DPDK patches and discussions
 help / color / mirror / Atom feed
* [dpdk-dev] [PATCH] net/iavf: enable AVX2 for iavf
@ 2019-09-04 10:06 Leyi Rong
  2019-09-04 13:29 ` Aaron Conole
                   ` (4 more replies)
  0 siblings, 5 replies; 11+ messages in thread
From: Leyi Rong @ 2019-09-04 10:06 UTC (permalink / raw)
  To: wenzhuo.lu, qi.z.zhang, xiaolong.ye; +Cc: dev, Leyi Rong

This patch enables AVX data path for iavf PMD.

Signed-off-by: Leyi Rong <leyi.rong@intel.com>
---
 drivers/net/iavf/Makefile               |  21 +
 drivers/net/iavf/iavf_rxtx.c            |  62 +-
 drivers/net/iavf/iavf_rxtx.h            |  10 +
 drivers/net/iavf/iavf_rxtx_vec_avx2.c   | 867 ++++++++++++++++++++++++
 drivers/net/iavf/iavf_rxtx_vec_common.h |  72 ++
 drivers/net/iavf/iavf_rxtx_vec_sse.c    |  12 +
 drivers/net/iavf/meson.build            |  17 +
 7 files changed, 1041 insertions(+), 20 deletions(-)
 create mode 100644 drivers/net/iavf/iavf_rxtx_vec_avx2.c

diff --git a/drivers/net/iavf/Makefile b/drivers/net/iavf/Makefile
index cd74e14ab..cf0ed99f0 100644
--- a/drivers/net/iavf/Makefile
+++ b/drivers/net/iavf/Makefile
@@ -48,4 +48,25 @@ ifeq ($(CONFIG_RTE_ARCH_X86), y)
 SRCS-$(CONFIG_RTE_LIBRTE_IAVF_INC_VECTOR) += iavf_rxtx_vec_sse.c
 endif
 
+ifeq ($(CONFIG_RTE_LIBRTE_IAVF_INC_VECTOR), y)
+	ifeq ($(findstring RTE_MACHINE_CPUFLAG_AVX2,$(CFLAGS)),RTE_MACHINE_CPUFLAG_AVX2)
+		CC_AVX2_SUPPORT=1
+	else
+		CC_AVX2_SUPPORT=\
+		$(shell $(CC) -march=core-avx2 -dM -E - </dev/null 2>&1 | \
+		grep -q AVX2 && echo 1)
+		ifeq ($(CC_AVX2_SUPPORT), 1)
+			ifeq ($(CONFIG_RTE_TOOLCHAIN_ICC),y)
+				CFLAGS_iavf_rxtx_vec_avx2.o += -march=core-avx2
+			else
+				CFLAGS_iavf_rxtx_vec_avx2.o += -mavx2
+			endif
+		endif
+	endif
+endif
+
+ifeq ($(CC_AVX2_SUPPORT), 1)
+	SRCS-$(CONFIG_RTE_LIBRTE_IAVF_PMD) += iavf_rxtx_vec_avx2.c
+endif
+
 include $(RTE_SDK)/mk/rte.lib.mk
diff --git a/drivers/net/iavf/iavf_rxtx.c b/drivers/net/iavf/iavf_rxtx.c
index 22d7bb612..2eb496d5f 100644
--- a/drivers/net/iavf/iavf_rxtx.c
+++ b/drivers/net/iavf/iavf_rxtx.c
@@ -1743,23 +1743,36 @@ iavf_set_rx_function(struct rte_eth_dev *dev)
 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 	struct iavf_rx_queue *rxq;
 	int i;
+	bool use_avx2 = false;
 
-	if (adapter->rx_vec_allowed) {
-		if (dev->data->scattered_rx) {
-			PMD_DRV_LOG(DEBUG, "Using Vector Scattered Rx callback"
-				    " (port=%d).", dev->data->port_id);
-			dev->rx_pkt_burst = iavf_recv_scattered_pkts_vec;
-		} else {
-			PMD_DRV_LOG(DEBUG, "Using Vector Rx callback"
-				    " (port=%d).", dev->data->port_id);
-			dev->rx_pkt_burst = iavf_recv_pkts_vec;
-		}
+	if (!iavf_rx_vec_dev_check(dev)) {
 		for (i = 0; i < dev->data->nb_rx_queues; i++) {
 			rxq = dev->data->rx_queues[i];
-			if (!rxq)
-				continue;
-			iavf_rxq_vec_setup(rxq);
+			(void)iavf_rxq_vec_setup(rxq);
 		}
+
+		if (rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1 ||
+		    rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX512F) == 1)
+			use_avx2 = true;
+
+		if (dev->data->scattered_rx) {
+			PMD_DRV_LOG(DEBUG,
+				    "Using %sVector Scattered Rx (port %d).",
+				    use_avx2 ? "avx2 " : "",
+				    dev->data->port_id);
+			dev->rx_pkt_burst = use_avx2 ?
+					    iavf_recv_scattered_pkts_vec_avx2 :
+					    iavf_recv_scattered_pkts_vec;
+		} else {
+			PMD_DRV_LOG(DEBUG, "Using %sVector Rx (port %d).",
+				    use_avx2 ? "avx2 " : "",
+				    dev->data->port_id);
+			dev->rx_pkt_burst = use_avx2 ?
+					    iavf_recv_pkts_vec_avx2 :
+					    iavf_recv_pkts_vec;
+		}
+
+		return;
 	} else if (dev->data->scattered_rx) {
 		PMD_DRV_LOG(DEBUG, "Using a Scattered Rx callback (port=%d).",
 			    dev->data->port_id);
@@ -1779,22 +1792,31 @@ iavf_set_rx_function(struct rte_eth_dev *dev)
 void
 iavf_set_tx_function(struct rte_eth_dev *dev)
 {
-	struct iavf_adapter *adapter =
-		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 	struct iavf_tx_queue *txq;
 	int i;
+	bool use_avx2 = false;
 
-	if (adapter->tx_vec_allowed) {
-		PMD_DRV_LOG(DEBUG, "Using Vector Tx callback (port=%d).",
-			    dev->data->port_id);
-		dev->tx_pkt_burst = iavf_xmit_pkts_vec;
-		dev->tx_pkt_prepare = NULL;
+	if (!iavf_tx_vec_dev_check(dev)) {
 		for (i = 0; i < dev->data->nb_tx_queues; i++) {
 			txq = dev->data->tx_queues[i];
 			if (!txq)
 				continue;
 			iavf_txq_vec_setup(txq);
 		}
+
+		if (rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1 ||
+		    rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX512F) == 1)
+			use_avx2 = true;
+
+		PMD_DRV_LOG(DEBUG, "Using %sVector Tx (port %d).",
+			    use_avx2 ? "avx2 " : "",
+			    dev->data->port_id);
+		dev->tx_pkt_burst = use_avx2 ?
+				    iavf_xmit_pkts_vec_avx2 :
+				    iavf_xmit_pkts_vec;
+		dev->tx_pkt_prepare = NULL;
+
+		return;
 	} else {
 		PMD_DRV_LOG(DEBUG, "Using Basic Tx callback (port=%d).",
 			    dev->data->port_id);
diff --git a/drivers/net/iavf/iavf_rxtx.h b/drivers/net/iavf/iavf_rxtx.h
index c86720bda..7dde392bf 100644
--- a/drivers/net/iavf/iavf_rxtx.h
+++ b/drivers/net/iavf/iavf_rxtx.h
@@ -19,6 +19,7 @@
 /* used for Vector PMD */
 #define IAVF_VPMD_RX_MAX_BURST    32
 #define IAVF_VPMD_TX_MAX_BURST    32
+#define IAVF_RXQ_REARM_THRESH     32
 #define IAVF_VPMD_DESCS_PER_LOOP  4
 #define IAVF_VPMD_TX_MAX_FREE_BUF 64
 
@@ -200,6 +201,15 @@ uint16_t iavf_recv_scattered_pkts_vec(void *rx_queue,
 				     uint16_t nb_pkts);
 uint16_t iavf_xmit_fixed_burst_vec(void *tx_queue, struct rte_mbuf **tx_pkts,
 				  uint16_t nb_pkts);
+uint16_t iavf_recv_pkts_vec_avx2(void *rx_queue, struct rte_mbuf **rx_pkts,
+				 uint16_t nb_pkts);
+uint16_t iavf_recv_scattered_pkts_vec_avx2(void *rx_queue,
+					   struct rte_mbuf **rx_pkts,
+					   uint16_t nb_pkts);
+uint16_t iavf_xmit_pkts_vec_avx2(void *tx_queue, struct rte_mbuf **tx_pkts,
+				 uint16_t nb_pkts);
+int iavf_rx_vec_dev_check(struct rte_eth_dev *dev);
+int iavf_tx_vec_dev_check(struct rte_eth_dev *dev);
 int iavf_rxq_vec_setup(struct iavf_rx_queue *rxq);
 int iavf_txq_vec_setup(struct iavf_tx_queue *txq);
 
diff --git a/drivers/net/iavf/iavf_rxtx_vec_avx2.c b/drivers/net/iavf/iavf_rxtx_vec_avx2.c
new file mode 100644
index 000000000..f0c00be56
--- /dev/null
+++ b/drivers/net/iavf/iavf_rxtx_vec_avx2.c
@@ -0,0 +1,867 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2019 Intel Corporation
+ */
+
+#include "base/iavf_prototype.h"
+#include "iavf_rxtx_vec_common.h"
+
+#include <x86intrin.h>
+
+#ifndef __INTEL_COMPILER
+#pragma GCC diagnostic ignored "-Wcast-qual"
+#endif
+
+static inline void
+iavf_rxq_rearm(struct iavf_rx_queue *rxq)
+{
+	int i;
+	uint16_t rx_id;
+	volatile union iavf_rx_desc *rxdp;
+	struct rte_mbuf **rxp = &rxq->sw_ring[rxq->rxrearm_start];
+
+	rxdp = rxq->rx_ring + rxq->rxrearm_start;
+
+	/* Pull 'n' more MBUFs into the software ring */
+	if (rte_mempool_get_bulk(rxq->mp,
+				 (void *)rxp,
+				 IAVF_RXQ_REARM_THRESH) < 0) {
+		if (rxq->rxrearm_nb + IAVF_RXQ_REARM_THRESH >=
+		    rxq->nb_rx_desc) {
+			__m128i dma_addr0;
+
+			dma_addr0 = _mm_setzero_si128();
+			for (i = 0; i < IAVF_VPMD_DESCS_PER_LOOP; i++) {
+				rxp[i] = &rxq->fake_mbuf;
+				_mm_store_si128((__m128i *)&rxdp[i].read,
+						dma_addr0);
+			}
+		}
+		rte_eth_devices[rxq->port_id].data->rx_mbuf_alloc_failed +=
+			IAVF_RXQ_REARM_THRESH;
+		return;
+	}
+
+#ifndef RTE_LIBRTE_IAVF_16BYTE_RX_DESC
+	struct rte_mbuf *mb0, *mb1;
+	__m128i dma_addr0, dma_addr1;
+	__m128i hdr_room = _mm_set_epi64x(RTE_PKTMBUF_HEADROOM,
+			RTE_PKTMBUF_HEADROOM);
+	/* Initialize the mbufs in vector, process 2 mbufs in one loop */
+	for (i = 0; i < IAVF_RXQ_REARM_THRESH; i += 2, rxp += 2) {
+		__m128i vaddr0, vaddr1;
+
+		mb0 = rxp[0];
+		mb1 = rxp[1];
+
+		/* load buf_addr(lo 64bit) and buf_physaddr(hi 64bit) */
+		RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, buf_physaddr) !=
+				offsetof(struct rte_mbuf, buf_addr) + 8);
+		vaddr0 = _mm_loadu_si128((__m128i *)&mb0->buf_addr);
+		vaddr1 = _mm_loadu_si128((__m128i *)&mb1->buf_addr);
+
+		/* convert pa to dma_addr hdr/data */
+		dma_addr0 = _mm_unpackhi_epi64(vaddr0, vaddr0);
+		dma_addr1 = _mm_unpackhi_epi64(vaddr1, vaddr1);
+
+		/* add headroom to pa values */
+		dma_addr0 = _mm_add_epi64(dma_addr0, hdr_room);
+		dma_addr1 = _mm_add_epi64(dma_addr1, hdr_room);
+
+		/* flush desc with pa dma_addr */
+		_mm_store_si128((__m128i *)&rxdp++->read, dma_addr0);
+		_mm_store_si128((__m128i *)&rxdp++->read, dma_addr1);
+	}
+#else
+	struct rte_mbuf *mb0, *mb1, *mb2, *mb3;
+	__m256i dma_addr0_1, dma_addr2_3;
+	__m256i hdr_room = _mm256_set1_epi64x(RTE_PKTMBUF_HEADROOM);
+	/* Initialize the mbufs in vector, process 4 mbufs in one loop */
+	for (i = 0; i < IAVF_RXQ_REARM_THRESH;
+			i += 4, rxp += 4, rxdp += 4) {
+		__m128i vaddr0, vaddr1, vaddr2, vaddr3;
+		__m256i vaddr0_1, vaddr2_3;
+
+		mb0 = rxp[0];
+		mb1 = rxp[1];
+		mb2 = rxp[2];
+		mb3 = rxp[3];
+
+		/* load buf_addr(lo 64bit) and buf_physaddr(hi 64bit) */
+		RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, buf_physaddr) !=
+				offsetof(struct rte_mbuf, buf_addr) + 8);
+		vaddr0 = _mm_loadu_si128((__m128i *)&mb0->buf_addr);
+		vaddr1 = _mm_loadu_si128((__m128i *)&mb1->buf_addr);
+		vaddr2 = _mm_loadu_si128((__m128i *)&mb2->buf_addr);
+		vaddr3 = _mm_loadu_si128((__m128i *)&mb3->buf_addr);
+
+		/**
+		 * merge 0 & 1, by casting 0 to 256-bit and inserting 1
+		 * into the high lanes. Similarly for 2 & 3
+		 */
+		vaddr0_1 =
+			_mm256_inserti128_si256(_mm256_castsi128_si256(vaddr0),
+						vaddr1, 1);
+		vaddr2_3 =
+			_mm256_inserti128_si256(_mm256_castsi128_si256(vaddr2),
+						vaddr3, 1);
+
+		/* convert pa to dma_addr hdr/data */
+		dma_addr0_1 = _mm256_unpackhi_epi64(vaddr0_1, vaddr0_1);
+		dma_addr2_3 = _mm256_unpackhi_epi64(vaddr2_3, vaddr2_3);
+
+		/* add headroom to pa values */
+		dma_addr0_1 = _mm256_add_epi64(dma_addr0_1, hdr_room);
+		dma_addr2_3 = _mm256_add_epi64(dma_addr2_3, hdr_room);
+
+		/* flush desc with pa dma_addr */
+		_mm256_store_si256((__m256i *)&rxdp->read, dma_addr0_1);
+		_mm256_store_si256((__m256i *)&(rxdp + 2)->read, dma_addr2_3);
+	}
+
+#endif
+
+	rxq->rxrearm_start += IAVF_RXQ_REARM_THRESH;
+	if (rxq->rxrearm_start >= rxq->nb_rx_desc)
+		rxq->rxrearm_start = 0;
+
+	rxq->rxrearm_nb -= IAVF_RXQ_REARM_THRESH;
+
+	rx_id = (uint16_t)((rxq->rxrearm_start == 0) ?
+			     (rxq->nb_rx_desc - 1) : (rxq->rxrearm_start - 1));
+
+	/* Update the tail pointer on the NIC */
+	IAVF_PCI_REG_WRITE(rxq->qrx_tail, rx_id);
+}
+
+#define PKTLEN_SHIFT     10
+
+static inline uint16_t
+_iavf_recv_raw_pkts_vec_avx2(struct iavf_rx_queue *rxq,
+			     struct rte_mbuf **rx_pkts,
+			     uint16_t nb_pkts, uint8_t *split_packet)
+{
+#define IAVF_DESCS_PER_LOOP_AVX 8
+
+	/* const uint32_t *ptype_tbl = rxq->vsi->adapter->ptype_tbl; */
+	static const uint32_t type_table[UINT8_MAX + 1] __rte_cache_aligned = {
+		/* [0] reserved */
+		[1] = RTE_PTYPE_L2_ETHER,
+		/* [2] - [21] reserved */
+		[22] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
+			RTE_PTYPE_L4_FRAG,
+		[23] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
+			RTE_PTYPE_L4_NONFRAG,
+		[24] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
+			RTE_PTYPE_L4_UDP,
+		/* [25] reserved */
+		[26] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
+			RTE_PTYPE_L4_TCP,
+		[27] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
+			RTE_PTYPE_L4_SCTP,
+		[28] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
+			RTE_PTYPE_L4_ICMP,
+		/* All others reserved */
+	};
+	const __m256i mbuf_init = _mm256_set_epi64x(0, 0,
+			0, rxq->mbuf_initializer);
+	/* struct iavf_rx_entry *sw_ring = &rxq->sw_ring[rxq->rx_tail]; */
+	struct rte_mbuf **sw_ring = &rxq->sw_ring[rxq->rx_tail];
+	volatile union iavf_rx_desc *rxdp = rxq->rx_ring + rxq->rx_tail;
+	const int avx_aligned = ((rxq->rx_tail & 1) == 0);
+
+	rte_prefetch0(rxdp);
+
+	/* nb_pkts has to be floor-aligned to IAVF_DESCS_PER_LOOP_AVX */
+	nb_pkts = RTE_ALIGN_FLOOR(nb_pkts, IAVF_DESCS_PER_LOOP_AVX);
+
+	/* See if we need to rearm the RX queue - gives the prefetch a bit
+	 * of time to act
+	 */
+	if (rxq->rxrearm_nb > IAVF_RXQ_REARM_THRESH)
+		iavf_rxq_rearm(rxq);
+
+	/* Before we start moving massive data around, check to see if
+	 * there is actually a packet available
+	 */
+	if (!(rxdp->wb.qword1.status_error_len &
+			rte_cpu_to_le_32(1 << IAVF_RX_DESC_STATUS_DD_SHIFT)))
+		return 0;
+
+	/* constants used in processing loop */
+	const __m256i crc_adjust =
+		_mm256_set_epi16
+			(/* first descriptor */
+			 0, 0, 0,       /* ignore non-length fields */
+			 -rxq->crc_len, /* sub crc on data_len */
+			 0,             /* ignore high-16bits of pkt_len */
+			 -rxq->crc_len, /* sub crc on pkt_len */
+			 0, 0,          /* ignore pkt_type field */
+			 /* second descriptor */
+			 0, 0, 0,       /* ignore non-length fields */
+			 -rxq->crc_len, /* sub crc on data_len */
+			 0,             /* ignore high-16bits of pkt_len */
+			 -rxq->crc_len, /* sub crc on pkt_len */
+			 0, 0           /* ignore pkt_type field */
+			);
+
+	/* 8 packets DD mask, LSB in each 32-bit value */
+	const __m256i dd_check = _mm256_set1_epi32(1);
+
+	/* 8 packets EOP mask, second-LSB in each 32-bit value */
+	const __m256i eop_check = _mm256_slli_epi32(dd_check,
+			IAVF_RX_DESC_STATUS_EOF_SHIFT);
+
+	/* mask to shuffle from desc. to mbuf (2 descriptors)*/
+	const __m256i shuf_msk =
+		_mm256_set_epi8
+			(/* first descriptor */
+			 7, 6, 5, 4,  /* octet 4~7, 32bits rss */
+			 3, 2,        /* octet 2~3, low 16 bits vlan_macip */
+			 15, 14,      /* octet 15~14, 16 bits data_len */
+			 0xFF, 0xFF,  /* skip high 16 bits pkt_len, zero out */
+			 15, 14,      /* octet 15~14, low 16 bits pkt_len */
+			 0xFF, 0xFF,  /* pkt_type set as unknown */
+			 0xFF, 0xFF,  /*pkt_type set as unknown */
+			 /* second descriptor */
+			 7, 6, 5, 4,  /* octet 4~7, 32bits rss */
+			 3, 2,        /* octet 2~3, low 16 bits vlan_macip */
+			 15, 14,      /* octet 15~14, 16 bits data_len */
+			 0xFF, 0xFF,  /* skip high 16 bits pkt_len, zero out */
+			 15, 14,      /* octet 15~14, low 16 bits pkt_len */
+			 0xFF, 0xFF,  /* pkt_type set as unknown */
+			 0xFF, 0xFF   /*pkt_type set as unknown */
+			);
+	/**
+	 * compile-time check the above crc and shuffle layout is correct.
+	 * NOTE: the first field (lowest address) is given last in set_epi
+	 * calls above.
+	 */
+	RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, pkt_len) !=
+			offsetof(struct rte_mbuf, rx_descriptor_fields1) + 4);
+	RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, data_len) !=
+			offsetof(struct rte_mbuf, rx_descriptor_fields1) + 8);
+	RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, vlan_tci) !=
+			offsetof(struct rte_mbuf, rx_descriptor_fields1) + 10);
+	RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, hash) !=
+			offsetof(struct rte_mbuf, rx_descriptor_fields1) + 12);
+
+	/* Status/Error flag masks */
+	/**
+	 * mask everything except RSS, flow director and VLAN flags
+	 * bit2 is for VLAN tag, bit11 for flow director indication
+	 * bit13:12 for RSS indication. Bits 3-5 of error
+	 * field (bits 22-24) are for IP/L4 checksum errors
+	 */
+	const __m256i flags_mask =
+		 _mm256_set1_epi32((1 << 2) | (1 << 11) |
+				   (3 << 12) | (7 << 22));
+	/**
+	 * data to be shuffled by result of flag mask. If VLAN bit is set,
+	 * (bit 2), then position 4 in this array will be used in the
+	 * destination
+	 */
+	const __m256i vlan_flags_shuf =
+		_mm256_set_epi32(0, 0, PKT_RX_VLAN | PKT_RX_VLAN_STRIPPED, 0,
+				 0, 0, PKT_RX_VLAN | PKT_RX_VLAN_STRIPPED, 0);
+	/**
+	 * data to be shuffled by result of flag mask, shifted down 11.
+	 * If RSS/FDIR bits are set, shuffle moves appropriate flags in
+	 * place.
+	 */
+	const __m256i rss_flags_shuf =
+		_mm256_set_epi8(0, 0, 0, 0, 0, 0, 0, 0,
+				PKT_RX_RSS_HASH | PKT_RX_FDIR, PKT_RX_RSS_HASH,
+				0, 0, 0, 0, PKT_RX_FDIR, 0,/* end up 128-bits */
+				0, 0, 0, 0, 0, 0, 0, 0,
+				PKT_RX_RSS_HASH | PKT_RX_FDIR, PKT_RX_RSS_HASH,
+				0, 0, 0, 0, PKT_RX_FDIR, 0);
+
+	/**
+	 * data to be shuffled by the result of the flags mask shifted by 22
+	 * bits.  This gives use the l3_l4 flags.
+	 */
+	const __m256i l3_l4_flags_shuf = _mm256_set_epi8(0, 0, 0, 0, 0, 0, 0, 0,
+			/* shift right 1 bit to make sure it not exceed 255 */
+			(PKT_RX_EIP_CKSUM_BAD | PKT_RX_L4_CKSUM_BAD |
+			 PKT_RX_IP_CKSUM_BAD) >> 1,
+			(PKT_RX_IP_CKSUM_GOOD | PKT_RX_EIP_CKSUM_BAD |
+			 PKT_RX_L4_CKSUM_BAD) >> 1,
+			(PKT_RX_EIP_CKSUM_BAD | PKT_RX_IP_CKSUM_BAD) >> 1,
+			(PKT_RX_IP_CKSUM_GOOD | PKT_RX_EIP_CKSUM_BAD) >> 1,
+			(PKT_RX_L4_CKSUM_BAD | PKT_RX_IP_CKSUM_BAD) >> 1,
+			(PKT_RX_IP_CKSUM_GOOD | PKT_RX_L4_CKSUM_BAD) >> 1,
+			PKT_RX_IP_CKSUM_BAD >> 1,
+			(PKT_RX_IP_CKSUM_GOOD | PKT_RX_L4_CKSUM_GOOD) >> 1,
+			/* second 128-bits */
+			0, 0, 0, 0, 0, 0, 0, 0,
+			(PKT_RX_EIP_CKSUM_BAD | PKT_RX_L4_CKSUM_BAD |
+			 PKT_RX_IP_CKSUM_BAD) >> 1,
+			(PKT_RX_IP_CKSUM_GOOD | PKT_RX_EIP_CKSUM_BAD |
+			 PKT_RX_L4_CKSUM_BAD) >> 1,
+			(PKT_RX_EIP_CKSUM_BAD | PKT_RX_IP_CKSUM_BAD) >> 1,
+			(PKT_RX_IP_CKSUM_GOOD | PKT_RX_EIP_CKSUM_BAD) >> 1,
+			(PKT_RX_L4_CKSUM_BAD | PKT_RX_IP_CKSUM_BAD) >> 1,
+			(PKT_RX_IP_CKSUM_GOOD | PKT_RX_L4_CKSUM_BAD) >> 1,
+			PKT_RX_IP_CKSUM_BAD >> 1,
+			(PKT_RX_IP_CKSUM_GOOD | PKT_RX_L4_CKSUM_GOOD) >> 1);
+
+	const __m256i cksum_mask =
+		 _mm256_set1_epi32(PKT_RX_IP_CKSUM_GOOD | PKT_RX_IP_CKSUM_BAD |
+				   PKT_RX_L4_CKSUM_GOOD | PKT_RX_L4_CKSUM_BAD |
+				   PKT_RX_EIP_CKSUM_BAD);
+
+	RTE_SET_USED(avx_aligned); /* for 32B descriptors we don't use this */
+
+	uint16_t i, received;
+
+	for (i = 0, received = 0; i < nb_pkts;
+	     i += IAVF_DESCS_PER_LOOP_AVX,
+	     rxdp += IAVF_DESCS_PER_LOOP_AVX) {
+		/* step 1, copy over 8 mbuf pointers to rx_pkts array */
+		_mm256_storeu_si256((void *)&rx_pkts[i],
+				    _mm256_loadu_si256((void *)&sw_ring[i]));
+#ifdef RTE_ARCH_X86_64
+		_mm256_storeu_si256
+			((void *)&rx_pkts[i + 4],
+			 _mm256_loadu_si256((void *)&sw_ring[i + 4]));
+#endif
+
+		__m256i raw_desc0_1, raw_desc2_3, raw_desc4_5, raw_desc6_7;
+#ifdef RTE_LIBRTE_IAVF_16BYTE_RX_DESC
+		/* for AVX we need alignment otherwise loads are not atomic */
+		if (avx_aligned) {
+			/* load in descriptors, 2 at a time, in reverse order */
+			raw_desc6_7 = _mm256_load_si256((void *)(rxdp + 6));
+			rte_compiler_barrier();
+			raw_desc4_5 = _mm256_load_si256((void *)(rxdp + 4));
+			rte_compiler_barrier();
+			raw_desc2_3 = _mm256_load_si256((void *)(rxdp + 2));
+			rte_compiler_barrier();
+			raw_desc0_1 = _mm256_load_si256((void *)(rxdp + 0));
+		} else
+#endif
+		{
+			const __m128i raw_desc7 =
+				_mm_load_si128((void *)(rxdp + 7));
+			rte_compiler_barrier();
+			const __m128i raw_desc6 =
+				_mm_load_si128((void *)(rxdp + 6));
+			rte_compiler_barrier();
+			const __m128i raw_desc5 =
+				_mm_load_si128((void *)(rxdp + 5));
+			rte_compiler_barrier();
+			const __m128i raw_desc4 =
+				_mm_load_si128((void *)(rxdp + 4));
+			rte_compiler_barrier();
+			const __m128i raw_desc3 =
+				_mm_load_si128((void *)(rxdp + 3));
+			rte_compiler_barrier();
+			const __m128i raw_desc2 =
+				_mm_load_si128((void *)(rxdp + 2));
+			rte_compiler_barrier();
+			const __m128i raw_desc1 =
+				_mm_load_si128((void *)(rxdp + 1));
+			rte_compiler_barrier();
+			const __m128i raw_desc0 =
+				_mm_load_si128((void *)(rxdp + 0));
+
+			raw_desc6_7 =
+				_mm256_inserti128_si256
+					(_mm256_castsi128_si256(raw_desc6),
+					 raw_desc7, 1);
+			raw_desc4_5 =
+				_mm256_inserti128_si256
+					(_mm256_castsi128_si256(raw_desc4),
+					 raw_desc5, 1);
+			raw_desc2_3 =
+				_mm256_inserti128_si256
+					(_mm256_castsi128_si256(raw_desc2),
+					 raw_desc3, 1);
+			raw_desc0_1 =
+				_mm256_inserti128_si256
+					(_mm256_castsi128_si256(raw_desc0),
+					 raw_desc1, 1);
+		}
+
+		if (split_packet) {
+			int j;
+
+			for (j = 0; j < IAVF_DESCS_PER_LOOP_AVX; j++)
+				rte_mbuf_prefetch_part2(rx_pkts[i + j]);
+		}
+
+		/**
+		 * convert descriptors 4-7 into mbufs, adjusting length and
+		 * re-arranging fields. Then write into the mbuf
+		 */
+		const __m256i len6_7 = _mm256_slli_epi32(raw_desc6_7,
+							 PKTLEN_SHIFT);
+		const __m256i len4_5 = _mm256_slli_epi32(raw_desc4_5,
+							 PKTLEN_SHIFT);
+		const __m256i desc6_7 = _mm256_blend_epi16(raw_desc6_7,
+							   len6_7, 0x80);
+		const __m256i desc4_5 = _mm256_blend_epi16(raw_desc4_5,
+							   len4_5, 0x80);
+		__m256i mb6_7 = _mm256_shuffle_epi8(desc6_7, shuf_msk);
+		__m256i mb4_5 = _mm256_shuffle_epi8(desc4_5, shuf_msk);
+
+		mb6_7 = _mm256_add_epi16(mb6_7, crc_adjust);
+		mb4_5 = _mm256_add_epi16(mb4_5, crc_adjust);
+		/**
+		 * to get packet types, shift 64-bit values down 30 bits
+		 * and so ptype is in lower 8-bits in each
+		 */
+		const __m256i ptypes6_7 = _mm256_srli_epi64(desc6_7, 30);
+		const __m256i ptypes4_5 = _mm256_srli_epi64(desc4_5, 30);
+		const uint8_t ptype7 = _mm256_extract_epi8(ptypes6_7, 24);
+		const uint8_t ptype6 = _mm256_extract_epi8(ptypes6_7, 8);
+		const uint8_t ptype5 = _mm256_extract_epi8(ptypes4_5, 24);
+		const uint8_t ptype4 = _mm256_extract_epi8(ptypes4_5, 8);
+
+		mb6_7 = _mm256_insert_epi32(mb6_7, type_table[ptype7], 4);
+		mb6_7 = _mm256_insert_epi32(mb6_7, type_table[ptype6], 0);
+		mb4_5 = _mm256_insert_epi32(mb4_5, type_table[ptype5], 4);
+		mb4_5 = _mm256_insert_epi32(mb4_5, type_table[ptype4], 0);
+		/* merge the status bits into one register */
+		const __m256i status4_7 = _mm256_unpackhi_epi32(desc6_7,
+				desc4_5);
+
+		/**
+		 * convert descriptors 0-3 into mbufs, adjusting length and
+		 * re-arranging fields. Then write into the mbuf
+		 */
+		const __m256i len2_3 = _mm256_slli_epi32(raw_desc2_3,
+							 PKTLEN_SHIFT);
+		const __m256i len0_1 = _mm256_slli_epi32(raw_desc0_1,
+							 PKTLEN_SHIFT);
+		const __m256i desc2_3 = _mm256_blend_epi16(raw_desc2_3,
+							   len2_3, 0x80);
+		const __m256i desc0_1 = _mm256_blend_epi16(raw_desc0_1,
+							   len0_1, 0x80);
+		__m256i mb2_3 = _mm256_shuffle_epi8(desc2_3, shuf_msk);
+		__m256i mb0_1 = _mm256_shuffle_epi8(desc0_1, shuf_msk);
+
+		mb2_3 = _mm256_add_epi16(mb2_3, crc_adjust);
+		mb0_1 = _mm256_add_epi16(mb0_1, crc_adjust);
+		/* get the packet types */
+		const __m256i ptypes2_3 = _mm256_srli_epi64(desc2_3, 30);
+		const __m256i ptypes0_1 = _mm256_srli_epi64(desc0_1, 30);
+		const uint8_t ptype3 = _mm256_extract_epi8(ptypes2_3, 24);
+		const uint8_t ptype2 = _mm256_extract_epi8(ptypes2_3, 8);
+		const uint8_t ptype1 = _mm256_extract_epi8(ptypes0_1, 24);
+		const uint8_t ptype0 = _mm256_extract_epi8(ptypes0_1, 8);
+
+		mb2_3 = _mm256_insert_epi32(mb2_3, type_table[ptype3], 4);
+		mb2_3 = _mm256_insert_epi32(mb2_3, type_table[ptype2], 0);
+		mb0_1 = _mm256_insert_epi32(mb0_1, type_table[ptype1], 4);
+		mb0_1 = _mm256_insert_epi32(mb0_1, type_table[ptype0], 0);
+		/* merge the status bits into one register */
+		const __m256i status0_3 = _mm256_unpackhi_epi32(desc2_3,
+								desc0_1);
+
+		/**
+		 * take the two sets of status bits and merge to one
+		 * After merge, the packets status flags are in the
+		 * order (hi->lo): [1, 3, 5, 7, 0, 2, 4, 6]
+		 */
+		__m256i status0_7 = _mm256_unpacklo_epi64(status4_7,
+							  status0_3);
+
+		/* now do flag manipulation */
+
+		/* get only flag/error bits we want */
+		const __m256i flag_bits =
+			_mm256_and_si256(status0_7, flags_mask);
+		/* set vlan and rss flags */
+		const __m256i vlan_flags =
+			_mm256_shuffle_epi8(vlan_flags_shuf, flag_bits);
+		const __m256i rss_flags =
+			_mm256_shuffle_epi8(rss_flags_shuf,
+					    _mm256_srli_epi32(flag_bits, 11));
+		/**
+		 * l3_l4_error flags, shuffle, then shift to correct adjustment
+		 * of flags in flags_shuf, and finally mask out extra bits
+		 */
+		__m256i l3_l4_flags = _mm256_shuffle_epi8(l3_l4_flags_shuf,
+				_mm256_srli_epi32(flag_bits, 22));
+		l3_l4_flags = _mm256_slli_epi32(l3_l4_flags, 1);
+		l3_l4_flags = _mm256_and_si256(l3_l4_flags, cksum_mask);
+
+		/* merge flags */
+		const __m256i mbuf_flags = _mm256_or_si256(l3_l4_flags,
+				_mm256_or_si256(rss_flags, vlan_flags));
+		/**
+		 * At this point, we have the 8 sets of flags in the low 16-bits
+		 * of each 32-bit value in vlan0.
+		 * We want to extract these, and merge them with the mbuf init
+		 * data so we can do a single write to the mbuf to set the flags
+		 * and all the other initialization fields. Extracting the
+		 * appropriate flags means that we have to do a shift and blend
+		 * for each mbuf before we do the write. However, we can also
+		 * add in the previously computed rx_descriptor fields to
+		 * make a single 256-bit write per mbuf
+		 */
+		/* check the structure matches expectations */
+		RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, ol_flags) !=
+				 offsetof(struct rte_mbuf, rearm_data) + 8);
+		RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, rearm_data) !=
+				 RTE_ALIGN(offsetof(struct rte_mbuf,
+						    rearm_data),
+					   16));
+		/* build up data and do writes */
+		__m256i rearm0, rearm1, rearm2, rearm3, rearm4, rearm5,
+			rearm6, rearm7;
+		rearm6 = _mm256_blend_epi32(mbuf_init,
+					    _mm256_slli_si256(mbuf_flags, 8),
+					    0x04);
+		rearm4 = _mm256_blend_epi32(mbuf_init,
+					    _mm256_slli_si256(mbuf_flags, 4),
+					    0x04);
+		rearm2 = _mm256_blend_epi32(mbuf_init, mbuf_flags, 0x04);
+		rearm0 = _mm256_blend_epi32(mbuf_init,
+					    _mm256_srli_si256(mbuf_flags, 4),
+					    0x04);
+		/* permute to add in the rx_descriptor e.g. rss fields */
+		rearm6 = _mm256_permute2f128_si256(rearm6, mb6_7, 0x20);
+		rearm4 = _mm256_permute2f128_si256(rearm4, mb4_5, 0x20);
+		rearm2 = _mm256_permute2f128_si256(rearm2, mb2_3, 0x20);
+		rearm0 = _mm256_permute2f128_si256(rearm0, mb0_1, 0x20);
+		/* write to mbuf */
+		_mm256_storeu_si256((__m256i *)&rx_pkts[i + 6]->rearm_data,
+				    rearm6);
+		_mm256_storeu_si256((__m256i *)&rx_pkts[i + 4]->rearm_data,
+				    rearm4);
+		_mm256_storeu_si256((__m256i *)&rx_pkts[i + 2]->rearm_data,
+				    rearm2);
+		_mm256_storeu_si256((__m256i *)&rx_pkts[i + 0]->rearm_data,
+				    rearm0);
+
+		/* repeat for the odd mbufs */
+		const __m256i odd_flags =
+			_mm256_castsi128_si256
+				(_mm256_extracti128_si256(mbuf_flags, 1));
+		rearm7 = _mm256_blend_epi32(mbuf_init,
+					    _mm256_slli_si256(odd_flags, 8),
+					    0x04);
+		rearm5 = _mm256_blend_epi32(mbuf_init,
+					    _mm256_slli_si256(odd_flags, 4),
+					    0x04);
+		rearm3 = _mm256_blend_epi32(mbuf_init, odd_flags, 0x04);
+		rearm1 = _mm256_blend_epi32(mbuf_init,
+					    _mm256_srli_si256(odd_flags, 4),
+					    0x04);
+		/* since odd mbufs are already in hi 128-bits use blend */
+		rearm7 = _mm256_blend_epi32(rearm7, mb6_7, 0xF0);
+		rearm5 = _mm256_blend_epi32(rearm5, mb4_5, 0xF0);
+		rearm3 = _mm256_blend_epi32(rearm3, mb2_3, 0xF0);
+		rearm1 = _mm256_blend_epi32(rearm1, mb0_1, 0xF0);
+		/* again write to mbufs */
+		_mm256_storeu_si256((__m256i *)&rx_pkts[i + 7]->rearm_data,
+				    rearm7);
+		_mm256_storeu_si256((__m256i *)&rx_pkts[i + 5]->rearm_data,
+				    rearm5);
+		_mm256_storeu_si256((__m256i *)&rx_pkts[i + 3]->rearm_data,
+				    rearm3);
+		_mm256_storeu_si256((__m256i *)&rx_pkts[i + 1]->rearm_data,
+				    rearm1);
+
+		/* extract and record EOP bit */
+		if (split_packet) {
+			const __m128i eop_mask =
+				_mm_set1_epi16(1 << IAVF_RX_DESC_STATUS_EOF_SHIFT);
+			const __m256i eop_bits256 = _mm256_and_si256(status0_7,
+								     eop_check);
+			/* pack status bits into a single 128-bit register */
+			const __m128i eop_bits =
+				_mm_packus_epi32
+					(_mm256_castsi256_si128(eop_bits256),
+					 _mm256_extractf128_si256(eop_bits256,
+								  1));
+			/**
+			 * flip bits, and mask out the EOP bit, which is now
+			 * a split-packet bit i.e. !EOP, rather than EOP one.
+			 */
+			__m128i split_bits = _mm_andnot_si128(eop_bits,
+					eop_mask);
+			/**
+			 * eop bits are out of order, so we need to shuffle them
+			 * back into order again. In doing so, only use low 8
+			 * bits, which acts like another pack instruction
+			 * The original order is (hi->lo): 1,3,5,7,0,2,4,6
+			 * [Since we use epi8, the 16-bit positions are
+			 * multiplied by 2 in the eop_shuffle value.]
+			 */
+			__m128i eop_shuffle =
+				_mm_set_epi8(/* zero hi 64b */
+					     0xFF, 0xFF, 0xFF, 0xFF,
+					     0xFF, 0xFF, 0xFF, 0xFF,
+					     /* move values to lo 64b */
+					     8, 0, 10, 2,
+					     12, 4, 14, 6);
+			split_bits = _mm_shuffle_epi8(split_bits, eop_shuffle);
+			*(uint64_t *)split_packet =
+				_mm_cvtsi128_si64(split_bits);
+			split_packet += IAVF_DESCS_PER_LOOP_AVX;
+		}
+
+		/* perform dd_check */
+		status0_7 = _mm256_and_si256(status0_7, dd_check);
+		status0_7 = _mm256_packs_epi32(status0_7,
+					       _mm256_setzero_si256());
+
+		uint64_t burst = __builtin_popcountll
+					(_mm_cvtsi128_si64
+						(_mm256_extracti128_si256
+							(status0_7, 1)));
+		burst += __builtin_popcountll
+				(_mm_cvtsi128_si64
+					(_mm256_castsi256_si128(status0_7)));
+		received += burst;
+		if (burst != IAVF_DESCS_PER_LOOP_AVX)
+			break;
+	}
+
+	/* update tail pointers */
+	rxq->rx_tail += received;
+	rxq->rx_tail &= (rxq->nb_rx_desc - 1);
+	if ((rxq->rx_tail & 1) == 1 && received > 1) { /* keep avx2 aligned */
+		rxq->rx_tail--;
+		received--;
+	}
+	rxq->rxrearm_nb += received;
+	return received;
+}
+
+/**
+ * Notice:
+ * - nb_pkts < IAVF_DESCS_PER_LOOP, just return no packet
+ */
+uint16_t
+iavf_recv_pkts_vec_avx2(void *rx_queue, struct rte_mbuf **rx_pkts,
+			uint16_t nb_pkts)
+{
+	return _iavf_recv_raw_pkts_vec_avx2(rx_queue, rx_pkts, nb_pkts, NULL);
+}
+
+/**
+ * vPMD receive routine that reassembles single burst of 32 scattered packets
+ * Notice:
+ * - nb_pkts < IAVF_DESCS_PER_LOOP, just return no packet
+ */
+static uint16_t
+iavf_recv_scattered_burst_vec_avx2(void *rx_queue, struct rte_mbuf **rx_pkts,
+				   uint16_t nb_pkts)
+{
+	struct iavf_rx_queue *rxq = rx_queue;
+	uint8_t split_flags[IAVF_VPMD_RX_MAX_BURST] = {0};
+
+	/* get some new buffers */
+	uint16_t nb_bufs = _iavf_recv_raw_pkts_vec_avx2(rxq, rx_pkts, nb_pkts,
+						       split_flags);
+	if (nb_bufs == 0)
+		return 0;
+
+	/* happy day case, full burst + no packets to be joined */
+	const uint64_t *split_fl64 = (uint64_t *)split_flags;
+
+	if (!rxq->pkt_first_seg &&
+	    split_fl64[0] == 0 && split_fl64[1] == 0 &&
+	    split_fl64[2] == 0 && split_fl64[3] == 0)
+		return nb_bufs;
+
+	/* reassemble any packets that need reassembly*/
+	unsigned int i = 0;
+
+	if (!rxq->pkt_first_seg) {
+		/* find the first split flag, and only reassemble then*/
+		while (i < nb_bufs && !split_flags[i])
+			i++;
+		if (i == nb_bufs)
+			return nb_bufs;
+		rxq->pkt_first_seg = rx_pkts[i];
+	}
+	return i + reassemble_packets(rxq, &rx_pkts[i], nb_bufs - i,
+					     &split_flags[i]);
+}
+
+/**
+ * vPMD receive routine that reassembles scattered packets.
+ * Main receive routine that can handle arbitrary burst sizes
+ * Notice:
+ * - nb_pkts < IAVF_DESCS_PER_LOOP, just return no packet
+ */
+uint16_t
+iavf_recv_scattered_pkts_vec_avx2(void *rx_queue, struct rte_mbuf **rx_pkts,
+				  uint16_t nb_pkts)
+{
+	uint16_t retval = 0;
+
+	while (nb_pkts > IAVF_VPMD_RX_MAX_BURST) {
+		uint16_t burst = iavf_recv_scattered_burst_vec_avx2(rx_queue,
+				rx_pkts + retval, IAVF_VPMD_RX_MAX_BURST);
+		retval += burst;
+		nb_pkts -= burst;
+		if (burst < IAVF_VPMD_RX_MAX_BURST)
+			return retval;
+	}
+	return retval + iavf_recv_scattered_burst_vec_avx2(rx_queue,
+				rx_pkts + retval, nb_pkts);
+}
+
+static inline void
+iavf_vtx1(volatile struct iavf_tx_desc *txdp,
+	  struct rte_mbuf *pkt, uint64_t flags)
+{
+	uint64_t high_qw =
+		(IAVF_TX_DESC_DTYPE_DATA |
+		 ((uint64_t)flags  << IAVF_TXD_QW1_CMD_SHIFT) |
+		 ((uint64_t)pkt->data_len << IAVF_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
+iavf_vtx(volatile struct iavf_tx_desc *txdp,
+	 struct rte_mbuf **pkt, uint16_t nb_pkts,  uint64_t flags)
+{
+	const uint64_t hi_qw_tmpl = (IAVF_TX_DESC_DTYPE_DATA |
+			((uint64_t)flags  << IAVF_TXD_QW1_CMD_SHIFT));
+
+	/* if unaligned on 32-bit boundary, do one to align */
+	if (((uintptr_t)txdp & 0x1F) != 0 && nb_pkts != 0) {
+		iavf_vtx1(txdp, *pkt, flags);
+		nb_pkts--, txdp++, pkt++;
+	}
+
+	/* do two at a time while possible, in bursts */
+	for (; nb_pkts > 3; txdp += 4, pkt += 4, nb_pkts -= 4) {
+		uint64_t hi_qw3 =
+			hi_qw_tmpl |
+			((uint64_t)pkt[3]->data_len <<
+			 IAVF_TXD_QW1_TX_BUF_SZ_SHIFT);
+		uint64_t hi_qw2 =
+			hi_qw_tmpl |
+			((uint64_t)pkt[2]->data_len <<
+			 IAVF_TXD_QW1_TX_BUF_SZ_SHIFT);
+		uint64_t hi_qw1 =
+			hi_qw_tmpl |
+			((uint64_t)pkt[1]->data_len <<
+			 IAVF_TXD_QW1_TX_BUF_SZ_SHIFT);
+		uint64_t hi_qw0 =
+			hi_qw_tmpl |
+			((uint64_t)pkt[0]->data_len <<
+			 IAVF_TXD_QW1_TX_BUF_SZ_SHIFT);
+
+		__m256i desc2_3 =
+			_mm256_set_epi64x
+				(hi_qw3,
+				 pkt[3]->buf_physaddr + pkt[3]->data_off,
+				 hi_qw2,
+				 pkt[2]->buf_physaddr + pkt[2]->data_off);
+		__m256i desc0_1 =
+			_mm256_set_epi64x
+				(hi_qw1,
+				 pkt[1]->buf_physaddr + pkt[1]->data_off,
+				 hi_qw0,
+				 pkt[0]->buf_physaddr + pkt[0]->data_off);
+		_mm256_store_si256((void *)(txdp + 2), desc2_3);
+		_mm256_store_si256((void *)txdp, desc0_1);
+	}
+
+	/* do any last ones */
+	while (nb_pkts) {
+		iavf_vtx1(txdp, *pkt, flags);
+		txdp++, pkt++, nb_pkts--;
+	}
+}
+
+static inline uint16_t
+iavf_xmit_fixed_burst_vec_avx2(void *tx_queue, struct rte_mbuf **tx_pkts,
+			       uint16_t nb_pkts)
+{
+	struct iavf_tx_queue *txq = (struct iavf_tx_queue *)tx_queue;
+	volatile struct iavf_tx_desc *txdp;
+	struct iavf_tx_entry *txep;
+	uint16_t n, nb_commit, tx_id;
+	uint64_t flags = IAVF_TX_DESC_CMD_EOP;
+	uint64_t rs = IAVF_TX_DESC_CMD_RS | IAVF_TX_DESC_CMD_EOP;
+
+	/* cross rx_thresh boundary is not allowed */
+	nb_pkts = RTE_MIN(nb_pkts, txq->rs_thresh);
+
+	if (txq->nb_free < txq->free_thresh)
+		iavf_tx_free_bufs(txq);
+
+	nb_commit = nb_pkts = (uint16_t)RTE_MIN(txq->nb_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_free = (uint16_t)(txq->nb_free - nb_pkts);
+
+	n = (uint16_t)(txq->nb_tx_desc - tx_id);
+	if (nb_commit >= n) {
+		tx_backlog_entry(txep, tx_pkts, n);
+
+		iavf_vtx(txdp, tx_pkts, n - 1, flags);
+		tx_pkts += (n - 1);
+		txdp += (n - 1);
+
+		iavf_vtx1(txdp, *tx_pkts++, rs);
+
+		nb_commit = (uint16_t)(nb_commit - n);
+
+		tx_id = 0;
+		txq->next_rs = (uint16_t)(txq->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);
+
+	iavf_vtx(txdp, tx_pkts, nb_commit, flags);
+
+	tx_id = (uint16_t)(tx_id + nb_commit);
+	if (tx_id > txq->next_rs) {
+		txq->tx_ring[txq->next_rs].cmd_type_offset_bsz |=
+			rte_cpu_to_le_64(((uint64_t)IAVF_TX_DESC_CMD_RS) <<
+					 IAVF_TXD_QW1_CMD_SHIFT);
+		txq->next_rs =
+			(uint16_t)(txq->next_rs + txq->rs_thresh);
+	}
+
+	txq->tx_tail = tx_id;
+
+	IAVF_PCI_REG_WRITE(txq->qtx_tail, txq->tx_tail);
+
+	return nb_pkts;
+}
+
+uint16_t
+iavf_xmit_pkts_vec_avx2(void *tx_queue, struct rte_mbuf **tx_pkts,
+			uint16_t nb_pkts)
+{
+	uint16_t nb_tx = 0;
+	struct iavf_tx_queue *txq = (struct iavf_tx_queue *)tx_queue;
+
+	while (nb_pkts) {
+		uint16_t ret, num;
+
+		num = (uint16_t)RTE_MIN(nb_pkts, txq->rs_thresh);
+		ret = iavf_xmit_fixed_burst_vec_avx2(tx_queue, &tx_pkts[nb_tx],
+						     num);
+		nb_tx += ret;
+		nb_pkts -= ret;
+		if (ret < num)
+			break;
+	}
+
+	return nb_tx;
+}
diff --git a/drivers/net/iavf/iavf_rxtx_vec_common.h b/drivers/net/iavf/iavf_rxtx_vec_common.h
index db509d71f..fff0555d2 100644
--- a/drivers/net/iavf/iavf_rxtx_vec_common.h
+++ b/drivers/net/iavf/iavf_rxtx_vec_common.h
@@ -207,4 +207,76 @@ iavf_rxq_vec_setup_default(struct iavf_rx_queue *rxq)
 	rxq->mbuf_initializer = *(uint64_t *)p;
 	return 0;
 }
+
+static inline int
+iavf_rx_vec_queue_default(struct iavf_rx_queue *rxq)
+{
+	if (!rxq)
+		return -1;
+
+	if (!rte_is_power_of_2(rxq->nb_rx_desc))
+		return -1;
+
+	if (rxq->rx_free_thresh < IAVF_VPMD_RX_MAX_BURST)
+		return -1;
+
+	if (rxq->nb_rx_desc % rxq->rx_free_thresh)
+		return -1;
+
+	return 0;
+}
+
+#define IAVF_NO_VECTOR_FLAGS (				 \
+		DEV_TX_OFFLOAD_MULTI_SEGS |		 \
+		DEV_TX_OFFLOAD_VLAN_INSERT |		 \
+		DEV_TX_OFFLOAD_SCTP_CKSUM |		 \
+		DEV_TX_OFFLOAD_UDP_CKSUM |		 \
+		DEV_TX_OFFLOAD_TCP_CKSUM)
+
+static inline int
+iavf_tx_vec_queue_default(struct iavf_tx_queue *txq)
+{
+	if (!txq)
+		return -1;
+
+	if (txq->offloads & IAVF_NO_VECTOR_FLAGS)
+		return -1;
+
+	if (txq->rs_thresh < IAVF_VPMD_TX_MAX_BURST ||
+	    txq->rs_thresh > IAVF_VPMD_TX_MAX_FREE_BUF)
+		return -1;
+
+	return 0;
+}
+
+static inline int
+iavf_rx_vec_dev_check_default(struct rte_eth_dev *dev)
+{
+	int i;
+	struct iavf_rx_queue *rxq;
+
+	for (i = 0; i < dev->data->nb_rx_queues; i++) {
+		rxq = dev->data->rx_queues[i];
+		if (iavf_rx_vec_queue_default(rxq))
+			return -1;
+	}
+
+	return 0;
+}
+
+static inline int
+iavf_tx_vec_dev_check_default(struct rte_eth_dev *dev)
+{
+	int i;
+	struct iavf_tx_queue *txq;
+
+	for (i = 0; i < dev->data->nb_tx_queues; i++) {
+		txq = dev->data->tx_queues[i];
+		if (iavf_tx_vec_queue_default(txq))
+			return -1;
+	}
+
+	return 0;
+}
+
 #endif
diff --git a/drivers/net/iavf/iavf_rxtx_vec_sse.c b/drivers/net/iavf/iavf_rxtx_vec_sse.c
index cc71f23a5..15c52a4c2 100644
--- a/drivers/net/iavf/iavf_rxtx_vec_sse.c
+++ b/drivers/net/iavf/iavf_rxtx_vec_sse.c
@@ -655,3 +655,15 @@ iavf_rxq_vec_setup(struct iavf_rx_queue *rxq)
 	rxq->ops = &sse_vec_rxq_ops;
 	return iavf_rxq_vec_setup_default(rxq);
 }
+
+int __attribute__((cold))
+iavf_rx_vec_dev_check(struct rte_eth_dev *dev)
+{
+	return iavf_rx_vec_dev_check_default(dev);
+}
+
+int __attribute__((cold))
+iavf_tx_vec_dev_check(struct rte_eth_dev *dev)
+{
+	return iavf_tx_vec_dev_check_default(dev);
+}
diff --git a/drivers/net/iavf/meson.build b/drivers/net/iavf/meson.build
index e5a2f5553..c2401af86 100644
--- a/drivers/net/iavf/meson.build
+++ b/drivers/net/iavf/meson.build
@@ -17,4 +17,21 @@ sources = files(
 if arch_subdir == 'x86'
 	dpdk_conf.set('RTE_LIBRTE_IAVF_INC_VECTOR', 1)
 	sources += files('iavf_rxtx_vec_sse.c')
+
+	# compile AVX2 version if either:
+	# a. we have AVX supported in minimum instruction set baseline
+	# b. it's not minimum instruction set, but supported by compiler
+	if dpdk_conf.has('RTE_MACHINE_CPUFLAG_AVX2')
+		cflags += ['-DCC_AVX2_SUPPORT']
+		sources += files('iavf_rxtx_vec_avx2.c')
+	elif cc.has_argument('-mavx2')
+		cflags += ['-DCC_AVX2_SUPPORT']
+		iavf_avx2_lib = static_library('iavf_avx2_lib',
+				'iavf_rxtx_vec_avx2.c',
+				dependencies: [static_rte_ethdev,
+					static_rte_kvargs, static_rte_hash],
+				include_directories: includes,
+				c_args: [cflags, '-mavx2'])
+		objs += iavf_avx2_lib.extract_objects('iavf_rxtx_vec_avx2.c')
+	endif
 endif
-- 
2.17.1


