#define NUM_TEST 3
unsigned int core_cnt[NUM_TEST] = {2, 4, 8};
-unsigned int slave_core_ids[RTE_MAX_LCORE];
+unsigned int worker_core_ids[RTE_MAX_LCORE];
struct perf {
uint32_t single_read;
uint32_t single_write;
struct rte_hash *h;
} tbl_rw_test_param;
-static rte_atomic64_t gcycles;
-static rte_atomic64_t ginsertions;
+static uint64_t gcycles;
+static uint64_t ginsertions;
-static rte_atomic64_t gread_cycles;
-static rte_atomic64_t gwrite_cycles;
+static uint64_t gread_cycles;
+static uint64_t gwrite_cycles;
-static rte_atomic64_t greads;
-static rte_atomic64_t gwrites;
+static uint64_t greads;
+static uint64_t gwrites;
static int
-test_hash_readwrite_worker(__attribute__((unused)) void *arg)
+test_hash_readwrite_worker(__rte_unused void *arg)
{
uint64_t i, offset;
uint32_t lcore_id = rte_lcore_id();
ret = rte_malloc(NULL, sizeof(int) *
tbl_rw_test_param.num_insert, 0);
for (i = 0; i < rte_lcore_count(); i++) {
- if (slave_core_ids[i] == lcore_id)
+ if (worker_core_ids[i] == lcore_id)
break;
}
offset = tbl_rw_test_param.num_insert * i;
}
cycles = rte_rdtsc_precise() - begin;
- rte_atomic64_add(&gcycles, cycles);
- rte_atomic64_add(&ginsertions, i - offset);
+ __atomic_fetch_add(&gcycles, cycles, __ATOMIC_RELAXED);
+ __atomic_fetch_add(&ginsertions, i - offset, __ATOMIC_RELAXED);
for (; i < offset + tbl_rw_test_param.num_insert; i++)
tbl_rw_test_param.keys[i] = RTE_RWTEST_FAIL;
}
static int
-init_params(int use_ext, int use_htm, int use_jhash)
+init_params(int use_ext, int use_htm, int rw_lf, int use_jhash)
{
unsigned int i;
else
hash_params.hash_func = rte_hash_crc;
+ hash_params.extra_flag = RTE_HASH_EXTRA_FLAGS_MULTI_WRITER_ADD;
if (use_htm)
- hash_params.extra_flag =
- RTE_HASH_EXTRA_FLAGS_TRANS_MEM_SUPPORT |
- RTE_HASH_EXTRA_FLAGS_RW_CONCURRENCY |
- RTE_HASH_EXTRA_FLAGS_MULTI_WRITER_ADD;
+ hash_params.extra_flag |=
+ RTE_HASH_EXTRA_FLAGS_TRANS_MEM_SUPPORT;
+ if (rw_lf)
+ hash_params.extra_flag |=
+ RTE_HASH_EXTRA_FLAGS_RW_CONCURRENCY_LF;
else
- hash_params.extra_flag =
- RTE_HASH_EXTRA_FLAGS_RW_CONCURRENCY |
- RTE_HASH_EXTRA_FLAGS_MULTI_WRITER_ADD;
+ hash_params.extra_flag |=
+ RTE_HASH_EXTRA_FLAGS_RW_CONCURRENCY;
if (use_ext)
hash_params.extra_flag |=
}
static int
-test_hash_readwrite_functional(int use_ext, int use_htm)
+test_hash_readwrite_functional(int use_htm, int use_rw_lf, int use_ext)
{
unsigned int i;
const void *next_key;
uint32_t duplicated_keys = 0;
uint32_t lost_keys = 0;
int use_jhash = 1;
- int slave_cnt = rte_lcore_count() - 1;
+ int worker_cnt = rte_lcore_count() - 1;
uint32_t tot_insert = 0;
- rte_atomic64_init(&gcycles);
- rte_atomic64_clear(&gcycles);
+ __atomic_store_n(&gcycles, 0, __ATOMIC_RELAXED);
+ __atomic_store_n(&ginsertions, 0, __ATOMIC_RELAXED);
- rte_atomic64_init(&ginsertions);
