
r = test_hash_scaling(LOCK_ELISION);
return r;
}
static struct test_command hash_scaling_cmd = {
.command = "hash_scaling_autotest",
.callback = test_hash_scaling_main,
};
REGISTER_TEST_COMMAND(hash_scaling_cmd);