From: Zhe Tao <zhe.tao@intel.com>
To: dev@dpdk.org
Subject: [dpdk-dev] [PATCH 2/8 v2] add vector PMD TX for FVL
Date: Fri, 30 Oct 2015 18:52:10 +0800 [thread overview]
Message-ID: <1446202336-8723-3-git-send-email-zhe.tao@intel.com> (raw)
In-Reply-To: <1446202336-8723-1-git-send-email-zhe.tao@intel.com>
The way to increase the performance of the vPMD TX is to use some fast mbuf
release method compares to the scalar TX.
Signed-off-by: Zhe Tao <zhe.tao@intel.com>
---
drivers/net/i40e/i40e_rxtx.c | 8 ++
drivers/net/i40e/i40e_rxtx.h | 3 +
drivers/net/i40e/i40e_rxtx_vec.c | 162 +++++++++++++++++++++++++++++++++++++++
3 files changed, 173 insertions(+)
diff --git a/drivers/net/i40e/i40e_rxtx.c b/drivers/net/i40e/i40e_rxtx.c
index dfdc7d5..8fac220 100644
--- a/drivers/net/i40e/i40e_rxtx.c
+++ b/drivers/net/i40e/i40e_rxtx.c
@@ -3085,3 +3085,11 @@ i40e_rx_queue_release_mbufs_vec(struct i40e_rx_queue __rte_unused*rxq)
{
return;
}
+
+uint16_t __attribute__((weak))
+i40e_xmit_pkts_vec(void __rte_unused *tx_queue,
+ struct rte_mbuf __rte_unused **tx_pkts,
+ uint16_t __rte_unused nb_pkts)
+{
+ return 0;
+}
diff --git a/drivers/net/i40e/i40e_rxtx.h b/drivers/net/i40e/i40e_rxtx.h
index 61d9a0e..3b0af9a 100644
--- a/drivers/net/i40e/i40e_rxtx.h
+++ b/drivers/net/i40e/i40e_rxtx.h
@@ -224,6 +224,9 @@ int i40e_dev_rx_descriptor_done(void *rx_queue, uint16_t offset);
uint16_t i40e_recv_pkts_vec(void *rx_queue, struct rte_mbuf **rx_pkts,
uint16_t nb_pkts);
int i40e_rxq_vec_setup(struct i40e_rx_queue *rxq);
+int i40e_txq_vec_setup(struct i40e_tx_queue *txq);
void i40e_rx_queue_release_mbufs_vec(struct i40e_rx_queue *rxq);
+uint16_t i40e_xmit_pkts_vec(void *tx_queue, struct rte_mbuf **tx_pkts,
+ uint16_t nb_pkts);
#endif /* _I40E_RXTX_H_ */
diff --git a/drivers/net/i40e/i40e_rxtx_vec.c b/drivers/net/i40e/i40e_rxtx_vec.c
index 470e6fa..4ae334d 100644
--- a/drivers/net/i40e/i40e_rxtx_vec.c
+++ b/drivers/net/i40e/i40e_rxtx_vec.c
@@ -447,6 +447,162 @@ i40e_recv_pkts_vec(void *rx_queue, struct rte_mbuf **rx_pkts,
return _recv_raw_pkts_vec(rx_queue, rx_pkts, nb_pkts, NULL);
}
+static inline void
+vtx1(volatile struct i40e_tx_desc *txdp,
+ struct rte_mbuf *pkt, uint64_t flags)
+{
+ uint64_t high_qw = (I40E_TX_DESC_DTYPE_DATA |
+ ((uint64_t)flags << I40E_TXD_QW1_CMD_SHIFT) |
+ ((uint64_t)pkt->data_len << I40E_TXD_QW1_TX_BUF_SZ_SHIFT));
+
+ __m128i descriptor = _mm_set_epi64x(high_qw,
+ pkt->buf_physaddr + pkt->data_off);
+ _mm_store_si128((__m128i *)txdp, descriptor);
+}
+
+static inline void
+vtx(volatile struct i40e_tx_desc *txdp,
+ struct rte_mbuf **pkt, uint16_t nb_pkts, uint64_t flags)
+{
+ int i;
+
+ for (i = 0; i < nb_pkts; ++i, ++txdp, ++pkt)
+ vtx1(txdp, *pkt, flags);
+}
+
+static inline int __attribute__((always_inline))
+i40e_tx_free_bufs(struct i40e_tx_queue *txq)
+{
+ struct i40e_tx_entry *txep;
+ uint32_t n;
+ uint32_t i;
+ int nb_free = 0;
+ struct rte_mbuf *m, *free[RTE_I40E_TX_MAX_FREE_BUF_SZ];
+ /* check DD bits on threshold descriptor */
+ if ((txq->tx_ring[txq->tx_next_dd].cmd_type_offset_bsz &
+ rte_cpu_to_le_64(I40E_TXD_QW1_DTYPE_MASK)) !=
+ rte_cpu_to_le_64(I40E_TX_DESC_DTYPE_DESC_DONE))
