Enabled driver support to dequeue inference requests from
internal queue. Dequeue checks for request completion by
polling the status field of the job request.
Signed-off-by: Srikanth Yalavarthi <syalavarthi@marvell.com>
---
drivers/ml/cnxk/cn10k_ml_ops.c | 61 ++++++++++++++++++++++++++++++++++
drivers/ml/cnxk/cn10k_ml_ops.h | 2 ++
2 files changed, 63 insertions(+)
@@ -473,6 +473,7 @@ cn10k_ml_dev_configure(struct rte_ml_dev *dev, const struct rte_ml_dev_config *c
rte_spinlock_init(&ocm->lock);
dev->enqueue_burst = cn10k_ml_enqueue_burst;
+ dev->dequeue_burst = cn10k_ml_dequeue_burst;
mldev->nb_models_loaded = 0;
mldev->state = ML_CN10K_DEV_STATE_CONFIGURED;
@@ -1421,6 +1422,23 @@ queue_free_count(uint64_t head, uint64_t tail, uint64_t nb_desc)
return nb_desc - queue_pending_count(head, tail, nb_desc) - 1;
}
+static __rte_always_inline void
+cn10k_ml_result_update(struct rte_ml_dev *dev, int qp_id, struct cn10k_ml_result *result,
+ struct rte_ml_op *op)
+{
+ PLT_SET_USED(dev);
+ PLT_SET_USED(qp_id);
+
+ op->impl_opaque = result->error_code;
+
+ if (likely(result->error_code == 0))
+ op->status = RTE_ML_OP_STATUS_SUCCESS;
+ else
+ op->status = RTE_ML_OP_STATUS_ERROR;
+
+ op->user_ptr = result->user_ptr;
+}
+
__rte_hot uint16_t
cn10k_ml_enqueue_burst(struct rte_ml_dev *dev, uint16_t qp_id, struct rte_ml_op **ops,
uint16_t nb_ops)
@@ -1475,6 +1493,49 @@ cn10k_ml_enqueue_burst(struct rte_ml_dev *dev, uint16_t qp_id, struct rte_ml_op
return count;
}
+__rte_hot uint16_t
+cn10k_ml_dequeue_burst(struct rte_ml_dev *dev, uint16_t qp_id, struct rte_ml_op **ops,
+ uint16_t nb_ops)
+{
+ struct cn10k_ml_queue *queue;
+ struct cn10k_ml_req *req;
+ struct cn10k_ml_qp *qp;
+
+ uint64_t status;
+ uint16_t count;
+ uint64_t tail;
+
+ qp = dev->data->queue_pairs[qp_id];
+ queue = &qp->queue;
+
+ tail = queue->tail;
+ nb_ops = PLT_MIN(nb_ops, queue_pending_count(queue->head, tail, qp->nb_desc));
+ count = 0;
+
+ if (unlikely(nb_ops == 0))
+ goto empty_or_active;
+
+dequeue_req:
+ req = &queue->reqs[tail];
+ status = plt_read64(&req->status);
+ if (unlikely(status != ML_CN10K_POLL_JOB_FINISH))
+ goto empty_or_active;
+
+ cn10k_ml_result_update(dev, qp_id, &req->result, req->op);
+ ops[count] = req->op;
+
+ queue_index_advance(&tail, qp->nb_desc);
+ count++;
+
+ if (count < nb_ops)
+ goto dequeue_req;
+
+empty_or_active:
+ queue->tail = tail;
+
+ return count;
+}
+
struct rte_ml_dev_ops cn10k_ml_ops = {
/* Device control ops */
.dev_info_get = cn10k_ml_dev_info_get,
@@ -73,5 +73,7 @@ int cn10k_ml_model_stop(struct rte_ml_dev *dev, int16_t model_id);
/* Fast-path ops */
__rte_hot uint16_t cn10k_ml_enqueue_burst(struct rte_ml_dev *dev, uint16_t qp_id,
struct rte_ml_op **ops, uint16_t nb_ops);
+__rte_hot uint16_t cn10k_ml_dequeue_burst(struct rte_ml_dev *dev, uint16_t qp_id,
+ struct rte_ml_op **ops, uint16_t nb_ops);
#endif /* _CN10K_ML_OPS_H_ */