^ permalink raw reply	[flat|nested] 11+ messages in thread

* Re: [dpdk-dev] [PATCH] net/iavf: enable AVX2 for iavf
  2019-09-04 10:06 [dpdk-dev] [PATCH] net/iavf: enable AVX2 for iavf Leyi Rong
@ 2019-09-04 13:29 ` Aaron Conole
  2019-09-06  3:28 ` [dpdk-dev] [PATCH v2] " Leyi Rong
                   ` (3 subsequent siblings)
  4 siblings, 0 replies; 11+ messages in thread
From: Aaron Conole @ 2019-09-04 13:29 UTC (permalink / raw)
  To: Leyi Rong; +Cc: wenzhuo.lu, qi.z.zhang, xiaolong.ye, dev

Leyi Rong <leyi.rong@intel.com> writes:

> This patch enables AVX data path for iavf PMD.
>
> Signed-off-by: Leyi Rong <leyi.rong@intel.com>
> ---
>  drivers/net/iavf/Makefile               |  21 +
>  drivers/net/iavf/iavf_rxtx.c            |  62 +-
>  drivers/net/iavf/iavf_rxtx.h            |  10 +
>  drivers/net/iavf/iavf_rxtx_vec_avx2.c   | 867 ++++++++++++++++++++++++
>  drivers/net/iavf/iavf_rxtx_vec_common.h |  72 ++
>  drivers/net/iavf/iavf_rxtx_vec_sse.c    |  12 +
>  drivers/net/iavf/meson.build            |  17 +
>  7 files changed, 1041 insertions(+), 20 deletions(-)
>  create mode 100644 drivers/net/iavf/iavf_rxtx_vec_avx2.c
>
> diff --git a/drivers/net/iavf/Makefile b/drivers/net/iavf/Makefile
> index cd74e14ab..cf0ed99f0 100644
> --- a/drivers/net/iavf/Makefile
> +++ b/drivers/net/iavf/Makefile
> @@ -48,4 +48,25 @@ ifeq ($(CONFIG_RTE_ARCH_X86), y)
>  SRCS-$(CONFIG_RTE_LIBRTE_IAVF_INC_VECTOR) += iavf_rxtx_vec_sse.c
>  endif
>  
> +ifeq ($(CONFIG_RTE_LIBRTE_IAVF_INC_VECTOR), y)
> +	ifeq ($(findstring RTE_MACHINE_CPUFLAG_AVX2,$(CFLAGS)),RTE_MACHINE_CPUFLAG_AVX2)
> +		CC_AVX2_SUPPORT=1
> +	else
> +		CC_AVX2_SUPPORT=\
> +		$(shell $(CC) -march=core-avx2 -dM -E - </dev/null 2>&1 | \
> +		grep -q AVX2 && echo 1)
> +		ifeq ($(CC_AVX2_SUPPORT), 1)
> +			ifeq ($(CONFIG_RTE_TOOLCHAIN_ICC),y)
> +				CFLAGS_iavf_rxtx_vec_avx2.o += -march=core-avx2
> +			else
> +				CFLAGS_iavf_rxtx_vec_avx2.o += -mavx2
> +			endif
> +		endif
> +	endif
> +endif
> +
> +ifeq ($(CC_AVX2_SUPPORT), 1)
> +	SRCS-$(CONFIG_RTE_LIBRTE_IAVF_PMD) += iavf_rxtx_vec_avx2.c
> +endif
> +
>  include $(RTE_SDK)/mk/rte.lib.mk
> diff --git a/drivers/net/iavf/iavf_rxtx.c b/drivers/net/iavf/iavf_rxtx.c
> index 22d7bb612..2eb496d5f 100644
> --- a/drivers/net/iavf/iavf_rxtx.c
> +++ b/drivers/net/iavf/iavf_rxtx.c
> @@ -1743,23 +1743,36 @@ iavf_set_rx_function(struct rte_eth_dev *dev)
>  		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
>  	struct iavf_rx_queue *rxq;
>  	int i;
> +	bool use_avx2 = false;
>  
> -	if (adapter->rx_vec_allowed) {
> -		if (dev->data->scattered_rx) {
> -			PMD_DRV_LOG(DEBUG, "Using Vector Scattered Rx callback"
> -				    " (port=%d).", dev->data->port_id);
> -			dev->rx_pkt_burst = iavf_recv_scattered_pkts_vec;
> -		} else {
> -			PMD_DRV_LOG(DEBUG, "Using Vector Rx callback"
> -				    " (port=%d).", dev->data->port_id);
> -			dev->rx_pkt_burst = iavf_recv_pkts_vec;
> -		}
> +	if (!iavf_rx_vec_dev_check(dev)) {
>  		for (i = 0; i < dev->data->nb_rx_queues; i++) {
>  			rxq = dev->data->rx_queues[i];
> -			if (!rxq)
> -				continue;
> -			iavf_rxq_vec_setup(rxq);
> +			(void)iavf_rxq_vec_setup(rxq);
>  		}
> +
> +		if (rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1 ||
> +		    rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX512F) == 1)
> +			use_avx2 = true;

This will cause a build failure on aarch64 systems.

ex:

../drivers/net/iavf/iavf_rxtx.c:1754:7: error: implicit declaration of function ‘rte_cpu_get_flag_enabled’ [-Werror=implicit-function-declaration]

   if (rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1 ||
       ^

../drivers/net/iavf/iavf_rxtx.c:1754:32: error: ‘RTE_CPUFLAG_AVX2’ undeclared (first use in this function)

   if (rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1 ||
                                ^

../drivers/net/iavf/iavf_rxtx.c:1754:32: note: each undeclared identifier is reported only once for each function it appears in

../drivers/net/iavf/iavf_rxtx.c:1755:32: error: ‘RTE_CPUFLAG_AVX512F’ undeclared (first use in this function)

       rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX512F) == 1)
                                ^

../drivers/net/iavf/iavf_rxtx.c: In function ‘iavf_set_tx_function’:

../drivers/net/iavf/iavf_rxtx.c:1807:32: error: ‘RTE_CPUFLAG_AVX2’ undeclared (first use in this function)

   if (rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1 ||
                                ^

../drivers/net/iavf/iavf_rxtx.c:1808:32: error: ‘RTE_CPUFLAG_AVX512F’ undeclared (first use in this function)

       rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX512F) == 1)
                                ^