- rte_atomic64_clear(&ginsertions);
-
- if (init_params(use_ext, use_htm, use_jhash) != 0)
+ if (init_params(use_ext, use_htm, use_rw_lf, use_jhash) != 0)
goto err;
if (use_ext)
tot_insert = TOTAL_INSERT;
tbl_rw_test_param.num_insert =
- tot_insert / slave_cnt;
+ tot_insert / worker_cnt;
tbl_rw_test_param.rounded_tot_insert =
- tbl_rw_test_param.num_insert
- * slave_cnt;
+ tbl_rw_test_param.num_insert * worker_cnt;
+ printf("\nHTM = %d, RW-LF = %d, EXT-Table = %d\n",
+ use_htm, use_rw_lf, use_ext);
printf("++++++++Start function tests:+++++++++\n");
/* Fire all threads. */
rte_eal_mp_remote_launch(test_hash_readwrite_worker,
- NULL, SKIP_MASTER);
+ NULL, SKIP_MAIN);
rte_eal_mp_wait_lcore();
while (rte_hash_iterate(tbl_rw_test_param.h, &next_key,
printf("No key corrupted during read-write test.\n");
unsigned long long int cycles_per_insertion =
- rte_atomic64_read(&gcycles) /
- rte_atomic64_read(&ginsertions);
+ __atomic_load_n(&gcycles, __ATOMIC_RELAXED) /
+ __atomic_load_n(&ginsertions, __ATOMIC_RELAXED);
printf("cycles per insertion and lookup: %llu\n", cycles_per_insertion);
}
cycles = rte_rdtsc_precise() - begin;
- rte_atomic64_add(&gread_cycles, cycles);
- rte_atomic64_add(&greads, i);
+ __atomic_fetch_add(&gread_cycles, cycles, __ATOMIC_RELAXED);
+ __atomic_fetch_add(&greads, i, __ATOMIC_RELAXED);
return 0;
}
uint64_t offset;
for (i = 0; i < rte_lcore_count(); i++) {
- if (slave_core_ids[i] == lcore_id)
+ if (worker_core_ids[i] == lcore_id)
break;
}
}
cycles = rte_rdtsc_precise() - begin;
- rte_atomic64_add(&gwrite_cycles, cycles);
- rte_atomic64_add(&gwrites, tbl_rw_test_param.num_insert);
+ __atomic_fetch_add(&gwrite_cycles, cycles, __ATOMIC_RELAXED);
+ __atomic_fetch_add(&gwrites, tbl_rw_test_param.num_insert,
+ __ATOMIC_RELAXED);
return 0;
}
uint64_t start = 0, end = 0;
- rte_atomic64_init(&greads);
- rte_atomic64_init(&gwrites);
- rte_atomic64_clear(&gwrites);
- rte_atomic64_clear(&greads);
+ __atomic_store_n(&gwrites, 0, __ATOMIC_RELAXED);
+ __atomic_store_n(&greads, 0, __ATOMIC_RELAXED);
- rte_atomic64_init(&gread_cycles);
- rte_atomic64_clear(&gread_cycles);
- rte_atomic64_init(&gwrite_cycles);
- rte_atomic64_clear(&gwrite_cycles);
+ __atomic_store_n(&gread_cycles, 0, __ATOMIC_RELAXED);
+ __atomic_store_n(&gwrite_cycles, 0, __ATOMIC_RELAXED);
- if (init_params(0, use_htm, use_jhash) != 0)
+ if (init_params(0, use_htm, 0, use_jhash) != 0)
goto err;
/*
perf_results->single_read = end / i;
for (n = 0; n < NUM_TEST; n++) {
- unsigned int tot_slave_lcore = rte_lcore_count() - 1;
- if (tot_slave_lcore < core_cnt[n] * 2)
+ unsigned int tot_worker_lcore = rte_lcore_count() - 1;
+ if (tot_worker_lcore < core_cnt[n] * 2)
goto finish;
- rte_atomic64_clear(&greads);
- rte_atomic64_clear(&gread_cycles);
- rte_atomic64_clear(&gwrites);
