diff mbox series

[v2,3/3] app/testpmd: further improve MAC swap performance for x86

Message ID 20181122173805.79555-4-qi.z.zhang@intel.com (mailing list archive)
State Superseded, archived
Delegated to: Ferruh Yigit
Headers
Series improve MAC swap performance |

Checks

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

Commit Message

Qi Zhang Nov. 22, 2018, 5:38 p.m. UTC 
  Do four packets macswap in same loop iterate to squeeze more
CPU cycles.

Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
---
 app/test-pmd/macswap_sse.h | 65 ++++++++++++++++++++++++++++++++++++++--------
 1 file changed, 54 insertions(+), 11 deletions(-)
diff mbox series

Patch

diff --git a/app/test-pmd/macswap_sse.h b/app/test-pmd/macswap_sse.h
index d5b0f6a21..0649539c2 100644
--- a/app/test-pmd/macswap_sse.h
+++ b/app/test-pmd/macswap_sse.h
@@ -10,11 +10,12 @@  static inline void
 do_macswap(struct rte_mbuf *pkts[], uint16_t nb,
 		struct rte_port *txp)
 {
-	struct ether_hdr *eth_hdr;
-	struct rte_mbuf *mb;
+	struct ether_hdr *eth_hdr[4];
+	struct rte_mbuf *mb[4];
 	uint64_t ol_flags;
 	int i;
-	__m128i addr;
+	int r;
+	__m128i addr0, addr1, addr2, addr3;
 	__m128i shfl_msk = _mm_set_epi8(15, 14, 13, 12,
 					5, 4, 3, 2,
 					1, 0, 11, 10,
@@ -22,19 +23,61 @@  do_macswap(struct rte_mbuf *pkts[], uint16_t nb,
 
 	ol_flags = ol_flags_init(txp->dev_conf.txmode.offloads);
 
-	for (i = 0; i < nb; i++) {
-		if (likely(i < nb - 1))
+	i = 0;
+	r = nb;
+
+	while (r >= 4) {
+		mb[0] = pkts[i++];
+		eth_hdr[0] = rte_pktmbuf_mtod(mb[0], struct ether_hdr *);
+		addr0 = _mm_loadu_si128((__m128i *)eth_hdr[0]);
+
+		mb[1] = pkts[i++];
+		eth_hdr[1] = rte_pktmbuf_mtod(mb[1], struct ether_hdr *);
+		addr1 = _mm_loadu_si128((__m128i *)eth_hdr[1]);
+
+
+		mb[2] = pkts[i++];
+		eth_hdr[2] = rte_pktmbuf_mtod(mb[2], struct ether_hdr *);
+		addr2 = _mm_loadu_si128((__m128i *)eth_hdr[2]);
+
+		mb[3] = pkts[i++];
+		eth_hdr[3] = rte_pktmbuf_mtod(mb[3], struct ether_hdr *);
+		addr3 = _mm_loadu_si128((__m128i *)eth_hdr[3]);
+
+		addr0 = _mm_shuffle_epi8(addr0, shfl_msk);
+		addr1 = _mm_shuffle_epi8(addr1, shfl_msk);
+		addr2 = _mm_shuffle_epi8(addr2, shfl_msk);
+		addr3 = _mm_shuffle_epi8(addr3, shfl_msk);
+
+		_mm_storeu_si128((__m128i *)eth_hdr[0], addr0);
+		_mm_storeu_si128((__m128i *)eth_hdr[1], addr1);
+		_mm_storeu_si128((__m128i *)eth_hdr[2], addr2);
+		_mm_storeu_si128((__m128i *)eth_hdr[3], addr3);
+
+		mbuf_field_set(mb[0], ol_flags, txp->tx_vlan_id,
+				txp->tx_vlan_id_outer);
+		mbuf_field_set(mb[1], ol_flags, txp->tx_vlan_id,
+				txp->tx_vlan_id_outer);
+		mbuf_field_set(mb[2], ol_flags, txp->tx_vlan_id,
+				txp->tx_vlan_id_outer);
+		mbuf_field_set(mb[3], ol_flags, txp->tx_vlan_id,
+				txp->tx_vlan_id_outer);
+		r -= 4;
+	}
+
+	for ( ; i < nb; i++) {
+		if (i < nb - 1)
 			rte_prefetch0(rte_pktmbuf_mtod(pkts[i+1], void *));
-		mb = pkts[i];
+		mb[0] = pkts[i];
 
-		eth_hdr = rte_pktmbuf_mtod(mb, struct ether_hdr *);
+		eth_hdr[0] = rte_pktmbuf_mtod(mb[0], struct ether_hdr *);
 
 		/* Swap dest and src mac addresses. */
-		addr = _mm_loadu_si128((__m128i *)eth_hdr);
-		addr = _mm_shuffle_epi8(addr, shfl_msk);
-		_mm_storeu_si128((__m128i *)eth_hdr, addr);
+		addr0 = _mm_loadu_si128((__m128i *)eth_hdr);
+		addr0 = _mm_shuffle_epi8(addr0, shfl_msk);
+		_mm_storeu_si128((__m128i *)eth_hdr[0], addr0);
 
-		mbuf_field_set(mb, ol_flags, txp->tx_vlan_id,
+		mbuf_field_set(mb[0], ol_flags, txp->tx_vlan_id,
 				txp->tx_vlan_id_outer);
 	}
 }