> +		if (dev->data->scattered_rx) {
> +			PMD_DRV_LOG(DEBUG,
> +				    "Using %sVector Scattered Rx (port %d).",
> +				    use_avx2 ? "avx2 " : "",
> +				    dev->data->port_id);
> +			dev->rx_pkt_burst = use_avx2 ?
> +					    iavf_recv_scattered_pkts_vec_avx2 :
> +					    iavf_recv_scattered_pkts_vec;
> +		} else {
> +			PMD_DRV_LOG(DEBUG, "Using %sVector Rx (port %d).",
> +				    use_avx2 ? "avx2 " : "",
> +				    dev->data->port_id);
> +			dev->rx_pkt_burst = use_avx2 ?
> +					    iavf_recv_pkts_vec_avx2 :
> +					    iavf_recv_pkts_vec;
> +		}
> +
> +		return;
>  	} else if (dev->data->scattered_rx) {
>  		PMD_DRV_LOG(DEBUG, "Using a Scattered Rx callback (port=%d).",
>  			    dev->data->port_id);
> @@ -1779,22 +1792,31 @@ iavf_set_rx_function(struct rte_eth_dev *dev)
>  void
>  iavf_set_tx_function(struct rte_eth_dev *dev)
>  {
> -	struct iavf_adapter *adapter =
> -		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
>  	struct iavf_tx_queue *txq;
>  	int i;
> +	bool use_avx2 = false;
>  
> -	if (adapter->tx_vec_allowed) {
> -		PMD_DRV_LOG(DEBUG, "Using Vector Tx callback (port=%d).",
> -			    dev->data->port_id);
> -		dev->tx_pkt_burst = iavf_xmit_pkts_vec;
> -		dev->tx_pkt_prepare = NULL;
> +	if (!iavf_tx_vec_dev_check(dev)) {
>  		for (i = 0; i < dev->data->nb_tx_queues; i++) {
>  			txq = dev->data->tx_queues[i];
>  			if (!txq)
>  				continue;
>  			iavf_txq_vec_setup(txq);
>  		}
> +
> +		if (rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1 ||
> +		    rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX512F) == 1)
> +			use_avx2 = true;
> +
> +		PMD_DRV_LOG(DEBUG, "Using %sVector Tx (port %d).",
> +			    use_avx2 ? "avx2 " : "",
> +			    dev->data->port_id);
> +		dev->tx_pkt_burst = use_avx2 ?
> +				    iavf_xmit_pkts_vec_avx2 :
> +				    iavf_xmit_pkts_vec;
> +		dev->tx_pkt_prepare = NULL;
> +
> +		return;
>  	} else {
>  		PMD_DRV_LOG(DEBUG, "Using Basic Tx callback (port=%d).",
>  			    dev->data->port_id);
> diff --git a/drivers/net/iavf/iavf_rxtx.h b/drivers/net/iavf/iavf_rxtx.h
> index c86720bda..7dde392bf 100644
> --- a/drivers/net/iavf/iavf_rxtx.h
> +++ b/drivers/net/iavf/iavf_rxtx.h
> @@ -19,6 +19,7 @@
>  /* used for Vector PMD */
>  #define IAVF_VPMD_RX_MAX_BURST    32
>  #define IAVF_VPMD_TX_MAX_BURST    32
> +#define IAVF_RXQ_REARM_THRESH     32
>  #define IAVF_VPMD_DESCS_PER_LOOP  4
>  #define IAVF_VPMD_TX_MAX_FREE_BUF 64
>  
> @@ -200,6 +201,15 @@ uint16_t iavf_recv_scattered_pkts_vec(void *rx_queue,
>  				     uint16_t nb_pkts);
>  uint16_t iavf_xmit_fixed_burst_vec(void *tx_queue, struct rte_mbuf **tx_pkts,
>  				  uint16_t nb_pkts);
> +uint16_t iavf_recv_pkts_vec_avx2(void *rx_queue, struct rte_mbuf **rx_pkts,
> +				 uint16_t nb_pkts);
> +uint16_t iavf_recv_scattered_pkts_vec_avx2(void *rx_queue,
> +					   struct rte_mbuf **rx_pkts,
> +					   uint16_t nb_pkts);
> +uint16_t iavf_xmit_pkts_vec_avx2(void *tx_queue, struct rte_mbuf **tx_pkts,
> +				 uint16_t nb_pkts);
> +int iavf_rx_vec_dev_check(struct rte_eth_dev *dev);
> +int iavf_tx_vec_dev_check(struct rte_eth_dev *dev);
>  int iavf_rxq_vec_setup(struct iavf_rx_queue *rxq);
>  int iavf_txq_vec_setup(struct iavf_tx_queue *txq);
>  
> diff --git a/drivers/net/iavf/iavf_rxtx_vec_avx2.c b/drivers/net/iavf/iavf_rxtx_vec_avx2.c
> new file mode 100644
> index 000000000..f0c00be56
> --- /dev/null
> +++ b/drivers/net/iavf/iavf_rxtx_vec_avx2.c
> @@ -0,0 +1,867 @@
> +/* SPDX-License-Identifier: BSD-3-Clause
> + * Copyright(c) 2019 Intel Corporation
> + */
> +
> +#include "base/iavf_prototype.h"
> +#include "iavf_rxtx_vec_common.h"
> +
> +#include <x86intrin.h>
> +
> +#ifndef __INTEL_COMPILER
> +#pragma GCC diagnostic ignored "-Wcast-qual"
> +#endif
> +
> +static inline void
> +iavf_rxq_rearm(struct iavf_rx_queue *rxq)
> +{
> +	int i;
> +	uint16_t rx_id;
> +	volatile union iavf_rx_desc *rxdp;
> +	struct rte_mbuf **rxp = &rxq->sw_ring[rxq->rxrearm_start];
> +
> +	rxdp = rxq->rx_ring + rxq->rxrearm_start;
> +
> +	/* Pull 'n' more MBUFs into the software ring */
> +	if (rte_mempool_get_bulk(rxq->mp,
> +				 (void *)rxp,
> +				 IAVF_RXQ_REARM_THRESH) < 0) {
> +		if (rxq->rxrearm_nb + IAVF_RXQ_REARM_THRESH >=
> +		    rxq->nb_rx_desc) {
> +			__m128i dma_addr0;
> +
> +			dma_addr0 = _mm_setzero_si128();
> +			for (i = 0; i < IAVF_VPMD_DESCS_PER_LOOP; i++) {
> +				rxp[i] = &rxq->fake_mbuf;
> +				_mm_store_si128((__m128i *)&rxdp[i].read,
> +						dma_addr0);
> +			}
> +		}
> +		rte_eth_devices[rxq->port_id].data->rx_mbuf_alloc_failed +=
> +			IAVF_RXQ_REARM_THRESH;
> +		return;
> +	}
> +
> +#ifndef RTE_LIBRTE_IAVF_16BYTE_RX_DESC
> +	struct rte_mbuf *mb0, *mb1;
> +	__m128i dma_addr0, dma_addr1;
> +	__m128i hdr_room = _mm_set_epi64x(RTE_PKTMBUF_HEADROOM,
> +			RTE_PKTMBUF_HEADROOM);
> +	/* Initialize the mbufs in vector, process 2 mbufs in one loop */
> +	for (i = 0; i < IAVF_RXQ_REARM_THRESH; i += 2, rxp += 2) {
> +		__m128i vaddr0, vaddr1;
> +
> +		mb0 = rxp[0];
> +		mb1 = rxp[1];
> +
> +		/* load buf_addr(lo 64bit) and buf_physaddr(hi 64bit) */
> +		RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, buf_physaddr) !=
> +				offsetof(struct rte_mbuf, buf_addr) + 8);
> +		vaddr0 = _mm_loadu_si128((__m128i *)&mb0->buf_addr);
> +		vaddr1 = _mm_loadu_si128((__m128i *)&mb1->buf_addr);
> +
> +		/* convert pa to dma_addr hdr/data */
> +		dma_addr0 = _mm_unpackhi_epi64(vaddr0, vaddr0);
> +		dma_addr1 = _mm_unpackhi_epi64(vaddr1, vaddr1);
> +
> +		/* add headroom to pa values */
> +		dma_addr0 = _mm_add_epi64(dma_addr0, hdr_room);
> +		dma_addr1 = _mm_add_epi64(dma_addr1, hdr_room);
> +
> +		/* flush desc with pa dma_addr */
> +		_mm_store_si128((__m128i *)&rxdp++->read, dma_addr0);
> +		_mm_store_si128((__m128i *)&rxdp++->read, dma_addr1);
> +	}
> +#else
> +	struct rte_mbuf *mb0, *mb1, *mb2, *mb3;
> +	__m256i dma_addr0_1, dma_addr2_3;
> +	__m256i hdr_room = _mm256_set1_epi64x(RTE_PKTMBUF_HEADROOM);
> +	/* Initialize the mbufs in vector, process 4 mbufs in one loop */
> +	for (i = 0; i < IAVF_RXQ_REARM_THRESH;
> +			i += 4, rxp += 4, rxdp += 4) {
> +		__m128i vaddr0, vaddr1, vaddr2, vaddr3;
> +		__m256i vaddr0_1, vaddr2_3;
> +
> +		mb0 = rxp[0];
> +		mb1 = rxp[1];
> +		mb2 = rxp[2];
> +		mb3 = rxp[3];
> +
> +		/* load buf_addr(lo 64bit) and buf_physaddr(hi 64bit) */
> +		RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, buf_physaddr) !=
> +				offsetof(struct rte_mbuf, buf_addr) + 8);
> +		vaddr0 = _mm_loadu_si128((__m128i *)&mb0->buf_addr);
> +		vaddr1 = _mm_loadu_si128((__m128i *)&mb1->buf_addr);
> +		vaddr2 = _mm_loadu_si128((__m128i *)&mb2->buf_addr);
> +		vaddr3 = _mm_loadu_si128((__m128i *)&mb3->buf_addr);
> +
> +		/**
> +		 * merge 0 & 1, by casting 0 to 256-bit and inserting 1
> +		 * into the high lanes. Similarly for 2 & 3
> +		 */
> +		vaddr0_1 =
> +			_mm256_inserti128_si256(_mm256_castsi128_si256(vaddr0),
> +						vaddr1, 1);
> +		vaddr2_3 =
> +			_mm256_inserti128_si256(_mm256_castsi128_si256(vaddr2),
> +						vaddr3, 1);
> +
> +		/* convert pa to dma_addr hdr/data */
> +		dma_addr0_1 = _mm256_unpackhi_epi64(vaddr0_1, vaddr0_1);
> +		dma_addr2_3 = _mm256_unpackhi_epi64(vaddr2_3, vaddr2_3);
> +
> +		/* add headroom to pa values */
> +		dma_addr0_1 = _mm256_add_epi64(dma_addr0_1, hdr_room);
> +		dma_addr2_3 = _mm256_add_epi64(dma_addr2_3, hdr_room);
> +
> +		/* flush desc with pa dma_addr */
> +		_mm256_store_si256((__m256i *)&rxdp->read, dma_addr0_1);
> +		_mm256_store_si256((__m256i *)&(rxdp + 2)->read, dma_addr2_3);
> +	}
> +
> +#endif
> +
> +	rxq->rxrearm_start += IAVF_RXQ_REARM_THRESH;
> +	if (rxq->rxrearm_start >= rxq->nb_rx_desc)
> +		rxq->rxrearm_start = 0;
> +
> +	rxq->rxrearm_nb -= IAVF_RXQ_REARM_THRESH;
> +
> +	rx_id = (uint16_t)((rxq->rxrearm_start == 0) ?
> +			     (rxq->nb_rx_desc - 1) : (rxq->rxrearm_start - 1));
> +
> +	/* Update the tail pointer on the NIC */
> +	IAVF_PCI_REG_WRITE(rxq->qrx_tail, rx_id);
> +}
> +
> +#define PKTLEN_SHIFT     10
> +
> +static inline uint16_t
> +_iavf_recv_raw_pkts_vec_avx2(struct iavf_rx_queue *rxq,
> +			     struct rte_mbuf **rx_pkts,
> +			     uint16_t nb_pkts, uint8_t *split_packet)
> +{
> +#define IAVF_DESCS_PER_LOOP_AVX 8
> +
> +	/* const uint32_t *ptype_tbl = rxq->vsi->adapter->ptype_tbl; */
> +	static const uint32_t type_table[UINT8_MAX + 1] __rte_cache_aligned = {
> +		/* [0] reserved */
> +		[1] = RTE_PTYPE_L2_ETHER,
> +		/* [2] - [21] reserved */
> +		[22] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
> +			RTE_PTYPE_L4_FRAG,
> +		[23] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
> +			RTE_PTYPE_L4_NONFRAG,
> +		[24] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
> +			RTE_PTYPE_L4_UDP,
> +		/* [25] reserved */
> +		[26] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
> +			RTE_PTYPE_L4_TCP,
> +		[27] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
> +			RTE_PTYPE_L4_SCTP,
> +		[28] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
> +			RTE_PTYPE_L4_ICMP,
> +		/* All others reserved */
> +	};
> +	const __m256i mbuf_init = _mm256_set_epi64x(0, 0,
> +			0, rxq->mbuf_initializer);
> +	/* struct iavf_rx_entry *sw_ring = &rxq->sw_ring[rxq->rx_tail]; */
> +	struct rte_mbuf **sw_ring = &rxq->sw_ring[rxq->rx_tail];
> +	volatile union iavf_rx_desc *rxdp = rxq->rx_ring + rxq->rx_tail;
> +	const int avx_aligned = ((rxq->rx_tail & 1) == 0);
> +
> +	rte_prefetch0(rxdp);
> +
> +	/* nb_pkts has to be floor-aligned to IAVF_DESCS_PER_LOOP_AVX */
> +	nb_pkts = RTE_ALIGN_FLOOR(nb_pkts, IAVF_DESCS_PER_LOOP_AVX);
> +
> +	/* See if we need to rearm the RX queue - gives the prefetch a bit
> +	 * of time to act
> +	 */
> +	if (rxq->rxrearm_nb > IAVF_RXQ_REARM_THRESH)
> +		iavf_rxq_rearm(rxq);
> +
> +	/* Before we start moving massive data around, check to see if
> +	 * there is actually a packet available
> +	 */
> +	if (!(rxdp->wb.qword1.status_error_len &
> +			rte_cpu_to_le_32(1 << IAVF_RX_DESC_STATUS_DD_SHIFT)))
> +		return 0;
> +
> +	/* constants used in processing loop */
> +	const __m256i crc_adjust =
> +		_mm256_set_epi16
> +			(/* first descriptor */
> +			 0, 0, 0,       /* ignore non-length fields */
> +			 -rxq->crc_len, /* sub crc on data_len */
> +			 0,             /* ignore high-16bits of pkt_len */
> +			 -rxq->crc_len, /* sub crc on pkt_len */
> +			 0, 0,          /* ignore pkt_type field */
> +			 /* second descriptor */
> +			 0, 0, 0,       /* ignore non-length fields */
> +			 -rxq->crc_len, /* sub crc on data_len */
> +			 0,             /* ignore high-16bits of pkt_len */
> +			 -rxq->crc_len, /* sub crc on pkt_len */
> +			 0, 0           /* ignore pkt_type field */
> +			);
> +
> +	/* 8 packets DD mask, LSB in each 32-bit value */
> +	const __m256i dd_check = _mm256_set1_epi32(1);
> +
> +	/* 8 packets EOP mask, second-LSB in each 32-bit value */
> +	const __m256i eop_check = _mm256_slli_epi32(dd_check,
> +			IAVF_RX_DESC_STATUS_EOF_SHIFT);
> +
> +	/* mask to shuffle from desc. to mbuf (2 descriptors)*/
> +	const __m256i shuf_msk =
> +		_mm256_set_epi8
> +			(/* first descriptor */
> +			 7, 6, 5, 4,  /* octet 4~7, 32bits rss */
> +			 3, 2,        /* octet 2~3, low 16 bits vlan_macip */
> +			 15, 14,      /* octet 15~14, 16 bits data_len */
> +			 0xFF, 0xFF,  /* skip high 16 bits pkt_len, zero out */
> +			 15, 14,      /* octet 15~14, low 16 bits pkt_len */
> +			 0xFF, 0xFF,  /* pkt_type set as unknown */
> +			 0xFF, 0xFF,  /*pkt_type set as unknown */
> +			 /* second descriptor */
> +			 7, 6, 5, 4,  /* octet 4~7, 32bits rss */
> +			 3, 2,        /* octet 2~3, low 16 bits vlan_macip */
> +			 15, 14,      /* octet 15~14, 16 bits data_len */
> +			 0xFF, 0xFF,  /* skip high 16 bits pkt_len, zero out */
> +			 15, 14,      /* octet 15~14, low 16 bits pkt_len */
> +			 0xFF, 0xFF,  /* pkt_type set as unknown */
> +			 0xFF, 0xFF   /*pkt_type set as unknown */
> +			);
> +	/**
> +	 * compile-time check the above crc and shuffle layout is correct.
> +	 * NOTE: the first field (lowest address) is given last in set_epi
> +	 * calls above.
> +	 */
> +	RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, pkt_len) !=
> +			offsetof(struct rte_mbuf, rx_descriptor_fields1) + 4);
> +	RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, data_len) !=
> +			offsetof(struct rte_mbuf, rx_descriptor_fields1) + 8);
> +	RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, vlan_tci) !=
> +			offsetof(struct rte_mbuf, rx_descriptor_fields1) + 10);
> +	RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, hash) !=
> +			offsetof(struct rte_mbuf, rx_descriptor_fields1) + 12);
> +
> +	/* Status/Error flag masks */
> +	/**
> +	 * mask everything except RSS, flow director and VLAN flags
> +	 * bit2 is for VLAN tag, bit11 for flow director indication
> +	 * bit13:12 for RSS indication. Bits 3-5 of error
> +	 * field (bits 22-24) are for IP/L4 checksum errors
> +	 */
> +	const __m256i flags_mask =
> +		 _mm256_set1_epi32((1 << 2) | (1 << 11) |
> +				   (3 << 12) | (7 << 22));
> +	/**
> +	 * data to be shuffled by result of flag mask. If VLAN bit is set,
> +	 * (bit 2), then position 4 in this array will be used in the
> +	 * destination
> +	 */
> +	const __m256i vlan_flags_shuf =
> +		_mm256_set_epi32(0, 0, PKT_RX_VLAN | PKT_RX_VLAN_STRIPPED, 0,
> +				 0, 0, PKT_RX_VLAN | PKT_RX_VLAN_STRIPPED, 0);
> +	/**
> +	 * data to be shuffled by result of flag mask, shifted down 11.
> +	 * If RSS/FDIR bits are set, shuffle moves appropriate flags in
> +	 * place.
> +	 */
> +	const __m256i rss_flags_shuf =
> +		_mm256_set_epi8(0, 0, 0, 0, 0, 0, 0, 0,
> +				PKT_RX_RSS_HASH | PKT_RX_FDIR, PKT_RX_RSS_HASH,
> +				0, 0, 0, 0, PKT_RX_FDIR, 0,/* end up 128-bits */
> +				0, 0, 0, 0, 0, 0, 0, 0,
> +				PKT_RX_RSS_HASH | PKT_RX_FDIR, PKT_RX_RSS_HASH,
> +				0, 0, 0, 0, PKT_RX_FDIR, 0);
> +
> +	/**
> +	 * data to be shuffled by the result of the flags mask shifted by 22
> +	 * bits.  This gives use the l3_l4 flags.
> +	 */
> +	const __m256i l3_l4_flags_shuf = _mm256_set_epi8(0, 0, 0, 0, 0, 0, 0, 0,
> +			/* shift right 1 bit to make sure it not exceed 255 */
> +			(PKT_RX_EIP_CKSUM_BAD | PKT_RX_L4_CKSUM_BAD |
> +			 PKT_RX_IP_CKSUM_BAD) >> 1,
> +			(PKT_RX_IP_CKSUM_GOOD | PKT_RX_EIP_CKSUM_BAD |
> +			 PKT_RX_L4_CKSUM_BAD) >> 1,
> +			(PKT_RX_EIP_CKSUM_BAD | PKT_RX_IP_CKSUM_BAD) >> 1,
> +			(PKT_RX_IP_CKSUM_GOOD | PKT_RX_EIP_CKSUM_BAD) >> 1,
> +			(PKT_RX_L4_CKSUM_BAD | PKT_RX_IP_CKSUM_BAD) >> 1,
> +			(PKT_RX_IP_CKSUM_GOOD | PKT_RX_L4_CKSUM_BAD) >> 1,
> +			PKT_RX_IP_CKSUM_BAD >> 1,
> +			(PKT_RX_IP_CKSUM_GOOD | PKT_RX_L4_CKSUM_GOOD) >> 1,
> +			/* second 128-bits */
> +			0, 0, 0, 0, 0, 0, 0, 0,
> +			(PKT_RX_EIP_CKSUM_BAD | PKT_RX_L4_CKSUM_BAD |
> +			 PKT_RX_IP_CKSUM_BAD) >> 1,
> +			(PKT_RX_IP_CKSUM_GOOD | PKT_RX_EIP_CKSUM_BAD |
> +			 PKT_RX_L4_CKSUM_BAD) >> 1,
> +			(PKT_RX_EIP_CKSUM_BAD | PKT_RX_IP_CKSUM_BAD) >> 1,
> +			(PKT_RX_IP_CKSUM_GOOD | PKT_RX_EIP_CKSUM_BAD) >> 1,
> +			(PKT_RX_L4_CKSUM_BAD | PKT_RX_IP_CKSUM_BAD) >> 1,
> +			(PKT_RX_IP_CKSUM_GOOD | PKT_RX_L4_CKSUM_BAD) >> 1,
> +			PKT_RX_IP_CKSUM_BAD >> 1,
> +			(PKT_RX_IP_CKSUM_GOOD | PKT_RX_L4_CKSUM_GOOD) >> 1);
> +
> +	const __m256i cksum_mask =
> +		 _mm256_set1_epi32(PKT_RX_IP_CKSUM_GOOD | PKT_RX_IP_CKSUM_BAD |
> +				   PKT_RX_L4_CKSUM_GOOD | PKT_RX_L4_CKSUM_BAD |
> +				   PKT_RX_EIP_CKSUM_BAD);
> +
> +	RTE_SET_USED(avx_aligned); /* for 32B descriptors we don't use this */
> +
> +	uint16_t i, received;
> +
> +	for (i = 0, received = 0; i < nb_pkts;
> +	     i += IAVF_DESCS_PER_LOOP_AVX,
> +	     rxdp += IAVF_DESCS_PER_LOOP_AVX) {
> +		/* step 1, copy over 8 mbuf pointers to rx_pkts array */
> +		_mm256_storeu_si256((void *)&rx_pkts[i],
> +				    _mm256_loadu_si256((void *)&sw_ring[i]));
> +#ifdef RTE_ARCH_X86_64
> +		_mm256_storeu_si256
> +			((void *)&rx_pkts[i + 4],
> +			 _mm256_loadu_si256((void *)&sw_ring[i + 4]));
> +#endif
> +
> +		__m256i raw_desc0_1, raw_desc2_3, raw_desc4_5, raw_desc6_7;
> +#ifdef RTE_LIBRTE_IAVF_16BYTE_RX_DESC
> +		/* for AVX we need alignment otherwise loads are not atomic */
> +		if (avx_aligned) {
> +			/* load in descriptors, 2 at a time, in reverse order */
> +			raw_desc6_7 = _mm256_load_si256((void *)(rxdp + 6));
> +			rte_compiler_barrier();
> +			raw_desc4_5 = _mm256_load_si256((void *)(rxdp + 4));
> +			rte_compiler_barrier();
> +			raw_desc2_3 = _mm256_load_si256((void *)(rxdp + 2));
> +			rte_compiler_barrier();
> +			raw_desc0_1 = _mm256_load_si256((void *)(rxdp + 0));
> +		} else
> +#endif
> +		{
> +			const __m128i raw_desc7 =
> +				_mm_load_si128((void *)(rxdp + 7));
> +			rte_compiler_barrier();
> +			const __m128i raw_desc6 =
> +				_mm_load_si128((void *)(rxdp + 6));
> +			rte_compiler_barrier();
> +			const __m128i raw_desc5 =
> +				_mm_load_si128((void *)(rxdp + 5));
> +			rte_compiler_barrier();
> +			const __m128i raw_desc4 =
> +				_mm_load_si128((void *)(rxdp + 4));
> +			rte_compiler_barrier();
> +			const __m128i raw_desc3 =
> +				_mm_load_si128((void *)(rxdp + 3));
> +			rte_compiler_barrier();
> +			const __m128i raw_desc2 =
> +				_mm_load_si128((void *)(rxdp + 2));
> +			rte_compiler_barrier();
> +			const __m128i raw_desc1 =
> +				_mm_load_si128((void *)(rxdp + 1));
> +			rte_compiler_barrier();
> +			const __m128i raw_desc0 =
> +				_mm_load_si128((void *)(rxdp + 0));
> +
> +			raw_desc6_7 =
> +				_mm256_inserti128_si256
> +					(_mm256_castsi128_si256(raw_desc6),
> +					 raw_desc7, 1);
> +			raw_desc4_5 =
> +				_mm256_inserti128_si256
> +					(_mm256_castsi128_si256(raw_desc4),
> +					 raw_desc5, 1);
> +			raw_desc2_3 =
> +				_mm256_inserti128_si256
> +					(_mm256_castsi128_si256(raw_desc2),
> +					 raw_desc3, 1);
> +			raw_desc0_1 =
> +				_mm256_inserti128_si256
> +					(_mm256_castsi128_si256(raw_desc0),
> +					 raw_desc1, 1);
> +		}
> +
> +		if (split_packet) {
> +			int j;
> +
> +			for (j = 0; j < IAVF_DESCS_PER_LOOP_AVX; j++)
> +				rte_mbuf_prefetch_part2(rx_pkts[i + j]);
> +		}
> +
> +		/**
> +		 * convert descriptors 4-7 into mbufs, adjusting length and
> +		 * re-arranging fields. Then write into the mbuf
> +		 */
> +		const __m256i len6_7 = _mm256_slli_epi32(raw_desc6_7,
> +							 PKTLEN_SHIFT);
> +		const __m256i len4_5 = _mm256_slli_epi32(raw_desc4_5,
> +							 PKTLEN_SHIFT);
> +		const __m256i desc6_7 = _mm256_blend_epi16(raw_desc6_7,
> +							   len6_7, 0x80);
> +		const __m256i desc4_5 = _mm256_blend_epi16(raw_desc4_5,
> +							   len4_5, 0x80);
> +		__m256i mb6_7 = _mm256_shuffle_epi8(desc6_7, shuf_msk);
> +		__m256i mb4_5 = _mm256_shuffle_epi8(desc4_5, shuf_msk);
> +
> +		mb6_7 = _mm256_add_epi16(mb6_7, crc_adjust);
> +		mb4_5 = _mm256_add_epi16(mb4_5, crc_adjust);
> +		/**
> +		 * to get packet types, shift 64-bit values down 30 bits
> +		 * and so ptype is in lower 8-bits in each
> +		 */
> +		const __m256i ptypes6_7 = _mm256_srli_epi64(desc6_7, 30);
> +		const __m256i ptypes4_5 = _mm256_srli_epi64(desc4_5, 30);
> +		const uint8_t ptype7 = _mm256_extract_epi8(ptypes6_7, 24);
> +		const uint8_t ptype6 = _mm256_extract_epi8(ptypes6_7, 8);
> +		const uint8_t ptype5 = _mm256_extract_epi8(ptypes4_5, 24);
> +		const uint8_t ptype4 = _mm256_extract_epi8(ptypes4_5, 8);
> +
> +		mb6_7 = _mm256_insert_epi32(mb6_7, type_table[ptype7], 4);
> +		mb6_7 = _mm256_insert_epi32(mb6_7, type_table[ptype6], 0);
> +		mb4_5 = _mm256_insert_epi32(mb4_5, type_table[ptype5], 4);
> +		mb4_5 = _mm256_insert_epi32(mb4_5, type_table[ptype4], 0);
> +		/* merge the status bits into one register */
> +		const __m256i status4_7 = _mm256_unpackhi_epi32(desc6_7,
> +				desc4_5);
> +
> +		/**
> +		 * convert descriptors 0-3 into mbufs, adjusting length and
> +		 * re-arranging fields. Then write into the mbuf
> +		 */
> +		const __m256i len2_3 = _mm256_slli_epi32(raw_desc2_3,
> +							 PKTLEN_SHIFT);
> +		const __m256i len0_1 = _mm256_slli_epi32(raw_desc0_1,
> +							 PKTLEN_SHIFT);
> +		const __m256i desc2_3 = _mm256_blend_epi16(raw_desc2_3,
> +							   len2_3, 0x80);
> +		const __m256i desc0_1 = _mm256_blend_epi16(raw_desc0_1,
> +							   len0_1, 0x80);
> +		__m256i mb2_3 = _mm256_shuffle_epi8(desc2_3, shuf_msk);
> +		__m256i mb0_1 = _mm256_shuffle_epi8(desc0_1, shuf_msk);
> +
> +		mb2_3 = _mm256_add_epi16(mb2_3, crc_adjust);
> +		mb0_1 = _mm256_add_epi16(mb0_1, crc_adjust);
> +		/* get the packet types */
> +		const __m256i ptypes2_3 = _mm256_srli_epi64(desc2_3, 30);
> +		const __m256i ptypes0_1 = _mm256_srli_epi64(desc0_1, 30);
> +		const uint8_t ptype3 = _mm256_extract_epi8(ptypes2_3, 24);
> +		const uint8_t ptype2 = _mm256_extract_epi8(ptypes2_3, 8);
> +		const uint8_t ptype1 = _mm256_extract_epi8(ptypes0_1, 24);
> +		const uint8_t ptype0 = _mm256_extract_epi8(ptypes0_1, 8);
> +
> +		mb2_3 = _mm256_insert_epi32(mb2_3, type_table[ptype3], 4);
> +		mb2_3 = _mm256_insert_epi32(mb2_3, type_table[ptype2], 0);
> +		mb0_1 = _mm256_insert_epi32(mb0_1, type_table[ptype1], 4);
> +		mb0_1 = _mm256_insert_epi32(mb0_1, type_table[ptype0], 0);
> +		/* merge the status bits into one register */
> +		const __m256i status0_3 = _mm256_unpackhi_epi32(desc2_3,
> +								desc0_1);
> +
> +		/**
> +		 * take the two sets of status bits and merge to one
> +		 * After merge, the packets status flags are in the
> +		 * order (hi->lo): [1, 3, 5, 7, 0, 2, 4, 6]
> +		 */
> +		__m256i status0_7 = _mm256_unpacklo_epi64(status4_7,
> +							  status0_3);
> +
> +		/* now do flag manipulation */
> +
> +		/* get only flag/error bits we want */
> +		const __m256i flag_bits =
> +			_mm256_and_si256(status0_7, flags_mask);
> +		/* set vlan and rss flags */
> +		const __m256i vlan_flags =
> +			_mm256_shuffle_epi8(vlan_flags_shuf, flag_bits);
> +		const __m256i rss_flags =
> +			_mm256_shuffle_epi8(rss_flags_shuf,
> +					    _mm256_srli_epi32(flag_bits, 11));
> +		/**
> +		 * l3_l4_error flags, shuffle, then shift to correct adjustment
> +		 * of flags in flags_shuf, and finally mask out extra bits
> +		 */
> +		__m256i l3_l4_flags = _mm256_shuffle_epi8(l3_l4_flags_shuf,
> +				_mm256_srli_epi32(flag_bits, 22));
> +		l3_l4_flags = _mm256_slli_epi32(l3_l4_flags, 1);
> +		l3_l4_flags = _mm256_and_si256(l3_l4_flags, cksum_mask);
> +
> +		/* merge flags */
> +		const __m256i mbuf_flags = _mm256_or_si256(l3_l4_flags,
> +				_mm256_or_si256(rss_flags, vlan_flags));
> +		/**
> +		 * At this point, we have the 8 sets of flags in the low 16-bits
> +		 * of each 32-bit value in vlan0.
> +		 * We want to extract these, and merge them with the mbuf init
> +		 * data so we can do a single write to the mbuf to set the flags
> +		 * and all the other initialization fields. Extracting the
> +		 * appropriate flags means that we have to do a shift and blend
> +		 * for each mbuf before we do the write. However, we can also
> +		 * add in the previously computed rx_descriptor fields to
> +		 * make a single 256-bit write per mbuf
> +		 */
> +		/* check the structure matches expectations */
> +		RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, ol_flags) !=
> +				 offsetof(struct rte_mbuf, rearm_data) + 8);
> +		RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, rearm_data) !=
> +				 RTE_ALIGN(offsetof(struct rte_mbuf,
> +						    rearm_data),
> +					   16));
> +		/* build up data and do writes */
> +		__m256i rearm0, rearm1, rearm2, rearm3, rearm4, rearm5,
> +			rearm6, rearm7;
> +		rearm6 = _mm256_blend_epi32(mbuf_init,
> +					    _mm256_slli_si256(mbuf_flags, 8),
> +					    0x04);
> +		rearm4 = _mm256_blend_epi32(mbuf_init,
> +					    _mm256_slli_si256(mbuf_flags, 4),
> +					    0x04);
> +		rearm2 = _mm256_blend_epi32(mbuf_init, mbuf_flags, 0x04);
> +		rearm0 = _mm256_blend_epi32(mbuf_init,
> +					    _mm256_srli_si256(mbuf_flags, 4),
> +					    0x04);
> +		/* permute to add in the rx_descriptor e.g. rss fields */
> +		rearm6 = _mm256_permute2f128_si256(rearm6, mb6_7, 0x20);
> +		rearm4 = _mm256_permute2f128_si256(rearm4, mb4_5, 0x20);
> +		rearm2 = _mm256_permute2f128_si256(rearm2, mb2_3, 0x20);
> +		rearm0 = _mm256_permute2f128_si256(rearm0, mb0_1, 0x20);
> +		/* write to mbuf */
> +		_mm256_storeu_si256((__m256i *)&rx_pkts[i + 6]->rearm_data,
> +				    rearm6);
> +		_mm256_storeu_si256((__m256i *)&rx_pkts[i + 4]->rearm_data,
> +				    rearm4);
> +		_mm256_storeu_si256((__m256i *)&rx_pkts[i + 2]->rearm_data,
> +				    rearm2);
> +		_mm256_storeu_si256((__m256i *)&rx_pkts[i + 0]->rearm_data,
> +				    rearm0);
> +
> +		/* repeat for the odd mbufs */
> +		const __m256i odd_flags =
> +			_mm256_castsi128_si256
> +				(_mm256_extracti128_si256(mbuf_flags, 1));
> +		rearm7 = _mm256_blend_epi32(mbuf_init,
> +					    _mm256_slli_si256(odd_flags, 8),
> +					    0x04);
> +		rearm5 = _mm256_blend_epi32(mbuf_init,
> +					    _mm256_slli_si256(odd_flags, 4),
> +					    0x04);
> +		rearm3 = _mm256_blend_epi32(mbuf_init, odd_flags, 0x04);
> +		rearm1 = _mm256_blend_epi32(mbuf_init,
> +					    _mm256_srli_si256(odd_flags, 4),
> +					    0x04);
> +		/* since odd mbufs are already in hi 128-bits use blend */
> +		rearm7 = _mm256_blend_epi32(rearm7, mb6_7, 0xF0);
> +		rearm5 = _mm256_blend_epi32(rearm5, mb4_5, 0xF0);
> +		rearm3 = _mm256_blend_epi32(rearm3, mb2_3, 0xF0);
> +		rearm1 = _mm256_blend_epi32(rearm1, mb0_1, 0xF0);
> +		/* again write to mbufs */
> +		_mm256_storeu_si256((__m256i *)&rx_pkts[i + 7]->rearm_data,
> +				    rearm7);
> +		_mm256_storeu_si256((__m256i *)&rx_pkts[i + 5]->rearm_data,
> +				    rearm5);
> +		_mm256_storeu_si256((__m256i *)&rx_pkts[i + 3]->rearm_data,
> +				    rearm3);
> +		_mm256_storeu_si256((__m256i *)&rx_pkts[i + 1]->rearm_data,
> +				    rearm1);
> +
> +		/* extract and record EOP bit */
> +		if (split_packet) {
> +			const __m128i eop_mask =
> +				_mm_set1_epi16(1 << IAVF_RX_DESC_STATUS_EOF_SHIFT);
> +			const __m256i eop_bits256 = _mm256_and_si256(status0_7,
> +								     eop_check);
> +			/* pack status bits into a single 128-bit register */
> +			const __m128i eop_bits =
> +				_mm_packus_epi32
> +					(_mm256_castsi256_si128(eop_bits256),
> +					 _mm256_extractf128_si256(eop_bits256,
> +								  1));
> +			/**
> +			 * flip bits, and mask out the EOP bit, which is now
> +			 * a split-packet bit i.e. !EOP, rather than EOP one.
> +			 */
> +			__m128i split_bits = _mm_andnot_si128(eop_bits,
> +					eop_mask);
> +			/**
> +			 * eop bits are out of order, so we need to shuffle them
> +			 * back into order again. In doing so, only use low 8
> +			 * bits, which acts like another pack instruction
> +			 * The original order is (hi->lo): 1,3,5,7,0,2,4,6
> +			 * [Since we use epi8, the 16-bit positions are
> +			 * multiplied by 2 in the eop_shuffle value.]
> +			 */
> +			__m128i eop_shuffle =
> +				_mm_set_epi8(/* zero hi 64b */
> +					     0xFF, 0xFF, 0xFF, 0xFF,
> +					     0xFF, 0xFF, 0xFF, 0xFF,
> +					     /* move values to lo 64b */
> +					     8, 0, 10, 2,
> +					     12, 4, 14, 6);
> +			split_bits = _mm_shuffle_epi8(split_bits, eop_shuffle);
> +			*(uint64_t *)split_packet =
> +				_mm_cvtsi128_si64(split_bits);
> +			split_packet += IAVF_DESCS_PER_LOOP_AVX;
> +		}
> +
> +		/* perform dd_check */
> +		status0_7 = _mm256_and_si256(status0_7, dd_check);
> +		status0_7 = _mm256_packs_epi32(status0_7,
> +					       _mm256_setzero_si256());
> +
> +		uint64_t burst = __builtin_popcountll
> +					(_mm_cvtsi128_si64
> +						(_mm256_extracti128_si256
> +							(status0_7, 1)));
> +		burst += __builtin_popcountll
> +				(_mm_cvtsi128_si64
> +					(_mm256_castsi256_si128(status0_7)));
> +		received += burst;
> +		if (burst != IAVF_DESCS_PER_LOOP_AVX)
> +			break;
> +	}
> +
> +	/* update tail pointers */
> +	rxq->rx_tail += received;
> +	rxq->rx_tail &= (rxq->nb_rx_desc - 1);
> +	if ((rxq->rx_tail & 1) == 1 && received > 1) { /* keep avx2 aligned */
> +		rxq->rx_tail--;
> +		received--;
> +	}
> +	rxq->rxrearm_nb += received;
> +	return received;
> +}
> +
> +/**
> + * Notice:
> + * - nb_pkts < IAVF_DESCS_PER_LOOP, just return no packet
> + */
> +uint16_t
> +iavf_recv_pkts_vec_avx2(void *rx_queue, struct rte_mbuf **rx_pkts,
> +			uint16_t nb_pkts)
> +{
> +	return _iavf_recv_raw_pkts_vec_avx2(rx_queue, rx_pkts, nb_pkts, NULL);
> +}
> +
> +/**
> + * vPMD receive routine that reassembles single burst of 32 scattered packets
> + * Notice:
> + * - nb_pkts < IAVF_DESCS_PER_LOOP, just return no packet
> + */
> +static uint16_t
> +iavf_recv_scattered_burst_vec_avx2(void *rx_queue, struct rte_mbuf **rx_pkts,
> +				   uint16_t nb_pkts)
> +{
> +	struct iavf_rx_queue *rxq = rx_queue;
> +	uint8_t split_flags[IAVF_VPMD_RX_MAX_BURST] = {0};
> +
> +	/* get some new buffers */
> +	uint16_t nb_bufs = _iavf_recv_raw_pkts_vec_avx2(rxq, rx_pkts, nb_pkts,
> +						       split_flags);
> +	if (nb_bufs == 0)
> +		return 0;
> +
> +	/* happy day case, full burst + no packets to be joined */
> +	const uint64_t *split_fl64 = (uint64_t *)split_flags;
> +
> +	if (!rxq->pkt_first_seg &&
> +	    split_fl64[0] == 0 && split_fl64[1] == 0 &&
> +	    split_fl64[2] == 0 && split_fl64[3] == 0)
> +		return nb_bufs;
> +
> +	/* reassemble any packets that need reassembly*/
> +	unsigned int i = 0;
> +
> +	if (!rxq->pkt_first_seg) {
> +		/* find the first split flag, and only reassemble then*/
> +		while (i < nb_bufs && !split_flags[i])
> +			i++;
> +		if (i == nb_bufs)
> +			return nb_bufs;
> +		rxq->pkt_first_seg = rx_pkts[i];
> +	}
> +	return i + reassemble_packets(rxq, &rx_pkts[i], nb_bufs - i,
> +					     &split_flags[i]);
> +}
> +
> +/**
> + * vPMD receive routine that reassembles scattered packets.
> + * Main receive routine that can handle arbitrary burst sizes
> + * Notice:
> + * - nb_pkts < IAVF_DESCS_PER_LOOP, just return no packet
> + */
> +uint16_t
> +iavf_recv_scattered_pkts_vec_avx2(void *rx_queue, struct rte_mbuf **rx_pkts,
> +				  uint16_t nb_pkts)
> +{
> +	uint16_t retval = 0;
> +
> +	while (nb_pkts > IAVF_VPMD_RX_MAX_BURST) {
> +		uint16_t burst = iavf_recv_scattered_burst_vec_avx2(rx_queue,
> +				rx_pkts + retval, IAVF_VPMD_RX_MAX_BURST);
> +		retval += burst;
> +		nb_pkts -= burst;
> +		if (burst < IAVF_VPMD_RX_MAX_BURST)
> +			return retval;
> +	}
> +	return retval + iavf_recv_scattered_burst_vec_avx2(rx_queue,
> +				rx_pkts + retval, nb_pkts);
> +}
> +
> +static inline void
> +iavf_vtx1(volatile struct iavf_tx_desc *txdp,
> +	  struct rte_mbuf *pkt, uint64_t flags)
> +{
> +	uint64_t high_qw =
> +		(IAVF_TX_DESC_DTYPE_DATA |
> +		 ((uint64_t)flags  << IAVF_TXD_QW1_CMD_SHIFT) |
> +		 ((uint64_t)pkt->data_len << IAVF_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
> +iavf_vtx(volatile struct iavf_tx_desc *txdp,
> +	 struct rte_mbuf **pkt, uint16_t nb_pkts,  uint64_t flags)
> +{
> +	const uint64_t hi_qw_tmpl = (IAVF_TX_DESC_DTYPE_DATA |
> +			((uint64_t)flags  << IAVF_TXD_QW1_CMD_SHIFT));
> +
> +	/* if unaligned on 32-bit boundary, do one to align */
> +	if (((uintptr_t)txdp & 0x1F) != 0 && nb_pkts != 0) {
> +		iavf_vtx1(txdp, *pkt, flags);
> +		nb_pkts--, txdp++, pkt++;
> +	}
> +
> +	/* do two at a time while possible, in bursts */
> +	for (; nb_pkts > 3; txdp += 4, pkt += 4, nb_pkts -= 4) {
> +		uint64_t hi_qw3 =
> +			hi_qw_tmpl |
> +			((uint64_t)pkt[3]->data_len <<
> +			 IAVF_TXD_QW1_TX_BUF_SZ_SHIFT);
> +		uint64_t hi_qw2 =
> +			hi_qw_tmpl |
> +			((uint64_t)pkt[2]->data_len <<
> +			 IAVF_TXD_QW1_TX_BUF_SZ_SHIFT);
> +		uint64_t hi_qw1 =
> +			hi_qw_tmpl |
> +			((uint64_t)pkt[1]->data_len <<
> +			 IAVF_TXD_QW1_TX_BUF_SZ_SHIFT);
> +		uint64_t hi_qw0 =
> +			hi_qw_tmpl |
> +			((uint64_t)pkt[0]->data_len <<
> +			 IAVF_TXD_QW1_TX_BUF_SZ_SHIFT);
> +
> +		__m256i desc2_3 =
> +			_mm256_set_epi64x
> +				(hi_qw3,
> +				 pkt[3]->buf_physaddr + pkt[3]->data_off,
> +				 hi_qw2,
> +				 pkt[2]->buf_physaddr + pkt[2]->data_off);
> +		__m256i desc0_1 =
> +			_mm256_set_epi64x
> +				(hi_qw1,
> +				 pkt[1]->buf_physaddr + pkt[1]->data_off,
> +				 hi_qw0,
> +				 pkt[0]->buf_physaddr + pkt[0]->data_off);
> +		_mm256_store_si256((void *)(txdp + 2), desc2_3);
> +		_mm256_store_si256((void *)txdp, desc0_1);
> +	}
> +
> +	/* do any last ones */
> +	while (nb_pkts) {
> +		iavf_vtx1(txdp, *pkt, flags);
> +		txdp++, pkt++, nb_pkts--;
> +	}
> +}
> +
> +static inline uint16_t
> +iavf_xmit_fixed_burst_vec_avx2(void *tx_queue, struct rte_mbuf **tx_pkts,
> +			       uint16_t nb_pkts)
> +{
> +	struct iavf_tx_queue *txq = (struct iavf_tx_queue *)tx_queue;
> +	volatile struct iavf_tx_desc *txdp;
> +	struct iavf_tx_entry *txep;
> +	uint16_t n, nb_commit, tx_id;
> +	uint64_t flags = IAVF_TX_DESC_CMD_EOP;
> +	uint64_t rs = IAVF_TX_DESC_CMD_RS | IAVF_TX_DESC_CMD_EOP;
> +
> +	/* cross rx_thresh boundary is not allowed */
> +	nb_pkts = RTE_MIN(nb_pkts, txq->rs_thresh);
> +
> +	if (txq->nb_free < txq->free_thresh)
> +		iavf_tx_free_bufs(txq);
> +
> +	nb_commit = nb_pkts = (uint16_t)RTE_MIN(txq->nb_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_free = (uint16_t)(txq->nb_free - nb_pkts);
> +
> +	n = (uint16_t)(txq->nb_tx_desc - tx_id);
> +	if (nb_commit >= n) {
> +		tx_backlog_entry(txep, tx_pkts, n);
> +
> +		iavf_vtx(txdp, tx_pkts, n - 1, flags);
> +		tx_pkts += (n - 1);
> +		txdp += (n - 1);
> +
> +		iavf_vtx1(txdp, *tx_pkts++, rs);
> +
> +		nb_commit = (uint16_t)(nb_commit - n);
> +
> +		tx_id = 0;
> +		txq->next_rs = (uint16_t)(txq->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);
> +
> +	iavf_vtx(txdp, tx_pkts, nb_commit, flags);
> +
> +	tx_id = (uint16_t)(tx_id + nb_commit);
> +	if (tx_id > txq->next_rs) {
> +		txq->tx_ring[txq->next_rs].cmd_type_offset_bsz |=
> +			rte_cpu_to_le_64(((uint64_t)IAVF_TX_DESC_CMD_RS) <<
> +					 IAVF_TXD_QW1_CMD_SHIFT);
> +		txq->next_rs =
> +			(uint16_t)(txq->next_rs + txq->rs_thresh);
> +	}
> +
> +	txq->tx_tail = tx_id;
> +
> +	IAVF_PCI_REG_WRITE(txq->qtx_tail, txq->tx_tail);
> +
> +	return nb_pkts;
> +}
> +
> +uint16_t
> +iavf_xmit_pkts_vec_avx2(void *tx_queue, struct rte_mbuf **tx_pkts,
> +			uint16_t nb_pkts)
> +{
> +	uint16_t nb_tx = 0;
> +	struct iavf_tx_queue *txq = (struct iavf_tx_queue *)tx_queue;
> +
> +	while (nb_pkts) {
> +		uint16_t ret, num;
> +
> +		num = (uint16_t)RTE_MIN(nb_pkts, txq->rs_thresh);
> +		ret = iavf_xmit_fixed_burst_vec_avx2(tx_queue, &tx_pkts[nb_tx],
> +						     num);
> +		nb_tx += ret;
> +		nb_pkts -= ret;
> +		if (ret < num)
> +			break;
> +	}
> +
> +	return nb_tx;
> +}
> diff --git a/drivers/net/iavf/iavf_rxtx_vec_common.h b/drivers/net/iavf/iavf_rxtx_vec_common.h
> index db509d71f..fff0555d2 100644
> --- a/drivers/net/iavf/iavf_rxtx_vec_common.h
> +++ b/drivers/net/iavf/iavf_rxtx_vec_common.h
> @@ -207,4 +207,76 @@ iavf_rxq_vec_setup_default(struct iavf_rx_queue *rxq)
>  	rxq->mbuf_initializer = *(uint64_t *)p;
>  	return 0;
>  }
> +
> +static inline int
> +iavf_rx_vec_queue_default(struct iavf_rx_queue *rxq)
> +{
> +	if (!rxq)
> +		return -1;
> +
> +	if (!rte_is_power_of_2(rxq->nb_rx_desc))
> +		return -1;
> +
> +	if (rxq->rx_free_thresh < IAVF_VPMD_RX_MAX_BURST)
> +		return -1;
> +
> +	if (rxq->nb_rx_desc % rxq->rx_free_thresh)
> +		return -1;
> +
> +	return 0;
> +}
> +
> +#define IAVF_NO_VECTOR_FLAGS (				 \
> +		DEV_TX_OFFLOAD_MULTI_SEGS |		 \
> +		DEV_TX_OFFLOAD_VLAN_INSERT |		 \
> +		DEV_TX_OFFLOAD_SCTP_CKSUM |		 \
> +		DEV_TX_OFFLOAD_UDP_CKSUM |		 \
> +		DEV_TX_OFFLOAD_TCP_CKSUM)
> +
> +static inline int
> +iavf_tx_vec_queue_default(struct iavf_tx_queue *txq)
> +{
> +	if (!txq)
> +		return -1;
> +
> +	if (txq->offloads & IAVF_NO_VECTOR_FLAGS)
> +		return -1;
> +
> +	if (txq->rs_thresh < IAVF_VPMD_TX_MAX_BURST ||
> +	    txq->rs_thresh > IAVF_VPMD_TX_MAX_FREE_BUF)
> +		return -1;
> +
> +	return 0;
> +}
> +
> +static inline int
> +iavf_rx_vec_dev_check_default(struct rte_eth_dev *dev)
> +{
> +	int i;
> +	struct iavf_rx_queue *rxq;
> +
> +	for (i = 0; i < dev->data->nb_rx_queues; i++) {
> +		rxq = dev->data->rx_queues[i];
> +		if (iavf_rx_vec_queue_default(rxq))
> +			return -1;
> +	}
> +
> +	return 0;
> +}
> +
> +static inline int
> +iavf_tx_vec_dev_check_default(struct rte_eth_dev *dev)
> +{
> +	int i;
> +	struct iavf_tx_queue *txq;
> +
> +	for (i = 0; i < dev->data->nb_tx_queues; i++) {
> +		txq = dev->data->tx_queues[i];
> +		if (iavf_tx_vec_queue_default(txq))
> +			return -1;
> +	}
> +
> +	return 0;
> +}
> +
>  #endif
> diff --git a/drivers/net/iavf/iavf_rxtx_vec_sse.c b/drivers/net/iavf/iavf_rxtx_vec_sse.c
> index cc71f23a5..15c52a4c2 100644
> --- a/drivers/net/iavf/iavf_rxtx_vec_sse.c
> +++ b/drivers/net/iavf/iavf_rxtx_vec_sse.c
> @@ -655,3 +655,15 @@ iavf_rxq_vec_setup(struct iavf_rx_queue *rxq)
>  	rxq->ops = &sse_vec_rxq_ops;
>  	return iavf_rxq_vec_setup_default(rxq);
>  }
> +
> +int __attribute__((cold))
> +iavf_rx_vec_dev_check(struct rte_eth_dev *dev)
> +{
> +	return iavf_rx_vec_dev_check_default(dev);
> +}
> +
> +int __attribute__((cold))
> +iavf_tx_vec_dev_check(struct rte_eth_dev *dev)
> +{
> +	return iavf_tx_vec_dev_check_default(dev);
> +}
> diff --git a/drivers/net/iavf/meson.build b/drivers/net/iavf/meson.build
> index e5a2f5553..c2401af86 100644
> --- a/drivers/net/iavf/meson.build
> +++ b/drivers/net/iavf/meson.build
> @@ -17,4 +17,21 @@ sources = files(
>  if arch_subdir == 'x86'
>  	dpdk_conf.set('RTE_LIBRTE_IAVF_INC_VECTOR', 1)
>  	sources += files('iavf_rxtx_vec_sse.c')
> +
> +	# compile AVX2 version if either:
> +	# a. we have AVX supported in minimum instruction set baseline
> +	# b. it's not minimum instruction set, but supported by compiler
> +	if dpdk_conf.has('RTE_MACHINE_CPUFLAG_AVX2')
> +		cflags += ['-DCC_AVX2_SUPPORT']
> +		sources += files('iavf_rxtx_vec_avx2.c')
> +	elif cc.has_argument('-mavx2')
> +		cflags += ['-DCC_AVX2_SUPPORT']
> +		iavf_avx2_lib = static_library('iavf_avx2_lib',
> +				'iavf_rxtx_vec_avx2.c',
> +				dependencies: [static_rte_ethdev,
> +					static_rte_kvargs, static_rte_hash],
> +				include_directories: includes,
> +				c_args: [cflags, '-mavx2'])
> +		objs += iavf_avx2_lib.extract_objects('iavf_rxtx_vec_avx2.c')
> +	endif
>  endif

^ permalink raw reply	[flat|nested] 11+ messages in thread

* [dpdk-dev] [PATCH v2] net/iavf: enable AVX2 for iavf
  2019-09-04 10:06 [dpdk-dev] [PATCH] net/iavf: enable AVX2 for iavf Leyi Rong
  2019-09-04 13:29 ` Aaron Conole
@ 2019-09-06  3:28 ` Leyi Rong
  2019-09-19  8:44 ` [dpdk-dev] [PATCH v3] " Leyi Rong
                   ` (2 subsequent siblings)
  4 siblings, 0 replies; 11+ messages in thread
From: Leyi Rong @ 2019-09-06  3:28 UTC (permalink / raw)
  To: wenzhuo.lu, qi.z.zhang, xiaolong.ye; +Cc: dev, Leyi Rong

This patch enables AVX data path for iavf PMD.