+ return 0;
+
+ n = txq->tx_rs_thresh;
+
+ /* first buffer to free from S/W ring is at index
+ * tx_next_dd - (tx_rs_thresh-1)
+ */
+ txep = &txq->sw_ring[txq->tx_next_dd - (n - 1)];
+ m = __rte_pktmbuf_prefree_seg(txep[0].mbuf);
+ if (likely(m != NULL)) {
+ free[0] = m;
+ nb_free = 1;
+ for (i = 1; i < n; i++) {
+ m = __rte_pktmbuf_prefree_seg(txep[i].mbuf);
+ if (likely(m != NULL)) {
+ if (likely(m->pool == free[0]->pool)) {
+ free[nb_free++] = m;
+ } else {
+ rte_mempool_put_bulk(free[0]->pool,
+ (void *)free,
+ nb_free);
+ free[0] = m;
+ nb_free = 1;
+ }
+ }
+ }
+ rte_mempool_put_bulk(free[0]->pool, (void **)free, nb_free);
+ } else {
+ for (i = 1; i < n; i++) {
+ m = __rte_pktmbuf_prefree_seg(txep[i].mbuf);
+ if (m != NULL)
+ rte_mempool_put(m->pool, m);
+ }
+ }
+
+ /* buffers were freed, update counters */
+ txq->nb_tx_free = (uint16_t)(txq->nb_tx_free + txq->tx_rs_thresh);
+ txq->tx_next_dd = (uint16_t)(txq->tx_next_dd + txq->tx_rs_thresh);
+ if (txq->tx_next_dd >= txq->nb_tx_desc)
+ txq->tx_next_dd = (uint16_t)(txq->tx_rs_thresh - 1);
+
+ return txq->tx_rs_thresh;
+}
+
+static inline void __attribute__((always_inline))
+tx_backlog_entry(struct i40e_tx_entry *txep,
+ struct rte_mbuf **tx_pkts, uint16_t nb_pkts)
+{
+ int i;
+
+ for (i = 0; i < (int)nb_pkts; ++i)
+ txep[i].mbuf = tx_pkts[i];
+}
+
+uint16_t
+i40e_xmit_pkts_vec(void *tx_queue, struct rte_mbuf **tx_pkts,
+ uint16_t nb_pkts)
+{
+ struct i40e_tx_queue *txq = (struct i40e_tx_queue *)tx_queue;
+ volatile struct i40e_tx_desc *txdp;
+ struct i40e_tx_entry *txep;
+ uint16_t n, nb_commit, tx_id;
+ uint64_t flags = I40E_TD_CMD;
+ uint64_t rs = I40E_TX_DESC_CMD_RS | I40E_TD_CMD;
+ int i;
+
+ /* cross rx_thresh boundary is not allowed */
+ nb_pkts = RTE_MIN(nb_pkts, txq->tx_rs_thresh);
+
+ if (txq->nb_tx_free < txq->tx_free_thresh)
+ i40e_tx_free_bufs(txq);
+
+ nb_commit = nb_pkts = (uint16_t)RTE_MIN(txq->nb_tx_free, nb_pkts);
+ if (unlikely(nb_pkts == 0))
+ return 0;
+
+ tx_id = txq->tx_tail;
+ txdp = &txq->tx_ring[tx_id];
+ txep = &txq->sw_ring[tx_id];
+
+ txq->nb_tx_free = (uint16_t)(txq->nb_tx_free - nb_pkts);
+
+ n = (uint16_t)(txq->nb_tx_desc - tx_id);
+ if (nb_commit >= n) {
+ tx_backlog_entry(txep, tx_pkts, n);
+
+ for (i = 0; i < n - 1; ++i, ++tx_pkts, ++txdp)
+ vtx1(txdp, *tx_pkts, flags);
+
+ vtx1(txdp, *tx_pkts++, rs);
+
+ nb_commit = (uint16_t)(nb_commit - n);
+
+ tx_id = 0;
+ txq->tx_next_rs = (uint16_t)(txq->tx_rs_thresh - 1);
+
+ /* avoid reach the end of ring */
+ txdp = &txq->tx_ring[tx_id];
+ txep = &txq->sw_ring[tx_id];
+ }
+
+ tx_backlog_entry(txep, tx_pkts, nb_commit);
+
+ vtx(txdp, tx_pkts, nb_commit, flags);
+
+ tx_id = (uint16_t)(tx_id + nb_commit);
+ if (tx_id > txq->tx_next_rs) {
+ txq->tx_ring[txq->tx_next_rs].cmd_type_offset_bsz |=
+ rte_cpu_to_le_64(((uint64_t)I40E_TX_DESC_CMD_RS) <<
+ I40E_TXD_QW1_CMD_SHIFT);
+ txq->tx_next_rs =
+ (uint16_t)(txq->tx_next_rs + txq->tx_rs_thresh);
+ }
+
+ txq->tx_tail = tx_id;
+
+ I40E_PCI_REG_WRITE(txq->qtx_tail, txq->tx_tail);
+
+ return nb_pkts;
+}
+
void __attribute__((cold))
i40e_rx_queue_release_mbufs_vec(struct i40e_rx_queue *rxq)
