[dpdk-dev] [PATCH v1 0/8] use compiler atomic builtins for examples

[dpdk-dev] [PATCH v1 0/8] use compiler atomic builtins for examples

[Joyce Kong]

Since atomic operations have been adopted in DPDK now[1],
change rte_atomicNN_xxx APIs to compiler's atomic built-ins
in examples module[2].

[1] https://www.dpdk.org/blog/2021/03/26/dpdk-adopts-the-c11-memory-model/
[2] https://doc.dpdk.org/guides/rel_notes/deprecation.html

Joyce Kong (8):
  examples/bbdev_app: use compiler atomics for flag sync
  examples/multi_process: use compiler atomics for sync
  examples/kni: use compiler atomics for status sync
  examples/performance-thread: use compiler atomics for sync
  examples/l2fwd-jobstats: use compiler atomics for stats sync
  examples/vm_power_manager: use compiler atomics for sync
  examples/server_node_efd: use compiler atomics for sync
  examples: remove unnecessary include of atomic

 examples/bbdev_app/main.c                     | 13 ++++---
 examples/bond/main.c                          |  1 -
 examples/ip_fragmentation/main.c              |  1 -
 examples/ip_reassembly/main.c                 |  1 -
 examples/ipsec-secgw/ipsec-secgw.c            |  1 -
 examples/ipv4_multicast/main.c                |  1 -
 examples/kni/main.c                           | 27 +++++++--------
 examples/l2fwd-crypto/main.c                  |  1 -
 examples/l2fwd-event/l2fwd_common.h           |  1 -
 examples/l2fwd-event/l2fwd_event.c            |  1 -
 examples/l2fwd-jobstats/main.c                | 11 +++---
 examples/l2fwd-keepalive/main.c               |  1 -
 examples/l2fwd/main.c                         |  1 -
 examples/l3fwd-acl/main.c                     |  1 -
 examples/l3fwd-power/main.c                   |  1 -
 examples/l3fwd/main.c                         |  1 -
 examples/link_status_interrupt/main.c         |  1 -
 .../client_server_mp/mp_client/client.c       |  1 -
 .../client_server_mp/mp_server/init.c         |  1 -
 .../client_server_mp/mp_server/main.c         |  7 ++--
 examples/multi_process/simple_mp/main.c       |  1 -
 .../multi_process/simple_mp/mp_commands.c     |  1 -
 examples/multi_process/symmetric_mp/main.c    |  1 -
 examples/performance-thread/common/lthread.c  | 10 +++---
 .../performance-thread/common/lthread_diag.h  | 10 +++---
 .../performance-thread/common/lthread_int.h   |  1 -
 .../performance-thread/common/lthread_mutex.c | 26 +++++++-------
 .../performance-thread/common/lthread_mutex.h |  2 +-
 .../performance-thread/common/lthread_sched.c | 34 ++++++++-----------
 .../performance-thread/common/lthread_tls.c   |  5 +--
 .../performance-thread/l3fwd-thread/main.c    | 22 +++++-------
 examples/server_node_efd/node/node.c          |  1 -
 examples/server_node_efd/server/init.c        |  1 -
 examples/server_node_efd/server/main.c        |  7 ++--
 examples/vhost_blk/blk.c                      |  1 -
 examples/vhost_blk/vhost_blk.c                |  1 -
 examples/vm_power_manager/channel_manager.c   |  1 -
 examples/vm_power_manager/channel_manager.h   |  1 -
 examples/vm_power_manager/channel_monitor.c   | 11 +++---
 examples/vmdq/main.c                          |  1 -
 examples/vmdq_dcb/main.c                      |  1 -
 41 files changed, 91 insertions(+), 122 deletions(-)

-- 
2.17.1

[dpdk-dev] [PATCH v1 1/8] examples/bbdev_app: use compiler atomics for flag sync

[Joyce Kong]

Convert rte_atomic usages to compiler atomic built-ins
for global_exit_flag sync.

Signed-off-by: Joyce Kong <joyce.kong@arm.com>
Reviewed-by: Ruifeng Wang <ruifeng.wang@arm.com>
---
 examples/bbdev_app/main.c | 13 ++++++-------
 1 file changed, 6 insertions(+), 7 deletions(-)

diff --git a/examples/bbdev_app/main.c b/examples/bbdev_app/main.c
index 5251db0b16..75c620ea75 100644
--- a/examples/bbdev_app/main.c
+++ b/examples/bbdev_app/main.c
@@ -18,7 +18,6 @@
 #include <getopt.h>
 #include <signal.h>
 
-#include <rte_atomic.h>
 #include <rte_common.h>
 #include <rte_eal.h>
 #include <rte_cycles.h>
@@ -167,7 +166,7 @@ static const struct app_config_params def_app_config = {
 	.num_dec_cores = 1,
 };
 
-static rte_atomic16_t global_exit_flag;
+static uint16_t global_exit_flag;
 
 /* display usage */
 static inline void
@@ -279,7 +278,7 @@ static void
 signal_handler(int signum)
 {
 	printf("\nSignal %d received\n", signum);
-	rte_atomic16_set(&global_exit_flag, 1);
+	__atomic_store_n(&global_exit_flag, 1, __ATOMIC_RELAXED);
 }
 
 static void
@@ -328,7 +327,7 @@ check_port_link_status(uint16_t port_id)
 	fflush(stdout);
 
 	for (count = 0; count <= MAX_CHECK_TIME &&
-			!rte_atomic16_read(&global_exit_flag); count++) {
+			!__atomic_load_n(&global_exit_flag, __ATOMIC_RELAXED); count++) {
 		memset(&link, 0, sizeof(link));
 		link_get_err = rte_eth_link_get_nowait(port_id, &link);
 
@@ -682,7 +681,7 @@ stats_loop(void *arg)
 {
 	struct stats_lcore_params *stats_lcore = arg;
 
-	while (!rte_atomic16_read(&global_exit_flag)) {
+	while (!__atomic_load_n(&global_exit_flag, __ATOMIC_RELAXED)) {
 		print_stats(stats_lcore);
 		rte_delay_ms(500);
 	}
@@ -928,7 +927,7 @@ processing_loop(void *arg)
 	const bool run_decoder = (lcore_conf->core_type &
 			(1 << RTE_BBDEV_OP_TURBO_DEC));
 
-	while (!rte_atomic16_read(&global_exit_flag)) {
+	while (!__atomic_load_n(&global_exit_flag, __ATOMIC_RELAXED)) {
 		if (run_encoder)
 			run_encoding(lcore_conf);
 		if (run_decoder)
@@ -1062,7 +1061,7 @@ main(int argc, char **argv)
 		.align = __alignof__(struct rte_mbuf *),
 	};
 
-	rte_atomic16_init(&global_exit_flag);
+	__atomic_store_n(&global_exit_flag, 0, __ATOMIC_RELAXED);
 
 	sigret = signal(SIGTERM, signal_handler);
 	if (sigret == SIG_ERR)
-- 
2.17.1

[dpdk-dev] [PATCH v1 2/8] examples/multi_process: use compiler atomics for sync

[Joyce Kong]

Convert rte_atomic32_test_and_set usage to compiler atomic
CAS operation for display_stats sync.