Signed-off-by: Leyi Rong <leyi.rong@intel.com>
---
 drivers/net/iavf/Makefile               |  21 +
 drivers/net/iavf/iavf_rxtx.c            |  99 +--
 drivers/net/iavf/iavf_rxtx.h            |  12 +
 drivers/net/iavf/iavf_rxtx_vec_avx2.c   | 867 ++++++++++++++++++++++++
 drivers/net/iavf/iavf_rxtx_vec_common.h |  72 ++
 drivers/net/iavf/iavf_rxtx_vec_sse.c    |  33 +
 drivers/net/iavf/meson.build            |  17 +
 7 files changed, 1075 insertions(+), 46 deletions(-)
 create mode 100644 drivers/net/iavf/iavf_rxtx_vec_avx2.c

diff --git a/drivers/net/iavf/Makefile b/drivers/net/iavf/Makefile
index cd74e14ab..cf0ed99f0 100644
--- a/drivers/net/iavf/Makefile
+++ b/drivers/net/iavf/Makefile
@@ -48,4 +48,25 @@ ifeq ($(CONFIG_RTE_ARCH_X86), y)
 SRCS-$(CONFIG_RTE_LIBRTE_IAVF_INC_VECTOR) += iavf_rxtx_vec_sse.c
 endif
 
+ifeq ($(CONFIG_RTE_LIBRTE_IAVF_INC_VECTOR), y)
+	ifeq ($(findstring RTE_MACHINE_CPUFLAG_AVX2,$(CFLAGS)),RTE_MACHINE_CPUFLAG_AVX2)
+		CC_AVX2_SUPPORT=1
+	else
+		CC_AVX2_SUPPORT=\
+		$(shell $(CC) -march=core-avx2 -dM -E - </dev/null 2>&1 | \
+		grep -q AVX2 && echo 1)
+		ifeq ($(CC_AVX2_SUPPORT), 1)
+			ifeq ($(CONFIG_RTE_TOOLCHAIN_ICC),y)
+				CFLAGS_iavf_rxtx_vec_avx2.o += -march=core-avx2
+			else
+				CFLAGS_iavf_rxtx_vec_avx2.o += -mavx2
+			endif
+		endif
+	endif
+endif
+
+ifeq ($(CC_AVX2_SUPPORT), 1)
+	SRCS-$(CONFIG_RTE_LIBRTE_IAVF_PMD) += iavf_rxtx_vec_avx2.c
+endif
+
 include $(RTE_SDK)/mk/rte.lib.mk
diff --git a/drivers/net/iavf/iavf_rxtx.c b/drivers/net/iavf/iavf_rxtx.c
index 22d7bb612..e36f035c5 100644
--- a/drivers/net/iavf/iavf_rxtx.c
+++ b/drivers/net/iavf/iavf_rxtx.c
@@ -1666,27 +1666,6 @@ iavf_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts, uint16_t nb_pkts)
 	return nb_tx;
 }
 
-static uint16_t
-iavf_xmit_pkts_vec(void *tx_queue, struct rte_mbuf **tx_pkts,
-		  uint16_t nb_pkts)
-{
-	uint16_t nb_tx = 0;
-	struct iavf_tx_queue *txq = (struct iavf_tx_queue *)tx_queue;
-
-	while (nb_pkts) {
-		uint16_t ret, num;
-
-		num = (uint16_t)RTE_MIN(nb_pkts, txq->rs_thresh);
-		ret = iavf_xmit_fixed_burst_vec(tx_queue, &tx_pkts[nb_tx], num);
-		nb_tx += ret;
-		nb_pkts -= ret;
-		if (ret < num)
-			break;
-	}
-
-	return nb_tx;
-}
-
 /* TX prep functions */
 uint16_t
 iavf_prep_pkts(__rte_unused void *tx_queue, struct rte_mbuf **tx_pkts,
@@ -1741,26 +1720,43 @@ iavf_set_rx_function(struct rte_eth_dev *dev)
 {
 	struct iavf_adapter *adapter =
 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
+#ifdef RTE_ARCH_X86
 	struct iavf_rx_queue *rxq;
 	int i;
+	bool use_avx2 = false;
 
-	if (adapter->rx_vec_allowed) {
-		if (dev->data->scattered_rx) {
-			PMD_DRV_LOG(DEBUG, "Using Vector Scattered Rx callback"
-				    " (port=%d).", dev->data->port_id);
-			dev->rx_pkt_burst = iavf_recv_scattered_pkts_vec;
-		} else {
-			PMD_DRV_LOG(DEBUG, "Using Vector Rx callback"
-				    " (port=%d).", dev->data->port_id);
-			dev->rx_pkt_burst = iavf_recv_pkts_vec;
-		}
+	if (!iavf_rx_vec_dev_check(dev)) {
 		for (i = 0; i < dev->data->nb_rx_queues; i++) {
 			rxq = dev->data->rx_queues[i];
-			if (!rxq)
-				continue;
-			iavf_rxq_vec_setup(rxq);
+			(void)iavf_rxq_vec_setup(rxq);
 		}
-	} else if (dev->data->scattered_rx) {
+
+		if (rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1 ||
+		    rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX512F) == 1)
+			use_avx2 = true;
+
+		if (dev->data->scattered_rx) {
+			PMD_DRV_LOG(DEBUG,
+				    "Using %sVector Scattered Rx (port %d).",
+				    use_avx2 ? "avx2 " : "",
+				    dev->data->port_id);
+			dev->rx_pkt_burst = use_avx2 ?
+					    iavf_recv_scattered_pkts_vec_avx2 :
+					    iavf_recv_scattered_pkts_vec;
+		} else {
+			PMD_DRV_LOG(DEBUG, "Using %sVector Rx (port %d).",
+				    use_avx2 ? "avx2 " : "",
+				    dev->data->port_id);
+			dev->rx_pkt_burst = use_avx2 ?
+					    iavf_recv_pkts_vec_avx2 :
+					    iavf_recv_pkts_vec;
+		}
+
+		return;
+	}
+#endif
+
+	if (dev->data->scattered_rx) {
 		PMD_DRV_LOG(DEBUG, "Using a Scattered Rx callback (port=%d).",
 			    dev->data->port_id);
 		dev->rx_pkt_burst = iavf_recv_scattered_pkts;
@@ -1779,28 +1775,39 @@ iavf_set_rx_function(struct rte_eth_dev *dev)
 void
 iavf_set_tx_function(struct rte_eth_dev *dev)
 {
-	struct iavf_adapter *adapter =
-		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
+#ifdef RTE_ARCH_X86
 	struct iavf_tx_queue *txq;
 	int i;
+	bool use_avx2 = false;
 
-	if (adapter->tx_vec_allowed) {
-		PMD_DRV_LOG(DEBUG, "Using Vector Tx callback (port=%d).",
-			    dev->data->port_id);
-		dev->tx_pkt_burst = iavf_xmit_pkts_vec;
-		dev->tx_pkt_prepare = NULL;
+	if (!iavf_tx_vec_dev_check(dev)) {
 		for (i = 0; i < dev->data->nb_tx_queues; i++) {
 			txq = dev->data->tx_queues[i];
 			if (!txq)
 				continue;
 			iavf_txq_vec_setup(txq);
 		}
-	} else {
-		PMD_DRV_LOG(DEBUG, "Using Basic Tx callback (port=%d).",
+
+		if (rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1 ||
+		    rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX512F) == 1)
+			use_avx2 = true;
+
+		PMD_DRV_LOG(DEBUG, "Using %sVector Tx (port %d).",
+			    use_avx2 ? "avx2 " : "",
 			    dev->data->port_id);
-		dev->tx_pkt_burst = iavf_xmit_pkts;
-		dev->tx_pkt_prepare = iavf_prep_pkts;
+		dev->tx_pkt_burst = use_avx2 ?
+				    iavf_xmit_pkts_vec_avx2 :
+				    iavf_xmit_pkts_vec;
+		dev->tx_pkt_prepare = NULL;
+
+		return;
 	}
+#endif
+
+	PMD_DRV_LOG(DEBUG, "Using Basic Tx callback (port=%d).",
+		    dev->data->port_id);
+	dev->tx_pkt_burst = iavf_xmit_pkts;
+	dev->tx_pkt_prepare = iavf_prep_pkts;
 }
 
 void
diff --git a/drivers/net/iavf/iavf_rxtx.h b/drivers/net/iavf/iavf_rxtx.h
index c86720bda..225a0c4c4 100644
--- a/drivers/net/iavf/iavf_rxtx.h
+++ b/drivers/net/iavf/iavf_rxtx.h
@@ -19,6 +19,7 @@
 /* used for Vector PMD */
 #define IAVF_VPMD_RX_MAX_BURST    32
 #define IAVF_VPMD_TX_MAX_BURST    32
+#define IAVF_RXQ_REARM_THRESH     32
 #define IAVF_VPMD_DESCS_PER_LOOP  4
 #define IAVF_VPMD_TX_MAX_FREE_BUF 64
 
@@ -200,6 +201,17 @@ uint16_t iavf_recv_scattered_pkts_vec(void *rx_queue,
 				     uint16_t nb_pkts);
 uint16_t iavf_xmit_fixed_burst_vec(void *tx_queue, struct rte_mbuf **tx_pkts,
 				  uint16_t nb_pkts);
+uint16_t iavf_recv_pkts_vec_avx2(void *rx_queue, struct rte_mbuf **rx_pkts,
+				 uint16_t nb_pkts);
+uint16_t iavf_recv_scattered_pkts_vec_avx2(void *rx_queue,
+					   struct rte_mbuf **rx_pkts,
+					   uint16_t nb_pkts);
+uint16_t iavf_xmit_pkts_vec(void *tx_queue, struct rte_mbuf **tx_pkts,
+			    uint16_t nb_pkts);
+uint16_t iavf_xmit_pkts_vec_avx2(void *tx_queue, struct rte_mbuf **tx_pkts,
+				 uint16_t nb_pkts);
+int iavf_rx_vec_dev_check(struct rte_eth_dev *dev);
+int iavf_tx_vec_dev_check(struct rte_eth_dev *dev);
 int iavf_rxq_vec_setup(struct iavf_rx_queue *rxq);
 int iavf_txq_vec_setup(struct iavf_tx_queue *txq);
 
diff --git a/drivers/net/iavf/iavf_rxtx_vec_avx2.c b/drivers/net/iavf/iavf_rxtx_vec_avx2.c
new file mode 100644
index 000000000..f0c00be56
--- /dev/null
+++ b/drivers/net/iavf/iavf_rxtx_vec_avx2.c
@@ -0,0 +1,867 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2019 Intel Corporation
+ */
+
+#include "base/iavf_prototype.h"
+#include "iavf_rxtx_vec_common.h"
+
+#include <x86intrin.h>
+
+#ifndef __INTEL_COMPILER
+#pragma GCC diagnostic ignored "-Wcast-qual"
+#endif
+
+static inline void
+iavf_rxq_rearm(struct iavf_rx_queue *rxq)
+{
+	int i;
+	uint16_t rx_id;
+	volatile union iavf_rx_desc *rxdp;
+	struct rte_mbuf **rxp = &rxq->sw_ring[rxq->rxrearm_start];
+
+	rxdp = rxq->rx_ring + rxq->rxrearm_start;
+
+	/* Pull 'n' more MBUFs into the software ring */
+	if (rte_mempool_get_bulk(rxq->mp,
+				 (void *)rxp,
+				 IAVF_RXQ_REARM_THRESH) < 0) {
+		if (rxq->rxrearm_nb + IAVF_RXQ_REARM_THRESH >=
+		    rxq->nb_rx_desc) {
+			__m128i dma_addr0;
+
+			dma_addr0 = _mm_setzero_si128();
+			for (i = 0; i < IAVF_VPMD_DESCS_PER_LOOP; i++) {
+				rxp[i] = &rxq->fake_mbuf;
+				_mm_store_si128((__m128i *)&rxdp[i].read,
+						dma_addr0);
+			}
+		}
+		rte_eth_devices[rxq->port_id].data->rx_mbuf_alloc_failed +=
+			IAVF_RXQ_REARM_THRESH;
+		return;
+	}
+
+#ifndef RTE_LIBRTE_IAVF_16BYTE_RX_DESC
+	struct rte_mbuf *mb0, *mb1;
+	__m128i dma_addr0, dma_addr1;
+	__m128i hdr_room = _mm_set_epi64x(RTE_PKTMBUF_HEADROOM,
+			RTE_PKTMBUF_HEADROOM);
+	/* Initialize the mbufs in vector, process 2 mbufs in one loop */
+	for (i = 0; i < IAVF_RXQ_REARM_THRESH; i += 2, rxp += 2) {
+		__m128i vaddr0, vaddr1;
+
+		mb0 = rxp[0];
+		mb1 = rxp[1];
+
+		/* load buf_addr(lo 64bit) and buf_physaddr(hi 64bit) */
+		RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, buf_physaddr) !=
+				offsetof(struct rte_mbuf, buf_addr) + 8);
+		vaddr0 = _mm_loadu_si128((__m128i *)&mb0->buf_addr);
+		vaddr1 = _mm_loadu_si128((__m128i *)&mb1->buf_addr);
+
+		/* convert pa to dma_addr hdr/data */
+		dma_addr0 = _mm_unpackhi_epi64(vaddr0, vaddr0);
+		dma_addr1 = _mm_unpackhi_epi64(vaddr1, vaddr1);
+
+		/* add headroom to pa values */
+		dma_addr0 = _mm_add_epi64(dma_addr0, hdr_room);
+		dma_addr1 = _mm_add_epi64(dma_addr1, hdr_room);
+
+		/* flush desc with pa dma_addr */
+		_mm_store_si128((__m128i *)&rxdp++->read, dma_addr0);
+		_mm_store_si128((__m128i *)&rxdp++->read, dma_addr1);
+	}
+#else
+	struct rte_mbuf *mb0, *mb1, *mb2, *mb3;
+	__m256i dma_addr0_1, dma_addr2_3;
+	__m256i hdr_room = _mm256_set1_epi64x(RTE_PKTMBUF_HEADROOM);
+	/* Initialize the mbufs in vector, process 4 mbufs in one loop */
+	for (i = 0; i < IAVF_RXQ_REARM_THRESH;
+			i += 4, rxp += 4, rxdp += 4) {
+		__m128i vaddr0, vaddr1, vaddr2, vaddr3;
+		__m256i vaddr0_1, vaddr2_3;
+
+		mb0 = rxp[0];
+		mb1 = rxp[1];
+		mb2 = rxp[2];
+		mb3 = rxp[3];
+
+		/* load buf_addr(lo 64bit) and buf_physaddr(hi 64bit) */
+		RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, buf_physaddr) !=
+				offsetof(struct rte_mbuf, buf_addr) + 8);
+		vaddr0 = _mm_loadu_si128((__m128i *)&mb0->buf_addr);
+		vaddr1 = _mm_loadu_si128((__m128i *)&mb1->buf_addr);
+		vaddr2 = _mm_loadu_si128((__m128i *)&mb2->buf_addr);
+		vaddr3 = _mm_loadu_si128((__m128i *)&mb3->buf_addr);
+
+		/**
+		 * merge 0 & 1, by casting 0 to 256-bit and inserting 1
+		 * into the high lanes. Similarly for 2 & 3
+		 */
+		vaddr0_1 =
+			_mm256_inserti128_si256(_mm256_castsi128_si256(vaddr0),
+						vaddr1, 1);
+		vaddr2_3 =
+			_mm256_inserti128_si256(_mm256_castsi128_si256(vaddr2),
+						vaddr3, 1);
+
+		/* convert pa to dma_addr hdr/data */
+		dma_addr0_1 = _mm256_unpackhi_epi64(vaddr0_1, vaddr0_1);
+		dma_addr2_3 = _mm256_unpackhi_epi64(vaddr2_3, vaddr2_3);
+
+		/* add headroom to pa values */
+		dma_addr0_1 = _mm256_add_epi64(dma_addr0_1, hdr_room);
+		dma_addr2_3 = _mm256_add_epi64(dma_addr2_3, hdr_room);
+
+		/* flush desc with pa dma_addr */
+		_mm256_store_si256((__m256i *)&rxdp->read, dma_addr0_1);
+		_mm256_store_si256((__m256i *)&(rxdp + 2)->read, dma_addr2_3);
+	}
+
+#endif
+
+	rxq->rxrearm_start += IAVF_RXQ_REARM_THRESH;
+	if (rxq->rxrearm_start >= rxq->nb_rx_desc)
+		rxq->rxrearm_start = 0;
+
+	rxq->rxrearm_nb -= IAVF_RXQ_REARM_THRESH;
+
+	rx_id = (uint16_t)((rxq->rxrearm_start == 0) ?
+			     (rxq->nb_rx_desc - 1) : (rxq->rxrearm_start - 1));
+
+	/* Update the tail pointer on the NIC */
+	IAVF_PCI_REG_WRITE(rxq->qrx_tail, rx_id);
+}
+
+#define PKTLEN_SHIFT     10
+
+static inline uint16_t
+_iavf_recv_raw_pkts_vec_avx2(struct iavf_rx_queue *rxq,
+			     struct rte_mbuf **rx_pkts,
+			     uint16_t nb_pkts, uint8_t *split_packet)
+{
+#define IAVF_DESCS_PER_LOOP_AVX 8
+
+	/* const uint32_t *ptype_tbl = rxq->vsi->adapter->ptype_tbl; */
+	static const uint32_t type_table[UINT8_MAX + 1] __rte_cache_aligned = {
+		/* [0] reserved */
+		[1] = RTE_PTYPE_L2_ETHER,
+		/* [2] - [21] reserved */
+		[22] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
+			RTE_PTYPE_L4_FRAG,
+		[23] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
+			RTE_PTYPE_L4_NONFRAG,
+		[24] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
+			RTE_PTYPE_L4_UDP,
+		/* [25] reserved */
+		[26] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
+			RTE_PTYPE_L4_TCP,
+		[27] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
+			RTE_PTYPE_L4_SCTP,
+		[28] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
+			RTE_PTYPE_L4_ICMP,
+		/* All others reserved */
+	};
+	const __m256i mbuf_init = _mm256_set_epi64x(0, 0,
+			0, rxq->mbuf_initializer);
+	/* struct iavf_rx_entry *sw_ring = &rxq->sw_ring[rxq->rx_tail]; */
+	struct rte_mbuf **sw_ring = &rxq->sw_ring[rxq->rx_tail];
+	volatile union iavf_rx_desc *rxdp = rxq->rx_ring + rxq->rx_tail;
+	const int avx_aligned = ((rxq->rx_tail & 1) == 0);
+
+	rte_prefetch0(rxdp);
+
+	/* nb_pkts has to be floor-aligned to IAVF_DESCS_PER_LOOP_AVX */
+	nb_pkts = RTE_ALIGN_FLOOR(nb_pkts, IAVF_DESCS_PER_LOOP_AVX);
+
+	/* See if we need to rearm the RX queue - gives the prefetch a bit
+	 * of time to act
+	 */
+	if (rxq->rxrearm_nb > IAVF_RXQ_REARM_THRESH)
+		iavf_rxq_rearm(rxq);
+
+	/* Before we start moving massive data around, check to see if
+	 * there is actually a packet available
+	 */
+	if (!(rxdp->wb.qword1.status_error_len &
+			rte_cpu_to_le_32(1 << IAVF_RX_DESC_STATUS_DD_SHIFT)))
+		return 0;
+
+	/* constants used in processing loop */
+	const __m256i crc_adjust =
+		_mm256_set_epi16
+			(/* first descriptor */
+			 0, 0, 0,       /* ignore non-length fields */
+			 -rxq->crc_len, /* sub crc on data_len */
+			 0,             /* ignore high-16bits of pkt_len */
+			 -rxq->crc_len, /* sub crc on pkt_len */
+			 0, 0,          /* ignore pkt_type field */
+			 /* second descriptor */
+			 0, 0, 0,       /* ignore non-length fields */
+			 -rxq->crc_len, /* sub crc on data_len */
+			 0,             /* ignore high-16bits of pkt_len */
+			 -rxq->crc_len, /* sub crc on pkt_len */
+			 0, 0           /* ignore pkt_type field */
+			);
+
+	/* 8 packets DD mask, LSB in each 32-bit value */
+	const __m256i dd_check = _mm256_set1_epi32(1);
+
+	/* 8 packets EOP mask, second-LSB in each 32-bit value */
+	const __m256i eop_check = _mm256_slli_epi32(dd_check,
+			IAVF_RX_DESC_STATUS_EOF_SHIFT);
+
+	/* mask to shuffle from desc. to mbuf (2 descriptors)*/
+	const __m256i shuf_msk =
+		_mm256_set_epi8
+			(/* first descriptor */
+			 7, 6, 5, 4,  /* octet 4~7, 32bits rss */
+			 3, 2,        /* octet 2~3, low 16 bits vlan_macip */
+			 15, 14,      /* octet 15~14, 16 bits data_len */
+			 0xFF, 0xFF,  /* skip high 16 bits pkt_len, zero out */
+			 15, 14,      /* octet 15~14, low 16 bits pkt_len */
+			 0xFF, 0xFF,  /* pkt_type set as unknown */
+			 0xFF, 0xFF,  /*pkt_type set as unknown */
+			 /* second descriptor */
+			 7, 6, 5, 4,  /* octet 4~7, 32bits rss */
+			 3, 2,        /* octet 2~3, low 16 bits vlan_macip */
+			 15, 14,      /* octet 15~14, 16 bits data_len */
+			 0xFF, 0xFF,  /* skip high 16 bits pkt_len, zero out */
+			 15, 14,      /* octet 15~14, low 16 bits pkt_len */
+			 0xFF, 0xFF,  /* pkt_type set as unknown */
+			 0xFF, 0xFF   /*pkt_type set as unknown */
+			);
+	/**
+	 * compile-time check the above crc and shuffle layout is correct.
+	 * NOTE: the first field (lowest address) is given last in set_epi
+	 * calls above.
+	 */
+	RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, pkt_len) !=
+			offsetof(struct rte_mbuf, rx_descriptor_fields1) + 4);
+	RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, data_len) !=
+			offsetof(struct rte_mbuf, rx_descriptor_fields1) + 8);
+	RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, vlan_tci) !=
+			offsetof(struct rte_mbuf, rx_descriptor_fields1) + 10);
+	RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, hash) !=
+			offsetof(struct rte_mbuf, rx_descriptor_fields1) + 12);
+
+	/* Status/Error flag masks */
+	/**
+	 * mask everything except RSS, flow director and VLAN flags
+	 * bit2 is for VLAN tag, bit11 for flow director indication
+	 * bit13:12 for RSS indication. Bits 3-5 of error
+	 * field (bits 22-24) are for IP/L4 checksum errors
+	 */
+	const __m256i flags_mask =
+		 _mm256_set1_epi32((1 << 2) | (1 << 11) |
+				   (3 << 12) | (7 << 22));
+	/**
+	 * data to be shuffled by result of flag mask. If VLAN bit is set,
+	 * (bit 2), then position 4 in this array will be used in the
+	 * destination
+	 */
+	const __m256i vlan_flags_shuf =
+		_mm256_set_epi32(0, 0, PKT_RX_VLAN | PKT_RX_VLAN_STRIPPED, 0,
+				 0, 0, PKT_RX_VLAN | PKT_RX_VLAN_STRIPPED, 0);
+	/**
+	 * data to be shuffled by result of flag mask, shifted down 11.
+	 * If RSS/FDIR bits are set, shuffle moves appropriate flags in
+	 * place.
+	 */
+	const __m256i rss_flags_shuf =
+		_mm256_set_epi8(0, 0, 0, 0, 0, 0, 0, 0,
+				PKT_RX_RSS_HASH | PKT_RX_FDIR, PKT_RX_RSS_HASH,
+				0, 0, 0, 0, PKT_RX_FDIR, 0,/* end up 128-bits */
+				0, 0, 0, 0, 0, 0, 0, 0,
+				PKT_RX_RSS_HASH | PKT_RX_FDIR, PKT_RX_RSS_HASH,
+				0, 0, 0, 0, PKT_RX_FDIR, 0);
+
+	/**
+	 * data to be shuffled by the result of the flags mask shifted by 22
+	 * bits.  This gives use the l3_l4 flags.
+	 */
+	const __m256i l3_l4_flags_shuf = _mm256_set_epi8(0, 0, 0, 0, 0, 0, 0, 0,
+			/* shift right 1 bit to make sure it not exceed 255 */
+			(PKT_RX_EIP_CKSUM_BAD | PKT_RX_L4_CKSUM_BAD |
+			 PKT_RX_IP_CKSUM_BAD) >> 1,
+			(PKT_RX_IP_CKSUM_GOOD | PKT_RX_EIP_CKSUM_BAD |
+			 PKT_RX_L4_CKSUM_BAD) >> 1,
+			(PKT_RX_EIP_CKSUM_BAD | PKT_RX_IP_CKSUM_BAD) >> 1,
+			(PKT_RX_IP_CKSUM_GOOD | PKT_RX_EIP_CKSUM_BAD) >> 1,
+			(PKT_RX_L4_CKSUM_BAD | PKT_RX_IP_CKSUM_BAD) >> 1,
+			(PKT_RX_IP_CKSUM_GOOD | PKT_RX_L4_CKSUM_BAD) >> 1,
+			PKT_RX_IP_CKSUM_BAD >> 1,
+			(PKT_RX_IP_CKSUM_GOOD | PKT_RX_L4_CKSUM_GOOD) >> 1,
+			/* second 128-bits */
+			0, 0, 0, 0, 0, 0, 0, 0,
+			(PKT_RX_EIP_CKSUM_BAD | PKT_RX_L4_CKSUM_BAD |
+			 PKT_RX_IP_CKSUM_BAD) >> 1,
+			(PKT_RX_IP_CKSUM_GOOD | PKT_RX_EIP_CKSUM_BAD |
+			 PKT_RX_L4_CKSUM_BAD) >> 1,
+			(PKT_RX_EIP_CKSUM_BAD | PKT_RX_IP_CKSUM_BAD) >> 1,
+			(PKT_RX_IP_CKSUM_GOOD | PKT_RX_EIP_CKSUM_BAD) >> 1,
+			(PKT_RX_L4_CKSUM_BAD | PKT_RX_IP_CKSUM_BAD) >> 1,
+			(PKT_RX_IP_CKSUM_GOOD | PKT_RX_L4_CKSUM_BAD) >> 1,
+			PKT_RX_IP_CKSUM_BAD >> 1,
+			(PKT_RX_IP_CKSUM_GOOD | PKT_RX_L4_CKSUM_GOOD) >> 1);
+
+	const __m256i cksum_mask =
+		 _mm256_set1_epi32(PKT_RX_IP_CKSUM_GOOD | PKT_RX_IP_CKSUM_BAD |
+				   PKT_RX_L4_CKSUM_GOOD | PKT_RX_L4_CKSUM_BAD |
+				   PKT_RX_EIP_CKSUM_BAD);
+
+	RTE_SET_USED(avx_aligned); /* for 32B descriptors we don't use this */
+
+	uint16_t i, received;
+
+	for (i = 0, received = 0; i < nb_pkts;
+	     i += IAVF_DESCS_PER_LOOP_AVX,
+	     rxdp += IAVF_DESCS_PER_LOOP_AVX) {
+		/* step 1, copy over 8 mbuf pointers to rx_pkts array */
+		_mm256_storeu_si256((void *)&rx_pkts[i],
+				    _mm256_loadu_si256((void *)&sw_ring[i]));
+#ifdef RTE_ARCH_X86_64
+		_mm256_storeu_si256
+			((void *)&rx_pkts[i + 4],
+			 _mm256_loadu_si256((void *)&sw_ring[i + 4]));
+#endif
+
+		__m256i raw_desc0_1, raw_desc2_3, raw_desc4_5, raw_desc6_7;
+#ifdef RTE_LIBRTE_IAVF_16BYTE_RX_DESC
+		/* for AVX we need alignment otherwise loads are not atomic */
+		if (avx_aligned) {
+			/* load in descriptors, 2 at a time, in reverse order */
+			raw_desc6_7 = _mm256_load_si256((void *)(rxdp + 6));
+			rte_compiler_barrier();
+			raw_desc4_5 = _mm256_load_si256((void *)(rxdp + 4));
+			rte_compiler_barrier();
+			raw_desc2_3 = _mm256_load_si256((void *)(rxdp + 2));
+			rte_compiler_barrier();
+			raw_desc0_1 = _mm256_load_si256((void *)(rxdp + 0));
+		} else
+#endif
+		{
+			const __m128i raw_desc7 =
+				_mm_load_si128((void *)(rxdp + 7));
+			rte_compiler_barrier();
+			const __m128i raw_desc6 =
+				_mm_load_si128((void *)(rxdp + 6));
+			rte_compiler_barrier();
+			const __m128i raw_desc5 =
+				_mm_load_si128((void *)(rxdp + 5));
+			rte_compiler_barrier();
+			const __m128i raw_desc4 =
+				_mm_load_si128((void *)(rxdp + 4));
+			rte_compiler_barrier();
+			const __m128i raw_desc3 =
+				_mm_load_si128((void *)(rxdp + 3));
+			rte_compiler_barrier();
+			const __m128i raw_desc2 =
+				_mm_load_si128((void *)(rxdp + 2));
+			rte_compiler_barrier();
+			const __m128i raw_desc1 =
+				_mm_load_si128((void *)(rxdp + 1));
+			rte_compiler_barrier();
+			const __m128i raw_desc0 =
+				_mm_load_si128((void *)(rxdp + 0));
+
+			raw_desc6_7 =
+				_mm256_inserti128_si256
+					(_mm256_castsi128_si256(raw_desc6),
+					 raw_desc7, 1);
+			raw_desc4_5 =
+				_mm256_inserti128_si256
+					(_mm256_castsi128_si256(raw_desc4),
+					 raw_desc5, 1);
+			raw_desc2_3 =
+				_mm256_inserti128_si256
+					(_mm256_castsi128_si256(raw_desc2),
+					 raw_desc3, 1);
+			raw_desc0_1 =
+				_mm256_inserti128_si256
+					(_mm256_castsi128_si256(raw_desc0),
+					 raw_desc1, 1);
+		}
+
+		if (split_packet) {
+			int j;
+
+			for (j = 0; j < IAVF_DESCS_PER_LOOP_AVX; j++)
+				rte_mbuf_prefetch_part2(rx_pkts[i + j]);
+		}
+
+		/**
+		 * convert descriptors 4-7 into mbufs, adjusting length and
+		 * re-arranging fields. Then write into the mbuf
+		 */
+		const __m256i len6_7 = _mm256_slli_epi32(raw_desc6_7,
+							 PKTLEN_SHIFT);
+		const __m256i len4_5 = _mm256_slli_epi32(raw_desc4_5,
+							 PKTLEN_SHIFT);
+		const __m256i desc6_7 = _mm256_blend_epi16(raw_desc6_7,
+							   len6_7, 0x80);
+		const __m256i desc4_5 = _mm256_blend_epi16(raw_desc4_5,
+							   len4_5, 0x80);
+		__m256i mb6_7 = _mm256_shuffle_epi8(desc6_7, shuf_msk);
+		__m256i mb4_5 = _mm256_shuffle_epi8(desc4_5, shuf_msk);
+
+		mb6_7 = _mm256_add_epi16(mb6_7, crc_adjust);
+		mb4_5 = _mm256_add_epi16(mb4_5, crc_adjust);
+		/**
+		 * to get packet types, shift 64-bit values down 30 bits
+		 * and so ptype is in lower 8-bits in each
+		 */
+		const __m256i ptypes6_7 = _mm256_srli_epi64(desc6_7, 30);
+		const __m256i ptypes4_5 = _mm256_srli_epi64(desc4_5, 30);
+		const uint8_t ptype7 = _mm256_extract_epi8(ptypes6_7, 24);
+		const uint8_t ptype6 = _mm256_extract_epi8(ptypes6_7, 8);
+		const uint8_t ptype5 = _mm256_extract_epi8(ptypes4_5, 24);
+		const uint8_t ptype4 = _mm256_extract_epi8(ptypes4_5, 8);
+
+		mb6_7 = _mm256_insert_epi32(mb6_7, type_table[ptype7], 4);
+		mb6_7 = _mm256_insert_epi32(mb6_7, type_table[ptype6], 0);
+		mb4_5 = _mm256_insert_epi32(mb4_5, type_table[ptype5], 4);
+		mb4_5 = _mm256_insert_epi32(mb4_5, type_table[ptype4], 0);
+		/* merge the status bits into one register */
+		const __m256i status4_7 = _mm256_unpackhi_epi32(desc6_7,
+				desc4_5);
+
+		/**
+		 * convert descriptors 0-3 into mbufs, adjusting length and
+		 * re-arranging fields. Then write into the mbuf
+		 */
+		const __m256i len2_3 = _mm256_slli_epi32(raw_desc2_3,
+							 PKTLEN_SHIFT);
+		const __m256i len0_1 = _mm256_slli_epi32(raw_desc0_1,
+							 PKTLEN_SHIFT);
+		const __m256i desc2_3 = _mm256_blend_epi16(raw_desc2_3,
+							   len2_3, 0x80);
+		const __m256i desc0_1 = _mm256_blend_epi16(raw_desc0_1,
+							   len0_1, 0x80);
+		__m256i mb2_3 = _mm256_shuffle_epi8(desc2_3, shuf_msk);
+		__m256i mb0_1 = _mm256_shuffle_epi8(desc0_1, shuf_msk);
+
+		mb2_3 = _mm256_add_epi16(mb2_3, crc_adjust);
+		mb0_1 = _mm256_add_epi16(mb0_1, crc_adjust);
+		/* get the packet types */
+		const __m256i ptypes2_3 = _mm256_srli_epi64(desc2_3, 30);
+		const __m256i ptypes0_1 = _mm256_srli_epi64(desc0_1, 30);
+		const uint8_t ptype3 = _mm256_extract_epi8(ptypes2_3, 24);
+		const uint8_t ptype2 = _mm256_extract_epi8(ptypes2_3, 8);
+		const uint8_t ptype1 = _mm256_extract_epi8(ptypes0_1, 24);
+		const uint8_t ptype0 = _mm256_extract_epi8(ptypes0_1, 8);
+
+		mb2_3 = _mm256_insert_epi32(mb2_3, type_table[ptype3], 4);
+		mb2_3 = _mm256_insert_epi32(mb2_3, type_table[ptype2], 0);
+		mb0_1 = _mm256_insert_epi32(mb0_1, type_table[ptype1], 4);
+		mb0_1 = _mm256_insert_epi32(mb0_1, type_table[ptype0], 0);
+		/* merge the status bits into one register */
+		const __m256i status0_3 = _mm256_unpackhi_epi32(desc2_3,
+								desc0_1);
+
+		/**
+		 * take the two sets of status bits and merge to one
+		 * After merge, the packets status flags are in the
+		 * order (hi->lo): [1, 3, 5, 7, 0, 2, 4, 6]
+		 */
+		__m256i status0_7 = _mm256_unpacklo_epi64(status4_7,
+							  status0_3);
+
+		/* now do flag manipulation */
+
+		/* get only flag/error bits we want */
+		const __m256i flag_bits =
+			_mm256_and_si256(status0_7, flags_mask);
+		/* set vlan and rss flags */
+		const __m256i vlan_flags =
+			_mm256_shuffle_epi8(vlan_flags_shuf, flag_bits);
+		const __m256i rss_flags =
+			_mm256_shuffle_epi8(rss_flags_shuf,
+					    _mm256_srli_epi32(flag_bits, 11));
+		/**
+		 * l3_l4_error flags, shuffle, then shift to correct adjustment
+		 * of flags in flags_shuf, and finally mask out extra bits
+		 */
+		__m256i l3_l4_flags = _mm256_shuffle_epi8(l3_l4_flags_shuf,
+				_mm256_srli_epi32(flag_bits, 22));
+		l3_l4_flags = _mm256_slli_epi32(l3_l4_flags, 1);
+		l3_l4_flags = _mm256_and_si256(l3_l4_flags, cksum_mask);
+
+		/* merge flags */
+		const __m256i mbuf_flags = _mm256_or_si256(l3_l4_flags,
+				_mm256_or_si256(rss_flags, vlan_flags));
+		/**
+		 * At this point, we have the 8 sets of flags in the low 16-bits
+		 * of each 32-bit value in vlan0.
+		 * We want to extract these, and merge them with the mbuf init
+		 * data so we can do a single write to the mbuf to set the flags
+		 * and all the other initialization fields. Extracting the
+		 * appropriate flags means that we have to do a shift and blend
+		 * for each mbuf before we do the write. However, we can also
+		 * add in the previously computed rx_descriptor fields to
+		 * make a single 256-bit write per mbuf
+		 */
+		/* check the structure matches expectations */
+		RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, ol_flags) !=
+				 offsetof(struct rte_mbuf, rearm_data) + 8);
+		RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, rearm_data) !=
+				 RTE_ALIGN(offsetof(struct rte_mbuf,
+						    rearm_data),
+					   16));
+		/* build up data and do writes */
+		__m256i rearm0, rearm1, rearm2, rearm3, rearm4, rearm5,
+			rearm6, rearm7;
+		rearm6 = _mm256_blend_epi32(mbuf_init,
+					    _mm256_slli_si256(mbuf_flags, 8),
+					    0x04);
+		rearm4 = _mm256_blend_epi32(mbuf_init,
+					    _mm256_slli_si256(mbuf_flags, 4),
+					    0x04);
+		rearm2 = _mm256_blend_epi32(mbuf_init, mbuf_flags, 0x04);
+		rearm0 = _mm256_blend_epi32(mbuf_init,
+					    _mm256_srli_si256(mbuf_flags, 4),
+					    0x04);
+		/* permute to add in the rx_descriptor e.g. rss fields */
+		rearm6 = _mm256_permute2f128_si256(rearm6, mb6_7, 0x20);
+		rearm4 = _mm256_permute2f128_si256(rearm4, mb4_5, 0x20);
+		rearm2 = _mm256_permute2f128_si256(rearm2, mb2_3, 0x20);
+		rearm0 = _mm256_permute2f128_si256(rearm0, mb0_1, 0x20);
+		/* write to mbuf */
+		_mm256_storeu_si256((__m256i *)&rx_pkts[i + 6]->rearm_data,
+				    rearm6);
+		_mm256_storeu_si256((__m256i *)&rx_pkts[i + 4]->rearm_data,
+				    rearm4);
+		_mm256_storeu_si256((__m256i *)&rx_pkts[i + 2]->rearm_data,
+				    rearm2);
+		_mm256_storeu_si256((__m256i *)&rx_pkts[i + 0]->rearm_data,
+				    rearm0);
+
+		/* repeat for the odd mbufs */
+		const __m256i odd_flags =
+			_mm256_castsi128_si256
+				(_mm256_extracti128_si256(mbuf_flags, 1));
+		rearm7 = _mm256_blend_epi32(mbuf_init,
+					    _mm256_slli_si256(odd_flags, 8),
+					    0x04);
+		rearm5 = _mm256_blend_epi32(mbuf_init,
+					    _mm256_slli_si256(odd_flags, 4),
+					    0x04);
+		rearm3 = _mm256_blend_epi32(mbuf_init, odd_flags, 0x04);
+		rearm1 = _mm256_blend_epi32(mbuf_init,
+					    _mm256_srli_si256(odd_flags, 4),
+					    0x04);
+		/* since odd mbufs are already in hi 128-bits use blend */
+		rearm7 = _mm256_blend_epi32(rearm7, mb6_7, 0xF0);
+		rearm5 = _mm256_blend_epi32(rearm5, mb4_5, 0xF0);
+		rearm3 = _mm256_blend_epi32(rearm3, mb2_3, 0xF0);
+		rearm1 = _mm256_blend_epi32(rearm1, mb0_1, 0xF0);
+		/* again write to mbufs */
+		_mm256_storeu_si256((__m256i *)&rx_pkts[i + 7]->rearm_data,
+				    rearm7);
+		_mm256_storeu_si256((__m256i *)&rx_pkts[i + 5]->rearm_data,
+				    rearm5);
+		_mm256_storeu_si256((__m256i *)&rx_pkts[i + 3]->rearm_data,
+				    rearm3);
+		_mm256_storeu_si256((__m256i *)&rx_pkts[i + 1]->rearm_data,
+				    rearm1);
+
+		/* extract and record EOP bit */
+		if (split_packet) {
+			const __m128i eop_mask =
+				_mm_set1_epi16(1 << IAVF_RX_DESC_STATUS_EOF_SHIFT);
+			const __m256i eop_bits256 = _mm256_and_si256(status0_7,
+								     eop_check);
+			/* pack status bits into a single 128-bit register */
+			const __m128i eop_bits =
+				_mm_packus_epi32
+					(_mm256_castsi256_si128(eop_bits256),
+					 _mm256_extractf128_si256(eop_bits256,
+								  1));
+			/**
+			 * flip bits, and mask out the EOP bit, which is now
+			 * a split-packet bit i.e. !EOP, rather than EOP one.
+			 */
+			__m128i split_bits = _mm_andnot_si128(eop_bits,
+					eop_mask);
+			/**
+			 * eop bits are out of order, so we need to shuffle them
+			 * back into order again. In doing so, only use low 8
+			 * bits, which acts like another pack instruction
+			 * The original order is (hi->lo): 1,3,5,7,0,2,4,6
+			 * [Since we use epi8, the 16-bit positions are
+			 * multiplied by 2 in the eop_shuffle value.]
+			 */
+			__m128i eop_shuffle =
+				_mm_set_epi8(/* zero hi 64b */
+					     0xFF, 0xFF, 0xFF, 0xFF,
+					     0xFF, 0xFF, 0xFF, 0xFF,
+					     /* move values to lo 64b */
+					     8, 0, 10, 2,
+					     12, 4, 14, 6);
+			split_bits = _mm_shuffle_epi8(split_bits, eop_shuffle);
+			*(uint64_t *)split_packet =
+				_mm_cvtsi128_si64(split_bits);
+			split_packet += IAVF_DESCS_PER_LOOP_AVX;
+		}
+
+		/* perform dd_check */
+		status0_7 = _mm256_and_si256(status0_7, dd_check);
+		status0_7 = _mm256_packs_epi32(status0_7,
+					       _mm256_setzero_si256());
+
+		uint64_t burst = __builtin_popcountll
+					(_mm_cvtsi128_si64
+						(_mm256_extracti128_si256
+							(status0_7, 1)));
+		burst += __builtin_popcountll
+				(_mm_cvtsi128_si64
+					(_mm256_castsi256_si128(status0_7)));
+		received += burst;
+		if (burst != IAVF_DESCS_PER_LOOP_AVX)
+			break;
+	}
+
+	/* update tail pointers */
+	rxq->rx_tail += received;
+	rxq->rx_tail &= (rxq->nb_rx_desc - 1);
+	if ((rxq->rx_tail & 1) == 1 && received > 1) { /* keep avx2 aligned */
+		rxq->rx_tail--;
+		received--;
+	}
+	rxq->rxrearm_nb += received;
+	return received;
+}
+
+/**
+ * Notice:
+ * - nb_pkts < IAVF_DESCS_PER_LOOP, just return no packet
+ */
+uint16_t
+iavf_recv_pkts_vec_avx2(void *rx_queue, struct rte_mbuf **rx_pkts,
+			uint16_t nb_pkts)
+{
+	return _iavf_recv_raw_pkts_vec_avx2(rx_queue, rx_pkts, nb_pkts, NULL);
+}
+
+/**
+ * vPMD receive routine that reassembles single burst of 32 scattered packets
+ * Notice:
+ * - nb_pkts < IAVF_DESCS_PER_LOOP, just return no packet
+ */
+static uint16_t
+iavf_recv_scattered_burst_vec_avx2(void *rx_queue, struct rte_mbuf **rx_pkts,
+				   uint16_t nb_pkts)
+{
+	struct iavf_rx_queue *rxq = rx_queue;
+	uint8_t split_flags[IAVF_VPMD_RX_MAX_BURST] = {0};
+
+	/* get some new buffers */
+	uint16_t nb_bufs = _iavf_recv_raw_pkts_vec_avx2(rxq, rx_pkts, nb_pkts,
+						       split_flags);
+	if (nb_bufs == 0)
+		return 0;
+
+	/* happy day case, full burst + no packets to be joined */
+	const uint64_t *split_fl64 = (uint64_t *)split_flags;
+
+	if (!rxq->pkt_first_seg &&
+	    split_fl64[0] == 0 && split_fl64[1] == 0 &&
+	    split_fl64[2] == 0 && split_fl64[3] == 0)
+		return nb_bufs;
+
+	/* reassemble any packets that need reassembly*/
+	unsigned int i = 0;
+
+	if (!rxq->pkt_first_seg) {
+		/* find the first split flag, and only reassemble then*/
+		while (i < nb_bufs && !split_flags[i])
+			i++;
+		if (i == nb_bufs)
+			return nb_bufs;
+		rxq->pkt_first_seg = rx_pkts[i];
+	}
+	return i + reassemble_packets(rxq, &rx_pkts[i], nb_bufs - i,
+					     &split_flags[i]);
+}
+
+/**
+ * vPMD receive routine that reassembles scattered packets.
+ * Main receive routine that can handle arbitrary burst sizes
+ * Notice:
+ * - nb_pkts < IAVF_DESCS_PER_LOOP, just return no packet
+ */
+uint16_t
+iavf_recv_scattered_pkts_vec_avx2(void *rx_queue, struct rte_mbuf **rx_pkts,
+				  uint16_t nb_pkts)
+{
+	uint16_t retval = 0;
+
+	while (nb_pkts > IAVF_VPMD_RX_MAX_BURST) {
+		uint16_t burst = iavf_recv_scattered_burst_vec_avx2(rx_queue,
+				rx_pkts + retval, IAVF_VPMD_RX_MAX_BURST);
+		retval += burst;
+		nb_pkts -= burst;
+		if (burst < IAVF_VPMD_RX_MAX_BURST)
+			return retval;
+	}
+	return retval + iavf_recv_scattered_burst_vec_avx2(rx_queue,
+				rx_pkts + retval, nb_pkts);
+}
+
+static inline void
+iavf_vtx1(volatile struct iavf_tx_desc *txdp,
+	  struct rte_mbuf *pkt, uint64_t flags)
+{
+	uint64_t high_qw =
+		(IAVF_TX_DESC_DTYPE_DATA |
+		 ((uint64_t)flags  << IAVF_TXD_QW1_CMD_SHIFT) |
+		 ((uint64_t)pkt->data_len << IAVF_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
+iavf_vtx(volatile struct iavf_tx_desc *txdp,
+	 struct rte_mbuf **pkt, uint16_t nb_pkts,  uint64_t flags)
+{
+	const uint64_t hi_qw_tmpl = (IAVF_TX_DESC_DTYPE_DATA |
+			((uint64_t)flags  << IAVF_TXD_QW1_CMD_SHIFT));
+
+	/* if unaligned on 32-bit boundary, do one to align */
+	if (((uintptr_t)txdp & 0x1F) != 0 && nb_pkts != 0) {
+		iavf_vtx1(txdp, *pkt, flags);
+		nb_pkts--, txdp++, pkt++;
+	}
+
+	/* do two at a time while possible, in bursts */
+	for (; nb_pkts > 3; txdp += 4, pkt += 4, nb_pkts -= 4) {
+		uint64_t hi_qw3 =
+			hi_qw_tmpl |
+			((uint64_t)pkt[3]->data_len <<
+			 IAVF_TXD_QW1_TX_BUF_SZ_SHIFT);
+		uint64_t hi_qw2 =
+			hi_qw_tmpl |
+			((uint64_t)pkt[2]->data_len <<
+			 IAVF_TXD_QW1_TX_BUF_SZ_SHIFT);
+		uint64_t hi_qw1 =
+			hi_qw_tmpl |
+			((uint64_t)pkt[1]->data_len <<
+			 IAVF_TXD_QW1_TX_BUF_SZ_SHIFT);
+		uint64_t hi_qw0 =
+			hi_qw_tmpl |
+			((uint64_t)pkt[0]->data_len <<
+			 IAVF_TXD_QW1_TX_BUF_SZ_SHIFT);
+
+		__m256i desc2_3 =
+			_mm256_set_epi64x
+				(hi_qw3,
+				 pkt[3]->buf_physaddr + pkt[3]->data_off,
+				 hi_qw2,
+				 pkt[2]->buf_physaddr + pkt[2]->data_off);
+		__m256i desc0_1 =
+			_mm256_set_epi64x
+				(hi_qw1,
+				 pkt[1]->buf_physaddr + pkt[1]->data_off,
+				 hi_qw0,
+				 pkt[0]->buf_physaddr + pkt[0]->data_off);
+		_mm256_store_si256((void *)(txdp + 2), desc2_3);
+		_mm256_store_si256((void *)txdp, desc0_1);
+	}
+
+	/* do any last ones */
+	while (nb_pkts) {
+		iavf_vtx1(txdp, *pkt, flags);
+		txdp++, pkt++, nb_pkts--;
+	}
+}
+
+static inline uint16_t
+iavf_xmit_fixed_burst_vec_avx2(void *tx_queue, struct rte_mbuf **tx_pkts,
+			       uint16_t nb_pkts)
+{
+	struct iavf_tx_queue *txq = (struct iavf_tx_queue *)tx_queue;
+	volatile struct iavf_tx_desc *txdp;
+	struct iavf_tx_entry *txep;
+	uint16_t n, nb_commit, tx_id;
+	uint64_t flags = IAVF_TX_DESC_CMD_EOP;
+	uint64_t rs = IAVF_TX_DESC_CMD_RS | IAVF_TX_DESC_CMD_EOP;
+
+	/* cross rx_thresh boundary is not allowed */
+	nb_pkts = RTE_MIN(nb_pkts, txq->rs_thresh);
+
+	if (txq->nb_free < txq->free_thresh)
+		iavf_tx_free_bufs(txq);
+
+	nb_commit = nb_pkts = (uint16_t)RTE_MIN(txq->nb_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_free = (uint16_t)(txq->nb_free - nb_pkts);
+
+	n = (uint16_t)(txq->nb_tx_desc - tx_id);
+	if (nb_commit >= n) {
+		tx_backlog_entry(txep, tx_pkts, n);
+
+		iavf_vtx(txdp, tx_pkts, n - 1, flags);
+		tx_pkts += (n - 1);
+		txdp += (n - 1);
+
+		iavf_vtx1(txdp, *tx_pkts++, rs);
+
+		nb_commit = (uint16_t)(nb_commit - n);
+
+		tx_id = 0;
+		txq->next_rs = (uint16_t)(txq->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);
+
+	iavf_vtx(txdp, tx_pkts, nb_commit, flags);
+
+	tx_id = (uint16_t)(tx_id + nb_commit);
+	if (tx_id > txq->next_rs) {
+		txq->tx_ring[txq->next_rs].cmd_type_offset_bsz |=
+			rte_cpu_to_le_64(((uint64_t)IAVF_TX_DESC_CMD_RS) <<
+					 IAVF_TXD_QW1_CMD_SHIFT);
+		txq->next_rs =
+			(uint16_t)(txq->next_rs + txq->rs_thresh);
+	}
+
+	txq->tx_tail = tx_id;
+
+	IAVF_PCI_REG_WRITE(txq->qtx_tail, txq->tx_tail);
+
+	return nb_pkts;
+}
+
+uint16_t
+iavf_xmit_pkts_vec_avx2(void *tx_queue, struct rte_mbuf **tx_pkts,
+			uint16_t nb_pkts)
+{
+	uint16_t nb_tx = 0;
+	struct iavf_tx_queue *txq = (struct iavf_tx_queue *)tx_queue;
+
+	while (nb_pkts) {
+		uint16_t ret, num;
+
+		num = (uint16_t)RTE_MIN(nb_pkts, txq->rs_thresh);
+		ret = iavf_xmit_fixed_burst_vec_avx2(tx_queue, &tx_pkts[nb_tx],
+						     num);
+		nb_tx += ret;
+		nb_pkts -= ret;
+		if (ret < num)
+			break;
+	}
+
+	return nb_tx;
+}
diff --git a/drivers/net/iavf/iavf_rxtx_vec_common.h b/drivers/net/iavf/iavf_rxtx_vec_common.h
index db509d71f..fff0555d2 100644
--- a/drivers/net/iavf/iavf_rxtx_vec_common.h
+++ b/drivers/net/iavf/iavf_rxtx_vec_common.h
@@ -207,4 +207,76 @@ iavf_rxq_vec_setup_default(struct iavf_rx_queue *rxq)
 	rxq->mbuf_initializer = *(uint64_t *)p;
 	return 0;
 }
+
+static inline int
+iavf_rx_vec_queue_default(struct iavf_rx_queue *rxq)
+{
+	if (!rxq)
+		return -1;
+
+	if (!rte_is_power_of_2(rxq->nb_rx_desc))
+		return -1;
+
+	if (rxq->rx_free_thresh < IAVF_VPMD_RX_MAX_BURST)
+		return -1;
+
+	if (rxq->nb_rx_desc % rxq->rx_free_thresh)
+		return -1;
+
+	return 0;
+}
+
+#define IAVF_NO_VECTOR_FLAGS (				 \
+		DEV_TX_OFFLOAD_MULTI_SEGS |		 \
+		DEV_TX_OFFLOAD_VLAN_INSERT |		 \
+		DEV_TX_OFFLOAD_SCTP_CKSUM |		 \
+		DEV_TX_OFFLOAD_UDP_CKSUM |		 \
+		DEV_TX_OFFLOAD_TCP_CKSUM)
+
+static inline int
+iavf_tx_vec_queue_default(struct iavf_tx_queue *txq)
+{
+	if (!txq)
+		return -1;
+
+	if (txq->offloads & IAVF_NO_VECTOR_FLAGS)
+		return -1;
+
+	if (txq->rs_thresh < IAVF_VPMD_TX_MAX_BURST ||
+	    txq->rs_thresh > IAVF_VPMD_TX_MAX_FREE_BUF)
+		return -1;
+
+	return 0;
+}
+
+static inline int
+iavf_rx_vec_dev_check_default(struct rte_eth_dev *dev)
+{
+	int i;
+	struct iavf_rx_queue *rxq;
+
+	for (i = 0; i < dev->data->nb_rx_queues; i++) {
+		rxq = dev->data->rx_queues[i];
+		if (iavf_rx_vec_queue_default(rxq))
+			return -1;
+	}
+
+	return 0;
+}
+
+static inline int
+iavf_tx_vec_dev_check_default(struct rte_eth_dev *dev)
+{
+	int i;
+	struct iavf_tx_queue *txq;
+
+	for (i = 0; i < dev->data->nb_tx_queues; i++) {
+		txq = dev->data->tx_queues[i];
+		if (iavf_tx_vec_queue_default(txq))
+			return -1;
+	}
+
+	return 0;
+}
+
 #endif
diff --git a/drivers/net/iavf/iavf_rxtx_vec_sse.c b/drivers/net/iavf/iavf_rxtx_vec_sse.c
index cc71f23a5..2b16dc1b5 100644
--- a/drivers/net/iavf/iavf_rxtx_vec_sse.c
+++ b/drivers/net/iavf/iavf_rxtx_vec_sse.c
@@ -622,6 +622,27 @@ iavf_xmit_fixed_burst_vec(void *tx_queue, struct rte_mbuf **tx_pkts,
 	return nb_pkts;
 }
 