- rte_atomic64_clear(&gwrite_cycles);
+ __atomic_store_n(&greads, 0, __ATOMIC_RELAXED);
+ __atomic_store_n(&gread_cycles, 0, __ATOMIC_RELAXED);
+ __atomic_store_n(&gwrites, 0, __ATOMIC_RELAXED);
+ __atomic_store_n(&gwrite_cycles, 0, __ATOMIC_RELAXED);
rte_hash_reset(tbl_rw_test_param.h);
for (i = 0; i < core_cnt[n]; i++)
rte_eal_remote_launch(test_rw_reader,
(void *)(uintptr_t)read_cnt,
- slave_core_ids[i]);
+ worker_core_ids[i]);
rte_eal_mp_wait_lcore();
for (; i < core_cnt[n] * 2; i++)
rte_eal_remote_launch(test_rw_writer,
(void *)((uintptr_t)start_coreid),
- slave_core_ids[i]);
+ worker_core_ids[i]);
rte_eal_mp_wait_lcore();
if (reader_faster) {
unsigned long long int cycles_per_insertion =
- rte_atomic64_read(&gread_cycles) /
- rte_atomic64_read(&greads);
+ __atomic_load_n(&gread_cycles, __ATOMIC_RELAXED) /
+ __atomic_load_n(&greads, __ATOMIC_RELAXED);
perf_results->read_only[n] = cycles_per_insertion;
printf("Reader only: cycles per lookup: %llu\n",
cycles_per_insertion);
else {
unsigned long long int cycles_per_insertion =
- rte_atomic64_read(&gwrite_cycles) /
- rte_atomic64_read(&gwrites);
+ __atomic_load_n(&gwrite_cycles, __ATOMIC_RELAXED) /
+ __atomic_load_n(&gwrites, __ATOMIC_RELAXED);
perf_results->write_only[n] = cycles_per_insertion;
printf("Writer only: cycles per writes: %llu\n",
cycles_per_insertion);
}
- rte_atomic64_clear(&greads);
- rte_atomic64_clear(&gread_cycles);
- rte_atomic64_clear(&gwrites);
- rte_atomic64_clear(&gwrite_cycles);
+ __atomic_store_n(&greads, 0, __ATOMIC_RELAXED);
+ __atomic_store_n(&gread_cycles, 0, __ATOMIC_RELAXED);
+ __atomic_store_n(&gwrites, 0, __ATOMIC_RELAXED);
+ __atomic_store_n(&gwrite_cycles, 0, __ATOMIC_RELAXED);
rte_hash_reset(tbl_rw_test_param.h);
for (i = core_cnt[n]; i < core_cnt[n] * 2; i++)
rte_eal_remote_launch(test_rw_writer,
(void *)((uintptr_t)start_coreid),
- slave_core_ids[i]);
+ worker_core_ids[i]);
for (i = 0; i < core_cnt[n]; i++)
rte_eal_remote_launch(test_rw_reader,
(void *)(uintptr_t)read_cnt,
- slave_core_ids[i]);
+ worker_core_ids[i]);
} else {
for (i = 0; i < core_cnt[n]; i++)
rte_eal_remote_launch(test_rw_reader,
(void *)(uintptr_t)read_cnt,
- slave_core_ids[i]);
+ worker_core_ids[i]);
for (; i < core_cnt[n] * 2; i++)
rte_eal_remote_launch(test_rw_writer,
(void *)((uintptr_t)start_coreid),
- slave_core_ids[i]);
+ worker_core_ids[i]);
}
rte_eal_mp_wait_lcore();
if (reader_faster) {
unsigned long long int cycles_per_insertion =
- rte_atomic64_read(&gread_cycles) /
- rte_atomic64_read(&greads);
+ __atomic_load_n(&gread_cycles, __ATOMIC_RELAXED) /
+ __atomic_load_n(&greads, __ATOMIC_RELAXED);
perf_results->read_write_r[n] = cycles_per_insertion;
printf("Read-write cycles per lookup: %llu\n",
cycles_per_insertion);
else {