Signed-off-by: Joyce Kong <joyce.kong@arm.com>
Reviewed-by: Ruifeng Wang <ruifeng.wang@arm.com>
---
 examples/multi_process/client_server_mp/mp_server/main.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/examples/multi_process/client_server_mp/mp_server/main.c b/examples/multi_process/client_server_mp/mp_server/main.c
index 9bcee460fd..b4761ebc7b 100644
--- a/examples/multi_process/client_server_mp/mp_server/main.c
+++ b/examples/multi_process/client_server_mp/mp_server/main.c
@@ -20,7 +20,6 @@
 #include <rte_per_lcore.h>
 #include <rte_lcore.h>
 #include <rte_branch_prediction.h>
-#include <rte_atomic.h>
 #include <rte_ring.h>
 #include <rte_log.h>
 #include <rte_debug.h>
@@ -158,10 +157,12 @@ static int
 sleep_lcore(__rte_unused void *dummy)
 {
 	/* Used to pick a display thread - static, so zero-initialised */
-	static rte_atomic32_t display_stats;
+	static uint32_t display_stats;
 
+	uint32_t status = 0;
 	/* Only one core should display stats */
-	if (rte_atomic32_test_and_set(&display_stats)) {
+	if (__atomic_compare_exchange_n(&display_stats, &status, 1, 0,
+			__ATOMIC_RELAXED, __ATOMIC_RELAXED)) {
 		const unsigned sleeptime = 1;
 		printf("Core %u displaying statistics\n", rte_lcore_id());
 
-- 
2.17.1

[dpdk-dev] [PATCH v1 3/8] examples/kni: use compiler atomics for status sync

[Joyce Kong]

Convert rte_atomic usages to compiler atomic builit-ins
for kni_stop and kni_pause sync.

Signed-off-by: Joyce Kong <joyce.kong@arm.com>
Reviewed-by: Ruifeng Wang <ruifeng.wang@arm.com>
---
 examples/kni/main.c | 27 +++++++++++++--------------
 1 file changed, 13 insertions(+), 14 deletions(-)

diff --git a/examples/kni/main.c b/examples/kni/main.c
index beabb3c848..ad1f569e18 100644
--- a/examples/kni/main.c
+++ b/examples/kni/main.c
@@ -27,7 +27,6 @@
 #include <rte_eal.h>
 #include <rte_per_lcore.h>
 #include <rte_launch.h>
-#include <rte_atomic.h>
 #include <rte_lcore.h>
 #include <rte_branch_prediction.h>
 #include <rte_interrupts.h>
@@ -131,8 +130,8 @@ static int kni_change_mtu(uint16_t port_id, unsigned int new_mtu);
 static int kni_config_network_interface(uint16_t port_id, uint8_t if_up);
 static int kni_config_mac_address(uint16_t port_id, uint8_t mac_addr[]);
 
-static rte_atomic32_t kni_stop = RTE_ATOMIC32_INIT(0);
-static rte_atomic32_t kni_pause = RTE_ATOMIC32_INIT(0);
+static uint32_t kni_stop;
+static uint32_t kni_pause;
 
 /* Print out statistics on packets handled */
 static void
@@ -185,7 +184,7 @@ signal_handler(int signum)
 	if (signum == SIGRTMIN || signum == SIGINT || signum == SIGTERM) {
 		printf("\nSIGRTMIN/SIGINT/SIGTERM received. "
 			"KNI processing stopping.\n");
-		rte_atomic32_inc(&kni_stop);
+		__atomic_fetch_add(&kni_stop, 1, __ATOMIC_RELAXED);
 		return;
         }
 }
@@ -311,8 +310,8 @@ main_loop(__rte_unused void *arg)
 					kni_port_params_array[i]->lcore_rx,
 					kni_port_params_array[i]->port_id);
 		while (1) {
-			f_stop = rte_atomic32_read(&kni_stop);
-			f_pause = rte_atomic32_read(&kni_pause);
+			f_stop = __atomic_load_n(&kni_stop, __ATOMIC_RELAXED);
+			f_pause = __atomic_load_n(&kni_pause, __ATOMIC_RELAXED);
 			if (f_stop)
 				break;
 			if (f_pause)
@@ -324,8 +323,8 @@ main_loop(__rte_unused void *arg)
 					kni_port_params_array[i]->lcore_tx,
 					kni_port_params_array[i]->port_id);
 		while (1) {
-			f_stop = rte_atomic32_read(&kni_stop);
-			f_pause = rte_atomic32_read(&kni_pause);
+			f_stop = __atomic_load_n(&kni_stop, __ATOMIC_RELAXED);
+			f_pause = __atomic_load_n(&kni_pause, __ATOMIC_RELAXED);
 			if (f_stop)
 				break;
 			if (f_pause)
@@ -856,9 +855,9 @@ kni_change_mtu(uint16_t port_id, unsigned int new_mtu)
 {
 	int ret;
 
-	rte_atomic32_inc(&kni_pause);
+	__atomic_fetch_add(&kni_pause, 1, __ATOMIC_RELAXED);
 	ret =  kni_change_mtu_(port_id, new_mtu);
-	rte_atomic32_dec(&kni_pause);
+	__atomic_fetch_sub(&kni_pause, 1, __ATOMIC_RELAXED);
 
 	return ret;
 }
@@ -877,14 +876,14 @@ kni_config_network_interface(uint16_t port_id, uint8_t if_up)
 	RTE_LOG(INFO, APP, "Configure network interface of %d %s\n",
 					port_id, if_up ? "up" : "down");
 
-	rte_atomic32_inc(&kni_pause);
+	__atomic_fetch_add(&kni_pause, 1, __ATOMIC_RELAXED);
 
 	if (if_up != 0) { /* Configure network interface up */
 		ret = rte_eth_dev_stop(port_id);
 		if (ret != 0) {
 			RTE_LOG(ERR, APP, "Failed to stop port %d: %s\n",
 				port_id, rte_strerror(-ret));
-			rte_atomic32_dec(&kni_pause);
+			__atomic_fetch_sub(&kni_pause, 1, __ATOMIC_RELAXED);
 			return ret;
 		}
 		ret = rte_eth_dev_start(port_id);
@@ -893,12 +892,12 @@ kni_config_network_interface(uint16_t port_id, uint8_t if_up)
 		if (ret != 0) {
 			RTE_LOG(ERR, APP, "Failed to stop port %d: %s\n",
 				port_id, rte_strerror(-ret));
-			rte_atomic32_dec(&kni_pause);
+			__atomic_fetch_sub(&kni_pause, 1, __ATOMIC_RELAXED);
 			return ret;
 		}
 	}
 
-	rte_atomic32_dec(&kni_pause);
+	__atomic_fetch_sub(&kni_pause, 1, __ATOMIC_RELAXED);
 
 	if (ret < 0)
 		RTE_LOG(ERR, APP, "Failed to start port %d\n", port_id);
-- 
2.17.1

[dpdk-dev] [PATCH v1 4/8] examples/performance-thread: use compiler atomics for sync

[Joyce Kong]

Convert rte_atomic usages to compiler atomic built-ins
for thread sync.