+uint16_t
+iavf_xmit_pkts_vec(void *tx_queue, struct rte_mbuf **tx_pkts,
+		   uint16_t nb_pkts)
+{
+	uint16_t nb_tx = 0;
+	struct iavf_tx_queue *txq = (struct iavf_tx_queue *)tx_queue;
+
+	while (nb_pkts) {
+		uint16_t ret, num;
+
+		num = (uint16_t)RTE_MIN(nb_pkts, txq->rs_thresh);
+		ret = iavf_xmit_fixed_burst_vec(tx_queue, &tx_pkts[nb_tx], num);
+		nb_tx += ret;
+		nb_pkts -= ret;
+		if (ret < num)
+			break;
+	}
+
+	return nb_tx;
+}
+
 static void __attribute__((cold))
 iavf_rx_queue_release_mbufs_sse(struct iavf_rx_queue *rxq)
 {
@@ -655,3 +676,15 @@ iavf_rxq_vec_setup(struct iavf_rx_queue *rxq)
 	rxq->ops = &sse_vec_rxq_ops;
 	return iavf_rxq_vec_setup_default(rxq);
 }
+
+int __attribute__((cold))
+iavf_rx_vec_dev_check(struct rte_eth_dev *dev)
+{
+	return iavf_rx_vec_dev_check_default(dev);
+}
+
+int __attribute__((cold))
+iavf_tx_vec_dev_check(struct rte_eth_dev *dev)
+{
+	return iavf_tx_vec_dev_check_default(dev);
+}
diff --git a/drivers/net/iavf/meson.build b/drivers/net/iavf/meson.build
index e5a2f5553..c2401af86 100644
--- a/drivers/net/iavf/meson.build
+++ b/drivers/net/iavf/meson.build
@@ -17,4 +17,21 @@ sources = files(
 if arch_subdir == 'x86'
 	dpdk_conf.set('RTE_LIBRTE_IAVF_INC_VECTOR', 1)
 	sources += files('iavf_rxtx_vec_sse.c')
+
+	# compile AVX2 version if either:
+	# a. we have AVX supported in minimum instruction set baseline
+	# b. it's not minimum instruction set, but supported by compiler
+	if dpdk_conf.has('RTE_MACHINE_CPUFLAG_AVX2')
+		cflags += ['-DCC_AVX2_SUPPORT']
+		sources += files('iavf_rxtx_vec_avx2.c')
+	elif cc.has_argument('-mavx2')
+		cflags += ['-DCC_AVX2_SUPPORT']
+		iavf_avx2_lib = static_library('iavf_avx2_lib',
+				'iavf_rxtx_vec_avx2.c',
+				dependencies: [static_rte_ethdev,
+					static_rte_kvargs, static_rte_hash],
+				include_directories: includes,
+				c_args: [cflags, '-mavx2'])
+		objs += iavf_avx2_lib.extract_objects('iavf_rxtx_vec_avx2.c')
+	endif
 endif
-- 
2.17.1


^ permalink raw reply	[flat|nested] 11+ messages in thread

* [dpdk-dev] [PATCH v3] net/iavf: enable AVX2 for iavf
  2019-09-04 10:06 [dpdk-dev] [PATCH] net/iavf: enable AVX2 for iavf Leyi Rong
  2019-09-04 13:29 ` Aaron Conole
  2019-09-06  3:28 ` [dpdk-dev] [PATCH v2] " Leyi Rong
@ 2019-09-19  8:44 ` Leyi Rong
  2019-09-19 12:55   ` Zhang, Qi Z
  2019-09-23  6:02 ` [dpdk-dev] [PATCH v4] " Leyi Rong
  2019-10-16  8:14 ` [dpdk-dev] [PATCH v5] " Leyi Rong
  4 siblings, 1 reply; 11+ messages in thread
From: Leyi Rong @ 2019-09-19  8:44 UTC (permalink / raw)
  To: wenzhuo.lu, qi.z.zhang, xiaolong.ye; +Cc: dev, Leyi Rong

This patch enables AVX data path for iavf PMD.

Signed-off-by: Leyi Rong <leyi.rong@intel.com>

---
v3:
- Adds release note.

v2:
- fix build error on aarch64.
---
 doc/guides/rel_notes/release_19_11.rst  |   3 +
 drivers/net/iavf/Makefile               |  21 +
 drivers/net/iavf/iavf_rxtx.c            |  99 +--
 drivers/net/iavf/iavf_rxtx.h            |  12 +
 drivers/net/iavf/iavf_rxtx_vec_avx2.c   | 867 ++++++++++++++++++++++++
 drivers/net/iavf/iavf_rxtx_vec_common.h |  72 ++
 drivers/net/iavf/iavf_rxtx_vec_sse.c    |  33 +
 drivers/net/iavf/meson.build            |  17 +
 8 files changed, 1078 insertions(+), 46 deletions(-)
 create mode 100644 drivers/net/iavf/iavf_rxtx_vec_avx2.c

diff --git a/doc/guides/rel_notes/release_19_11.rst b/doc/guides/rel_notes/release_19_11.rst
index 8490d897c..e4e29e219 100644
--- a/doc/guides/rel_notes/release_19_11.rst
+++ b/doc/guides/rel_notes/release_19_11.rst
@@ -56,6 +56,9 @@ New Features
      Also, make sure to start the actual text at the margin.
      =========================================================
 
+* **Updated iavf PMD.**
+
+  Enabled AVX2 data path for iavf PMD.
 
 Removed Items
 -------------
diff --git a/drivers/net/iavf/Makefile b/drivers/net/iavf/Makefile
index cd74e14ab..cf0ed99f0 100644
--- a/drivers/net/iavf/Makefile
+++ b/drivers/net/iavf/Makefile
@@ -48,4 +48,25 @@ ifeq ($(CONFIG_RTE_ARCH_X86), y)
 SRCS-$(CONFIG_RTE_LIBRTE_IAVF_INC_VECTOR) += iavf_rxtx_vec_sse.c
 endif
 
+ifeq ($(CONFIG_RTE_LIBRTE_IAVF_INC_VECTOR), y)
+	ifeq ($(findstring RTE_MACHINE_CPUFLAG_AVX2,$(CFLAGS)),RTE_MACHINE_CPUFLAG_AVX2)
+		CC_AVX2_SUPPORT=1
+	else
+		CC_AVX2_SUPPORT=\
+		$(shell $(CC) -march=core-avx2 -dM -E - </dev/null 2>&1 | \
+		grep -q AVX2 && echo 1)
+		ifeq ($(CC_AVX2_SUPPORT), 1)
+			ifeq ($(CONFIG_RTE_TOOLCHAIN_ICC),y)
+				CFLAGS_iavf_rxtx_vec_avx2.o += -march=core-avx2
+			else
+				CFLAGS_iavf_rxtx_vec_avx2.o += -mavx2
+			endif
+		endif
+	endif
+endif
+
+ifeq ($(CC_AVX2_SUPPORT), 1)
+	SRCS-$(CONFIG_RTE_LIBRTE_IAVF_PMD) += iavf_rxtx_vec_avx2.c
+endif
+
 include $(RTE_SDK)/mk/rte.lib.mk
diff --git a/drivers/net/iavf/iavf_rxtx.c b/drivers/net/iavf/iavf_rxtx.c
index 22d7bb612..e36f035c5 100644
--- a/drivers/net/iavf/iavf_rxtx.c
+++ b/drivers/net/iavf/iavf_rxtx.c
@@ -1666,27 +1666,6 @@ iavf_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts, uint16_t nb_pkts)
 	return nb_tx;
 }
 
-static uint16_t
-iavf_xmit_pkts_vec(void *tx_queue, struct rte_mbuf **tx_pkts,
-		  uint16_t nb_pkts)
-{
-	uint16_t nb_tx = 0;
-	struct iavf_tx_queue *txq = (struct iavf_tx_queue *)tx_queue;
-
-	while (nb_pkts) {
-		uint16_t ret, num;
-
-		num = (uint16_t)RTE_MIN(nb_pkts, txq->rs_thresh);
-		ret = iavf_xmit_fixed_burst_vec(tx_queue, &tx_pkts[nb_tx], num);
-		nb_tx += ret;
-		nb_pkts -= ret;
-		if (ret < num)
-			break;
-	}
-
-	return nb_tx;
-}
-
 /* TX prep functions */
 uint16_t
 iavf_prep_pkts(__rte_unused void *tx_queue, struct rte_mbuf **tx_pkts,
@@ -1741,26 +1720,43 @@ iavf_set_rx_function(struct rte_eth_dev *dev)
 {
 	struct iavf_adapter *adapter =
 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
+#ifdef RTE_ARCH_X86
 	struct iavf_rx_queue *rxq;
 	int i;
+	bool use_avx2 = false;
 
-	if (adapter->rx_vec_allowed) {
-		if (dev->data->scattered_rx) {
-			PMD_DRV_LOG(DEBUG, "Using Vector Scattered Rx callback"
-				    " (port=%d).", dev->data->port_id);
-			dev->rx_pkt_burst = iavf_recv_scattered_pkts_vec;
-		} else {
-			PMD_DRV_LOG(DEBUG, "Using Vector Rx callback"
-				    " (port=%d).", dev->data->port_id);
-			dev->rx_pkt_burst = iavf_recv_pkts_vec;
-		}
+	if (!iavf_rx_vec_dev_check(dev)) {
 		for (i = 0; i < dev->data->nb_rx_queues; i++) {
 			rxq = dev->data->rx_queues[i];
-			if (!rxq)
-				continue;
-			iavf_rxq_vec_setup(rxq);
+			(void)iavf_rxq_vec_setup(rxq);
 		}
-	} else if (dev->data->scattered_rx) {
+
+		if (rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1 ||
+		    rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX512F) == 1)
+			use_avx2 = true;
+
+		if (dev->data->scattered_rx) {
+			PMD_DRV_LOG(DEBUG,
+				    "Using %sVector Scattered Rx (port %d).",
+				    use_avx2 ? "avx2 " : "",
+				    dev->data->port_id);
+			dev->rx_pkt_burst = use_avx2 ?
+					    iavf_recv_scattered_pkts_vec_avx2 :
+					    iavf_recv_scattered_pkts_vec;
+		} else {
+			PMD_DRV_LOG(DEBUG, "Using %sVector Rx (port %d).",
+				    use_avx2 ? "avx2 " : "",
+				    dev->data->port_id);
+			dev->rx_pkt_burst = use_avx2 ?
+					    iavf_recv_pkts_vec_avx2 :
+					    iavf_recv_pkts_vec;
+		}
+
+		return;
+	}
+#endif
+
+	if (dev->data->scattered_rx) {
 		PMD_DRV_LOG(DEBUG, "Using a Scattered Rx callback (port=%d).",
 			    dev->data->port_id);
 		dev->rx_pkt_burst = iavf_recv_scattered_pkts;
@@ -1779,28 +1775,39 @@ iavf_set_rx_function(struct rte_eth_dev *dev)
 void
 iavf_set_tx_function(struct rte_eth_dev *dev)
 {
-	struct iavf_adapter *adapter =
-		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
+#ifdef RTE_ARCH_X86
 	struct iavf_tx_queue *txq;
 	int i;
+	bool use_avx2 = false;
 
-	if (adapter->tx_vec_allowed) {
-		PMD_DRV_LOG(DEBUG, "Using Vector Tx callback (port=%d).",
-			    dev->data->port_id);
-		dev->tx_pkt_burst = iavf_xmit_pkts_vec;
-		dev->tx_pkt_prepare = NULL;
+	if (!iavf_tx_vec_dev_check(dev)) {
 		for (i = 0; i < dev->data->nb_tx_queues; i++) {
 			txq = dev->data->tx_queues[i];
 			if (!txq)
 				continue;
 			iavf_txq_vec_setup(txq);
 		}
-	} else {
-		PMD_DRV_LOG(DEBUG, "Using Basic Tx callback (port=%d).",
+
+		if (rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1 ||
+		    rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX512F) == 1)
+			use_avx2 = true;
+
+		PMD_DRV_LOG(DEBUG, "Using %sVector Tx (port %d).",
+			    use_avx2 ? "avx2 " : "",
 			    dev->data->port_id);
-		dev->tx_pkt_burst = iavf_xmit_pkts;
-		dev->tx_pkt_prepare = iavf_prep_pkts;
+		dev->tx_pkt_burst = use_avx2 ?
+				    iavf_xmit_pkts_vec_avx2 :
+				    iavf_xmit_pkts_vec;
+		dev->tx_pkt_prepare = NULL;
+
+		return;
 	}
+#endif
+
+	PMD_DRV_LOG(DEBUG, "Using Basic Tx callback (port=%d).",
+		    dev->data->port_id);
+	dev->tx_pkt_burst = iavf_xmit_pkts;
+	dev->tx_pkt_prepare = iavf_prep_pkts;
 }
 
 void
diff --git a/drivers/net/iavf/iavf_rxtx.h b/drivers/net/iavf/iavf_rxtx.h
index c86720bda..225a0c4c4 100644
--- a/drivers/net/iavf/iavf_rxtx.h
+++ b/drivers/net/iavf/iavf_rxtx.h
@@ -19,6 +19,7 @@
 /* used for Vector PMD */
 #define IAVF_VPMD_RX_MAX_BURST    32
 #define IAVF_VPMD_TX_MAX_BURST    32
+#define IAVF_RXQ_REARM_THRESH     32
 #define IAVF_VPMD_DESCS_PER_LOOP  4
 #define IAVF_VPMD_TX_MAX_FREE_BUF 64
 
@@ -200,6 +201,17 @@ uint16_t iavf_recv_scattered_pkts_vec(void *rx_queue,
 				     uint16_t nb_pkts);
 uint16_t iavf_xmit_fixed_burst_vec(void *tx_queue, struct rte_mbuf **tx_pkts,
 				  uint16_t nb_pkts);
+uint16_t iavf_recv_pkts_vec_avx2(void *rx_queue, struct rte_mbuf **rx_pkts,
+				 uint16_t nb_pkts);
+uint16_t iavf_recv_scattered_pkts_vec_avx2(void *rx_queue,
+					   struct rte_mbuf **rx_pkts,
+					   uint16_t nb_pkts);
+uint16_t iavf_xmit_pkts_vec(void *tx_queue, struct rte_mbuf **tx_pkts,
+			    uint16_t nb_pkts);
+uint16_t iavf_xmit_pkts_vec_avx2(void *tx_queue, struct rte_mbuf **tx_pkts,
+				 uint16_t nb_pkts);
+int iavf_rx_vec_dev_check(struct rte_eth_dev *dev);
+int iavf_tx_vec_dev_check(struct rte_eth_dev *dev);
 int iavf_rxq_vec_setup(struct iavf_rx_queue *rxq);
 int iavf_txq_vec_setup(struct iavf_tx_queue *txq);
 
