diff mbox

[dpdk-dev,11/13] examples/eventdev: add atq single stage pipeline worker

Message ID 20171207203705.25020-12-pbhagavatula@caviumnetworks.com (mailing list archive)
State Superseded, archived
Delegated to: Jerin Jacob
Headers show

Checks

Context Check Description
ci/checkpatch success coding style OK
ci/Intel-compilation success Compilation OK

Commit Message

Pavan Nikhilesh Dec. 7, 2017, 8:37 p.m. UTC
Add optimized eventdev pipeline when ethdev supports thread safe Tx,
number of configured stages is one and all type queue option is enabled.

Signed-off-by: Pavan Nikhilesh <pbhagavatula@caviumnetworks.com>
---
 .../eventdev_pipeline_sw_pmd/pipeline_worker_tx.c  | 88 +++++++++++++++++++++-
 1 file changed, 86 insertions(+), 2 deletions(-)

Comments

Van Haaren, Harry Dec. 19, 2017, 1:34 p.m. UTC | #1
> From: Pavan Nikhilesh [mailto:pbhagavatula@caviumnetworks.com]
> Sent: Thursday, December 7, 2017 8:37 PM
> To: Eads, Gage <gage.eads@intel.com>; jerin.jacobkollanukkaran@cavium.com;
> Van Haaren, Harry <harry.van.haaren@intel.com>; Rao, Nikhil
> <nikhil.rao@intel.com>; hemant.agrawal@nxp.com; Ma, Liang J
> <liang.j.ma@intel.com>
> Cc: dev@dpdk.org; Pavan Nikhilesh <pbhagavatula@caviumnetworks.com>
> Subject: [PATCH 11/13] examples/eventdev: add atq single stage pipeline
> worker
> 
> Add optimized eventdev pipeline when ethdev supports thread safe Tx,
> number of configured stages is one and all type queue option is enabled.
> 
> Signed-off-by: Pavan Nikhilesh <pbhagavatula@caviumnetworks.com>
> ---
>  .../eventdev_pipeline_sw_pmd/pipeline_worker_tx.c  | 88
> +++++++++++++++++++++-
>  1 file changed, 86 insertions(+), 2 deletions(-)


>  	if (cdata.num_stages == 1) {
> -		if (burst)
> +		if (burst && atq)
> +			caps->worker_loop = worker_do_tx_single_burst_atq;
> +		if (burst && !atq)
>  			caps->worker_loop = worker_do_tx_single_burst;
> -		if (!burst)
> +		if (!burst && atq)
> +			caps->worker_loop = worker_do_tx_single_atq;
> +		if (!burst && !atq)
>  			caps->worker_loop = worker_do_tx_single;
>  	} else {
>  		if (burst && atq)

As per previous notes, this doesn't scale.
diff mbox

Patch

diff --git a/examples/eventdev_pipeline_sw_pmd/pipeline_worker_tx.c b/examples/eventdev_pipeline_sw_pmd/pipeline_worker_tx.c
index 15df21b7e..24ffd431f 100644
--- a/examples/eventdev_pipeline_sw_pmd/pipeline_worker_tx.c
+++ b/examples/eventdev_pipeline_sw_pmd/pipeline_worker_tx.c
@@ -106,6 +106,41 @@  worker_do_tx_single(void *arg)
 	return 0;
 }
 
+static int
+worker_do_tx_single_atq(void *arg)
+{
+	struct worker_data *data = (struct worker_data *)arg;
+	const uint8_t dev = data->dev_id;
+	const uint8_t port = data->port_id;
+	size_t fwd = 0, received = 0, tx = 0;
+	struct rte_event ev;
+
+	while (!fdata->done) {
+
+		if (!rte_event_dequeue_burst(dev, port, &ev, 1, 0)) {
+			rte_pause();
+			continue;
+		}
+
+		received++;
+
+		if (ev.sched_type == RTE_SCHED_TYPE_ATOMIC) {
+			worker_tx_pkt(ev.mbuf);
+			tx++;
+			continue;
+		}
+		work(ev.mbuf);
+		worker_fwd_event(&ev, RTE_SCHED_TYPE_ATOMIC);
+		worker_event_enqueue(dev, port, &ev);
+		fwd++;
+	}
+
+	if (!cdata.quiet)
+		printf("  worker %u thread done. RX=%zu FWD=%zu TX=%zu\n",
+				rte_lcore_id(), received, fwd, tx);
+	return 0;
+}
+
 static int
 worker_do_tx_single_burst(void *arg)
 {
@@ -153,6 +188,51 @@  worker_do_tx_single_burst(void *arg)
 	return 0;
 }
 
+static int
+worker_do_tx_single_burst_atq(void *arg)
+{
+	struct rte_event ev[BATCH_SIZE + 1];
+
+	struct worker_data *data = (struct worker_data *)arg;
+	const uint8_t dev = data->dev_id;
+	const uint8_t port = data->port_id;
+	size_t fwd = 0, received = 0, tx = 0;
+
+	while (!fdata->done) {
+		uint16_t i;
+		uint16_t nb_rx = rte_event_dequeue_burst(dev, port, ev,
+				BATCH_SIZE, 0);
+
+		if (!nb_rx) {
+			rte_pause();
+			continue;
+		}
+
+		received += nb_rx;
+
+		for (i = 0; i < nb_rx; i++) {
+			rte_prefetch0(ev[i + 1].mbuf);
+			if (ev[i].sched_type == RTE_SCHED_TYPE_ATOMIC) {
+
+				worker_tx_pkt(ev[i].mbuf);
+				ev[i].op = RTE_EVENT_OP_RELEASE;
+				tx++;
+			} else
+				worker_fwd_event(&ev[i],
+						RTE_SCHED_TYPE_ATOMIC);
+			work(ev[i].mbuf);
+		}
+
+		worker_event_enqueue_burst(dev, port, ev, nb_rx);
+		fwd += nb_rx;
+	}
+
+	if (!cdata.quiet)
+		printf("  worker %u thread done. RX=%zu FWD=%zu TX=%zu\n",
+				rte_lcore_id(), received, fwd, tx);
+	return 0;
+}
+
 /* Multi stage Pipeline Workers */
 
 static int
@@ -697,9 +777,13 @@  set_worker_tx_setup_data(struct setup_data *caps, bool burst)
 	uint8_t atq = cdata.all_type_queues ? 1 : 0;
 
 	if (cdata.num_stages == 1) {
-		if (burst)
+		if (burst && atq)
+			caps->worker_loop = worker_do_tx_single_burst_atq;
+		if (burst && !atq)
 			caps->worker_loop = worker_do_tx_single_burst;
-		if (!burst)
+		if (!burst && atq)
+			caps->worker_loop = worker_do_tx_single_atq;
+		if (!burst && !atq)
 			caps->worker_loop = worker_do_tx_single;
 	} else {
 		if (burst && atq)