unsigned long long int cycles_per_insertion =
- rte_atomic64_read(&gwrite_cycles) /
- rte_atomic64_read(&gwrites);
+ __atomic_load_n(&gwrite_cycles, __ATOMIC_RELAXED) /
+ __atomic_load_n(&gwrites, __ATOMIC_RELAXED);
perf_results->read_write_w[n] = cycles_per_insertion;
printf("Read-write cycles per writes: %llu\n",
cycles_per_insertion);
return TEST_SKIPPED;
}
- RTE_LCORE_FOREACH_SLAVE(core_id) {
- slave_core_ids[i] = core_id;
+ RTE_LCORE_FOREACH_WORKER(core_id) {
+ worker_core_ids[i] = core_id;
i++;
}
printf("Results summary:\n");
printf("================\n");
- printf("single read: %u\n", htm_results.single_read);
- printf("single write: %u\n", htm_results.single_write);
+ printf("HTM:\n");
+ printf(" single read: %u\n", htm_results.single_read);
+ printf(" single write: %u\n", htm_results.single_write);
+ printf("non HTM:\n");
+ printf(" single read: %u\n", non_htm_results.single_read);
+ printf(" single write: %u\n", non_htm_results.single_write);
for (i = 0; i < NUM_TEST; i++) {
printf("+++ core_cnt: %u +++\n", core_cnt[i]);
printf("HTM:\n");
* than writer threads. This is to timing either reader threads or
* writer threads for performance numbers.
*/
- int use_htm, use_ext;
unsigned int i = 0, core_id = 0;
if (rte_lcore_count() < 3) {
return TEST_SKIPPED;
}
- RTE_LCORE_FOREACH_SLAVE(core_id) {
- slave_core_ids[i] = core_id;
+ RTE_LCORE_FOREACH_WORKER(core_id) {
+ worker_core_ids[i] = core_id;
i++;
}
printf("Test read-write with Hardware transactional memory\n");
- use_htm = 1;
- use_ext = 0;
+ /* htm = 1, rw_lf = 0, ext = 0 */
+ if (test_hash_readwrite_functional(1, 0, 0) < 0)
+ return -1;
- if (test_hash_readwrite_functional(use_ext, use_htm) < 0)
+ /* htm = 1, rw_lf = 1, ext = 0 */
+ if (test_hash_readwrite_functional(1, 1, 0) < 0)
return -1;
- use_ext = 1;
- if (test_hash_readwrite_functional(use_ext, use_htm) < 0)
+ /* htm = 1, rw_lf = 0, ext = 1 */
+ if (test_hash_readwrite_functional(1, 0, 1) < 0)
return -1;
+ /* htm = 1, rw_lf = 1, ext = 1 */
+ if (test_hash_readwrite_functional(1, 1, 1) < 0)
+ return -1;
} else {
printf("Hardware transactional memory (lock elision) "
"is NOT supported\n");
}
printf("Test read-write without Hardware transactional memory\n");
- use_htm = 0;
- use_ext = 0;
- if (test_hash_readwrite_functional(use_ext, use_htm) < 0)
+ /* htm = 0, rw_lf = 0, ext = 0 */
+ if (test_hash_readwrite_functional(0, 0, 0) < 0)
+ return -1;
+
+ /* htm = 0, rw_lf = 1, ext = 0 */
+ if (test_hash_readwrite_functional(0, 1, 0) < 0)
+ return -1;
+
+ /* htm = 0, rw_lf = 0, ext = 1 */
+ if (test_hash_readwrite_functional(0, 0, 1) < 0)
return -1;
- use_ext = 1;
- if (test_hash_readwrite_functional(use_ext, use_htm) < 0)
+ /* htm = 0, rw_lf = 1, ext = 1 */
+ if (test_hash_readwrite_functional(0, 1, 1) < 0)
return -1;
return 0;