diff --git a/drivers/net/iavf/iavf_rxtx_vec_avx2.c b/drivers/net/iavf/iavf_rxtx_vec_avx2.c
new file mode 100644
index 000000000..f0c00be56
--- /dev/null
+++ b/drivers/net/iavf/iavf_rxtx_vec_avx2.c
@@ -0,0 +1,867 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2019 Intel Corporation
+ */
+
+#include "base/iavf_prototype.h"
+#include "iavf_rxtx_vec_common.h"
+
+#include <x86intrin.h>
+
+#ifndef __INTEL_COMPILER
+#pragma GCC diagnostic ignored "-Wcast-qual"
+#endif
+
+static inline void
+iavf_rxq_rearm(struct iavf_rx_queue *rxq)
+{
+	int i;
+	uint16_t rx_id;
+	volatile union iavf_rx_desc *rxdp;
+	struct rte_mbuf **rxp = &rxq->sw_ring[rxq->rxrearm_start];
+
+	rxdp = rxq->rx_ring + rxq->rxrearm_start;
+
+	/* Pull 'n' more MBUFs into the software ring */
+	if (rte_mempool_get_bulk(rxq->mp,
+				 (void *)rxp,
+				 IAVF_RXQ_REARM_THRESH) < 0) {
+		if (rxq->rxrearm_nb + IAVF_RXQ_REARM_THRESH >=
+		    rxq->nb_rx_desc) {
+			__m128i dma_addr0;
+
+			dma_addr0 = _mm_setzero_si128();
+			for (i = 0; i < IAVF_VPMD_DESCS_PER_LOOP; i++) {
+				rxp[i] = &rxq->fake_mbuf;
+				_mm_store_si128((__m128i *)&rxdp[i].read,
+						dma_addr0);
+			}
+		}
+		rte_eth_devices[rxq->port_id].data->rx_mbuf_alloc_failed +=
+			IAVF_RXQ_REARM_THRESH;
+		return;
+	}
+
+#ifndef RTE_LIBRTE_IAVF_16BYTE_RX_DESC
+	struct rte_mbuf *mb0, *mb1;
+	__m128i dma_addr0, dma_addr1;
+	__m128i hdr_room = _mm_set_epi64x(RTE_PKTMBUF_HEADROOM,
+			RTE_PKTMBUF_HEADROOM);
+	/* Initialize the mbufs in vector, process 2 mbufs in one loop */
+	for (i = 0; i < IAVF_RXQ_REARM_THRESH; i += 2, rxp += 2) {
+		__m128i vaddr0, vaddr1;
+
+		mb0 = rxp[0];
+		mb1 = rxp[1];
+
+		/* load buf_addr(lo 64bit) and buf_physaddr(hi 64bit) */
+		RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, buf_physaddr) !=
+				offsetof(struct rte_mbuf, buf_addr) + 8);
+		vaddr0 = _mm_loadu_si128((__m128i *)&mb0->buf_addr);
+		vaddr1 = _mm_loadu_si128((__m128i *)&mb1->buf_addr);
+
+		/* convert pa to dma_addr hdr/data */
+		dma_addr0 = _mm_unpackhi_epi64(vaddr0, vaddr0);
+		dma_addr1 = _mm_unpackhi_epi64(vaddr1, vaddr1);
+
+		/* add headroom to pa values */
+		dma_addr0 = _mm_add_epi64(dma_addr0, hdr_room);
+		dma_addr1 = _mm_add_epi64(dma_addr1, hdr_room);
+
+		/* flush desc with pa dma_addr */
+		_mm_store_si128((__m128i *)&rxdp++->read, dma_addr0);
+		_mm_store_si128((__m128i *)&rxdp++->read, dma_addr1);
+	}
+#else
+	struct rte_mbuf *mb0, *mb1, *mb2, *mb3;
+	__m256i dma_addr0_1, dma_addr2_3;
+	__m256i hdr_room = _mm256_set1_epi64x(RTE_PKTMBUF_HEADROOM);
+	/* Initialize the mbufs in vector, process 4 mbufs in one loop */
+	for (i = 0; i < IAVF_RXQ_REARM_THRESH;
+			i += 4, rxp += 4, rxdp += 4) {
+		__m128i vaddr0, vaddr1, vaddr2, vaddr3;
+		__m256i vaddr0_1, vaddr2_3;
+
+		mb0 = rxp[0];
+		mb1 = rxp[1];
+		mb2 = rxp[2];
+		mb3 = rxp[3];
+
+		/* load buf_addr(lo 64bit) and buf_physaddr(hi 64bit) */
+		RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, buf_physaddr) !=
+				offsetof(struct rte_mbuf, buf_addr) + 8);
+		vaddr0 = _mm_loadu_si128((__m128i *)&mb0->buf_addr);
+		vaddr1 = _mm_loadu_si128((__m128i *)&mb1->buf_addr);
+		vaddr2 = _mm_loadu_si128((__m128i *)&mb2->buf_addr);
+		vaddr3 = _mm_loadu_si128((__m128i *)&mb3->buf_addr);
+
+		/**
+		 * merge 0 & 1, by casting 0 to 256-bit and inserting 1
+		 * into the high lanes. Similarly for 2 & 3
+		 */
+		vaddr0_1 =
+			_mm256_inserti128_si256(_mm256_castsi128_si256(vaddr0),
+						vaddr1, 1);
+		vaddr2_3 =
+			_mm256_inserti128_si256(_mm256_castsi128_si256(vaddr2),
+						vaddr3, 1);
+
+		/* convert pa to dma_addr hdr/data */
+		dma_addr0_1 = _mm256_unpackhi_epi64(vaddr0_1, vaddr0_1);
+		dma_addr2_3 = _mm256_unpackhi_epi64(vaddr2_3, vaddr2_3);
+
+		/* add headroom to pa values */
+		dma_addr0_1 = _mm256_add_epi64(dma_addr0_1, hdr_room);
+		dma_addr2_3 = _mm256_add_epi64(dma_addr2_3, hdr_room);
+
+		/* flush desc with pa dma_addr */
+		_mm256_store_si256((__m256i *)&rxdp->read, dma_addr0_1);
+		_mm256_store_si256((__m256i *)&(rxdp + 2)->read, dma_addr2_3);
+	}
+
+#endif
+
+	rxq->rxrearm_start += IAVF_RXQ_REARM_THRESH;
+	if (rxq->rxrearm_start >= rxq->nb_rx_desc)
+		rxq->rxrearm_start = 0;
+
+	rxq->rxrearm_nb -= IAVF_RXQ_REARM_THRESH;
+
+	rx_id = (uint16_t)((rxq->rxrearm_start == 0) ?
+			     (rxq->nb_rx_desc - 1) : (rxq->rxrearm_start - 1));
+
+	/* Update the tail pointer on the NIC */
+	IAVF_PCI_REG_WRITE(rxq->qrx_tail, rx_id);
+}
+
+#define PKTLEN_SHIFT     10
+
+static inline uint16_t
+_iavf_recv_raw_pkts_vec_avx2(struct iavf_rx_queue *rxq,
+			     struct rte_mbuf **rx_pkts,
+			     uint16_t nb_pkts, uint8_t *split_packet)
+{
+#define IAVF_DESCS_PER_LOOP_AVX 8
+
+	/* const uint32_t *ptype_tbl = rxq->vsi->adapter->ptype_tbl; */
+	static const uint32_t type_table[UINT8_MAX + 1] __rte_cache_aligned = {
+		/* [0] reserved */
+		[1] = RTE_PTYPE_L2_ETHER,
+		/* [2] - [21] reserved */
+		[22] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
+			RTE_PTYPE_L4_FRAG,
+		[23] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
+			RTE_PTYPE_L4_NONFRAG,
+		[24] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
+			RTE_PTYPE_L4_UDP,
+		/* [25] reserved */
+		[26] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
+			RTE_PTYPE_L4_TCP,
+		[27] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
+			RTE_PTYPE_L4_SCTP,
+		[28] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
+			RTE_PTYPE_L4_ICMP,
+		/* All others reserved */
+	};
+	const __m256i mbuf_init = _mm256_set_epi64x(0, 0,
+			0, rxq->mbuf_initializer);
+	/* struct iavf_rx_entry *sw_ring = &rxq->sw_ring[rxq->rx_tail]; */
+	struct rte_mbuf **sw_ring = &rxq->sw_ring[rxq->rx_tail];
+	volatile union iavf_rx_desc *rxdp = rxq->rx_ring + rxq->rx_tail;
+	const int avx_aligned = ((rxq->rx_tail & 1) == 0);
+
+	rte_prefetch0(rxdp);
+
+	/* nb_pkts has to be floor-aligned to IAVF_DESCS_PER_LOOP_AVX */
+	nb_pkts = RTE_ALIGN_FLOOR(nb_pkts, IAVF_DESCS_PER_LOOP_AVX);
+
+	/* See if we need to rearm the RX queue - gives the prefetch a bit
+	 * of time to act
+	 */
+	if (rxq->rxrearm_nb > IAVF_RXQ_REARM_THRESH)
+		iavf_rxq_rearm(rxq);
+
+	/* Before we start moving massive data around, check to see if
+	 * there is actually a packet available
+	 */
+	if (!(rxdp->wb.qword1.status_error_len &
+			rte_cpu_to_le_32(1 << IAVF_RX_DESC_STATUS_DD_SHIFT)))
+		return 0;
+
+	/* constants used in processing loop */
+	const __m256i crc_adjust =
+		_mm256_set_epi16
+			(/* first descriptor */
+			 0, 0, 0,       /* ignore non-length fields */
+			 -rxq->crc_len, /* sub crc on data_len */
+			 0,             /* ignore high-16bits of pkt_len */
+			 -rxq->crc_len, /* sub crc on pkt_len */
+			 0, 0,          /* ignore pkt_type field */
+			 /* second descriptor */
+			 0, 0, 0,       /* ignore non-length fields */
+			 -rxq->crc_len, /* sub crc on data_len */
+			 0,             /* ignore high-16bits of pkt_len */
+			 -rxq->crc_len, /* sub crc on pkt_len */
+			 0, 0           /* ignore pkt_type field */
+			);
+
+	/* 8 packets DD mask, LSB in each 32-bit value */
+	const __m256i dd_check = _mm256_set1_epi32(1);
+
+	/* 8 packets EOP mask, second-LSB in each 32-bit value */
+	const __m256i eop_check = _mm256_slli_epi32(dd_check,
+			IAVF_RX_DESC_STATUS_EOF_SHIFT);
+
+	/* mask to shuffle from desc. to mbuf (2 descriptors)*/
+	const __m256i shuf_msk =
+		_mm256_set_epi8
+			(/* first descriptor */
+			 7, 6, 5, 4,  /* octet 4~7, 32bits rss */
+			 3, 2,        /* octet 2~3, low 16 bits vlan_macip */
+			 15, 14,      /* octet 15~14, 16 bits data_len */
+			 0xFF, 0xFF,  /* skip high 16 bits pkt_len, zero out */
+			 15, 14,      /* octet 15~14, low 16 bits pkt_len */
+			 0xFF, 0xFF,  /* pkt_type set as unknown */
+			 0xFF, 0xFF,  /*pkt_type set as unknown */
+			 /* second descriptor */
+			 7, 6, 5, 4,  /* octet 4~7, 32bits rss */
+			 3, 2,        /* octet 2~3, low 16 bits vlan_macip */
+			 15, 14,      /* octet 15~14, 16 bits data_len */
+			 0xFF, 0xFF,  /* skip high 16 bits pkt_len, zero out */
+			 15, 14,      /* octet 15~14, low 16 bits pkt_len */
+			 0xFF, 0xFF,  /* pkt_type set as unknown */
+			 0xFF, 0xFF   /*pkt_type set as unknown */
+			);
+	/**
+	 * compile-time check the above crc and shuffle layout is correct.
+	 * NOTE: the first field (lowest address) is given last in set_epi
+	 * calls above.
+	 */
+	RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, pkt_len) !=
+			offsetof(struct rte_mbuf, rx_descriptor_fields1) + 4);
+	RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, data_len) !=
+			offsetof(struct rte_mbuf, rx_descriptor_fields1) + 8);
+	RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, vlan_tci) !=
+			offsetof(struct rte_mbuf, rx_descriptor_fields1) + 10);
+	RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, hash) !=
+			offsetof(struct rte_mbuf, rx_descriptor_fields1) + 12);
+
+	/* Status/Error flag masks */
+	/**
+	 * mask everything except RSS, flow director and VLAN flags
+	 * bit2 is for VLAN tag, bit11 for flow director indication
+	 * bit13:12 for RSS indication. Bits 3-5 of error
+	 * field (bits 22-24) are for IP/L4 checksum errors
+	 */
+	const __m256i flags_mask =
+		 _mm256_set1_epi32((1 << 2) | (1 << 11) |
+				   (3 << 12) | (7 << 22));
+	/**
+	 * data to be shuffled by result of flag mask. If VLAN bit is set,
+	 * (bit 2), then position 4 in this array will be used in the
+	 * destination
+	 */
+	const __m256i vlan_flags_shuf =
+		_mm256_set_epi32(0, 0, PKT_RX_VLAN | PKT_RX_VLAN_STRIPPED, 0,
+				 0, 0, PKT_RX_VLAN | PKT_RX_VLAN_STRIPPED, 0);
+	/**
+	 * data to be shuffled by result of flag mask, shifted down 11.
+	 * If RSS/FDIR bits are set, shuffle moves appropriate flags in
+	 * place.
+	 */
+	const __m256i rss_flags_shuf =
+		_mm256_set_epi8(0, 0, 0, 0, 0, 0, 0, 0,
+				PKT_RX_RSS_HASH | PKT_RX_FDIR, PKT_RX_RSS_HASH,
+				0, 0, 0, 0, PKT_RX_FDIR, 0,/* end up 128-bits */
+				0, 0, 0, 0, 0, 0, 0, 0,
+				PKT_RX_RSS_HASH | PKT_RX_FDIR, PKT_RX_RSS_HASH,
+				0, 0, 0, 0, PKT_RX_FDIR, 0);
+
+	/**
+	 * data to be shuffled by the result of the flags mask shifted by 22
+	 * bits.  This gives use the l3_l4 flags.
+	 */
+	const __m256i l3_l4_flags_shuf = _mm256_set_epi8(0, 0, 0, 0, 0, 0, 0, 0,
+			/* shift right 1 bit to make sure it not exceed 255 */
+			(PKT_RX_EIP_CKSUM_BAD | PKT_RX_L4_CKSUM_BAD |
+			 PKT_RX_IP_CKSUM_BAD) >> 1,
+			(PKT_RX_IP_CKSUM_GOOD | PKT_RX_EIP_CKSUM_BAD |
+			 PKT_RX_L4_CKSUM_BAD) >> 1,
+			(PKT_RX_EIP_CKSUM_BAD | PKT_RX_IP_CKSUM_BAD) >> 1,
+			(PKT_RX_IP_CKSUM_GOOD | PKT_RX_EIP_CKSUM_BAD) >> 1,
+			(PKT_RX_L4_CKSUM_BAD | PKT_RX_IP_CKSUM_BAD) >> 1,
+			(PKT_RX_IP_CKSUM_GOOD | PKT_RX_L4_CKSUM_BAD) >> 1,
+			PKT_RX_IP_CKSUM_BAD >> 1,
+			(PKT_RX_IP_CKSUM_GOOD | PKT_RX_L4_CKSUM_GOOD) >> 1,
+			/* second 128-bits */
+			0, 0, 0, 0, 0, 0, 0, 0,
+			(PKT_RX_EIP_CKSUM_BAD | PKT_RX_L4_CKSUM_BAD |
+			 PKT_RX_IP_CKSUM_BAD) >> 1,
+			(PKT_RX_IP_CKSUM_GOOD | PKT_RX_EIP_CKSUM_BAD |
+			 PKT_RX_L4_CKSUM_BAD) >> 1,
+			(PKT_RX_EIP_CKSUM_BAD | PKT_RX_IP_CKSUM_BAD) >> 1,
+			(PKT_RX_IP_CKSUM_GOOD | PKT_RX_EIP_CKSUM_BAD) >> 1,
+			(PKT_RX_L4_CKSUM_BAD | PKT_RX_IP_CKSUM_BAD) >> 1,
+			(PKT_RX_IP_CKSUM_GOOD | PKT_RX_L4_CKSUM_BAD) >> 1,
+			PKT_RX_IP_CKSUM_BAD >> 1,
+			(PKT_RX_IP_CKSUM_GOOD | PKT_RX_L4_CKSUM_GOOD) >> 1);
+
+	const __m256i cksum_mask =
+		 _mm256_set1_epi32(PKT_RX_IP_CKSUM_GOOD | PKT_RX_IP_CKSUM_BAD |
+				   PKT_RX_L4_CKSUM_GOOD | PKT_RX_L4_CKSUM_BAD |
+				   PKT_RX_EIP_CKSUM_BAD);
+
+	RTE_SET_USED(avx_aligned); /* for 32B descriptors we don't use this */
+
+	uint16_t i, received;
+
+	for (i = 0, received = 0; i < nb_pkts;
+	     i += IAVF_DESCS_PER_LOOP_AVX,
+	     rxdp += IAVF_DESCS_PER_LOOP_AVX) {
+		/* step 1, copy over 8 mbuf pointers to rx_pkts array */
+		_mm256_storeu_si256((void *)&rx_pkts[i],
+				    _mm256_loadu_si256((void *)&sw_ring[i]));
+#ifdef RTE_ARCH_X86_64
+		_mm256_storeu_si256
+			((void *)&rx_pkts[i + 4],
+			 _mm256_loadu_si256((void *)&sw_ring[i + 4]));
+#endif
+
+		__m256i raw_desc0_1, raw_desc2_3, raw_desc4_5, raw_desc6_7;
+#ifdef RTE_LIBRTE_IAVF_16BYTE_RX_DESC
+		/* for AVX we need alignment otherwise loads are not atomic */
+		if (avx_aligned) {
+			/* load in descriptors, 2 at a time, in reverse order */
+			raw_desc6_7 = _mm256_load_si256((void *)(rxdp + 6));
+			rte_compiler_barrier();
+			raw_desc4_5 = _mm256_load_si256((void *)(rxdp + 4));
+			rte_compiler_barrier();
+			raw_desc2_3 = _mm256_load_si256((void *)(rxdp + 2));
+			rte_compiler_barrier();
+			raw_desc0_1 = _mm256_load_si256((void *)(rxdp + 0));
+		} else
+#endif
+		{
+			const __m128i raw_desc7 =
+				_mm_load_si128((void *)(rxdp + 7));
+			rte_compiler_barrier();
+			const __m128i raw_desc6 =
+				_mm_load_si128((void *)(rxdp + 6));
+			rte_compiler_barrier();
+			const __m128i raw_desc5 =
+				_mm_load_si128((void *)(rxdp + 5));
+			rte_compiler_barrier();
+			const __m128i raw_desc4 =
+				_mm_load_si128((void *)(rxdp + 4));
+			rte_compiler_barrier();
+			const __m128i raw_desc3 =
+				_mm_load_si128((void *)(rxdp + 3));
+			rte_compiler_barrier();
+			const __m128i raw_desc2 =
+				_mm_load_si128((void *)(rxdp + 2));
+			rte_compiler_barrier();
+			const __m128i raw_desc1 =
+				_mm_load_si128((void *)(rxdp + 1));
+			rte_compiler_barrier();
+			const __m128i raw_desc0 =
+				_mm_load_si128((void *)(rxdp + 0));
+
+			raw_desc6_7 =
+				_mm256_inserti128_si256
+					(_mm256_castsi128_si256(raw_desc6),
+					 raw_desc7, 1);
+			raw_desc4_5 =
+				_mm256_inserti128_si256
+					(_mm256_castsi128_si256(raw_desc4),
+					 raw_desc5, 1);
+			raw_desc2_3 =
+				_mm256_inserti128_si256
+					(_mm256_castsi128_si256(raw_desc2),
+					 raw_desc3, 1);
+			raw_desc0_1 =
+				_mm256_inserti128_si256
+					(_mm256_castsi128_si256(raw_desc0),
+					 raw_desc1, 1);
+		}
+
+		if (split_packet) {
+			int j;
+
+			for (j = 0; j < IAVF_DESCS_PER_LOOP_AVX; j++)
+				rte_mbuf_prefetch_part2(rx_pkts[i + j]);
+		}
+
+		/**
+		 * convert descriptors 4-7 into mbufs, adjusting length and
+		 * re-arranging fields. Then write into the mbuf
+		 */
+		const __m256i len6_7 = _mm256_slli_epi32(raw_desc6_7,
+							 PKTLEN_SHIFT);
+		const __m256i len4_5 = _mm256_slli_epi32(raw_desc4_5,
+							 PKTLEN_SHIFT);
+		const __m256i desc6_7 = _mm256_blend_epi16(raw_desc6_7,
+							   len6_7, 0x80);
+		const __m256i desc4_5 = _mm256_blend_epi16(raw_desc4_5,
+							   len4_5, 0x80);
+		__m256i mb6_7 = _mm256_shuffle_epi8(desc6_7, shuf_msk);
+		__m256i mb4_5 = _mm256_shuffle_epi8(desc4_5, shuf_msk);
+
+		mb6_7 = _mm256_add_epi16(mb6_7, crc_adjust);
+		mb4_5 = _mm256_add_epi16(mb4_5, crc_adjust);
+		/**
+		 * to get packet types, shift 64-bit values down 30 bits
+		 * and so ptype is in lower 8-bits in each
+		 */
+		const __m256i ptypes6_7 = _mm256_srli_epi64(desc6_7, 30);
+		const __m256i ptypes4_5 = _mm256_srli_epi64(desc4_5, 30);
+		const uint8_t ptype7 = _mm256_extract_epi8(ptypes6_7, 24);
+		const uint8_t ptype6 = _mm256_extract_epi8(ptypes6_7, 8);
+		const uint8_t ptype5 = _mm256_extract_epi8(ptypes4_5, 24);
+		const uint8_t ptype4 = _mm256_extract_epi8(ptypes4_5, 8);
+
+		mb6_7 = _mm256_insert_epi32(mb6_7, type_table[ptype7], 4);
+		mb6_7 = _mm256_insert_epi32(mb6_7, type_table[ptype6], 0);
+		mb4_5 = _mm256_insert_epi32(mb4_5, type_table[ptype5], 4);
+		mb4_5 = _mm256_insert_epi32(mb4_5, type_table[ptype4], 0);
+		/* merge the status bits into one register */
+		const __m256i status4_7 = _mm256_unpackhi_epi32(desc6_7,
+				desc4_5);
+
+		/**
+		 * convert descriptors 0-3 into mbufs, adjusting length and
+		 * re-arranging fields. Then write into the mbuf
+		 */
+		const __m256i len2_3 = _mm256_slli_epi32(raw_desc2_3,
+							 PKTLEN_SHIFT);
+		const __m256i len0_1 = _mm256_slli_epi32(raw_desc0_1,
+							 PKTLEN_SHIFT);
+		const __m256i desc2_3 = _mm256_blend_epi16(raw_desc2_3,
+							   len2_3, 0x80);
+		const __m256i desc0_1 = _mm256_blend_epi16(raw_desc0_1,
+							   len0_1, 0x80);
+		__m256i mb2_3 = _mm256_shuffle_epi8(desc2_3, shuf_msk);
+		__m256i mb0_1 = _mm256_shuffle_epi8(desc0_1, shuf_msk);
+
+		mb2_3 = _mm256_add_epi16(mb2_3, crc_adjust);
+		mb0_1 = _mm256_add_epi16(mb0_1, crc_adjust);
+		/* get the packet types */
+		const __m256i ptypes2_3 = _mm256_srli_epi64(desc2_3, 30);
+		const __m256i ptypes0_1 = _mm256_srli_epi64(desc0_1, 30);
+		const uint8_t ptype3 = _mm256_extract_epi8(ptypes2_3, 24);
+		const uint8_t ptype2 = _mm256_extract_epi8(ptypes2_3, 8);
+		const uint8_t ptype1 = _mm256_extract_epi8(ptypes0_1, 24);
+		const uint8_t ptype0 = _mm256_extract_epi8(ptypes0_1, 8);
+
+		mb2_3 = _mm256_insert_epi32(mb2_3, type_table[ptype3], 4);
+		mb2_3 = _mm256_insert_epi32(mb2_3, type_table[ptype2], 0);
+		mb0_1 = _mm256_insert_epi32(mb0_1, type_table[ptype1], 4);
+		mb0_1 = _mm256_insert_epi32(mb0_1, type_table[ptype0], 0);
+		/* merge the status bits into one register */
+		const __m256i status0_3 = _mm256_unpackhi_epi32(desc2_3,
+								desc0_1);
+
+		/**
+		 * take the two sets of status bits and merge to one
+		 * After merge, the packets status flags are in the
+		 * order (hi->lo): [1, 3, 5, 7, 0, 2, 4, 6]
+		 */
+		__m256i status0_7 = _mm256_unpacklo_epi64(status4_7,
+							  status0_3);
+
+		/* now do flag manipulation */
+
+		/* get only flag/error bits we want */
+		const __m256i flag_bits =
+			_mm256_and_si256(status0_7, flags_mask);
+		/* set vlan and rss flags */
+		const __m256i vlan_flags =
+			_mm256_shuffle_epi8(vlan_flags_shuf, flag_bits);
+		const __m256i rss_flags =
+			_mm256_shuffle_epi8(rss_flags_shuf,
+					    _mm256_srli_epi32(flag_bits, 11));
+		/**
+		 * l3_l4_error flags, shuffle, then shift to correct adjustment
+		 * of flags in flags_shuf, and finally mask out extra bits
+		 */
+		__m256i l3_l4_flags = _mm256_shuffle_epi8(l3_l4_flags_shuf,
+				_mm256_srli_epi32(flag_bits, 22));
+		l3_l4_flags = _mm256_slli_epi32(l3_l4_flags, 1);
+		l3_l4_flags = _mm256_and_si256(l3_l4_flags, cksum_mask);
+
+		/* merge flags */
+		const __m256i mbuf_flags = _mm256_or_si256(l3_l4_flags,
+				_mm256_or_si256(rss_flags, vlan_flags));
+		/**
+		 * At this point, we have the 8 sets of flags in the low 16-bits
+		 * of each 32-bit value in vlan0.
+		 * We want to extract these, and merge them with the mbuf init
+		 * data so we can do a single write to the mbuf to set the flags
+		 * and all the other initialization fields. Extracting the
+		 * appropriate flags means that we have to do a shift and blend
+		 * for each mbuf before we do the write. However, we can also
+		 * add in the previously computed rx_descriptor fields to
+		 * make a single 256-bit write per mbuf
+		 */
+		/* check the structure matches expectations */
+		RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, ol_flags) !=
+				 offsetof(struct rte_mbuf, rearm_data) + 8);
+		RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, rearm_data) !=
+				 RTE_ALIGN(offsetof(struct rte_mbuf,
+						    rearm_data),
+					   16));
+		/* build up data and do writes */
+		__m256i rearm0, rearm1, rearm2, rearm3, rearm4, rearm5,
+			rearm6, rearm7;
+		rearm6 = _mm256_blend_epi32(mbuf_init,
+					    _mm256_slli_si256(mbuf_flags, 8),
+					    0x04);
+		rearm4 = _mm256_blend_epi32(mbuf_init,
+					    _mm256_slli_si256(mbuf_flags, 4),
+					    0x04);
+		rearm2 = _mm256_blend_epi32(mbuf_init, mbuf_flags, 0x04);
+		rearm0 = _mm256_blend_epi32(mbuf_init,
+					    _mm256_srli_si256(mbuf_flags, 4),
+					    0x04);
+		/* permute to add in the rx_descriptor e.g. rss fields */
+		rearm6 = _mm256_permute2f128_si256(rearm6, mb6_7, 0x20);
+		rearm4 = _mm256_permute2f128_si256(rearm4, mb4_5, 0x20);
+		rearm2 = _mm256_permute2f128_si256(rearm2, mb2_3, 0x20);
+		rearm0 = _mm256_permute2f128_si256(rearm0, mb0_1, 0x20);
+		/* write to mbuf */
+		_mm256_storeu_si256((__m256i *)&rx_pkts[i + 6]->rearm_data,
+				    rearm6);
+		_mm256_storeu_si256((__m256i *)&rx_pkts[i + 4]->rearm_data,
+				    rearm4);
+		_mm256_storeu_si256((__m256i *)&rx_pkts[i + 2]->rearm_data,
+				    rearm2);
+		_mm256_storeu_si256((__m256i *)&rx_pkts[i + 0]->rearm_data,
+				    rearm0);
+
+		/* repeat for the odd mbufs */
+		const __m256i odd_flags =
+			_mm256_castsi128_si256
+				(_mm256_extracti128_si256(mbuf_flags, 1));
+		rearm7 = _mm256_blend_epi32(mbuf_init,
+					    _mm256_slli_si256(odd_flags, 8),
+					    0x04);
+		rearm5 = _mm256_blend_epi32(mbuf_init,
+					    _mm256_slli_si256(odd_flags, 4),
+					    0x04);
+		rearm3 = _mm256_blend_epi32(mbuf_init, odd_flags, 0x04);
+		rearm1 = _mm256_blend_epi32(mbuf_init,
+					    _mm256_srli_si256(odd_flags, 4),
+					    0x04);
+		/* since odd mbufs are already in hi 128-bits use blend */
+		rearm7 = _mm256_blend_epi32(rearm7, mb6_7, 0xF0);
+		rearm5 = _mm256_blend_epi32(rearm5, mb4_5, 0xF0);
+		rearm3 = _mm256_blend_epi32(rearm3, mb2_3, 0xF0);
+		rearm1 = _mm256_blend_epi32(rearm1, mb0_1, 0xF0);
+		/* again write to mbufs */
+		_mm256_storeu_si256((__m256i *)&rx_pkts[i + 7]->rearm_data,
+				    rearm7);
+		_mm256_storeu_si256((__m256i *)&rx_pkts[i + 5]->rearm_data,
+				    rearm5);
+		_mm256_storeu_si256((__m256i *)&rx_pkts[i + 3]->rearm_data,
+				    rearm3);
+		_mm256_storeu_si256((__m256i *)&rx_pkts[i + 1]->rearm_data,
+				    rearm1);
+
+		/* extract and record EOP bit */
+		if (split_packet) {
+			const __m128i eop_mask =
+				_mm_set1_epi16(1 << IAVF_RX_DESC_STATUS_EOF_SHIFT);
+			const __m256i eop_bits256 = _mm256_and_si256(status0_7,
+								     eop_check);
+			/* pack status bits into a single 128-bit register */
+			const __m128i eop_bits =
+				_mm_packus_epi32
+					(_mm256_castsi256_si128(eop_bits256),
+					 _mm256_extractf128_si256(eop_bits256,
+								  1));
+			/**
+			 * flip bits, and mask out the EOP bit, which is now
+			 * a split-packet bit i.e. !EOP, rather than EOP one.
+			 */
+			__m128i split_bits = _mm_andnot_si128(eop_bits,
+					eop_mask);
+			/**
+			 * eop bits are out of order, so we need to shuffle them
+			 * back into order again. In doing so, only use low 8
+			 * bits, which acts like another pack instruction
+			 * The original order is (hi->lo): 1,3,5,7,0,2,4,6
+			 * [Since we use epi8, the 16-bit positions are
+			 * multiplied by 2 in the eop_shuffle value.]
+			 */
+			__m128i eop_shuffle =
+				_mm_set_epi8(/* zero hi 64b */
+					     0xFF, 0xFF, 0xFF, 0xFF,
+					     0xFF, 0xFF, 0xFF, 0xFF,
+					     /* move values to lo 64b */
+					     8, 0, 10, 2,
+					     12, 4, 14, 6);
+			split_bits = _mm_shuffle_epi8(split_bits, eop_shuffle);
+			*(uint64_t *)split_packet =
+				_mm_cvtsi128_si64(split_bits);
+			split_packet += IAVF_DESCS_PER_LOOP_AVX;
+		}
+
+		/* perform dd_check */
+		status0_7 = _mm256_and_si256(status0_7, dd_check);
+		status0_7 = _mm256_packs_epi32(status0_7,
+					       _mm256_setzero_si256());
+
+		uint64_t burst = __builtin_popcountll
+					(_mm_cvtsi128_si64
+						(_mm256_extracti128_si256
+							(status0_7, 1)));
+		burst += __builtin_popcountll
+				(_mm_cvtsi128_si64
+					(_mm256_castsi256_si128(status0_7)));
+		received += burst;
+		if (burst != IAVF_DESCS_PER_LOOP_AVX)
+			break;
+	}
+
+	/* update tail pointers */
+	rxq->rx_tail += received;
+	rxq->rx_tail &= (rxq->nb_rx_desc - 1);
+	if ((rxq->rx_tail & 1) == 1 && received > 1) { /* keep avx2 aligned */
+		rxq->rx_tail--;
+		received--;
+	}
+	rxq->rxrearm_nb += received;
+	return received;
+}
+
+/**
+ * Notice:
+ * - nb_pkts < IAVF_DESCS_PER_LOOP, just return no packet
+ */
+uint16_t
+iavf_recv_pkts_vec_avx2(void *rx_queue, struct rte_mbuf **rx_pkts,
+			uint16_t nb_pkts)
+{
+	return _iavf_recv_raw_pkts_vec_avx2(rx_queue, rx_pkts, nb_pkts, NULL);
+}
+
+/**
+ * vPMD receive routine that reassembles single burst of 32 scattered packets
+ * Notice:
+ * - nb_pkts < IAVF_DESCS_PER_LOOP, just return no packet
+ */
+static uint16_t
+iavf_recv_scattered_burst_vec_avx2(void *rx_queue, struct rte_mbuf **rx_pkts,
+				   uint16_t nb_pkts)
+{
+	struct iavf_rx_queue *rxq = rx_queue;
+	uint8_t split_flags[IAVF_VPMD_RX_MAX_BURST] = {0};
+
+	/* get some new buffers */
+	uint16_t nb_bufs = _iavf_recv_raw_pkts_vec_avx2(rxq, rx_pkts, nb_pkts,
+						       split_flags);
+	if (nb_bufs == 0)
+		return 0;
+
+	/* happy day case, full burst + no packets to be joined */
+	const uint64_t *split_fl64 = (uint64_t *)split_flags;
+
+	if (!rxq->pkt_first_seg &&
+	    split_fl64[0] == 0 && split_fl64[1] == 0 &&
+	    split_fl64[2] == 0 && split_fl64[3] == 0)
+		return nb_bufs;
+
+	/* reassemble any packets that need reassembly*/
+	unsigned int i = 0;
+
+	if (!rxq->pkt_first_seg) {
+		/* find the first split flag, and only reassemble then*/
+		while (i < nb_bufs && !split_flags[i])
+			i++;
+		if (i == nb_bufs)
+			return nb_bufs;
+		rxq->pkt_first_seg = rx_pkts[i];
+	}
+	return i + reassemble_packets(rxq, &rx_pkts[i], nb_bufs - i,
+					     &split_flags[i]);
+}
+
+/**
+ * vPMD receive routine that reassembles scattered packets.
+ * Main receive routine that can handle arbitrary burst sizes
+ * Notice:
+ * - nb_pkts < IAVF_DESCS_PER_LOOP, just return no packet
+ */
+uint16_t
+iavf_recv_scattered_pkts_vec_avx2(void *rx_queue, struct rte_mbuf **rx_pkts,
+				  uint16_t nb_pkts)
+{
+	uint16_t retval = 0;
+
+	while (nb_pkts > IAVF_VPMD_RX_MAX_BURST) {
+		uint16_t burst = iavf_recv_scattered_burst_vec_avx2(rx_queue,
+				rx_pkts + retval, IAVF_VPMD_RX_MAX_BURST);
+		retval += burst;
+		nb_pkts -= burst;
+		if (burst < IAVF_VPMD_RX_MAX_BURST)
+			return retval;
+	}
+	return retval + iavf_recv_scattered_burst_vec_avx2(rx_queue,
+				rx_pkts + retval, nb_pkts);
+}
+
+static inline void
+iavf_vtx1(volatile struct iavf_tx_desc *txdp,
+	  struct rte_mbuf *pkt, uint64_t flags)
+{
+	uint64_t high_qw =
+		(IAVF_TX_DESC_DTYPE_DATA |
+		 ((uint64_t)flags  << IAVF_TXD_QW1_CMD_SHIFT) |
+		 ((uint64_t)pkt->data_len << IAVF_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
+iavf_vtx(volatile struct iavf_tx_desc *txdp,
+	 struct rte_mbuf **pkt, uint16_t nb_pkts,  uint64_t flags)
+{
+	const uint64_t hi_qw_tmpl = (IAVF_TX_DESC_DTYPE_DATA |
+			((uint64_t)flags  << IAVF_TXD_QW1_CMD_SHIFT));
+
+	/* if unaligned on 32-bit boundary, do one to align */
+	if (((uintptr_t)txdp & 0x1F) != 0 && nb_pkts != 0) {
+		iavf_vtx1(txdp, *pkt, flags);
+		nb_pkts--, txdp++, pkt++;
+	}
+
+	/* do two at a time while possible, in bursts */
+	for (; nb_pkts > 3; txdp += 4, pkt += 4, nb_pkts -= 4) {
+		uint64_t hi_qw3 =
+			hi_qw_tmpl |
+			((uint64_t)pkt[3]->data_len <<
+			 IAVF_TXD_QW1_TX_BUF_SZ_SHIFT);
+		uint64_t hi_qw2 =
+			hi_qw_tmpl |
+			((uint64_t)pkt[2]->data_len <<
+			 IAVF_TXD_QW1_TX_BUF_SZ_SHIFT);
+		uint64_t hi_qw1 =
+			hi_qw_tmpl |
+			((uint64_t)pkt[1]->data_len <<
+			 IAVF_TXD_QW1_TX_BUF_SZ_SHIFT);
+		uint64_t hi_qw0 =
+			hi_qw_tmpl |
+			((uint64_t)pkt[0]->data_len <<
+			 IAVF_TXD_QW1_TX_BUF_SZ_SHIFT);
+
+		__m256i desc2_3 =
+			_mm256_set_epi64x
+				(hi_qw3,
+				 pkt[3]->buf_physaddr + pkt[3]->data_off,
+				 hi_qw2,
+				 pkt[2]->buf_physaddr + pkt[2]->data_off);
+		__m256i desc0_1 =
+			_mm256_set_epi64x
+				(hi_qw1,
+				 pkt[1]->buf_physaddr + pkt[1]->data_off,
+				 hi_qw0,
+				 pkt[0]->buf_physaddr + pkt[0]->data_off);
+		_mm256_store_si256((void *)(txdp + 2), desc2_3);
+		_mm256_store_si256((void *)txdp, desc0_1);
+	}
+
+	/* do any last ones */
+	while (nb_pkts) {
+		iavf_vtx1(txdp, *pkt, flags);
+		txdp++, pkt++, nb_pkts--;
+	}
+}
+
+static inline uint16_t
+iavf_xmit_fixed_burst_vec_avx2(void *tx_queue, struct rte_mbuf **tx_pkts,
+			       uint16_t nb_pkts)
+{
+	struct iavf_tx_queue *txq = (struct iavf_tx_queue *)tx_queue;
+	volatile struct iavf_tx_desc *txdp;
+	struct iavf_tx_entry *txep;
+	uint16_t n, nb_commit, tx_id;
+	uint64_t flags = IAVF_TX_DESC_CMD_EOP;
+	uint64_t rs = IAVF_TX_DESC_CMD_RS | IAVF_TX_DESC_CMD_EOP;
+
+	/* cross rx_thresh boundary is not allowed */
+	nb_pkts = RTE_MIN(nb_pkts, txq->rs_thresh);
+
+	if (txq->nb_free < txq->free_thresh)
+		iavf_tx_free_bufs(txq);
+
+	nb_commit = nb_pkts = (uint16_t)RTE_MIN(txq->nb_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_free = (uint16_t)(txq->nb_free - nb_pkts);
+
+	n = (uint16_t)(txq->nb_tx_desc - tx_id);
+	if (nb_commit >= n) {
+		tx_backlog_entry(txep, tx_pkts, n);
+
+		iavf_vtx(txdp, tx_pkts, n - 1, flags);
+		tx_pkts += (n - 1);
+		txdp += (n - 1);
+
+		iavf_vtx1(txdp, *tx_pkts++, rs);
+
+		nb_commit = (uint16_t)(nb_commit - n);
+
+		tx_id = 0;
+		txq->next_rs = (uint16_t)(txq->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);
+
+	iavf_vtx(txdp, tx_pkts, nb_commit, flags);
+
+	tx_id = (uint16_t)(tx_id + nb_commit);
+	if (tx_id > txq->next_rs) {
+		txq->tx_ring[txq->next_rs].cmd_type_offset_bsz |=
+			rte_cpu_to_le_64(((uint64_t)IAVF_TX_DESC_CMD_RS) <<
+					 IAVF_TXD_QW1_CMD_SHIFT);
+		txq->next_rs =
+			(uint16_t)(txq->next_rs + txq->rs_thresh);
+	}
+
+	txq->tx_tail = tx_id;
+
+	IAVF_PCI_REG_WRITE(txq->qtx_tail, txq->tx_tail);
+
+	return nb_pkts;
+}
+
+uint16_t
+iavf_xmit_pkts_vec_avx2(void *tx_queue, struct rte_mbuf **tx_pkts,
+			uint16_t nb_pkts)
+{
+	uint16_t nb_tx = 0;
+	struct iavf_tx_queue *txq = (struct iavf_tx_queue *)tx_queue;
+
+	while (nb_pkts) {
+		uint16_t ret, num;
+
+		num = (uint16_t)RTE_MIN(nb_pkts, txq->rs_thresh);
+		ret = iavf_xmit_fixed_burst_vec_avx2(tx_queue, &tx_pkts[nb_tx],
+						     num);
+		nb_tx += ret;
+		nb_pkts -= ret;
+		if (ret < num)
+			break;
+	}
+
+	return nb_tx;
+}
diff --git a/drivers/net/iavf/iavf_rxtx_vec_common.h b/drivers/net/iavf/iavf_rxtx_vec_common.h
index db509d71f..fff0555d2 100644
--- a/drivers/net/iavf/iavf_rxtx_vec_common.h
+++ b/drivers/net/iavf/iavf_rxtx_vec_common.h
@@ -207,4 +207,76 @@ iavf_rxq_vec_setup_default(struct iavf_rx_queue *rxq)
 	rxq->mbuf_initializer = *(uint64_t *)p;
 	return 0;
 }
+
+static inline int
+iavf_rx_vec_queue_default(struct iavf_rx_queue *rxq)
+{
+	if (!rxq)
+		return -1;
+
+	if (!rte_is_power_of_2(rxq->nb_rx_desc))
+		return -1;
+
+	if (rxq->rx_free_thresh < IAVF_VPMD_RX_MAX_BURST)
+		return -1;
+
+	if (rxq->nb_rx_desc % rxq->rx_free_thresh)
+		return -1;
+
+	return 0;
+}
+
+#define IAVF_NO_VECTOR_FLAGS (				 \
+		DEV_TX_OFFLOAD_MULTI_SEGS |		 \
+		DEV_TX_OFFLOAD_VLAN_INSERT |		 \
+		DEV_TX_OFFLOAD_SCTP_CKSUM |		 \
+		DEV_TX_OFFLOAD_UDP_CKSUM |		 \
+		DEV_TX_OFFLOAD_TCP_CKSUM)
+
+static inline int
+iavf_tx_vec_queue_default(struct iavf_tx_queue *txq)
+{
+	if (!txq)
+		return -1;
+
+	if (txq->offloads & IAVF_NO_VECTOR_FLAGS)
+		return -1;
+
+	if (txq->rs_thresh < IAVF_VPMD_TX_MAX_BURST ||
+	    txq->rs_thresh > IAVF_VPMD_TX_MAX_FREE_BUF)
+		return -1;
+
+	return 0;
+}
+
+static inline int
+iavf_rx_vec_dev_check_default(struct rte_eth_dev *dev)
+{
+	int i;
+	struct iavf_rx_queue *rxq;
+
+	for (i = 0; i < dev->data->nb_rx_queues; i++) {
+		rxq = dev->data->rx_queues[i];
+		if (iavf_rx_vec_queue_default(rxq))
+			return -1;
+	}
+
+	return 0;
+}
+
+static inline int
+iavf_tx_vec_dev_check_default(struct rte_eth_dev *dev)
+{
+	int i;
+	struct iavf_tx_queue *txq;
+
+	for (i = 0; i < dev->data->nb_tx_queues; i++) {
+		txq = dev->data->tx_queues[i];
+		if (iavf_tx_vec_queue_default(txq))
+			return -1;
+	}
+
+	return 0;
+}
+
 #endif
diff --git a/drivers/net/iavf/iavf_rxtx_vec_sse.c b/drivers/net/iavf/iavf_rxtx_vec_sse.c
index cc71f23a5..2b16dc1b5 100644
--- a/drivers/net/iavf/iavf_rxtx_vec_sse.c
+++ b/drivers/net/iavf/iavf_rxtx_vec_sse.c
@@ -622,6 +622,27 @@ iavf_xmit_fixed_burst_vec(void *tx_queue, struct rte_mbuf **tx_pkts,
 	return nb_pkts;
 }
 
+uint16_t
+iavf_xmit_pkts_vec(void *tx_queue, struct rte_mbuf **tx_pkts,
+		   uint16_t nb_pkts)
+{
+	uint16_t nb_tx = 0;
+	struct iavf_tx_queue *txq = (struct iavf_tx_queue *)tx_queue;
+
+	while (nb_pkts) {
+		uint16_t ret, num;
+
+		num = (uint16_t)RTE_MIN(nb_pkts, txq->rs_thresh);
+		ret = iavf_xmit_fixed_burst_vec(tx_queue, &tx_pkts[nb_tx], num);
+		nb_tx += ret;
+		nb_pkts -= ret;
+		if (ret < num)
+			break;
+	}
+
+	return nb_tx;
+}
+
 static void __attribute__((cold))
 iavf_rx_queue_release_mbufs_sse(struct iavf_rx_queue *rxq)
 {
@@ -655,3 +676,15 @@ iavf_rxq_vec_setup(struct iavf_rx_queue *rxq)
 	rxq->ops = &sse_vec_rxq_ops;
 	return iavf_rxq_vec_setup_default(rxq);
 }
+
+int __attribute__((cold))
+iavf_rx_vec_dev_check(struct rte_eth_dev *dev)
+{
+	return iavf_rx_vec_dev_check_default(dev);
+}
+
+int __attribute__((cold))
+iavf_tx_vec_dev_check(struct rte_eth_dev *dev)
+{
+	return iavf_tx_vec_dev_check_default(dev);
+}
diff --git a/drivers/net/iavf/meson.build b/drivers/net/iavf/meson.build
index e5a2f5553..c2401af86 100644
--- a/drivers/net/iavf/meson.build
+++ b/drivers/net/iavf/meson.build
@@ -17,4 +17,21 @@ sources = files(
 if arch_subdir == 'x86'
 	dpdk_conf.set('RTE_LIBRTE_IAVF_INC_VECTOR', 1)
 	sources += files('iavf_rxtx_vec_sse.c')
+
+	# compile AVX2 version if either:
+	# a. we have AVX supported in minimum instruction set baseline
+	# b. it's not minimum instruction set, but supported by compiler
+	if dpdk_conf.has('RTE_MACHINE_CPUFLAG_AVX2')
+		cflags += ['-DCC_AVX2_SUPPORT']
+		sources += files('iavf_rxtx_vec_avx2.c')
+	elif cc.has_argument('-mavx2')
+		cflags += ['-DCC_AVX2_SUPPORT']
+		iavf_avx2_lib = static_library('iavf_avx2_lib',
+				'iavf_rxtx_vec_avx2.c',
+				dependencies: [static_rte_ethdev,
+					static_rte_kvargs, static_rte_hash],
+				include_directories: includes,
+				c_args: [cflags, '-mavx2'])
+		objs += iavf_avx2_lib.extract_objects('iavf_rxtx_vec_avx2.c')
+	endif
 endif
-- 
2.17.1


^ permalink raw reply	[flat|nested] 11+ messages in thread

* Re: [dpdk-dev] [PATCH v3] net/iavf: enable AVX2 for iavf
  2019-09-19  8:44 ` [dpdk-dev] [PATCH v3] " Leyi Rong
@ 2019-09-19 12:55   ` Zhang, Qi Z
  0 siblings, 0 replies; 11+ messages in thread
From: Zhang, Qi Z @ 2019-09-19 12:55 UTC (permalink / raw)
  To: Rong, Leyi, Lu, Wenzhuo, Ye, Xiaolong; +Cc: dev

minor capture inline

> -----Original Message-----
> From: Rong, Leyi
> Sent: Thursday, September 19, 2019 4:45 PM
> To: Lu, Wenzhuo <wenzhuo.lu@intel.com>; Zhang, Qi Z
> <qi.z.zhang@intel.com>; Ye, Xiaolong <xiaolong.ye@intel.com>
> Cc: dev@dpdk.org; Rong, Leyi <leyi.rong@intel.com>
> Subject: [PATCH v3] net/iavf: enable AVX2 for iavf
> 
> This patch enables AVX data path for iavf PMD.
> 
> Signed-off-by: Leyi Rong <leyi.rong@intel.com>
> 
> ---
> v3:
> - Adds release note.
> 
> v2:
> - fix build error on aarch64.
> ---

<...>
> diff --git a/drivers/net/iavf/iavf_rxtx_vec_common.h
> b/drivers/net/iavf/iavf_rxtx_vec_common.h
> index db509d71f..fff0555d2 100644
> --- a/drivers/net/iavf/iavf_rxtx_vec_common.h
> +++ b/drivers/net/iavf/iavf_rxtx_vec_common.h
> @@ -207,4 +207,76 @@ iavf_rxq_vec_setup_default(struct iavf_rx_queue
> *rxq)
>  	rxq->mbuf_initializer = *(uint64_t *)p;
>  	return 0;
>  }
> +
> +static inline int
> +iavf_rx_vec_queue_default(struct iavf_rx_queue *rxq) {
> +	if (!rxq)
> +		return -1;
> +
> +	if (!rte_is_power_of_2(rxq->nb_rx_desc))
> +		return -1;
> +
> +	if (rxq->rx_free_thresh < IAVF_VPMD_RX_MAX_BURST)
> +		return -1;
> +
> +	if (rxq->nb_rx_desc % rxq->rx_free_thresh)
> +		return -1;
> +
> +	return 0;
> +}
> +
> +#define IAVF_NO_VECTOR_FLAGS (				 \
> +		DEV_TX_OFFLOAD_MULTI_SEGS |		 \
> +		DEV_TX_OFFLOAD_VLAN_INSERT |		 \
> +		DEV_TX_OFFLOAD_SCTP_CKSUM |		 \
> +		DEV_TX_OFFLOAD_UDP_CKSUM |		 \
> +		DEV_TX_OFFLOAD_TCP_CKSUM)

This macro looks redundant, it is already defined in ice_rxtx.h

<...>


^ permalink raw reply	[flat|nested] 11+ messages in thread

* [dpdk-dev] [PATCH v4] net/iavf: enable AVX2 for iavf
  2019-09-04 10:06 [dpdk-dev] [PATCH] net/iavf: enable AVX2 for iavf Leyi Rong
                   ` (2 preceding siblings ...)
  2019-09-19  8:44 ` [dpdk-dev] [PATCH v3] " Leyi Rong
@ 2019-09-23  6:02 ` Leyi Rong
  2019-09-23  6:49   ` Zhang, Qi Z
                     ` (2 more replies)
  2019-10-16  8:14 ` [dpdk-dev] [PATCH v5] " Leyi Rong
  4 siblings, 3 replies; 11+ messages in thread
From: Leyi Rong @ 2019-09-23  6:02 UTC (permalink / raw)
  To: wenzhuo.lu, qi.z.zhang, xiaolong.ye; +Cc: dev, Leyi Rong

This patch enables AVX data path for iavf PMD.