Signed-off-by: Joyce Kong <joyce.kong@arm.com>
Reviewed-by: Ruifeng Wang <ruifeng.wang@arm.com>
---
 examples/performance-thread/common/lthread.c  | 10 +++---
 .../performance-thread/common/lthread_diag.h  | 10 +++---
 .../performance-thread/common/lthread_int.h   |  1 -
 .../performance-thread/common/lthread_mutex.c | 26 +++++++-------
 .../performance-thread/common/lthread_mutex.h |  2 +-
 .../performance-thread/common/lthread_sched.c | 34 ++++++++-----------
 .../performance-thread/common/lthread_tls.c   |  5 +--
 .../performance-thread/l3fwd-thread/main.c    | 22 +++++-------
 8 files changed, 53 insertions(+), 57 deletions(-)

diff --git a/examples/performance-thread/common/lthread.c b/examples/performance-thread/common/lthread.c
index 3f1f48db43..98123f34f8 100644
--- a/examples/performance-thread/common/lthread.c
+++ b/examples/performance-thread/common/lthread.c
@@ -357,9 +357,10 @@ void lthread_exit(void *ptr)
 	 *  - if exit before join then we suspend and resume on join
 	 *  - if join before exit then we resume the joining thread
 	 */
+	uint64_t join_initial = LT_JOIN_INITIAL;
 	if ((lt->join == LT_JOIN_INITIAL)
-	    && rte_atomic64_cmpset(&lt->join, LT_JOIN_INITIAL,
-				   LT_JOIN_EXITING)) {
+	    && __atomic_compare_exchange_n(&lt->join, &join_initial,
+		LT_JOIN_EXITING, 0, __ATOMIC_RELAXED, __ATOMIC_RELAXED)) {
 
 		DIAG_EVENT(lt, LT_DIAG_LTHREAD_EXIT, 1, 0);
 		_suspend();
@@ -415,9 +416,10 @@ int lthread_join(struct lthread *lt, void **ptr)
 	 *  - if join before exit we suspend and will resume when exit is called
 	 *  - if exit before join we resume the exiting thread
 	 */
+	uint64_t join_initial = LT_JOIN_INITIAL;
 	if ((lt->join == LT_JOIN_INITIAL)
-	    && rte_atomic64_cmpset(&lt->join, LT_JOIN_INITIAL,
-				   LT_JOIN_THREAD_SET)) {
+	    && __atomic_compare_exchange_n(&lt->join, &join_initial,
+		LT_JOIN_THREAD_SET, 0, __ATOMIC_RELAXED, __ATOMIC_RELAXED)) {
 
 		DIAG_EVENT(current, LT_DIAG_LTHREAD_JOIN, lt, 1);
 		_suspend();
diff --git a/examples/performance-thread/common/lthread_diag.h b/examples/performance-thread/common/lthread_diag.h
index e876dda6da..7ee89eef38 100644
--- a/examples/performance-thread/common/lthread_diag.h
+++ b/examples/performance-thread/common/lthread_diag.h
@@ -78,11 +78,11 @@ extern uint64_t diag_mask;
 	}								\
 } while (0)
 
-#define DIAG_COUNT_DEFINE(x) rte_atomic64_t count_##x
-#define DIAG_COUNT_INIT(o, x) rte_atomic64_init(&((o)->count_##x))
-#define DIAG_COUNT_INC(o, x) rte_atomic64_inc(&((o)->count_##x))
-#define DIAG_COUNT_DEC(o, x) rte_atomic64_dec(&((o)->count_##x))
-#define DIAG_COUNT(o, x) rte_atomic64_read(&((o)->count_##x))
+#define DIAG_COUNT_DEFINE(x) uint64_t count_##x
+#define DIAG_COUNT_INIT(o, x) __atomic_store_n(&((o)->count_##x), 0, __ATOMIC_RELAXED)
+#define DIAG_COUNT_INC(o, x) __atomic_fetch_add(&((o)->count_##x), 1, __ATOMIC_RELAXED)
+#define DIAG_COUNT_DEC(o, x) __atomic_fetch_sub(&((o)->count_##x), 1, __ATOMIC_RELAXED)
+#define DIAG_COUNT(o, x) __atomic_load_n(&((o)->count_##x), __ATOMIC_RELAXED)
 
 #define DIAG_USED
 
diff --git a/examples/performance-thread/common/lthread_int.h b/examples/performance-thread/common/lthread_int.h
index a352f13b75..d010126f16 100644
--- a/examples/performance-thread/common/lthread_int.h
+++ b/examples/performance-thread/common/lthread_int.h
@@ -21,7 +21,6 @@ extern "C" {
 #include <rte_cycles.h>
 #include <rte_per_lcore.h>
 #include <rte_timer.h>
-#include <rte_atomic_64.h>
 #include <rte_spinlock.h>
 #include <ctx.h>
 
diff --git a/examples/performance-thread/common/lthread_mutex.c b/examples/performance-thread/common/lthread_mutex.c
index 01da6cad4f..43cc9bbfb9 100644
--- a/examples/performance-thread/common/lthread_mutex.c
+++ b/examples/performance-thread/common/lthread_mutex.c
@@ -60,7 +60,7 @@ lthread_mutex_init(char *name, struct lthread_mutex **mutex,
 	m->root_sched = THIS_SCHED;
 	m->owner = NULL;
 
-	rte_atomic64_init(&m->count);
+	__atomic_store_n(&m->count, 0, __ATOMIC_RELAXED);
 
 	DIAG_CREATE_EVENT(m, LT_DIAG_MUTEX_CREATE);
 	/* success */
@@ -115,10 +115,11 @@ int lthread_mutex_lock(struct lthread_mutex *m)
 	}
 
 	for (;;) {
-		rte_atomic64_inc(&m->count);
+		__atomic_fetch_add(&m->count, 1, __ATOMIC_RELAXED);
 		do {
-			if (rte_atomic64_cmpset
-			    ((uint64_t *) &m->owner, 0, (uint64_t) lt)) {
+			uint64_t lt_init = 0;
+			if (__atomic_compare_exchange_n(&m->owner, &lt_init, lt,
+					0, __ATOMIC_RELAXED, __ATOMIC_RELAXED)) {
 				/* happy days, we got the lock */
 				DIAG_EVENT(m, LT_DIAG_MUTEX_LOCK, m, 0);
 				return 0;
@@ -126,7 +127,7 @@ int lthread_mutex_lock(struct lthread_mutex *m)
 			/* spin due to race with unlock when
 			* nothing was blocked
 			*/
-		} while ((rte_atomic64_read(&m->count) == 1) &&
+		} while ((__atomic_load_n(&m->count, __ATOMIC_RELAXED) == 1) &&
 				(m->owner == NULL));
 
 		/* queue the current thread in the blocked queue
@@ -160,16 +161,17 @@ int lthread_mutex_trylock(struct lthread_mutex *m)
 		return POSIX_ERRNO(EDEADLK);
 	}
 
-	rte_atomic64_inc(&m->count);
-	if (rte_atomic64_cmpset
-	    ((uint64_t *) &m->owner, (uint64_t) NULL, (uint64_t) lt)) {
+	__atomic_fetch_add(&m->count, 1, __ATOMIC_RELAXED);
+	uint64_t lt_init = 0;
+	if (__atomic_compare_exchange_n(&m->owner, &lt_init, lt,
+			0, __ATOMIC_RELAXED, __ATOMIC_RELAXED)) {
 		/* got the lock */
 		DIAG_EVENT(m, LT_DIAG_MUTEX_TRYLOCK, m, 0);
 		return 0;
 	}
 