{
@@ -482,3 +638,9 @@ i40e_rxq_vec_setup(struct i40e_rx_queue *rxq)
rxq->mbuf_initializer = *(uint64_t *)p;
return 0;
}
+
+int __attribute__((cold))
+i40e_txq_vec_setup(struct i40e_tx_queue __rte_unused *txq)
+{
+ return 0;
+}
--
1.9.3
next prev parent reply other threads:[~2015-10-30 10:52 UTC|newest]
Thread overview: 36+ messages / expand[flat|nested] mbox.gz Atom feed top
2015-09-27 17:05 [dpdk-dev] [PATCH 0/4] i40e: add vector PMD support " Zhe Tao
2015-09-27 17:05 ` [dpdk-dev] [PATCH 1/4] add vector PMD RX " Zhe Tao
2015-09-27 23:15 ` Ananyev, Konstantin
[not found] ` <2601191342CEEE43887BDE71AB97725836A9CE35@irsmsx105.ger.corp.intel.com>
2015-09-29 13:06 ` [dpdk-dev] FW: " Ananyev, Konstantin
2015-09-29 14:27 ` [dpdk-dev] " Bruce Richardson
2015-09-27 17:05 ` [dpdk-dev] [PATCH 2/4] add vector PMD TX " Zhe Tao
2015-09-27 17:05 ` [dpdk-dev] [PATCH 3/4] add vector PMD scatter RX " Zhe Tao
2015-09-27 17:05 ` [dpdk-dev] [PATCH 4/4] add RX and TX selection function " Zhe Tao
2015-10-30 10:52 ` [dpdk-dev] [PATCH 0/8 v2] i40e: add vector PMD support " Zhe Tao
2015-10-30 10:52 ` [dpdk-dev] [PATCH 1/8 v2] add vector PMD RX " Zhe Tao
2015-10-30 10:52 ` Zhe Tao [this message]
2015-10-30 10:52 ` [dpdk-dev] [PATCH 3/8 v2] add vector PMD scatter " Zhe Tao
2015-10-30 10:52 ` [dpdk-dev] [PATCH 4/8 v2] add RX and TX selection function " Zhe Tao
2015-10-30 10:52 ` [dpdk-dev] [PATCH 5/8 v2] edit the comments Zhe Tao
2015-10-30 10:52 ` [dpdk-dev] [PATCH 6/8 v2] change the postion of data prefetch for splitter packets Zhe Tao
2015-10-30 10:52 ` [dpdk-dev] [PATCH 7/8 v2] move all the extra definition out of share code Zhe Tao
2015-10-30 10:52 ` [dpdk-dev] [PATCH 8/8 v2] update the release note Zhe Tao
2015-10-30 13:01 ` [dpdk-dev] [PATCH 0/4 v3] i40e: add vector PMD support for FVL Zhe Tao
2015-10-30 13:01 ` [dpdk-dev] [PATCH 1/4 v3] add vector PMD RX " Zhe Tao
2015-10-30 13:40 ` Liang, Cunming
2015-10-30 13:01 ` [dpdk-dev] [PATCH 2/4 v3] add vector PMD TX " Zhe Tao
2015-10-30 13:29 ` Liang, Cunming
2015-11-03 21:20 ` Stephen Hemminger
2015-10-30 13:01 ` [dpdk-dev] [PATCH 3/4 v3] add vector PMD scatter RX " Zhe Tao
2015-10-30 13:46 ` Liang, Cunming
2015-10-30 13:01 ` [dpdk-dev] [PATCH 4/4 v3] add RX and TX selection function " Zhe Tao
2015-10-30 14:16 ` [dpdk-dev] [PATCH 0/4 v4] i40e: add vector PMD support " Zhe Tao
2015-10-30 14:16 ` [dpdk-dev] [PATCH 1/4 v4] add vector PMD RX " Zhe Tao
2015-10-30 15:33 ` Thomas Monjalon
2015-10-30 14:16 ` [dpdk-dev] [PATCH 2/4 v4] add vector PMD TX " Zhe Tao
2015-10-30 14:16 ` [dpdk-dev] [PATCH 3/4 v4] add vector PMD scatter RX " Zhe Tao
2015-10-30 14:16 ` [dpdk-dev] [PATCH 4/4 v4] add RX and TX selection function " Zhe Tao
2015-10-30 15:59 ` Thomas Monjalon
2015-10-30 14:24 ` [dpdk-dev] [PATCH 0/4 v4] i40e: add vector PMD support " Liang, Cunming
2015-10-30 15:41 ` Thomas Monjalon
2015-10-30 16:02 ` Thomas Monjalon
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=1446202336-8723-3-git-send-email-zhe.tao@intel.com \
--to=zhe.tao@intel.com \
--cc=dev@dpdk.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).