Signed-off-by: Leyi Rong <leyi.rong@intel.com>

---
v4:
- Romoves redundant IAVF_NO_VECTOR_FLAGS define.

v3:
- Adds release note.

v2:
- Fixes build error on aarch64.
---
 doc/guides/rel_notes/release_19_11.rst  |   3 +
 drivers/net/iavf/Makefile               |  21 +
 drivers/net/iavf/iavf_rxtx.c            |  99 +--
 drivers/net/iavf/iavf_rxtx.h            |  12 +
 drivers/net/iavf/iavf_rxtx_vec_avx2.c   | 867 ++++++++++++++++++++++++
 drivers/net/iavf/iavf_rxtx_vec_common.h |  65 ++
 drivers/net/iavf/iavf_rxtx_vec_sse.c    |  33 +
 drivers/net/iavf/meson.build            |  17 +
 8 files changed, 1071 insertions(+), 46 deletions(-)
 create mode 100644 drivers/net/iavf/iavf_rxtx_vec_avx2.c

diff --git a/doc/guides/rel_notes/release_19_11.rst b/doc/guides/rel_notes/release_19_11.rst
index 8490d897c..e4e29e219 100644
--- a/doc/guides/rel_notes/release_19_11.rst
+++ b/doc/guides/rel_notes/release_19_11.rst
@@ -56,6 +56,9 @@ New Features
      Also, make sure to start the actual text at the margin.
      =========================================================
 
+* **Updated iavf PMD.**
+
+  Enabled AVX2 data path for iavf PMD.
 
 Removed Items
 -------------
diff --git a/drivers/net/iavf/Makefile b/drivers/net/iavf/Makefile
index cd74e14ab..cf0ed99f0 100644
--- a/drivers/net/iavf/Makefile
+++ b/drivers/net/iavf/Makefile
@@ -48,4 +48,25 @@ ifeq ($(CONFIG_RTE_ARCH_X86), y)
 SRCS-$(CONFIG_RTE_LIBRTE_IAVF_INC_VECTOR) += iavf_rxtx_vec_sse.c
 endif
 
+ifeq ($(CONFIG_RTE_LIBRTE_IAVF_INC_VECTOR), y)
+	ifeq ($(findstring RTE_MACHINE_CPUFLAG_AVX2,$(CFLAGS)),RTE_MACHINE_CPUFLAG_AVX2)
+		CC_AVX2_SUPPORT=1
+	else
+		CC_AVX2_SUPPORT=\
+		$(shell $(CC) -march=core-avx2 -dM -E - </dev/null 2>&1 | \
+		grep -q AVX2 && echo 1)
+		ifeq ($(CC_AVX2_SUPPORT), 1)
+			ifeq ($(CONFIG_RTE_TOOLCHAIN_ICC),y)
+				CFLAGS_iavf_rxtx_vec_avx2.o += -march=core-avx2
+			else
+				CFLAGS_iavf_rxtx_vec_avx2.o += -mavx2
+			endif
+		endif
+	endif
+endif
+
+ifeq ($(CC_AVX2_SUPPORT), 1)
+	SRCS-$(CONFIG_RTE_LIBRTE_IAVF_PMD) += iavf_rxtx_vec_avx2.c
+endif
+
 include $(RTE_SDK)/mk/rte.lib.mk
diff --git a/drivers/net/iavf/iavf_rxtx.c b/drivers/net/iavf/iavf_rxtx.c
index 22d7bb612..e36f035c5 100644
--- a/drivers/net/iavf/iavf_rxtx.c
+++ b/drivers/net/iavf/iavf_rxtx.c
@@ -1666,27 +1666,6 @@ iavf_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts, uint16_t nb_pkts)
 	return nb_tx;
 }
 
-static uint16_t
-iavf_xmit_pkts_vec(void *tx_queue, struct rte_mbuf **tx_pkts,
-		  uint16_t nb_pkts)
-{
-	uint16_t nb_tx = 0;
-	struct iavf_tx_queue *txq = (struct iavf_tx_queue *)tx_queue;
-
-	while (nb_pkts) {
-		uint16_t ret, num;
-
-		num = (uint16_t)RTE_MIN(nb_pkts, txq->rs_thresh);
-		ret = iavf_xmit_fixed_burst_vec(tx_queue, &tx_pkts[nb_tx], num);
-		nb_tx += ret;
-		nb_pkts -= ret;
-		if (ret < num)
-			break;
-	}
-
-	return nb_tx;
-}
-
 /* TX prep functions */
 uint16_t
 iavf_prep_pkts(__rte_unused void *tx_queue, struct rte_mbuf **tx_pkts,
@@ -1741,26 +1720,43 @@ iavf_set_rx_function(struct rte_eth_dev *dev)
 {
 	struct iavf_adapter *adapter =
 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
+#ifdef RTE_ARCH_X86
 	struct iavf_rx_queue *rxq;
 	int i;
+	bool use_avx2 = false;
 
-	if (adapter->rx_vec_allowed) {
-		if (dev->data->scattered_rx) {
-			PMD_DRV_LOG(DEBUG, "Using Vector Scattered Rx callback"
-				    " (port=%d).", dev->data->port_id);
-			dev->rx_pkt_burst = iavf_recv_scattered_pkts_vec;
-		} else {
-			PMD_DRV_LOG(DEBUG, "Using Vector Rx callback"
-				    " (port=%d).", dev->data->port_id);
-			dev->rx_pkt_burst = iavf_recv_pkts_vec;
-		}
+	if (!iavf_rx_vec_dev_check(dev)) {
 		for (i = 0; i < dev->data->nb_rx_queues; i++) {
 			rxq = dev->data->rx_queues[i];
-			if (!rxq)
-				continue;
-			iavf_rxq_vec_setup(rxq);
+			(void)iavf_rxq_vec_setup(rxq);
 		}
-	} else if (dev->data->scattered_rx) {
+
+		if (rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1 ||
+		    rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX512F) == 1)
+			use_avx2 = true;
+
+		if (dev->data->scattered_rx) {
+			PMD_DRV_LOG(DEBUG,
+				    "Using %sVector Scattered Rx (port %d).",
+				    use_avx2 ? "avx2 " : "",
+				    dev->data->port_id);
+			dev->rx_pkt_burst = use_avx2 ?
+					    iavf_recv_scattered_pkts_vec_avx2 :
+					    iavf_recv_scattered_pkts_vec;
+		} else {
+			PMD_DRV_LOG(DEBUG, "Using %sVector Rx (port %d).",
+				    use_avx2 ? "avx2 " : "",
+				    dev->data->port_id);
+			dev->rx_pkt_burst = use_avx2 ?
+					    iavf_recv_pkts_vec_avx2 :
+					    iavf_recv_pkts_vec;
+		}
+
+		return;
+	}
+#endif
+
+	if (dev->data->scattered_rx) {
 		PMD_DRV_LOG(DEBUG, "Using a Scattered Rx callback (port=%d).",
 			    dev->data->port_id);
 		dev->rx_pkt_burst = iavf_recv_scattered_pkts;
@@ -1779,28 +1775,39 @@ iavf_set_rx_function(struct rte_eth_dev *dev)
 void
 iavf_set_tx_function(struct rte_eth_dev *dev)
 {
-	struct iavf_adapter *adapter =
-		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
+#ifdef RTE_ARCH_X86
 	struct iavf_tx_queue *txq;
 	int i;
+	bool use_avx2 = false;
 
-	if (adapter->tx_vec_allowed) {
-		PMD_DRV_LOG(DEBUG, "Using Vector Tx callback (port=%d).",
-			    dev->data->port_id);
-		dev->tx_pkt_burst = iavf_xmit_pkts_vec;
-		dev->tx_pkt_prepare = NULL;
+	if (!iavf_tx_vec_dev_check(dev)) {
 		for (i = 0; i < dev->data->nb_tx_queues; i++) {
 			txq = dev->data->tx_queues[i];
 			if (!txq)
 				continue;
 			iavf_txq_vec_setup(txq);
 		}
-	} else {
-		PMD_DRV_LOG(DEBUG, "Using Basic Tx callback (port=%d).",
+
+		if (rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1 ||
+		    rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX512F) == 1)
+			use_avx2 = true;
+
+		PMD_DRV_LOG(DEBUG, "Using %sVector Tx (port %d).",
+			    use_avx2 ? "avx2 " : "",
 			    dev->data->port_id);
-		dev->tx_pkt_burst = iavf_xmit_pkts;
-		dev->tx_pkt_prepare = iavf_prep_pkts;
+		dev->tx_pkt_burst = use_avx2 ?
+				    iavf_xmit_pkts_vec_avx2 :
+				    iavf_xmit_pkts_vec;
+		dev->tx_pkt_prepare = NULL;
+
+		return;
 	}
+#endif
+
+	PMD_DRV_LOG(DEBUG, "Using Basic Tx callback (port=%d).",
+		    dev->data->port_id);
+	dev->tx_pkt_burst = iavf_xmit_pkts;
+	dev->tx_pkt_prepare = iavf_prep_pkts;
 }
 
 void
diff --git a/drivers/net/iavf/iavf_rxtx.h b/drivers/net/iavf/iavf_rxtx.h
index c86720bda..225a0c4c4 100644
--- a/drivers/net/iavf/iavf_rxtx.h
+++ b/drivers/net/iavf/iavf_rxtx.h
@@ -19,6 +19,7 @@
 /* used for Vector PMD */
 #define IAVF_VPMD_RX_MAX_BURST    32
 #define IAVF_VPMD_TX_MAX_BURST    32
+#define IAVF_RXQ_REARM_THRESH     32
 #define IAVF_VPMD_DESCS_PER_LOOP  4
 #define IAVF_VPMD_TX_MAX_FREE_BUF 64
 
@@ -200,6 +201,17 @@ uint16_t iavf_recv_scattered_pkts_vec(void *rx_queue,
 				     uint16_t nb_pkts);
 uint16_t iavf_xmit_fixed_burst_vec(void *tx_queue, struct rte_mbuf **tx_pkts,
 				  uint16_t nb_pkts);
+uint16_t iavf_recv_pkts_vec_avx2(void *rx_queue, struct rte_mbuf **rx_pkts,
+				 uint16_t nb_pkts);
+uint16_t iavf_recv_scattered_pkts_vec_avx2(void *rx_queue,
+					   struct rte_mbuf **rx_pkts,
+					   uint16_t nb_pkts);
+uint16_t iavf_xmit_pkts_vec(void *tx_queue, struct rte_mbuf **tx_pkts,
+			    uint16_t nb_pkts);
+uint16_t iavf_xmit_pkts_vec_avx2(void *tx_queue, struct rte_mbuf **tx_pkts,
+				 uint16_t nb_pkts);
+int iavf_rx_vec_dev_check(struct rte_eth_dev *dev);
+int iavf_tx_vec_dev_check(struct rte_eth_dev *dev);
 int iavf_rxq_vec_setup(struct iavf_rx_queue *rxq);
 int iavf_txq_vec_setup(struct iavf_tx_queue *txq);
 
diff --git a/drivers/net/iavf/iavf_rxtx_vec_avx2.c b/drivers/net/iavf/iavf_rxtx_vec_avx2.c
new file mode 100644
index 000000000..f0c00be56
--- /dev/null
+++ b/drivers/net/iavf/iavf_rxtx_vec_avx2.c
@@ -0,0 +1,867 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2019 Intel Corporation
+ */
+
+#include "base/iavf_prototype.h"
+#include "iavf_rxtx_vec_common.h"
+
+#include <x86intrin.h>
+
+#ifndef __INTEL_COMPILER
+#pragma GCC diagnostic ignored "-Wcast-qual"
+#endif
+
+static inline void
+iavf_rxq_rearm(struct iavf_rx_queue *rxq)
+{
+	int i;
+	uint16_t rx_id;
+	volatile union iavf_rx_desc *rxdp;
+	struct rte_mbuf **rxp = &rxq->sw_ring[rxq->rxrearm_start];
+
+	rxdp = rxq->rx_ring + rxq->rxrearm_start;
+
+	/* Pull 'n' more MBUFs into the software ring */
+	if (rte_mempool_get_bulk(rxq->mp,
+				 (void *)rxp,
+				 IAVF_RXQ_REARM_THRESH) < 0) {
+		if (rxq->rxrearm_nb + IAVF_RXQ_REARM_THRESH >=
+		    rxq->nb_rx_desc) {
+			__m128i dma_addr0;
+
+			dma_addr0 = _mm_setzero_si128();
+			for (i = 0; i < IAVF_VPMD_DESCS_PER_LOOP; i++) {
+				rxp[i] = &rxq->fake_mbuf;
+				_mm_store_si128((__m128i *)&rxdp[i].read,
+						dma_addr0);
+			}
+		}
+		rte_eth_devices[rxq->port_id].data->rx_mbuf_alloc_failed +=
+			IAVF_RXQ_REARM_THRESH;
+		return;
+	}
+
+#ifndef RTE_LIBRTE_IAVF_16BYTE_RX_DESC
+	struct rte_mbuf *mb0, *mb1;
+	__m128i dma_addr0, dma_addr1;
+	__m128i hdr_room = _mm_set_epi64x(RTE_PKTMBUF_HEADROOM,
+			RTE_PKTMBUF_HEADROOM);
+	/* Initialize the mbufs in vector, process 2 mbufs in one loop */
+	for (i = 0; i < IAVF_RXQ_REARM_THRESH; i += 2, rxp += 2) {
+		__m128i vaddr0, vaddr1;
+
+		mb0 = rxp[0];
+		mb1 = rxp[1];
+
+		/* load buf_addr(lo 64bit) and buf_physaddr(hi 64bit) */
+		RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, buf_physaddr) !=
+				offsetof(struct rte_mbuf, buf_addr) + 8);
+		vaddr0 = _mm_loadu_si128((__m128i *)&mb0->buf_addr);
+		vaddr1 = _mm_loadu_si128((__m128i *)&mb1->buf_addr);
+
+		/* convert pa to dma_addr hdr/data */
+		dma_addr0 = _mm_unpackhi_epi64(vaddr0, vaddr0);
+		dma_addr1 = _mm_unpackhi_epi64(vaddr1, vaddr1);
+
+		/* add headroom to pa values */
+		dma_addr0 = _mm_add_epi64(dma_addr0, hdr_room);
+		dma_addr1 = _mm_add_epi64(dma_addr1, hdr_room);
+
+		/* flush desc with pa dma_addr */
+		_mm_store_si128((__m128i *)&rxdp++->read, dma_addr0);
+		_mm_store_si128((__m128i *)&rxdp++->read, dma_addr1);
+	}
+#else
+	struct rte_mbuf *mb0, *mb1, *mb2, *mb3;
+	__m256i dma_addr0_1, dma_addr2_3;
+	__m256i hdr_room = _mm256_set1_epi64x(RTE_PKTMBUF_HEADROOM);
+	/* Initialize the mbufs in vector, process 4 mbufs in one loop */
+	for (i = 0; i < IAVF_RXQ_REARM_THRESH;
+			i += 4, rxp += 4, rxdp += 4) {
+		__m128i vaddr0, vaddr1, vaddr2, vaddr3;
+		__m256i vaddr0_1, vaddr2_3;
+
+		mb0 = rxp[0];
+		mb1 = rxp[1];
+		mb2 = rxp[2];
+		mb3 = rxp[3];
+
+		/* load buf_addr(lo 64bit) and buf_physaddr(hi 64bit) */
+		RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, buf_physaddr) !=
+				offsetof(struct rte_mbuf, buf_addr) + 8);
+		vaddr0 = _mm_loadu_si128((__m128i *)&mb0->buf_addr);
+		vaddr1 = _mm_loadu_si128((__m128i *)&mb1->buf_addr);
+		vaddr2 = _mm_loadu_si128((__m128i *)&mb2->buf_addr);
+		vaddr3 = _mm_loadu_si128((__m128i *)&mb3->buf_addr);
+
+		/**
+		 * merge 0 & 1, by casting 0 to 256-bit and inserting 1
+		 * into the high lanes. Similarly for 2 & 3
+		 */
+		vaddr0_1 =
+			_mm256_inserti128_si256(_mm256_castsi128_si256(vaddr0),
+						vaddr1, 1);
+		vaddr2_3 =
+			_mm256_inserti128_si256(_mm256_castsi128_si256(vaddr2),
+						vaddr3, 1);
+
+		/* convert pa to dma_addr hdr/data */
+		dma_addr0_1 = _mm256_unpackhi_epi64(vaddr0_1, vaddr0_1);
+		dma_addr2_3 = _mm256_unpackhi_epi64(vaddr2_3, vaddr2_3);
+
+		/* add headroom to pa values */
+		dma_addr0_1 = _mm256_add_epi64(dma_addr0_1, hdr_room);
+		dma_addr2_3 = _mm256_add_epi64(dma_addr2_3, hdr_room);
+
+		/* flush desc with pa dma_addr */
+		_mm256_store_si256((__m256i *)&rxdp->read, dma_addr0_1);
+		_mm256_store_si256((__m256i *)&(rxdp + 2)->read, dma_addr2_3);
+	}
+
+#endif
+
+	rxq->rxrearm_start += IAVF_RXQ_REARM_THRESH;
+	if (rxq->rxrearm_start >= rxq->nb_rx_desc)
+		rxq->rxrearm_start = 0;
+
+	rxq->rxrearm_nb -= IAVF_RXQ_REARM_THRESH;
+
+	rx_id = (uint16_t)((rxq->rxrearm_start == 0) ?
+			     (rxq->nb_rx_desc - 1) : (rxq->rxrearm_start - 1));
+
+	/* Update the tail pointer on the NIC */
+	IAVF_PCI_REG_WRITE(rxq->qrx_tail, rx_id);
+}
+
+#define PKTLEN_SHIFT     10
+
+static inline uint16_t
+_iavf_recv_raw_pkts_vec_avx2(struct iavf_rx_queue *rxq,
+			     struct rte_mbuf **rx_pkts,
+			     uint16_t nb_pkts, uint8_t *split_packet)
+{
+#define IAVF_DESCS_PER_LOOP_AVX 8
+
+	/* const uint32_t *ptype_tbl = rxq->vsi->adapter->ptype_tbl; */
+	static const uint32_t type_table[UINT8_MAX + 1] __rte_cache_aligned = {
+		/* [0] reserved */
+		[1] = RTE_PTYPE_L2_ETHER,
+		/* [2] - [21] reserved */
+		[22] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
+			RTE_PTYPE_L4_FRAG,
+		[23] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
+			RTE_PTYPE_L4_NONFRAG,
+		[24] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
+			RTE_PTYPE_L4_UDP,
+		/* [25] reserved */
+		[26] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
+			RTE_PTYPE_L4_TCP,
+		[27] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
+			RTE_PTYPE_L4_SCTP,
+		[28] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
+			RTE_PTYPE_L4_ICMP,
+		/* All others reserved */
+	};
+	const __m256i mbuf_init = _mm256_set_epi64x(0, 0,
+			0, rxq->mbuf_initializer);
+	/* struct iavf_rx_entry *sw_ring = &rxq->sw_ring[rxq->rx_tail]; */
+	struct rte_mbuf **sw_ring = &rxq->sw_ring[rxq->rx_tail];
+	volatile union iavf_rx_desc *rxdp = rxq->rx_ring + rxq->rx_tail;
+	const int avx_aligned = ((rxq->rx_tail & 1) == 0);
+
+	rte_prefetch0(rxdp);
+
+	/* nb_pkts has to be floor-aligned to IAVF_DESCS_PER_LOOP_AVX */
+	nb_pkts = RTE_ALIGN_FLOOR(nb_pkts, IAVF_DESCS_PER_LOOP_AVX);
+
+	/* See if we need to rearm the RX queue - gives the prefetch a bit
+	 * of time to act
+	 */
+	if (rxq->rxrearm_nb > IAVF_RXQ_REARM_THRESH)
+		iavf_rxq_rearm(rxq);
+
+	/* Before we start moving massive data around, check to see if
+	 * there is actually a packet available
+	 */
+	if (!(rxdp->wb.qword1.status_error_len &
+			rte_cpu_to_le_32(1 << IAVF_RX_DESC_STATUS_DD_SHIFT)))
+		return 0;
+
+	/* constants used in processing loop */
+	const __m256i crc_adjust =
+		_mm256_set_epi16
+			(/* first descriptor */
+			 0, 0, 0,       /* ignore non-length fields */
+			 -rxq->crc_len, /* sub crc on data_len */
+			 0,             /* ignore high-16bits of pkt_len */
+			 -rxq->crc_len, /* sub crc on pkt_len */
+			 0, 0,          /* ignore pkt_type field */
+			 /* second descriptor */
+			 0, 0, 0,       /* ignore non-length fields */
+			 -rxq->crc_len, /* sub crc on data_len */
+			 0,             /* ignore high-16bits of pkt_len */
+			 -rxq->crc_len, /* sub crc on pkt_len */
+			 0, 0           /* ignore pkt_type field */
+			);
+
+	/* 8 packets DD mask, LSB in each 32-bit value */
+	const __m256i dd_check = _mm256_set1_epi32(1);
+
+	/* 8 packets EOP mask, second-LSB in each 32-bit value */
+	const __m256i eop_check = _mm256_slli_epi32(dd_check,
+			IAVF_RX_DESC_STATUS_EOF_SHIFT);
+
+	/* mask to shuffle from desc. to mbuf (2 descriptors)*/
+	const __m256i shuf_msk =
+		_mm256_set_epi8
+			(/* first descriptor */
+			 7, 6, 5, 4,  /* octet 4~7, 32bits rss */
+			 3, 2,        /* octet 2~3, low 16 bits vlan_macip */
+			 15, 14,      /* octet 15~14, 16 bits data_len */
+			 0xFF, 0xFF,  /* skip high 16 bits pkt_len, zero out */
+			 15, 14,      /* octet 15~14, low 16 bits pkt_len */
+			 0xFF, 0xFF,  /* pkt_type set as unknown */
+			 0xFF, 0xFF,  /*pkt_type set as unknown */
+			 /* second descriptor */
+			 7, 6, 5, 4,  /* octet 4~7, 32bits rss */
+			 3, 2,        /* octet 2~3, low 16 bits vlan_macip */
+			 15, 14,      /* octet 15~14, 16 bits data_len */
+			 0xFF, 0xFF,  /* skip high 16 bits pkt_len, zero out */
+			 15, 14,      /* octet 15~14, low 16 bits pkt_len */
+			 0xFF, 0xFF,  /* pkt_type set as unknown */
+			 0xFF, 0xFF   /*pkt_type set as unknown */
+			);
+	/**
+	 * compile-time check the above crc and shuffle layout is correct.
+	 * NOTE: the first field (lowest address) is given last in set_epi
+	 * calls above.
+	 */
+	RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, pkt_len) !=
+			offsetof(struct rte_mbuf, rx_descriptor_fields1) + 4);
+	RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, data_len) !=
+			offsetof(struct rte_mbuf, rx_descriptor_fields1) + 8);
+	RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, vlan_tci) !=
+			offsetof(struct rte_mbuf, rx_descriptor_fields1) + 10);
+	RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, hash) !=
+			offsetof(struct rte_mbuf, rx_descriptor_fields1) + 12);
+
+	/* Status/Error flag masks */
+	/**
+	 * mask everything except RSS, flow director and VLAN flags
+	 * bit2 is for VLAN tag, bit11 for flow director indication
+	 * bit13:12 for RSS indication. Bits 3-5 of error
+	 * field (bits 22-24) are for IP/L4 checksum errors
+	 */
+	const __m256i flags_mask =
+		 _mm256_set1_epi32((1 << 2) | (1 << 11) |
+				   (3 << 12) | (7 << 22));
+	/**
+	 * data to be shuffled by result of flag mask. If VLAN bit is set,
+	 * (bit 2), then position 4 in this array will be used in the
+	 * destination
+	 */
+	const __m256i vlan_flags_shuf =
+		_mm256_set_epi32(0, 0, PKT_RX_VLAN | PKT_RX_VLAN_STRIPPED, 0,
+				 0, 0, PKT_RX_VLAN | PKT_RX_VLAN_STRIPPED, 0);
+	/**
+	 * data to be shuffled by result of flag mask, shifted down 11.
+	 * If RSS/FDIR bits are set, shuffle moves appropriate flags in
+	 * place.
+	 */
+	const __m256i rss_flags_shuf =
+		_mm256_set_epi8(0, 0, 0, 0, 0, 0, 0, 0,
+				PKT_RX_RSS_HASH | PKT_RX_FDIR, PKT_RX_RSS_HASH,
+				0, 0, 0, 0, PKT_RX_FDIR, 0,/* end up 128-bits */
+				0, 0, 0, 0, 0, 0, 0, 0,
+				PKT_RX_RSS_HASH | PKT_RX_FDIR, PKT_RX_RSS_HASH,
+				0, 0, 0, 0, PKT_RX_FDIR, 0);
+
+	/**
+	 * data to be shuffled by the result of the flags mask shifted by 22
+	 * bits.  This gives use the l3_l4 flags.
+	 */
+	const __m256i l3_l4_flags_shuf = _mm256_set_epi8(0, 0, 0, 0, 0, 0, 0, 0,
+			/* shift right 1 bit to make sure it not exceed 255 */
+			(PKT_RX_EIP_CKSUM_BAD | PKT_RX_L4_CKSUM_BAD |
+			 PKT_RX_IP_CKSUM_BAD) >> 1,
+			(PKT_RX_IP_CKSUM_GOOD | PKT_RX_EIP_CKSUM_BAD |
+			 PKT_RX_L4_CKSUM_BAD) >> 1,
+			(PKT_RX_EIP_CKSUM_BAD | PKT_RX_IP_CKSUM_BAD) >> 1,
+			(PKT_RX_IP_CKSUM_GOOD | PKT_RX_EIP_CKSUM_BAD) >> 1,
+			(PKT_RX_L4_CKSUM_BAD | PKT_RX_IP_CKSUM_BAD) >> 1,
+			(PKT_RX_IP_CKSUM_GOOD | PKT_RX_L4_CKSUM_BAD) >> 1,
+			PKT_RX_IP_CKSUM_BAD >> 1,
+			(PKT_RX_IP_CKSUM_GOOD | PKT_RX_L4_CKSUM_GOOD) >> 1,
+			/* second 128-bits */
+			0, 0, 0, 0, 0, 0, 0, 0,
+			(PKT_RX_EIP_CKSUM_BAD | PKT_RX_L4_CKSUM_BAD |
+			 PKT_RX_IP_CKSUM_BAD) >> 1,
+			(PKT_RX_IP_CKSUM_GOOD | PKT_RX_EIP_CKSUM_BAD |
+			 PKT_RX_L4_CKSUM_BAD) >> 1,
+			(PKT_RX_EIP_CKSUM_BAD | PKT_RX_IP_CKSUM_BAD) >> 1,
+			(PKT_RX_IP_CKSUM_GOOD | PKT_RX_EIP_CKSUM_BAD) >> 1,
+			(PKT_RX_L4_CKSUM_BAD | PKT_RX_IP_CKSUM_BAD) >> 1,
+			(PKT_RX_IP_CKSUM_GOOD | PKT_RX_L4_CKSUM_BAD) >> 1,
+			PKT_RX_IP_CKSUM_BAD >> 1,
+			(PKT_RX_IP_CKSUM_GOOD | PKT_RX_L4_CKSUM_GOOD) >> 1);
+
+	const __m256i cksum_mask =
+		 _mm256_set1_epi32(PKT_RX_IP_CKSUM_GOOD | PKT_RX_IP_CKSUM_BAD |
+				   PKT_RX_L4_CKSUM_GOOD | PKT_RX_L4_CKSUM_BAD |
+				   PKT_RX_EIP_CKSUM_BAD);
+
+	RTE_SET_USED(avx_aligned); /* for 32B descriptors we don't use this */
+
+	uint16_t i, received;
+
+	for (i = 0, received = 0; i < nb_pkts;
+	     i += IAVF_DESCS_PER_LOOP_AVX,
+	     rxdp += IAVF_DESCS_PER_LOOP_AVX) {
+		/* step 1, copy over 8 mbuf pointers to rx_pkts array */
+		_mm256_storeu_si256((void *)&rx_pkts[i],
+				    _mm256_loadu_si256((void *)&sw_ring[i]));
+#ifdef RTE_ARCH_X86_64
+		_mm256_storeu_si256
+			((void *)&rx_pkts[i + 4],
+			 _mm256_loadu_si256((void *)&sw_ring[i + 4]));
+#endif
+
+		__m256i raw_desc0_1, raw_desc2_3, raw_desc4_5, raw_desc6_7;
+#ifdef RTE_LIBRTE_IAVF_16BYTE_RX_DESC
+		/* for AVX we need alignment otherwise loads are not atomic */
+		if (avx_aligned) {
+			/* load in descriptors, 2 at a time, in reverse order */
+			raw_desc6_7 = _mm256_load_si256((void *)(rxdp + 6));
+			rte_compiler_barrier();
+			raw_desc4_5 = _mm256_load_si256((void *)(rxdp + 4));
+			rte_compiler_barrier();
+			raw_desc2_3 = _mm256_load_si256((void *)(rxdp + 2));
+			rte_compiler_barrier();
+			raw_desc0_1 = _mm256_load_si256((void *)(rxdp + 0));
+		} else
+#endif
+		{
+			const __m128i raw_desc7 =
+				_mm_load_si128((void *)(rxdp + 7));
+			rte_compiler_barrier();
+			const __m128i raw_desc6 =
+				_mm_load_si128((void *)(rxdp + 6));
+			rte_compiler_barrier();
+			const __m128i raw_desc5 =
+				_mm_load_si128((void *)(rxdp + 5));
+			rte_compiler_barrier();
+			const __m128i raw_desc4 =
+				_mm_load_si128((void *)(rxdp + 4));
+			rte_compiler_barrier();
+			const __m128i raw_desc3 =
+				_mm_load_si128((void *)(rxdp + 3));
+			rte_compiler_barrier();
+			const __m128i raw_desc2 =
+				_mm_load_si128((void *)(rxdp + 2));
+			rte_compiler_barrier();
+			const __m128i raw_desc1 =
+				_mm_load_si128((void *)(rxdp + 1));
+			rte_compiler_barrier();
+			const __m128i raw_desc0 =
+				_mm_load_si128((void *)(rxdp + 0));
+
+			raw_desc6_7 =
+				_mm256_inserti128_si256
+					(_mm256_castsi128_si256(raw_desc6),
+					 raw_desc7, 1);
+			raw_desc4_5 =
+				_mm256_inserti128_si256
+					(_mm256_castsi128_si256(raw_desc4),
+					 raw_desc5, 1);
+			raw_desc2_3 =
+				_mm256_inserti128_si256
+					(_mm256_castsi128_si256(raw_desc2),
+					 raw_desc3, 1);
+			raw_desc0_1 =
+				_mm256_inserti128_si256
+					(_mm256_castsi128_si256(raw_desc0),
+					 raw_desc1, 1);
+		}
+
+		if (split_packet) {
+			int j;
+
+			for (j = 0; j < IAVF_DESCS_PER_LOOP_AVX; j++)
+				rte_mbuf_prefetch_part2(rx_pkts[i + j]);
+		}
+
+		/**
+		 * convert descriptors 4-7 into mbufs, adjusting length and
+		 * re-arranging fields. Then write into the mbuf
+		 */
+		const __m256i len6_7 = _mm256_slli_epi32(raw_desc6_7,
+							 PKTLEN_SHIFT);
+		const __m256i len4_5 = _mm256_slli_epi32(raw_desc4_5,
+							 PKTLEN_SHIFT);
+		const __m256i desc6_7 = _mm256_blend_epi16(raw_desc6_7,
+							   len6_7, 0x80);
+		const __m256i desc4_5 = _mm256_blend_epi16(raw_desc4_5,
+							   len4_5, 0x80);
+		__m256i mb6_7 = _mm256_shuffle_epi8(desc6_7, shuf_msk);
+		__m256i mb4_5 = _mm256_shuffle_epi8(desc4_5, shuf_msk);
+
+		mb6_7 = _mm256_add_epi16(mb6_7, crc_adjust);
+		mb4_5 = _mm256_add_epi16(mb4_5, crc_adjust);
+		/**
+		 * to get packet types, shift 64-bit values down 30 bits
+		 * and so ptype is in lower 8-bits in each
+		 */
+		const __m256i ptypes6_7 = _mm256_srli_epi64(desc6_7, 30);
+		const __m256i ptypes4_5 = _mm256_srli_epi64(desc4_5, 30);
+		const uint8_t ptype7 = _mm256_extract_epi8(ptypes6_7, 24);
+		const uint8_t ptype6 = _mm256_extract_epi8(ptypes6_7, 8);
+		const uint8_t ptype5 = _mm256_extract_epi8(ptypes4_5, 24);
+		const uint8_t ptype4 = _mm256_extract_epi8(ptypes4_5, 8);
+
+		mb6_7 = _mm256_insert_epi32(mb6_7, type_table[ptype7], 4);
+		mb6_7 = _mm256_insert_epi32(mb6_7, type_table[ptype6], 0);
+		mb4_5 = _mm256_insert_epi32(mb4_5, type_table[ptype5], 4);
+		mb4_5 = _mm256_insert_epi32(mb4_5, type_table[ptype4], 0);
+		/* merge the status bits into one register */
+		const __m256i status4_7 = _mm256_unpackhi_epi32(desc6_7,
+				desc4_5);
+
+		/**
+		 * convert descriptors 0-3 into mbufs, adjusting length and
+		 * re-arranging fields. Then write into the mbuf
+		 */
+		const __m256i len2_3 = _mm256_slli_epi32(raw_desc2_3,
+							 PKTLEN_SHIFT);
+		const __m256i len0_1 = _mm256_slli_epi32(raw_desc0_1,
+							 PKTLEN_SHIFT);
+		const __m256i desc2_3 = _mm256_blend_epi16(raw_desc2_3,
+							   len2_3, 0x80);
+		const __m256i desc0_1 = _mm256_blend_epi16(raw_desc0_1,
+							   len0_1, 0x80);
+		__m256i mb2_3 = _mm256_shuffle_epi8(desc2_3, shuf_msk);
+		__m256i mb0_1 = _mm256_shuffle_epi8(desc0_1, shuf_msk);
+
+		mb2_3 = _mm256_add_epi16(mb2_3, crc_adjust);
+		mb0_1 = _mm256_add_epi16(mb0_1, crc_adjust);
+		/* get the packet types */
+		const __m256i ptypes2_3 = _mm256_srli_epi64(desc2_3, 30);
+		const __m256i ptypes0_1 = _mm256_srli_epi64(desc0_1, 30);
+		const uint8_t ptype3 = _mm256_extract_epi8(ptypes2_3, 24);
+		const uint8_t ptype2 = _mm256_extract_epi8(ptypes2_3, 8);
+		const uint8_t ptype1 = _mm256_extract_epi8(ptypes0_1, 24);
+		const uint8_t ptype0 = _mm256_extract_epi8(ptypes0_1, 8);
+
+		mb2_3 = _mm256_insert_epi32(mb2_3, type_table[ptype3], 4);
+		mb2_3 = _mm256_insert_epi32(mb2_3, type_table[ptype2], 0);
+		mb0_1 = _mm256_insert_epi32(mb0_1, type_table[ptype1], 4);
+		mb0_1 = _mm256_insert_epi32(mb0_1, type_table[ptype0], 0);
+		/* merge the status bits into one register */
+		const __m256i status0_3 = _mm256_unpackhi_epi32(desc2_3,
+								desc0_1);
+
+		/**
+		 * take the two sets of status bits and merge to one
+		 * After merge, the packets status flags are in the
+		 * order (hi->lo): [1, 3, 5, 7, 0, 2, 4, 6]
+		 */
+		__m256i status0_7 = _mm256_unpacklo_epi64(status4_7,
+							  status0_3);
+
+		/* now do flag manipulation */
+
+		/* get only flag/error bits we want */
+		const __m256i flag_bits =
+			_mm256_and_si256(status0_7, flags_mask);
+		/* set vlan and rss flags */
+		const __m256i vlan_flags =
+			_mm256_shuffle_epi8(vlan_flags_shuf, flag_bits);
+		const __m256i rss_flags =
+			_mm256_shuffle_epi8(rss_flags_shuf,
+					    _mm256_srli_epi32(flag_bits, 11));
+		/**
+		 * l3_l4_error flags, shuffle, then shift to correct adjustment
+		 * of flags in flags_shuf, and finally mask out extra bits
+		 */
+		__m256i l3_l4_flags = _mm256_shuffle_epi8(l3_l4_flags_shuf,
+				_mm256_srli_epi32(flag_bits, 22));
+		l3_l4_flags = _mm256_slli_epi32(l3_l4_flags, 1);
+		l3_l4_flags = _mm256_and_si256(l3_l4_flags, cksum_mask);
+
+		/* merge flags */
+		const __m256i mbuf_flags = _mm256_or_si256(l3_l4_flags,
+				_mm256_or_si256(rss_flags, vlan_flags));
+		/**
+		 * At this point, we have the 8 sets of flags in the low 16-bits
+		 * of each 32-bit value in vlan0.
+		 * We want to extract these, and merge them with the mbuf init
+		 * data so we can do a single write to the mbuf to set the flags
+		 * and all the other initialization fields. Extracting the
+		 * appropriate flags means that we have to do a shift and blend
+		 * for each mbuf before we do the write. However, we can also
+		 * add in the previously computed rx_descriptor fields to
+		 * make a single 256-bit write per mbuf
+		 */
+		/* check the structure matches expectations */
+		RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, ol_flags) !=
+				 offsetof(struct rte_mbuf, rearm_data) + 8);
+		RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, rearm_data) !=
+				 RTE_ALIGN(offsetof(struct rte_mbuf,
+						    rearm_data),
+					   16));
+		/* build up data and do writes */
+		__m256i rearm0, rearm1, rearm2, rearm3, rearm4, rearm5,
+			rearm6, rearm7;
+		rearm6 = _mm256_blend_epi32(mbuf_init,
+					    _mm256_slli_si256(mbuf_flags, 8),
+					    0x04);
+		rearm4 = _mm256_blend_epi32(mbuf_init,
+					    _mm256_slli_si256(mbuf_flags, 4),
+					    0x04);
+		rearm2 = _mm256_blend_epi32(mbuf_init, mbuf_flags, 0x04);
+		rearm0 = _mm256_blend_epi32(mbuf_init,
+					    _mm256_srli_si256(mbuf_flags, 4),
+					    0x04);
+		/* permute to add in the rx_descriptor e.g. rss fields */
+		rearm6 = _mm256_permute2f128_si256(rearm6, mb6_7, 0x20);
+		rearm4 = _mm256_permute2f128_si256(rearm4, mb4_5, 0x20);
+		rearm2 = _mm256_permute2f128_si256(rearm2, mb2_3, 0x20);
+		rearm0 = _mm256_permute2f128_si256(rearm0, mb0_1, 0x20);
+		/* write to mbuf */
+		_mm256_storeu_si256((__m256i *)&rx_pkts[i + 6]->rearm_data,
+				    rearm6);
+		_mm256_storeu_si256((__m256i *)&rx_pkts[i + 4]->rearm_data,
+				    rearm4);
+		_mm256_storeu_si256((__m256i *)&rx_pkts[i + 2]->rearm_data,
+				    rearm2);
+		_mm256_storeu_si256((__m256i *)&rx_pkts[i + 0]->rearm_data,
+				    rearm0);
+
+		/* repeat for the odd mbufs */
+		const __m256i odd_flags =
+			_mm256_castsi128_si256
+				(_mm256_extracti128_si256(mbuf_flags, 1));
+		rearm7 = _mm256_blend_epi32(mbuf_init,
+					    _mm256_slli_si256(odd_flags, 8),
+					    0x04);
+		rearm5 = _mm256_blend_epi32(mbuf_init,
+					    _mm256_slli_si256(odd_flags, 4),
+					    0x04);
+		rearm3 = _mm256_blend_epi32(mbuf_init, odd_flags, 0x04);
+		rearm1 = _mm256_blend_epi32(mbuf_init,
+					    _mm256_srli_si256(odd_flags, 4),
+					    0x04);
+		/* since odd mbufs are already in hi 128-bits use blend */
+		rearm7 = _mm256_blend_epi32(rearm7, mb6_7, 0xF0);
+		rearm5 = _mm256_blend_epi32(rearm5, mb4_5, 0xF0);
+		rearm3 = _mm256_blend_epi32(rearm3, mb2_3, 0xF0);
+		rearm1 = _mm256_blend_epi32(rearm1, mb0_1, 0xF0);
+		/* again write to mbufs */
+		_mm256_storeu_si256((__m256i *)&rx_pkts[i + 7]->rearm_data,
+				    rearm7);
+		_mm256_storeu_si256((__m256i *)&rx_pkts[i + 5]->rearm_data,
+				    rearm5);
+		_mm256_storeu_si256((__m256i *)&rx_pkts[i + 3]->rearm_data,
+				    rearm3);
+		_mm256_storeu_si256((__m256i *)&rx_pkts[i + 1]->rearm_data,
+				    rearm1);
+
+		/* extract and record EOP bit */
+		if (split_packet) {
+			const __m128i eop_mask =
+				_mm_set1_epi16(1 << IAVF_RX_DESC_STATUS_EOF_SHIFT);
+			const __m256i eop_bits256 = _mm256_and_si256(status0_7,
+								     eop_check);
+			/* pack status bits into a single 128-bit register */
+			const __m128i eop_bits =
+				_mm_packus_epi32
+					(_mm256_castsi256_si128(eop_bits256),
+					 _mm256_extractf128_si256(eop_bits256,
+								  1));
+			/**
+			 * flip bits, and mask out the EOP bit, which is now
+			 * a split-packet bit i.e. !EOP, rather than EOP one.
+			 */
+			__m128i split_bits = _mm_andnot_si128(eop_bits,
+					eop_mask);
+			/**
+			 * eop bits are out of order, so we need to shuffle them
+			 * back into order again. In doing so, only use low 8
+			 * bits, which acts like another pack instruction
+			 * The original order is (hi->lo): 1,3,5,7,0,2,4,6
+			 * [Since we use epi8, the 16-bit positions are
+			 * multiplied by 2 in the eop_shuffle value.]
+			 */
+			__m128i eop_shuffle =
+				_mm_set_epi8(/* zero hi 64b */
+					     0xFF, 0xFF, 0xFF, 0xFF,
+					     0xFF, 0xFF, 0xFF, 0xFF,
+					     /* move values to lo 64b */
+					     8, 0, 10, 2,
+					     12, 4, 14, 6);
+			split_bits = _mm_shuffle_epi8(split_bits, eop_shuffle);
+			*(uint64_t *)split_packet =
+				_mm_cvtsi128_si64(split_bits);
+			split_packet += IAVF_DESCS_PER_LOOP_AVX;
+		}
+
+		/* perform dd_check */
+		status0_7 = _mm256_and_si256(status0_7, dd_check);
+		status0_7 = _mm256_packs_epi32(status0_7,
+					       _mm256_setzero_si256());
+
+		uint64_t burst = __builtin_popcountll
+					(_mm_cvtsi128_si64
+						(_mm256_extracti128_si256
+							(status0_7, 1)));
+		burst += __builtin_popcountll
+				(_mm_cvtsi128_si64
+					(_mm256_castsi256_si128(status0_7)));
+		received += burst;
+		if (burst != IAVF_DESCS_PER_LOOP_AVX)
+			break;
+	}
+
+	/* update tail pointers */
+	rxq->rx_tail += received;
+	rxq->rx_tail &= (rxq->nb_rx_desc - 1);
+	if ((rxq->rx_tail & 1) == 1 && received > 1) { /* keep avx2 aligned */
+		rxq->rx_tail--;
+		received--;
+	}
+	rxq->rxrearm_nb += received;
+	return received;
+}
+
+/**
+ * Notice:
+ * - nb_pkts < IAVF_DESCS_PER_LOOP, just return no packet
+ */
+uint16_t
+iavf_recv_pkts_vec_avx2(void *rx_queue, struct rte_mbuf **rx_pkts,
+			uint16_t nb_pkts)
+{
+	return _iavf_recv_raw_pkts_vec_avx2(rx_queue, rx_pkts, nb_pkts, NULL);
+}
+
+/**
+ * vPMD receive routine that reassembles single burst of 32 scattered packets
+ * Notice:
+ * - nb_pkts < IAVF_DESCS_PER_LOOP, just return no packet
+ */
+static uint16_t
+iavf_recv_scattered_burst_vec_avx2(void *rx_queue, struct rte_mbuf **rx_pkts,
+				   uint16_t nb_pkts)
+{
+	struct iavf_rx_queue *rxq = rx_queue;
+	uint8_t split_flags[IAVF_VPMD_RX_MAX_BURST] = {0};
+
+	/* get some new buffers */
+	uint16_t nb_bufs = _iavf_recv_raw_pkts_vec_avx2(rxq, rx_pkts, nb_pkts,
+						       split_flags);
+	if (nb_bufs == 0)
+		return 0;
+
+	/* happy day case, full burst + no packets to be joined */
+	const uint64_t *split_fl64 = (uint64_t *)split_flags;
+
+	if (!rxq->pkt_first_seg &&
+	    split_fl64[0] == 0 && split_fl64[1] == 0 &&
+	    split_fl64[2] == 0 && split_fl64[3] == 0)
+		return nb_bufs;
+
+	/* reassemble any packets that need reassembly*/
+	unsigned int i = 0;
+
+	if (!rxq->pkt_first_seg) {
+		/* find the first split flag, and only reassemble then*/
+		while (i < nb_bufs && !split_flags[i])
+			i++;
+		if (i == nb_bufs)
+			return nb_bufs;
+		rxq->pkt_first_seg = rx_pkts[i];
+	}
+	return i + reassemble_packets(rxq, &rx_pkts[i], nb_bufs - i,
+					     &split_flags[i]);
+}
+
+/**
+ * vPMD receive routine that reassembles scattered packets.
+ * Main receive routine that can handle arbitrary burst sizes
+ * Notice:
+ * - nb_pkts < IAVF_DESCS_PER_LOOP, just return no packet
+ */
+uint16_t
+iavf_recv_scattered_pkts_vec_avx2(void *rx_queue, struct rte_mbuf **rx_pkts,
+				  uint16_t nb_pkts)
+{
+	uint16_t retval = 0;
+
+	while (nb_pkts > IAVF_VPMD_RX_MAX_BURST) {
+		uint16_t burst = iavf_recv_scattered_burst_vec_avx2(rx_queue,
+				rx_pkts + retval, IAVF_VPMD_RX_MAX_BURST);
+		retval += burst;
+		nb_pkts -= burst;
+		if (burst < IAVF_VPMD_RX_MAX_BURST)
+			return retval;
+	}
+	return retval + iavf_recv_scattered_burst_vec_avx2(rx_queue,
+				rx_pkts + retval, nb_pkts);
+}
+
+static inline void
+iavf_vtx1(volatile struct iavf_tx_desc *txdp,
+	  struct rte_mbuf *pkt, uint64_t flags)
+{
+	uint64_t high_qw =
+		(IAVF_TX_DESC_DTYPE_DATA |
+		 ((uint64_t)flags  << IAVF_TXD_QW1_CMD_SHIFT) |
+		 ((uint64_t)pkt->data_len << IAVF_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
+iavf_vtx(volatile struct iavf_tx_desc *txdp,
+	 struct rte_mbuf **pkt, uint16_t nb_pkts,  uint64_t flags)
+{
+	const uint64_t hi_qw_tmpl = (IAVF_TX_DESC_DTYPE_DATA |
+			((uint64_t)flags  << IAVF_TXD_QW1_CMD_SHIFT));
+
+	/* if unaligned on 32-bit boundary, do one to align */
+	if (((uintptr_t)txdp & 0x1F) != 0 && nb_pkts != 0) {
+		iavf_vtx1(txdp, *pkt, flags);
+		nb_pkts--, txdp++, pkt++;
+	}
+
+	/* do two at a time while possible, in bursts */
+	for (; nb_pkts > 3; txdp += 4, pkt += 4, nb_pkts -= 4) {
+		uint64_t hi_qw3 =
+			hi_qw_tmpl |
+			((uint64_t)pkt[3]->data_len <<
+			 IAVF_TXD_QW1_TX_BUF_SZ_SHIFT);
+		uint64_t hi_qw2 =
+			hi_qw_tmpl |
+			((uint64_t)pkt[2]->data_len <<
+			 IAVF_TXD_QW1_TX_BUF_SZ_SHIFT);
+		uint64_t hi_qw1 =
+			hi_qw_tmpl |
+			((uint64_t)pkt[1]->data_len <<
+			 IAVF_TXD_QW1_TX_BUF_SZ_SHIFT);
+		uint64_t hi_qw0 =
+			hi_qw_tmpl |
+			((uint64_t)pkt[0]->data_len <<
+			 IAVF_TXD_QW1_TX_BUF_SZ_SHIFT);
+
+		__m256i desc2_3 =
+			_mm256_set_epi64x
+				(hi_qw3,
+				 pkt[3]->buf_physaddr + pkt[3]->data_off,
+				 hi_qw2,
+				 pkt[2]->buf_physaddr + pkt[2]->data_off);
+		__m256i desc0_1 =
+			_mm256_set_epi64x
+				(hi_qw1,
+				 pkt[1]->buf_physaddr + pkt[1]->data_off,
+				 hi_qw0,
+				 pkt[0]->buf_physaddr + pkt[0]->data_off);
+		_mm256_store_si256((void *)(txdp + 2), desc2_3);
+		_mm256_store_si256((void *)txdp, desc0_1);
+	}
+
+	/* do any last ones */
+	while (nb_pkts) {
+		iavf_vtx1(txdp, *pkt, flags);
+		txdp++, pkt++, nb_pkts--;
+	}
+}
+
+static inline uint16_t
+iavf_xmit_fixed_burst_vec_avx2(void *tx_queue, struct rte_mbuf **tx_pkts,
+			       uint16_t nb_pkts)
+{
+	struct iavf_tx_queue *txq = (struct iavf_tx_queue *)tx_queue;
+	volatile struct iavf_tx_desc *txdp;
+	struct iavf_tx_entry *txep;
+	uint16_t n, nb_commit, tx_id;
+	uint64_t flags = IAVF_TX_DESC_CMD_EOP;
+	uint64_t rs = IAVF_TX_DESC_CMD_RS | IAVF_TX_DESC_CMD_EOP;
+
+	/* cross rx_thresh boundary is not allowed */
+	nb_pkts = RTE_MIN(nb_pkts, txq->rs_thresh);
+
+	if (txq->nb_free < txq->free_thresh)
+		iavf_tx_free_bufs(txq);
+
+	nb_commit = nb_pkts = (uint16_t)RTE_MIN(txq->nb_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_free = (uint16_t)(txq->nb_free - nb_pkts);
+
+	n = (uint16_t)(txq->nb_tx_desc - tx_id);
+	if (nb_commit >= n) {
+		tx_backlog_entry(txep, tx_pkts, n);
+
+		iavf_vtx(txdp, tx_pkts, n - 1, flags);
+		tx_pkts += (n - 1);
+		txdp += (n - 1);
+
+		iavf_vtx1(txdp, *tx_pkts++, rs);
+
+		nb_commit = (uint16_t)(nb_commit - n);
+
+		tx_id = 0;
+		txq->next_rs = (uint16_t)(txq->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);
+
+	iavf_vtx(txdp, tx_pkts, nb_commit, flags);
+
+	tx_id = (uint16_t)(tx_id + nb_commit);
+	if (tx_id > txq->next_rs) {
+		txq->tx_ring[txq->next_rs].cmd_type_offset_bsz |=
+			rte_cpu_to_le_64(((uint64_t)IAVF_TX_DESC_CMD_RS) <<
+					 IAVF_TXD_QW1_CMD_SHIFT);
+		txq->next_rs =
+			(uint16_t)(txq->next_rs + txq->rs_thresh);
+	}
+
+	txq->tx_tail = tx_id;
+
+	IAVF_PCI_REG_WRITE(txq->qtx_tail, txq->tx_tail);
+
+	return nb_pkts;
+}
+
+uint16_t
+iavf_xmit_pkts_vec_avx2(void *tx_queue, struct rte_mbuf **tx_pkts,
+			uint16_t nb_pkts)
+{
+	uint16_t nb_tx = 0;
+	struct iavf_tx_queue *txq = (struct iavf_tx_queue *)tx_queue;
+
+	while (nb_pkts) {
+		uint16_t ret, num;
+
+		num = (uint16_t)RTE_MIN(nb_pkts, txq->rs_thresh);
+		ret = iavf_xmit_fixed_burst_vec_avx2(tx_queue, &tx_pkts[nb_tx],
+						     num);
+		nb_tx += ret;
+		nb_pkts -= ret;
+		if (ret < num)
+			break;
+	}
+
+	return nb_tx;
+}
diff --git a/drivers/net/iavf/iavf_rxtx_vec_common.h b/drivers/net/iavf/iavf_rxtx_vec_common.h
index db509d71f..a6ba22758 100644
--- a/drivers/net/iavf/iavf_rxtx_vec_common.h
+++ b/drivers/net/iavf/iavf_rxtx_vec_common.h
@@ -207,4 +207,69 @@ iavf_rxq_vec_setup_default(struct iavf_rx_queue *rxq)
 	rxq->mbuf_initializer = *(uint64_t *)p;
 	return 0;
 }
+
+static inline int
+iavf_rx_vec_queue_default(struct iavf_rx_queue *rxq)
+{
+	if (!rxq)
+		return -1;
+
+	if (!rte_is_power_of_2(rxq->nb_rx_desc))
+		return -1;
+
+	if (rxq->rx_free_thresh < IAVF_VPMD_RX_MAX_BURST)
+		return -1;
+
+	if (rxq->nb_rx_desc % rxq->rx_free_thresh)
+		return -1;
+
+	return 0;
+}
+
+static inline int
+iavf_tx_vec_queue_default(struct iavf_tx_queue *txq)
+{
+	if (!txq)
+		return -1;
+
+	if (txq->offloads & IAVF_NO_VECTOR_FLAGS)
+		return -1;
+
+	if (txq->rs_thresh < IAVF_VPMD_TX_MAX_BURST ||
+	    txq->rs_thresh > IAVF_VPMD_TX_MAX_FREE_BUF)
+		return -1;
+
+	return 0;
+}
+
+static inline int
+iavf_rx_vec_dev_check_default(struct rte_eth_dev *dev)
+{
+	int i;
+	struct iavf_rx_queue *rxq;
+
+	for (i = 0; i < dev->data->nb_rx_queues; i++) {
+		rxq = dev->data->rx_queues[i];
+		if (iavf_rx_vec_queue_default(rxq))
+			return -1;
+	}
+
+	return 0;
+}
+
+static inline int
+iavf_tx_vec_dev_check_default(struct rte_eth_dev *dev)
+{
+	int i;
+	struct iavf_tx_queue *txq;
+
+	for (i = 0; i < dev->data->nb_tx_queues; i++) {
+		txq = dev->data->tx_queues[i];
+		if (iavf_tx_vec_queue_default(txq))
+			return -1;
+	}
+
+	return 0;
+}
+
 #endif
diff --git a/drivers/net/iavf/iavf_rxtx_vec_sse.c b/drivers/net/iavf/iavf_rxtx_vec_sse.c
index cc71f23a5..2b16dc1b5 100644
--- a/drivers/net/iavf/iavf_rxtx_vec_sse.c
+++ b/drivers/net/iavf/iavf_rxtx_vec_sse.c
@@ -622,6 +622,27 @@ iavf_xmit_fixed_burst_vec(void *tx_queue, struct rte_mbuf **tx_pkts,
 	return nb_pkts;
 }
 