 	/* failed so return busy */
-	rte_atomic64_dec(&m->count);
+	__atomic_fetch_sub(&m->count, 1, __ATOMIC_RELAXED);
 	DIAG_EVENT(m, LT_DIAG_MUTEX_TRYLOCK, m, POSIX_ERRNO(EBUSY));
 	return POSIX_ERRNO(EBUSY);
 }
@@ -193,13 +195,13 @@ int lthread_mutex_unlock(struct lthread_mutex *m)
 		return POSIX_ERRNO(EPERM);
 	}
 
-	rte_atomic64_dec(&m->count);
+	__atomic_fetch_sub(&m->count, 1, __ATOMIC_RELAXED);
 	/* if there are blocked threads then make one ready */
-	while (rte_atomic64_read(&m->count) > 0) {
+	while (__atomic_load_n(&m->count, __ATOMIC_RELAXED) > 0) {
 		unblocked = _lthread_queue_remove(m->blocked);
 
 		if (unblocked != NULL) {
-			rte_atomic64_dec(&m->count);
+			__atomic_fetch_sub(&m->count, 1, __ATOMIC_RELAXED);
 			DIAG_EVENT(m, LT_DIAG_MUTEX_UNLOCKED, m, unblocked);
 			RTE_ASSERT(unblocked->sched != NULL);
 			_ready_queue_insert((struct lthread_sched *)
diff --git a/examples/performance-thread/common/lthread_mutex.h b/examples/performance-thread/common/lthread_mutex.h
index cd866f87b8..730092bdf8 100644
--- a/examples/performance-thread/common/lthread_mutex.h
+++ b/examples/performance-thread/common/lthread_mutex.h
@@ -17,7 +17,7 @@ extern "C" {
 
 struct lthread_mutex {
 	struct lthread *owner;
-	rte_atomic64_t	count;
+	uint64_t count;
 	struct lthread_queue *blocked __rte_cache_aligned;
 	struct lthread_sched *root_sched;
 	char			name[MAX_MUTEX_NAME_SIZE];
diff --git a/examples/performance-thread/common/lthread_sched.c b/examples/performance-thread/common/lthread_sched.c
index 38ca0c45cb..3784b010c2 100644
--- a/examples/performance-thread/common/lthread_sched.c
+++ b/examples/performance-thread/common/lthread_sched.c
@@ -22,8 +22,6 @@
 
 #include <rte_prefetch.h>
 #include <rte_per_lcore.h>
-#include <rte_atomic.h>
-#include <rte_atomic_64.h>
 #include <rte_log.h>
 #include <rte_common.h>
 #include <rte_branch_prediction.h>
@@ -47,8 +45,8 @@
  * When a scheduler shuts down it is assumed that the application is terminating
  */
 
-static rte_atomic16_t num_schedulers;
-static rte_atomic16_t active_schedulers;
+static uint16_t num_schedulers;
+static uint16_t active_schedulers;
 
 /* one scheduler per lcore */
 RTE_DEFINE_PER_LCORE(struct lthread_sched *, this_sched) = NULL;
@@ -64,10 +62,8 @@ uint64_t diag_mask;
 RTE_INIT(lthread_sched_ctor)
 {
 	memset(schedcore, 0, sizeof(schedcore));
-	rte_atomic16_init(&num_schedulers);
-	rte_atomic16_set(&num_schedulers, 1);
-	rte_atomic16_init(&active_schedulers);
-	rte_atomic16_set(&active_schedulers, 0);
+	__atomic_store_n(&num_schedulers, 1, __ATOMIC_RELAXED);
+	__atomic_store_n(&active_schedulers, 0, __ATOMIC_RELAXED);
 	diag_cb = NULL;
 }
 
@@ -260,8 +256,8 @@ struct lthread_sched *_lthread_sched_create(size_t stack_size)
  */
 int lthread_num_schedulers_set(int num)
 {
-	rte_atomic16_set(&num_schedulers, num);
-	return (int)rte_atomic16_read(&num_schedulers);
+	__atomic_store_n(&num_schedulers, num, __ATOMIC_RELAXED);
+	return (int)__atomic_load_n(&num_schedulers, __ATOMIC_RELAXED);
 }
 
 /*
@@ -269,7 +265,7 @@ int lthread_num_schedulers_set(int num)
  */
 int lthread_active_schedulers(void)
 {
-	return (int)rte_atomic16_read(&active_schedulers);
+	return (int)__atomic_load_n(&active_schedulers, __ATOMIC_RELAXED);
 }
 
 
@@ -299,8 +295,8 @@ void lthread_scheduler_shutdown_all(void)
 	 * for the possibility of a pthread wrapper on lthread_yield(),
 	 * something that is not possible unless the scheduler is running.
 	 */
-	while (rte_atomic16_read(&active_schedulers) <
-	       rte_atomic16_read(&num_schedulers))
+	while (__atomic_load_n(&active_schedulers, __ATOMIC_RELAXED) <
+	       __atomic_load_n(&num_schedulers, __ATOMIC_RELAXED))
 		sched_yield();
 
 	for (i = 0; i < LTHREAD_MAX_LCORES; i++) {
@@ -415,15 +411,15 @@ static inline int _lthread_sched_isdone(struct lthread_sched *sched)
  */
 static inline void _lthread_schedulers_sync_start(void)
 {
-	rte_atomic16_inc(&active_schedulers);
+	__atomic_fetch_add(&active_schedulers, 1, __ATOMIC_RELAXED);
 
 	/* wait for lthread schedulers
 	 * Note we use sched_yield() rather than pthread_yield() to allow
 	 * for the possibility of a pthread wrapper on lthread_yield(),
 	 * something that is not possible unless the scheduler is running.
 	 */
-	while (rte_atomic16_read(&active_schedulers) <
-	       rte_atomic16_read(&num_schedulers))
+	while (__atomic_load_n(&active_schedulers, __ATOMIC_RELAXED) <
+	       __atomic_load_n(&num_schedulers, __ATOMIC_RELAXED))
 		sched_yield();
 
 }
@@ -433,15 +429,15 @@ static inline void _lthread_schedulers_sync_start(void)
  */
 static inline void _lthread_schedulers_sync_stop(void)
 {
-	rte_atomic16_dec(&active_schedulers);
-	rte_atomic16_dec(&num_schedulers);
+	__atomic_fetch_sub(&active_schedulers, 1, __ATOMIC_RELAXED);
+	__atomic_fetch_sub(&num_schedulers, 1, __ATOMIC_RELAXED);
 
 	/* wait for schedulers
 	 * Note we use sched_yield() rather than pthread_yield() to allow
 	 * for the possibility of a pthread wrapper on lthread_yield(),
 	 * something that is not possible unless the scheduler is running.
 	 */
-	while (rte_atomic16_read(&active_schedulers) > 0)
+	while (__atomic_load_n(&active_schedulers, __ATOMIC_RELAXED) > 0)
 		sched_yield();
 
