From: Nélio Laranjeiro <nelio.laranjeiro@6wind.com>
Date: Tue, 20 Sep 2016 08:53:48 +0000 (+0200)
Subject: net/mlx5: reduce memory overhead for CQE handling
X-Git-Tag: spdx-start~5843
X-Git-Url: http://git.droids-corp.org/?a=commitdiff_plain;h=e2f116ee3cac59598f804885f728c18482c8d31c;p=dpdk.git

net/mlx5: reduce memory overhead for CQE handling

PMD uses only power of two number of Completion Queue Elements (aka CQE),
storing the number of elements in log2 helps to reduce the size of the
container to store it.

Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
---

diff --git a/drivers/net/mlx5/mlx5_rxq.c b/drivers/net/mlx5/mlx5_rxq.c
index a784c8e68c..24ccbc83cd 100644
--- a/drivers/net/mlx5/mlx5_rxq.c
+++ b/drivers/net/mlx5/mlx5_rxq.c
@@ -881,7 +881,7 @@ rxq_setup(struct rxq_ctrl *tmpl)
 	if (elts == NULL)
 		return ENOMEM;
 	tmpl->rxq.rq_db = rwq->rq.db;
-	tmpl->rxq.cqe_n = ibcq->cqe + 1;
+	tmpl->rxq.cqe_n = log2above(ibcq->cqe);
 	tmpl->rxq.cq_ci = 0;
 	tmpl->rxq.rq_ci = 0;
 	tmpl->rxq.cq_db = cq->dbrec;
diff --git a/drivers/net/mlx5/mlx5_rxtx.c b/drivers/net/mlx5/mlx5_rxtx.c
index ebc094e811..e3b03b3847 100644
--- a/drivers/net/mlx5/mlx5_rxtx.c
+++ b/drivers/net/mlx5/mlx5_rxtx.c
@@ -166,8 +166,8 @@ txq_complete(struct txq *txq) __attribute__((always_inline));
 static inline void
 txq_complete(struct txq *txq)
 {
-	const unsigned int cqe_n = txq->cqe_n;
 	const unsigned int elts_n = 1 << txq->elts_n;
+	const unsigned int cqe_n = 1 << txq->cqe_n;
 	const unsigned int cqe_cnt = cqe_n - 1;
 	uint16_t elts_free = txq->elts_tail;
 	uint16_t elts_tail;
@@ -427,9 +427,9 @@ mlx5_tx_dbrec(struct txq *txq)
 static inline void
 tx_prefetch_cqe(struct txq *txq, uint16_t ci)
 {
-	volatile struct mlx5_cqe64 *cqe;
+	volatile struct mlx5_cqe *cqe;
 
-	cqe = &(*txq->cqes)[ci & (txq->cqe_n - 1)].cqe64;
+	cqe = &(*txq->cqes)[ci & ((1 << txq->cqe_n) - 1)];
 	rte_prefetch0(cqe);
 }
 
@@ -1272,8 +1272,8 @@ uint16_t
 mlx5_rx_burst(void *dpdk_rxq, struct rte_mbuf **pkts, uint16_t pkts_n)
 {
 	struct rxq *rxq = dpdk_rxq;
-	const unsigned int cqe_cnt = rxq->cqe_n - 1;
 	const unsigned int wqe_cnt = (1 << rxq->elts_n) - 1;
+	const unsigned int cqe_cnt = (1 << rxq->cqe_n) - 1;
 	const unsigned int sges_n = rxq->sges_n;
 	struct rte_mbuf *pkt = NULL;
 	struct rte_mbuf *seg = NULL;
diff --git a/drivers/net/mlx5/mlx5_rxtx.h b/drivers/net/mlx5/mlx5_rxtx.h
index 71a7d58e30..78a1f8a28b 100644
--- a/drivers/net/mlx5/mlx5_rxtx.h
+++ b/drivers/net/mlx5/mlx5_rxtx.h
@@ -110,13 +110,13 @@ struct rxq {
 	unsigned int vlan_strip:1; /* Enable VLAN stripping. */
 	unsigned int crc_present:1; /* CRC must be subtracted. */
 	unsigned int sges_n:2; /* Log 2 of SGEs (max buffers per packet). */
+	unsigned int cqe_n:4; /* Log 2 of CQ elements. */
 	unsigned int elts_n:4; /* Log 2 of Mbufs. */
 	unsigned int port_id:8;
 	volatile uint32_t *rq_db;
 	volatile uint32_t *cq_db;
 	uint16_t rq_ci;
 	uint16_t cq_ci;
-	uint16_t cqe_n; /* Number of CQ elements. */
 	volatile struct mlx5_wqe_data_seg(*wqes)[];
 	volatile struct mlx5_cqe(*cqes)[];
 	struct rxq_zip zip; /* Compressed context. */
@@ -245,10 +245,10 @@ struct txq {
 	uint16_t elts_tail; /* First element awaiting completion. */
 	uint16_t elts_comp; /* Counter since last completion request. */
 	uint16_t cq_ci; /* Consumer index for completion queue. */
-	uint16_t cqe_n; /* Number of CQ elements. */
 	uint16_t wqe_ci; /* Consumer index for work queue. */
 	uint16_t wqe_n; /* Number of WQ elements. */
 	uint16_t elts_n:4; /* (*elts)[] length (in log2). */
+	uint16_t cqe_n:4; /* Number of CQ elements (in log2). */
 	uint16_t bf_offset; /* Blueflame offset. */
 	uint16_t bf_buf_size; /* Blueflame size. */
 	uint16_t max_inline; /* Multiple of RTE_CACHE_LINE_SIZE to inline. */
diff --git a/drivers/net/mlx5/mlx5_txq.c b/drivers/net/mlx5/mlx5_txq.c
index d9e61ed9a0..3fc9e55f09 100644
--- a/drivers/net/mlx5/mlx5_txq.c
+++ b/drivers/net/mlx5/mlx5_txq.c
@@ -212,7 +212,7 @@ txq_setup(struct txq_ctrl *tmpl, struct txq_ctrl *txq_ctrl)
 		      "it should be set to %u", RTE_CACHE_LINE_SIZE);
 		return EINVAL;
 	}
-	tmpl->txq.cqe_n = ibcq->cqe + 1;
+	tmpl->txq.cqe_n = log2above(ibcq->cqe);
 	tmpl->txq.qp_num_8s = qp->ctrl_seg.qp_num << 8;
 	tmpl->txq.wqes =
 		(volatile struct mlx5_wqe64 (*)[])