test/mcslock: use compiler atomics for lcores sync
authorJoyce Kong <joyce.kong@arm.com>
Tue, 20 Jul 2021 03:51:21 +0000 (22:51 -0500)
committerThomas Monjalon <thomas@monjalon.net>
Fri, 30 Jul 2021 17:10:58 +0000 (19:10 +0200)
Convert rte_atomic usages to compiler atomic built-ins for lcores
sync in mcslock testcases.

Signed-off-by: Joyce Kong <joyce.kong@arm.com>
Reviewed-by: Ruifeng Wang <ruifeng.wang@arm.com>
Acked-by: Stephen Hemminger <stephen@networkplumber.org>
app/test/test_mcslock.c

index 80eaecc..52e45e7 100644 (file)
@@ -17,7 +17,6 @@
 #include <rte_lcore.h>
 #include <rte_cycles.h>
 #include <rte_mcslock.h>
-#include <rte_atomic.h>
 
 #include "test.h"
 
@@ -43,7 +42,7 @@ rte_mcslock_t *p_ml_perf;
 
 static unsigned int count;
 
-static rte_atomic32_t synchro;
+static uint32_t synchro;
 
 static int
 test_mcslock_per_core(__rte_unused void *arg)
@@ -76,8 +75,7 @@ load_loop_fn(void *func_param)
        rte_mcslock_t ml_perf_me;
 
        /* wait synchro */
-       while (rte_atomic32_read(&synchro) == 0)
-               ;
+       rte_wait_until_equal_32(&synchro, 1, __ATOMIC_RELAXED);
 
        begin = rte_get_timer_cycles();
        while (lcount < MAX_LOOP) {
@@ -102,15 +100,15 @@ test_mcslock_perf(void)
        const unsigned int lcore = rte_lcore_id();
 
        printf("\nTest with no lock on single core...\n");
-       rte_atomic32_set(&synchro, 1);
+       __atomic_store_n(&synchro, 1, __ATOMIC_RELAXED);
        load_loop_fn(&lock);
        printf("Core [%u] Cost Time = %"PRIu64" us\n",
                        lcore, time_count[lcore]);
        memset(time_count, 0, sizeof(time_count));
 
        printf("\nTest with lock on single core...\n");
+       __atomic_store_n(&synchro, 1, __ATOMIC_RELAXED);
        lock = 1;
-       rte_atomic32_set(&synchro, 1);
        load_loop_fn(&lock);
        printf("Core [%u] Cost Time = %"PRIu64" us\n",
                        lcore, time_count[lcore]);
@@ -118,11 +116,11 @@ test_mcslock_perf(void)
 
        printf("\nTest with lock on %u cores...\n", (rte_lcore_count()));
 
-       rte_atomic32_set(&synchro, 0);
+       __atomic_store_n(&synchro, 0, __ATOMIC_RELAXED);
        rte_eal_mp_remote_launch(load_loop_fn, &lock, SKIP_MAIN);
 
        /* start synchro and launch test on main */
-       rte_atomic32_set(&synchro, 1);
+       __atomic_store_n(&synchro, 1, __ATOMIC_RELAXED);
        load_loop_fn(&lock);
 
        rte_eal_mp_wait_lcore();