 }
diff --git a/examples/performance-thread/common/lthread_tls.c b/examples/performance-thread/common/lthread_tls.c
index 07de6cafab..4ab2e3558b 100644
--- a/examples/performance-thread/common/lthread_tls.c
+++ b/examples/performance-thread/common/lthread_tls.c
@@ -18,7 +18,6 @@
 #include <rte_malloc.h>
 #include <rte_log.h>
 #include <rte_ring.h>
-#include <rte_atomic_64.h>
 
 #include "lthread_tls.h"
 #include "lthread_queue.h"
@@ -52,8 +51,10 @@ void _lthread_key_pool_init(void)
 
 	bzero(key_table, sizeof(key_table));
 
+	uint64_t pool_init = 0;
 	/* only one lcore should do this */
-	if (rte_atomic64_cmpset(&key_pool_init, 0, 1)) {
+	if (__atomic_compare_exchange_n(&key_pool_init, &pool_init, 1, 0,
+			__ATOMIC_RELAXED, __ATOMIC_RELAXED)) {
 
 		snprintf(name,
 			MAX_LTHREAD_NAME_SIZE,
diff --git a/examples/performance-thread/l3fwd-thread/main.c b/examples/performance-thread/l3fwd-thread/main.c
index 2f593abf26..ba9ad034e6 100644
--- a/examples/performance-thread/l3fwd-thread/main.c
+++ b/examples/performance-thread/l3fwd-thread/main.c
@@ -26,7 +26,6 @@
 #include <rte_memcpy.h>
 #include <rte_eal.h>
 #include <rte_launch.h>
-#include <rte_atomic.h>
 #include <rte_cycles.h>
 #include <rte_prefetch.h>
 #include <rte_lcore.h>
@@ -570,8 +569,8 @@ RTE_DEFINE_PER_LCORE(struct lcore_conf *, lcore_conf);
  */
 static int lthreads_on = 1; /**< Use lthreads for processing*/
 
-rte_atomic16_t rx_counter;  /**< Number of spawned rx threads */
-rte_atomic16_t tx_counter;  /**< Number of spawned tx threads */
+uint16_t rx_counter;  /**< Number of spawned rx threads */
+uint16_t tx_counter;  /**< Number of spawned tx threads */
 
 struct thread_conf {
 	uint16_t lcore_id;      /**< Initial lcore for rx thread */
@@ -1910,11 +1909,8 @@ cpu_load_collector(__rte_unused void *arg) {
 	printf("Waiting for %d rx threads and %d tx threads\n", n_rx_thread,
 			n_tx_thread);
 
-	while (rte_atomic16_read(&rx_counter) < n_rx_thread)
-		rte_pause();
-
-	while (rte_atomic16_read(&tx_counter) < n_tx_thread)
-		rte_pause();
+	rte_wait_until_equal_16(&rx_counter, n_rx_thread, __ATOMIC_RELAXED);
+	rte_wait_until_equal_16(&tx_counter, n_tx_thread, __ATOMIC_RELAXED);
 
 	for (i = 0; i < n_rx_thread; i++) {
 
@@ -2036,7 +2032,7 @@ lthread_tx_per_ring(void *dummy)
 	RTE_LOG(INFO, L3FWD, "entering main tx loop on lcore %u\n", rte_lcore_id());
 
 	nb_rx = 0;
-	rte_atomic16_inc(&tx_counter);
+	__atomic_fetch_add(&tx_counter, 1, __ATOMIC_RELAXED);
 	while (1) {
 
 		/*
@@ -2161,7 +2157,7 @@ lthread_rx(void *dummy)
 	worker_id = 0;
 
 	rx_conf->conf.cpu_id = sched_getcpu();
-	rte_atomic16_inc(&rx_counter);
+	__atomic_fetch_add(&rx_counter, 1, __ATOMIC_RELAXED);
 	while (1) {
 
 		/*
@@ -2243,7 +2239,7 @@ lthread_spawner(__rte_unused void *arg)
 	 * scheduler as this lthread, yielding is required to let them to run and
 	 * prevent deadlock here.
 	 */
-	while (rte_atomic16_read(&rx_counter) < n_rx_thread)
+	while (__atomic_load_n(&rx_counter, __ATOMIC_RELAXED) < n_rx_thread)
 		lthread_sleep(100000);
 
 	/*
@@ -2323,7 +2319,7 @@ pthread_tx(void *dummy)
 	RTE_LOG(INFO, L3FWD, "Entering main Tx loop on lcore %u\n", rte_lcore_id());
 
 	tx_conf->conf.cpu_id = sched_getcpu();
-	rte_atomic16_inc(&tx_counter);
+	__atomic_fetch_add(&tx_counter, 1, __ATOMIC_RELAXED);
 	while (1) {
 
 		cur_tsc = rte_rdtsc();
@@ -2406,7 +2402,7 @@ pthread_rx(void *dummy)
 
 	worker_id = 0;
 	rx_conf->conf.cpu_id = sched_getcpu();
-	rte_atomic16_inc(&rx_counter);
+	__atomic_fetch_add(&rx_counter, 1, __ATOMIC_RELAXED);
 	while (1) {
 
 		/*
-- 
2.17.1

[dpdk-dev] [PATCH v1 5/8] examples/l2fwd-jobstats: use compiler atomics for stats sync

[Joyce Kong]

Convert rte_atomic usages to compiler atomic built-ins
for stats_read_pending sync in l2fwd_jobstats module.

Signed-off-by: Joyce Kong <joyce.kong@arm.com>
Reviewed-by: Ruifeng Wang <ruifeng.wang@arm.com>
---
 examples/l2fwd-jobstats/main.c | 11 +++++------
 1 file changed, 5 insertions(+), 6 deletions(-)

diff --git a/examples/l2fwd-jobstats/main.c b/examples/l2fwd-jobstats/main.c
index bbb4a27a6d..99a17de181 100644
--- a/examples/l2fwd-jobstats/main.c
+++ b/examples/l2fwd-jobstats/main.c
@@ -16,7 +16,6 @@
 #include <rte_memcpy.h>
 #include <rte_eal.h>
 #include <rte_launch.h>
-#include <rte_atomic.h>
 #include <rte_cycles.h>
 #include <rte_prefetch.h>
 #include <rte_lcore.h>
@@ -81,7 +80,7 @@ struct lcore_queue_conf {
 	struct rte_jobstats idle_job;
 	struct rte_jobstats_context jobs_context;
 
-	rte_atomic16_t stats_read_pending;
+	uint16_t stats_read_pending;
 	rte_spinlock_t lock;
 } __rte_cache_aligned;
 /* >8 End of list of queues to be polled for given lcore. */
@@ -155,9 +154,9 @@ show_lcore_stats(unsigned lcore_id)
 	uint64_t collection_time = rte_get_timer_cycles();
 
 	/* Ask forwarding thread to give us stats. */
-	rte_atomic16_set(&qconf->stats_read_pending, 1);
+	__atomic_store_n(&qconf->stats_read_pending, 1, __ATOMIC_RELAXED);
 	rte_spinlock_lock(&qconf->lock);
-	rte_atomic16_set(&qconf->stats_read_pending, 0);
+	__atomic_store_n(&qconf->stats_read_pending, 0, __ATOMIC_RELAXED);
 
 	/* Collect context statistics. */
 	stats_period = ctx->state_time - ctx->start_time;
@@ -526,8 +525,8 @@ l2fwd_main_loop(void)
 				repeats++;
 				need_manage = qconf->flush_timer.expire < now;
 				/* Check if we was esked to give a stats. */
-				stats_read_pending =
-						rte_atomic16_read(&qconf->stats_read_pending);
+				stats_read_pending = __atomic_load_n(&qconf->stats_read_pending,
+						__ATOMIC_RELAXED);
 				need_manage |= stats_read_pending;
 
 				for (i = 0; i < qconf->n_rx_port && !need_manage; i++)
-- 
2.17.1

[dpdk-dev] [PATCH v1 6/8] examples/vm_power_manager: use compiler atomics for sync

[Joyce Kong]

Convert rte_atomic32_cmpset to compiler atomic CAS
operation for channel status sync.

Signed-off-by: Joyce Kong <joyce.kong@arm.com>
Reviewed-by: Ruifeng Wang <ruifeng.wang@arm.com>
---
 examples/vm_power_manager/channel_monitor.c | 11 ++++++-----
 1 file changed, 6 insertions(+), 5 deletions(-)

diff --git a/examples/vm_power_manager/channel_monitor.c b/examples/vm_power_manager/channel_monitor.c
index 99f81544d7..3c20406f7c 100644
--- a/examples/vm_power_manager/channel_monitor.c
+++ b/examples/vm_power_manager/channel_monitor.c
@@ -25,7 +25,6 @@
 #include <rte_log.h>
 #include <rte_memory.h>
 #include <rte_malloc.h>
-#include <rte_atomic.h>
 #include <rte_cycles.h>
 #include <rte_ethdev.h>
 #ifdef RTE_NET_I40E
@@ -829,8 +828,9 @@ process_request(struct rte_power_channel_packet *pkt,
 	if (chan_info == NULL)
 		return -1;
 
-	if (rte_atomic32_cmpset(&(chan_info->status), CHANNEL_MGR_CHANNEL_CONNECTED,
-			CHANNEL_MGR_CHANNEL_PROCESSING) == 0)
+	uint32_t channel_connected = CHANNEL_MGR_CHANNEL_CONNECTED;
+	if (__atomic_compare_exchange_n(&(chan_info->status), &channel_connected,
+		CHANNEL_MGR_CHANNEL_PROCESSING, 0, __ATOMIC_RELAXED, __ATOMIC_RELAXED) == 0)
 		return -1;
 
 	if (pkt->command == RTE_POWER_CPU_POWER) {
@@ -934,8 +934,9 @@ process_request(struct rte_power_channel_packet *pkt,
 	 * Return is not checked as channel status may have been set to DISABLED
 	 * from management thread
 	 */
-	rte_atomic32_cmpset(&(chan_info->status), CHANNEL_MGR_CHANNEL_PROCESSING,
-			CHANNEL_MGR_CHANNEL_CONNECTED);
+	uint32_t channel_processing = CHANNEL_MGR_CHANNEL_PROCESSING;
+	__atomic_compare_exchange_n(&(chan_info->status), &channel_processing,
+		CHANNEL_MGR_CHANNEL_CONNECTED, 0, __ATOMIC_RELAXED, __ATOMIC_RELAXED);
 	return 0;
 
 }
-- 
2.17.1

[dpdk-dev] [PATCH v1 7/8] examples/server_node_efd: use compiler atomics for sync

[Joyce Kong]

Convert rte_atomic32_test_and_set to compiler CAS atomic
operation for display_stats sync.

Signed-off-by: Joyce Kong <joyce.kong@arm.com>
Reviewed-by: Ruifeng Wang <ruifeng.wang@arm.com>
---
 examples/server_node_efd/server/main.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/examples/server_node_efd/server/main.c b/examples/server_node_efd/server/main.c
index 7d07131dbf..b69beb012c 100644
--- a/examples/server_node_efd/server/main.c
+++ b/examples/server_node_efd/server/main.c
@@ -22,7 +22,6 @@
 #include <rte_per_lcore.h>
 #include <rte_lcore.h>
 #include <rte_branch_prediction.h>
-#include <rte_atomic.h>
 #include <rte_ring.h>
 #include <rte_log.h>
 #include <rte_debug.h>
@@ -180,10 +179,12 @@ static int
 sleep_lcore(__rte_unused void *dummy)
 {
 	/* Used to pick a display thread - static, so zero-initialised */
-	static rte_atomic32_t display_stats;
+	static uint32_t display_stats;
 
 	/* Only one core should display stats */
-	if (rte_atomic32_test_and_set(&display_stats)) {
+	uint32_t display_init = 0;
+	if (__atomic_compare_exchange_n(&display_stats, &display_init, 1, 0,
+			__ATOMIC_RELAXED, __ATOMIC_RELAXED)) {
 		const unsigned int sleeptime = 1;
 
 		printf("Core %u displaying statistics\n", rte_lcore_id());
-- 
2.17.1

[dpdk-dev] [PATCH v1 8/8] examples: remove unnecessary include of atomic

[Joyce Kong]

Remove the unnecessary header file rte_atomic.h
included in example module.

Signed-off-by: Joyce Kong <joyce.kong@arm.com>
Reviewed-by: Ruifeng Wang <ruifeng.wang@arm.com>
---
 examples/bond/main.c                                       | 1 -
 examples/ip_fragmentation/main.c                           | 1 -
 examples/ip_reassembly/main.c                              | 1 -
 examples/ipsec-secgw/ipsec-secgw.c                         | 1 -
 examples/ipv4_multicast/main.c                             | 1 -
 examples/l2fwd-crypto/main.c                               | 1 -
 examples/l2fwd-event/l2fwd_common.h                        | 1 -
 examples/l2fwd-event/l2fwd_event.c                         | 1 -
 examples/l2fwd-keepalive/main.c                            | 1 -
 examples/l2fwd/main.c                                      | 1 -
 examples/l3fwd-acl/main.c                                  | 1 -
 examples/l3fwd-power/main.c                                | 1 -
 examples/l3fwd/main.c                                      | 1 -
 examples/link_status_interrupt/main.c                      | 1 -
 examples/multi_process/client_server_mp/mp_client/client.c | 1 -
 examples/multi_process/client_server_mp/mp_server/init.c   | 1 -
 examples/multi_process/simple_mp/main.c                    | 1 -
 examples/multi_process/simple_mp/mp_commands.c             | 1 -
 examples/multi_process/symmetric_mp/main.c                 | 1 -
 examples/server_node_efd/node/node.c                       | 1 -
 examples/server_node_efd/server/init.c                     | 1 -
 examples/vhost_blk/blk.c                                   | 1 -
 examples/vhost_blk/vhost_blk.c                             | 1 -
 examples/vm_power_manager/channel_manager.c                | 1 -
 examples/vm_power_manager/channel_manager.h                | 1 -
 examples/vmdq/main.c                                       | 1 -
 examples/vmdq_dcb/main.c                                   | 1 -
 27 files changed, 27 deletions(-)

diff --git a/examples/bond/main.c b/examples/bond/main.c
index f48400e211..de1e995a8a 100644
--- a/examples/bond/main.c
+++ b/examples/bond/main.c
@@ -24,7 +24,6 @@
 #include <rte_memcpy.h>
 #include <rte_eal.h>
 #include <rte_launch.h>
-#include <rte_atomic.h>
 #include <rte_cycles.h>
 #include <rte_prefetch.h>
 #include <rte_lcore.h>
diff --git a/examples/ip_fragmentation/main.c b/examples/ip_fragmentation/main.c
index f245369720..f965624c8b 100644
--- a/examples/ip_fragmentation/main.c
+++ b/examples/ip_fragmentation/main.c
@@ -21,7 +21,6 @@
 #include <rte_memcpy.h>
 #include <rte_eal.h>
 #include <rte_launch.h>
-#include <rte_atomic.h>
 #include <rte_cycles.h>
 #include <rte_prefetch.h>
 #include <rte_lcore.h>
diff --git a/examples/ip_reassembly/main.c b/examples/ip_reassembly/main.c
index 8645ac790b..6a8e91673d 100644
--- a/examples/ip_reassembly/main.c
+++ b/examples/ip_reassembly/main.c
@@ -22,7 +22,6 @@
 #include <rte_memcpy.h>
 #include <rte_eal.h>
 #include <rte_launch.h>
-#include <rte_atomic.h>
 #include <rte_cycles.h>
 #include <rte_prefetch.h>
 #include <rte_lcore.h>
diff --git a/examples/ipsec-secgw/ipsec-secgw.c b/examples/ipsec-secgw/ipsec-secgw.c
index f252d34985..012b908410 100644
--- a/examples/ipsec-secgw/ipsec-secgw.c
+++ b/examples/ipsec-secgw/ipsec-secgw.c
@@ -24,7 +24,6 @@
 #include <rte_log.h>
 #include <rte_eal.h>
 #include <rte_launch.h>
-#include <rte_atomic.h>
 #include <rte_cycles.h>
 #include <rte_prefetch.h>
 #include <rte_lcore.h>
diff --git a/examples/ipv4_multicast/main.c b/examples/ipv4_multicast/main.c
index cc527d7f6b..5830ad61c1 100644
--- a/examples/ipv4_multicast/main.c
+++ b/examples/ipv4_multicast/main.c
@@ -20,7 +20,6 @@
 #include <rte_memcpy.h>
 #include <rte_eal.h>
 #include <rte_launch.h>
-#include <rte_atomic.h>
 #include <rte_cycles.h>
 #include <rte_prefetch.h>
 #include <rte_lcore.h>
diff --git a/examples/l2fwd-crypto/main.c b/examples/l2fwd-crypto/main.c
index 5f539c458c..e302ad022c 100644
--- a/examples/l2fwd-crypto/main.c
+++ b/examples/l2fwd-crypto/main.c
@@ -20,7 +20,6 @@
 #include <unistd.h>
 
 #include <rte_string_fns.h>
-#include <rte_atomic.h>
 #include <rte_branch_prediction.h>
 #include <rte_common.h>
 #include <rte_cryptodev.h>
diff --git a/examples/l2fwd-event/l2fwd_common.h b/examples/l2fwd-event/l2fwd_common.h
index 939221d45a..1a418e6a22 100644
--- a/examples/l2fwd-event/l2fwd_common.h
+++ b/examples/l2fwd-event/l2fwd_common.h
@@ -27,7 +27,6 @@
 #include <rte_memcpy.h>
 #include <rte_eal.h>
 #include <rte_launch.h>
-#include <rte_atomic.h>
 #include <rte_cycles.h>
 #include <rte_prefetch.h>
 #include <rte_lcore.h>
diff --git a/examples/l2fwd-event/l2fwd_event.c b/examples/l2fwd-event/l2fwd_event.c
index 7ba5311d66..acfacfa4fb 100644
--- a/examples/l2fwd-event/l2fwd_event.c
+++ b/examples/l2fwd-event/l2fwd_event.c
@@ -5,7 +5,6 @@
 #include <stdbool.h>
 #include <getopt.h>
 
-#include <rte_atomic.h>
 #include <rte_cycles.h>
 #include <rte_ethdev.h>
 #include <rte_eventdev.h>
diff --git a/examples/l2fwd-keepalive/main.c b/examples/l2fwd-keepalive/main.c
index 4e1a17cfe4..cb7f267259 100644
--- a/examples/l2fwd-keepalive/main.c
+++ b/examples/l2fwd-keepalive/main.c
@@ -24,7 +24,6 @@
 #include <rte_memcpy.h>
 #include <rte_eal.h>
 #include <rte_launch.h>
-#include <rte_atomic.h>
 #include <rte_cycles.h>
 #include <rte_prefetch.h>
 #include <rte_lcore.h>
diff --git a/examples/l2fwd/main.c b/examples/l2fwd/main.c
index 911e40c66e..56a0663111 100644
--- a/examples/l2fwd/main.c
+++ b/examples/l2fwd/main.c
@@ -24,7 +24,6 @@
 #include <rte_memcpy.h>
 #include <rte_eal.h>
 #include <rte_launch.h>
-#include <rte_atomic.h>
 #include <rte_cycles.h>
 #include <rte_prefetch.h>
 #include <rte_lcore.h>
diff --git a/examples/l3fwd-acl/main.c b/examples/l3fwd-acl/main.c
index a1f457b564..2a942731bc 100644
--- a/examples/l3fwd-acl/main.c
+++ b/examples/l3fwd-acl/main.c
@@ -20,7 +20,6 @@
 #include <rte_memcpy.h>
 #include <rte_eal.h>
 #include <rte_launch.h>
-#include <rte_atomic.h>
 #include <rte_cycles.h>
 #include <rte_prefetch.h>
 #include <rte_lcore.h>
diff --git a/examples/l3fwd-power/main.c b/examples/l3fwd-power/main.c
index aa7b8db44a..c29b4fe034 100644
--- a/examples/l3fwd-power/main.c
+++ b/examples/l3fwd-power/main.c
@@ -24,7 +24,6 @@
 #include <rte_memcpy.h>
 #include <rte_eal.h>
 #include <rte_launch.h>
-#include <rte_atomic.h>
 #include <rte_cycles.h>
 #include <rte_prefetch.h>
 #include <rte_lcore.h>
diff --git a/examples/l3fwd/main.c b/examples/l3fwd/main.c
index 00ac267af1..518fefe90b 100644
--- a/examples/l3fwd/main.c
+++ b/examples/l3fwd/main.c
@@ -24,7 +24,6 @@
 #include <rte_memcpy.h>
 #include <rte_eal.h>
 #include <rte_launch.h>
-#include <rte_atomic.h>
 #include <rte_cycles.h>
 #include <rte_prefetch.h>
 #include <rte_lcore.h>
diff --git a/examples/link_status_interrupt/main.c b/examples/link_status_interrupt/main.c
index 7470aa539a..302a267e0e 100644
--- a/examples/link_status_interrupt/main.c
+++ b/examples/link_status_interrupt/main.c
@@ -22,7 +22,6 @@
 #include <rte_memcpy.h>
 #include <rte_eal.h>
 #include <rte_launch.h>
-#include <rte_atomic.h>
 #include <rte_cycles.h>
 #include <rte_prefetch.h>
 #include <rte_lcore.h>
diff --git a/examples/multi_process/client_server_mp/mp_client/client.c b/examples/multi_process/client_server_mp/mp_client/client.c
index 6d4c246816..087c38ef73 100644
--- a/examples/multi_process/client_server_mp/mp_client/client.c
+++ b/examples/multi_process/client_server_mp/mp_client/client.c
@@ -17,7 +17,6 @@
 #include <rte_memory.h>
 #include <rte_memzone.h>
 #include <rte_eal.h>
-#include <rte_atomic.h>
 #include <rte_branch_prediction.h>
 #include <rte_log.h>
 #include <rte_per_lcore.h>
diff --git a/examples/multi_process/client_server_mp/mp_server/init.c b/examples/multi_process/client_server_mp/mp_server/init.c
index 1ad71ca7ec..be669c2bcc 100644
--- a/examples/multi_process/client_server_mp/mp_server/init.c
+++ b/examples/multi_process/client_server_mp/mp_server/init.c
@@ -15,7 +15,6 @@
 #include <rte_memzone.h>
 #include <rte_eal.h>
 #include <rte_byteorder.h>
-#include <rte_atomic.h>
 #include <rte_launch.h>
 #include <rte_per_lcore.h>
 #include <rte_lcore.h>
diff --git a/examples/multi_process/simple_mp/main.c b/examples/multi_process/simple_mp/main.c
index df996f0f84..5df2a39000 100644
--- a/examples/multi_process/simple_mp/main.c
+++ b/examples/multi_process/simple_mp/main.c
@@ -30,7 +30,6 @@
 #include <rte_per_lcore.h>
 #include <rte_lcore.h>
 #include <rte_debug.h>
-#include <rte_atomic.h>
 #include <rte_branch_prediction.h>
 #include <rte_ring.h>
 #include <rte_log.h>
diff --git a/examples/multi_process/simple_mp/mp_commands.c b/examples/multi_process/simple_mp/mp_commands.c
index 311d0fe775..a5f91b00be 100644
--- a/examples/multi_process/simple_mp/mp_commands.c
+++ b/examples/multi_process/simple_mp/mp_commands.c
@@ -14,7 +14,6 @@
 #include <rte_common.h>
 #include <rte_memory.h>
 #include <rte_eal.h>
-#include <rte_atomic.h>
 #include <rte_branch_prediction.h>
 #include <rte_launch.h>
 #include <rte_log.h>
diff --git a/examples/multi_process/symmetric_mp/main.c b/examples/multi_process/symmetric_mp/main.c
index 01dc3acf34..a66328ba0c 100644
--- a/examples/multi_process/symmetric_mp/main.c
+++ b/examples/multi_process/symmetric_mp/main.c
@@ -31,7 +31,6 @@
 #include <rte_eal.h>
 #include <rte_per_lcore.h>
 #include <rte_lcore.h>
-#include <rte_atomic.h>
 #include <rte_branch_prediction.h>
 #include <rte_debug.h>
 #include <rte_interrupts.h>
diff --git a/examples/server_node_efd/node/node.c b/examples/server_node_efd/node/node.c
index 4580a44e3e..ba1c7e5153 100644
--- a/examples/server_node_efd/node/node.c
+++ b/examples/server_node_efd/node/node.c
@@ -17,7 +17,6 @@
 #include <rte_memory.h>
 #include <rte_memzone.h>
 #include <rte_eal.h>
-#include <rte_atomic.h>
 #include <rte_branch_prediction.h>
 #include <rte_log.h>
 #include <rte_per_lcore.h>
diff --git a/examples/server_node_efd/server/init.c b/examples/server_node_efd/server/init.c
index 9ebd88bac2..a19934dbe0 100644
--- a/examples/server_node_efd/server/init.c
+++ b/examples/server_node_efd/server/init.c
@@ -15,7 +15,6 @@
 #include <rte_memzone.h>
 #include <rte_eal.h>
 #include <rte_byteorder.h>
-#include <rte_atomic.h>
 #include <rte_launch.h>
 #include <rte_per_lcore.h>
 #include <rte_lcore.h>
diff --git a/examples/vhost_blk/blk.c b/examples/vhost_blk/blk.c
index f8c8549b3a..d082ab3c94 100644
--- a/examples/vhost_blk/blk.c
+++ b/examples/vhost_blk/blk.c
@@ -15,7 +15,6 @@
 #include <string.h>
 #include <stddef.h>
 
-#include <rte_atomic.h>
 #include <rte_cycles.h>
 #include <rte_log.h>
 #include <rte_malloc.h>
diff --git a/examples/vhost_blk/vhost_blk.c b/examples/vhost_blk/vhost_blk.c
index fe2b4e4803..9e291dc442 100644
--- a/examples/vhost_blk/vhost_blk.c
+++ b/examples/vhost_blk/vhost_blk.c
@@ -17,7 +17,6 @@
 #include <linux/virtio_blk.h>
 #include <linux/virtio_ring.h>
 
-#include <rte_atomic.h>
 #include <rte_cycles.h>
 #include <rte_log.h>
 #include <rte_malloc.h>
diff --git a/examples/vm_power_manager/channel_manager.c b/examples/vm_power_manager/channel_manager.c
index fe91567854..838465ab4b 100644
--- a/examples/vm_power_manager/channel_manager.c
+++ b/examples/vm_power_manager/channel_manager.c
@@ -21,7 +21,6 @@
 #include <rte_memory.h>
 #include <rte_mempool.h>
 #include <rte_log.h>
-#include <rte_atomic.h>
 #include <rte_spinlock.h>
 
 #include <libvirt/libvirt.h>
diff --git a/examples/vm_power_manager/channel_manager.h b/examples/vm_power_manager/channel_manager.h
index e55376fcdb..7038e9d83b 100644
--- a/examples/vm_power_manager/channel_manager.h
+++ b/examples/vm_power_manager/channel_manager.h
@@ -11,7 +11,6 @@ extern "C" {
 
 #include <linux/limits.h>
 #include <linux/un.h>
-#include <rte_atomic.h>
 #include <stdbool.h>
 
 /* Maximum name length including '\0' terminator */
diff --git a/examples/vmdq/main.c b/examples/vmdq/main.c
index d3bc19f78e..a958ad2167 100644
--- a/examples/vmdq/main.c
+++ b/examples/vmdq/main.c
@@ -20,7 +20,6 @@
 #include <rte_memcpy.h>
 #include <rte_eal.h>
 #include <rte_launch.h>
-#include <rte_atomic.h>
 #include <rte_cycles.h>
 #include <rte_prefetch.h>
 #include <rte_lcore.h>
diff --git a/examples/vmdq_dcb/main.c b/examples/vmdq_dcb/main.c
index 685a03bdd1..52e930f83a 100644
--- a/examples/vmdq_dcb/main.c
+++ b/examples/vmdq_dcb/main.c
@@ -20,7 +20,6 @@
 #include <rte_memcpy.h>
 #include <rte_eal.h>
 #include <rte_launch.h>
-#include <rte_atomic.h>
 #include <rte_cycles.h>
 #include <rte_prefetch.h>
 #include <rte_lcore.h>
-- 
2.17.1