+uint16_t
+iavf_xmit_pkts_vec(void *tx_queue, struct rte_mbuf **tx_pkts,
+		   uint16_t nb_pkts)
+{
+	uint16_t nb_tx = 0;
+	struct iavf_tx_queue *txq = (struct iavf_tx_queue *)tx_queue;
+
+	while (nb_pkts) {
+		uint16_t ret, num;
+
+		num = (uint16_t)RTE_MIN(nb_pkts, txq->rs_thresh);
+		ret = iavf_xmit_fixed_burst_vec(tx_queue, &tx_pkts[nb_tx], num);
+		nb_tx += ret;
+		nb_pkts -= ret;
+		if (ret < num)
+			break;
+	}
+
+	return nb_tx;
+}
+
 static void __attribute__((cold))
 iavf_rx_queue_release_mbufs_sse(struct iavf_rx_queue *rxq)
 {
@@ -655,3 +676,15 @@ iavf_rxq_vec_setup(struct iavf_rx_queue *rxq)
 	rxq->ops = &sse_vec_rxq_ops;
 	return iavf_rxq_vec_setup_default(rxq);
 }
+
+int __attribute__((cold))
+iavf_rx_vec_dev_check(struct rte_eth_dev *dev)
+{
+	return iavf_rx_vec_dev_check_default(dev);
+}
+
+int __attribute__((cold))
+iavf_tx_vec_dev_check(struct rte_eth_dev *dev)
+{
+	return iavf_tx_vec_dev_check_default(dev);
+}
diff --git a/drivers/net/iavf/meson.build b/drivers/net/iavf/meson.build
index e5a2f5553..c2401af86 100644
--- a/drivers/net/iavf/meson.build
+++ b/drivers/net/iavf/meson.build
@@ -17,4 +17,21 @@ sources = files(
 if arch_subdir == 'x86'
 	dpdk_conf.set('RTE_LIBRTE_IAVF_INC_VECTOR', 1)
 	sources += files('iavf_rxtx_vec_sse.c')
+
+	# compile AVX2 version if either:
+	# a. we have AVX supported in minimum instruction set baseline
+	# b. it's not minimum instruction set, but supported by compiler
+	if dpdk_conf.has('RTE_MACHINE_CPUFLAG_AVX2')
+		cflags += ['-DCC_AVX2_SUPPORT']
+		sources += files('iavf_rxtx_vec_avx2.c')
+	elif cc.has_argument('-mavx2')
+		cflags += ['-DCC_AVX2_SUPPORT']
+		iavf_avx2_lib = static_library('iavf_avx2_lib',
+				'iavf_rxtx_vec_avx2.c',
+				dependencies: [static_rte_ethdev,
+					static_rte_kvargs, static_rte_hash],
+				include_directories: includes,
+				c_args: [cflags, '-mavx2'])
+		objs += iavf_avx2_lib.extract_objects('iavf_rxtx_vec_avx2.c')
+	endif
 endif
-- 
2.17.1


^ permalink raw reply	[flat|nested] 11+ messages in thread

* Re: [dpdk-dev] [PATCH v4] net/iavf: enable AVX2 for iavf
  2019-09-23  6:02 ` [dpdk-dev] [PATCH v4] " Leyi Rong
@ 2019-09-23  6:49   ` Zhang, Qi Z
  2019-09-30  9:47   ` Ye Xiaolong
  2019-09-30 13:08   ` Ferruh Yigit
  2 siblings, 0 replies; 11+ messages in thread
From: Zhang, Qi Z @ 2019-09-23  6:49 UTC (permalink / raw)
  To: Rong, Leyi, Lu, Wenzhuo, Ye, Xiaolong; +Cc: dev



> -----Original Message-----
> From: Rong, Leyi
> Sent: Monday, September 23, 2019 2:03 PM
> To: Lu, Wenzhuo <wenzhuo.lu@intel.com>; Zhang, Qi Z
> <qi.z.zhang@intel.com>; Ye, Xiaolong <xiaolong.ye@intel.com>
> Cc: dev@dpdk.org; Rong, Leyi <leyi.rong@intel.com>
> Subject: [PATCH v4] net/iavf: enable AVX2 for iavf
> 
> This patch enables AVX data path for iavf PMD.
> 
> Signed-off-by: Leyi Rong <leyi.rong@intel.com>

Acked-by: Qi Zhang <qi.z.zhang@intel.com>


^ permalink raw reply	[flat|nested] 11+ messages in thread

* Re: [dpdk-dev] [PATCH v4] net/iavf: enable AVX2 for iavf
  2019-09-23  6:02 ` [dpdk-dev] [PATCH v4] " Leyi Rong
  2019-09-23  6:49   ` Zhang, Qi Z
@ 2019-09-30  9:47   ` Ye Xiaolong
  2019-09-30 13:08   ` Ferruh Yigit
  2 siblings, 0 replies; 11+ messages in thread
From: Ye Xiaolong @ 2019-09-30  9:47 UTC (permalink / raw)
  To: Leyi Rong; +Cc: wenzhuo.lu, qi.z.zhang, dev

On 09/23, Leyi Rong wrote:
>This patch enables AVX data path for iavf PMD.
>
>Signed-off-by: Leyi Rong <leyi.rong@intel.com>
>
>---
>v4:
>- Romoves redundant IAVF_NO_VECTOR_FLAGS define.
>
>v3:
>- Adds release note.
>
>v2:
>- Fixes build error on aarch64.
>---
> doc/guides/rel_notes/release_19_11.rst  |   3 +
> drivers/net/iavf/Makefile               |  21 +
> drivers/net/iavf/iavf_rxtx.c            |  99 +--
> drivers/net/iavf/iavf_rxtx.h            |  12 +
> drivers/net/iavf/iavf_rxtx_vec_avx2.c   | 867 ++++++++++++++++++++++++
> drivers/net/iavf/iavf_rxtx_vec_common.h |  65 ++
> drivers/net/iavf/iavf_rxtx_vec_sse.c    |  33 +
> drivers/net/iavf/meson.build            |  17 +
> 8 files changed, 1071 insertions(+), 46 deletions(-)
> create mode 100644 drivers/net/iavf/iavf_rxtx_vec_avx2.c
>

Applied to dpdk-next-net-intel.

^ permalink raw reply	[flat|nested] 11+ messages in thread

* Re: [dpdk-dev] [PATCH v4] net/iavf: enable AVX2 for iavf
  2019-09-23  6:02 ` [dpdk-dev] [PATCH v4] " Leyi Rong
  2019-09-23  6:49   ` Zhang, Qi Z
  2019-09-30  9:47   ` Ye Xiaolong
@ 2019-09-30 13:08   ` Ferruh Yigit
  2 siblings, 0 replies; 11+ messages in thread
From: Ferruh Yigit @ 2019-09-30 13:08 UTC (permalink / raw)
  To: Leyi Rong, wenzhuo.lu, qi.z.zhang, xiaolong.ye; +Cc: dev

On 9/23/2019 7:02 AM, Leyi Rong wrote:
> This patch enables AVX data path for iavf PMD.
> 
> Signed-off-by: Leyi Rong <leyi.rong@intel.com>

<...>

> @@ -1741,26 +1720,43 @@ iavf_set_rx_function(struct rte_eth_dev *dev)
>  {
>  	struct iavf_adapter *adapter =
>  		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
> +#ifdef RTE_ARCH_X86
>  	struct iavf_rx_queue *rxq;
>  	int i;
> +	bool use_avx2 = false;
>  
> -	if (adapter->rx_vec_allowed) {
> -		if (dev->data->scattered_rx) {
> -			PMD_DRV_LOG(DEBUG, "Using Vector Scattered Rx callback"
> -				    " (port=%d).", dev->data->port_id);
> -			dev->rx_pkt_burst = iavf_recv_scattered_pkts_vec;
> -		} else {
> -			PMD_DRV_LOG(DEBUG, "Using Vector Rx callback"
> -				    " (port=%d).", dev->data->port_id);
> -			dev->rx_pkt_burst = iavf_recv_pkts_vec;
> -		}
> +	if (!iavf_rx_vec_dev_check(dev)) {
>  		for (i = 0; i < dev->data->nb_rx_queues; i++) {
>  			rxq = dev->data->rx_queues[i];
> -			if (!rxq)
> -				continue;
> -			iavf_rxq_vec_setup(rxq);
> +			(void)iavf_rxq_vec_setup(rxq);
>  		}
> -	} else if (dev->data->scattered_rx) {
> +
> +		if (rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1 ||
> +		    rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX512F) == 1)
> +			use_avx2 = true;
> +
> +		if (dev->data->scattered_rx) {
> +			PMD_DRV_LOG(DEBUG,
> +				    "Using %sVector Scattered Rx (port %d).",
> +				    use_avx2 ? "avx2 " : "",
> +				    dev->data->port_id);
> +			dev->rx_pkt_burst = use_avx2 ?
> +					    iavf_recv_scattered_pkts_vec_avx2 :
> +					    iavf_recv_scattered_pkts_vec;

Hi Leyi,

Will this build if config option 'CONFIG_RTE_LIBRTE_IAVF_INC_VECTOR' is disabled?
And indeed why not eliminate 'CONFIG_RTE_LIBRTE_IAVF_INC_VECTOR' at all? Please
check ice for the sample.

And another thing is to remove the __rte_weak, for the consistency, since it has
been removed from the i40e.

^ permalink raw reply	[flat|nested] 11+ messages in thread

* [dpdk-dev] [PATCH v5] net/iavf: enable AVX2 for iavf
  2019-09-04 10:06 [dpdk-dev] [PATCH] net/iavf: enable AVX2 for iavf Leyi Rong
                   ` (3 preceding siblings ...)
  2019-09-23  6:02 ` [dpdk-dev] [PATCH v4] " Leyi Rong
@ 2019-10-16  8:14 ` Leyi Rong
  2019-10-20  1:12   ` Ye Xiaolong
  4 siblings, 1 reply; 11+ messages in thread
From: Leyi Rong @ 2019-10-16  8:14 UTC (permalink / raw)
  To: wenzhuo.lu, qi.z.zhang, xiaolong.ye; +Cc: dev, Leyi Rong

This patch enables AVX data path for iavf PMD.

Signed-off-by: Leyi Rong <leyi.rong@intel.com>

---
v5:
- Eliminates CONFIG_RTE_LIBRTE_IAVF_INC_VECTOR and __rte_weak.

v4:
- Romoves redundant IAVF_NO_VECTOR_FLAGS define.

v3:
- Adds release note.

v2:
- Fixes build error on aarch64.
---
 config/common_base                      |   1 -
 doc/guides/rel_notes/release_19_11.rst  |   4 +
 drivers/net/iavf/Makefile               |  23 +-
 drivers/net/iavf/iavf_ethdev.c          |   5 -
 drivers/net/iavf/iavf_rxtx.c            | 141 ++--
 drivers/net/iavf/iavf_rxtx.h            |  12 +
 drivers/net/iavf/iavf_rxtx_vec_avx2.c   | 867 ++++++++++++++++++++++++
 drivers/net/iavf/iavf_rxtx_vec_common.h |  65 ++
 drivers/net/iavf/iavf_rxtx_vec_sse.c    |  33 +
 drivers/net/iavf/meson.build            |  18 +-
 10 files changed, 1074 insertions(+), 95 deletions(-)
 create mode 100644 drivers/net/iavf/iavf_rxtx_vec_avx2.c

diff --git a/config/common_base b/config/common_base
index 9c301b79c..60e008e76 100644
--- a/config/common_base
+++ b/config/common_base
@@ -332,7 +332,6 @@ CONFIG_RTE_LIBRTE_ICE_16BYTE_RX_DESC=n
 # Compile burst-oriented IAVF PMD driver
 #
 CONFIG_RTE_LIBRTE_IAVF_PMD=y
-CONFIG_RTE_LIBRTE_IAVF_INC_VECTOR=y
 CONFIG_RTE_LIBRTE_IAVF_DEBUG_TX=n
 CONFIG_RTE_LIBRTE_IAVF_DEBUG_TX_FREE=n
 CONFIG_RTE_LIBRTE_IAVF_DEBUG_RX=n
diff --git a/doc/guides/rel_notes/release_19_11.rst b/doc/guides/rel_notes/release_19_11.rst
index ef736718f..ac26fe669 100644
--- a/doc/guides/rel_notes/release_19_11.rst
+++ b/doc/guides/rel_notes/release_19_11.rst
@@ -92,6 +92,10 @@ New Features
   * Added support for protocol extraction on per Rx queue.
   * Added support for the ``RTE_ETH_DEV_CLOSE_REMOVE`` flag.
 
+* **Updated iavf PMD.**
+
+  Enable AVX2 data path for iavf PMD.
+
 * **Updated the enic driver.**
 
   * Added support for Geneve with options offload.
diff --git a/drivers/net/iavf/Makefile b/drivers/net/iavf/Makefile
index cd74e14ab..d3f9972f4 100644
--- a/drivers/net/iavf/Makefile
+++ b/drivers/net/iavf/Makefile
@@ -45,7 +45,28 @@ SRCS-$(CONFIG_RTE_LIBRTE_IAVF_PMD) += iavf_ethdev.c
 SRCS-$(CONFIG_RTE_LIBRTE_IAVF_PMD) += iavf_vchnl.c
 SRCS-$(CONFIG_RTE_LIBRTE_IAVF_PMD) += iavf_rxtx.c
 ifeq ($(CONFIG_RTE_ARCH_X86), y)
-SRCS-$(CONFIG_RTE_LIBRTE_IAVF_INC_VECTOR) += iavf_rxtx_vec_sse.c
+SRCS-$(CONFIG_RTE_LIBRTE_IAVF_PMD) += iavf_rxtx_vec_sse.c
+endif
+
+ifeq ($(CONFIG_RTE_LIBRTE_IAVF_PMD), y)
+	ifeq ($(findstring RTE_MACHINE_CPUFLAG_AVX2,$(CFLAGS)),RTE_MACHINE_CPUFLAG_AVX2)
+		CC_AVX2_SUPPORT=1
+	else
+		CC_AVX2_SUPPORT=\
+		$(shell $(CC) -march=core-avx2 -dM -E - </dev/null 2>&1 | \
+		grep -q AVX2 && echo 1)
+		ifeq ($(CC_AVX2_SUPPORT), 1)
+			ifeq ($(CONFIG_RTE_TOOLCHAIN_ICC),y)
+				CFLAGS_iavf_rxtx_vec_avx2.o += -march=core-avx2
+			else
+				CFLAGS_iavf_rxtx_vec_avx2.o += -mavx2
+			endif
+		endif
+	endif
+endif
+
+ifeq ($(CC_AVX2_SUPPORT), 1)
+	SRCS-$(CONFIG_RTE_LIBRTE_IAVF_PMD) += iavf_rxtx_vec_avx2.c
 endif
 
 include $(RTE_SDK)/mk/rte.lib.mk
diff --git a/drivers/net/iavf/iavf_ethdev.c b/drivers/net/iavf/iavf_ethdev.c
index 64037e11d..eebc49ade 100644
--- a/drivers/net/iavf/iavf_ethdev.c
+++ b/drivers/net/iavf/iavf_ethdev.c
@@ -141,16 +141,11 @@ iavf_dev_configure(struct rte_eth_dev *dev)
 	struct rte_eth_conf *dev_conf = &dev->data->dev_conf;
 
 	ad->rx_bulk_alloc_allowed = true;
-#ifdef RTE_LIBRTE_IAVF_INC_VECTOR
 	/* Initialize to TRUE. If any of Rx queues doesn't meet the
 	 * vector Rx/Tx preconditions, it will be reset.
 	 */
 	ad->rx_vec_allowed = true;
 	ad->tx_vec_allowed = true;
-#else
-	ad->rx_vec_allowed = false;
-	ad->tx_vec_allowed = false;
-#endif
 
 	/* Vlan stripping setting */
 	if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN) {
diff --git a/drivers/net/iavf/iavf_rxtx.c b/drivers/net/iavf/iavf_rxtx.c
index 22d7bb612..3ee92e839 100644
--- a/drivers/net/iavf/iavf_rxtx.c
+++ b/drivers/net/iavf/iavf_rxtx.c
@@ -92,7 +92,6 @@ check_tx_thresh(uint16_t nb_desc, uint16_t tx_rs_thresh,
 	return 0;
 }
 
-#ifdef RTE_LIBRTE_IAVF_INC_VECTOR
 static inline bool
 check_rx_vec_allow(struct iavf_rx_queue *rxq)
 {
@@ -118,7 +117,6 @@ check_tx_vec_allow(struct iavf_tx_queue *txq)
 	PMD_INIT_LOG(DEBUG, "Vector Tx cannot be enabled on this txq.");
 	return FALSE;
 }
-#endif
 
 static inline bool
 check_rx_bulk_allow(struct iavf_rx_queue *rxq)
@@ -413,10 +411,9 @@ iavf_dev_rx_queue_setup(struct rte_eth_dev *dev, uint16_t queue_idx,
 		ad->rx_bulk_alloc_allowed = false;
 	}
 
-#ifdef RTE_LIBRTE_IAVF_INC_VECTOR
 	if (check_rx_vec_allow(rxq) == FALSE)
 		ad->rx_vec_allowed = false;
-#endif
+
 	return 0;
 }
 
@@ -511,13 +508,11 @@ iavf_dev_tx_queue_setup(struct rte_eth_dev *dev,
 	txq->qtx_tail = hw->hw_addr + IAVF_QTX_TAIL1(queue_idx);
 	txq->ops = &def_txq_ops;
 
-#ifdef RTE_LIBRTE_IAVF_INC_VECTOR
 	if (check_tx_vec_allow(txq) == FALSE) {
 		struct iavf_adapter *ad =
 			IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 		ad->tx_vec_allowed = false;
 	}
-#endif
 
 	return 0;
 }
@@ -1666,27 +1661,6 @@ iavf_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts, uint16_t nb_pkts)
 	return nb_tx;
 }
 
-static uint16_t
-iavf_xmit_pkts_vec(void *tx_queue, struct rte_mbuf **tx_pkts,
-		  uint16_t nb_pkts)
-{
-	uint16_t nb_tx = 0;
-	struct iavf_tx_queue *txq = (struct iavf_tx_queue *)tx_queue;
-
-	while (nb_pkts) {
-		uint16_t ret, num;
-
-		num = (uint16_t)RTE_MIN(nb_pkts, txq->rs_thresh);
-		ret = iavf_xmit_fixed_burst_vec(tx_queue, &tx_pkts[nb_tx], num);
-		nb_tx += ret;
-		nb_pkts -= ret;
-		if (ret < num)
-			break;
-	}
-
-	return nb_tx;
-}
-
 /* TX prep functions */
 uint16_t
 iavf_prep_pkts(__rte_unused void *tx_queue, struct rte_mbuf **tx_pkts,
@@ -1741,26 +1715,43 @@ iavf_set_rx_function(struct rte_eth_dev *dev)
 {
 	struct iavf_adapter *adapter =
 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
+#ifdef RTE_ARCH_X86
 	struct iavf_rx_queue *rxq;
 	int i;
+	bool use_avx2 = false;
 
-	if (adapter->rx_vec_allowed) {
-		if (dev->data->scattered_rx) {
-			PMD_DRV_LOG(DEBUG, "Using Vector Scattered Rx callback"
-				    " (port=%d).", dev->data->port_id);
-			dev->rx_pkt_burst = iavf_recv_scattered_pkts_vec;
-		} else {
-			PMD_DRV_LOG(DEBUG, "Using Vector Rx callback"
-				    " (port=%d).", dev->data->port_id);
-			dev->rx_pkt_burst = iavf_recv_pkts_vec;
-		}
+	if (!iavf_rx_vec_dev_check(dev)) {
 		for (i = 0; i < dev->data->nb_rx_queues; i++) {
 			rxq = dev->data->rx_queues[i];
-			if (!rxq)
-				continue;
-			iavf_rxq_vec_setup(rxq);
+			(void)iavf_rxq_vec_setup(rxq);
+		}
+
+		if (rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1 ||
+		    rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX512F) == 1)
+			use_avx2 = true;
+
+		if (dev->data->scattered_rx) {
+			PMD_DRV_LOG(DEBUG,
+				    "Using %sVector Scattered Rx (port %d).",
+				    use_avx2 ? "avx2 " : "",
+				    dev->data->port_id);
+			dev->rx_pkt_burst = use_avx2 ?
+					    iavf_recv_scattered_pkts_vec_avx2 :
+					    iavf_recv_scattered_pkts_vec;
+		} else {
+			PMD_DRV_LOG(DEBUG, "Using %sVector Rx (port %d).",
+				    use_avx2 ? "avx2 " : "",
+				    dev->data->port_id);
+			dev->rx_pkt_burst = use_avx2 ?
+					    iavf_recv_pkts_vec_avx2 :
+					    iavf_recv_pkts_vec;
 		}
-	} else if (dev->data->scattered_rx) {
+
+		return;
+	}
+#endif
+
+	if (dev->data->scattered_rx) {
 		PMD_DRV_LOG(DEBUG, "Using a Scattered Rx callback (port=%d).",
 			    dev->data->port_id);
 		dev->rx_pkt_burst = iavf_recv_scattered_pkts;
@@ -1779,28 +1770,39 @@ iavf_set_rx_function(struct rte_eth_dev *dev)
 void
 iavf_set_tx_function(struct rte_eth_dev *dev)
 {
-	struct iavf_adapter *adapter =
-		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
+#ifdef RTE_ARCH_X86
 	struct iavf_tx_queue *txq;
 	int i;
+	bool use_avx2 = false;
 
-	if (adapter->tx_vec_allowed) {
-		PMD_DRV_LOG(DEBUG, "Using Vector Tx callback (port=%d).",
-			    dev->data->port_id);
-		dev->tx_pkt_burst = iavf_xmit_pkts_vec;
-		dev->tx_pkt_prepare = NULL;
+	if (!iavf_tx_vec_dev_check(dev)) {
 		for (i = 0; i < dev->data->nb_tx_queues; i++) {
 			txq = dev->data->tx_queues[i];
 			if (!txq)
 				continue;
 			iavf_txq_vec_setup(txq);
 		}
-	} else {
-		PMD_DRV_LOG(DEBUG, "Using Basic Tx callback (port=%d).",
+
+		if (rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1 ||
+		    rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX512F) == 1)
+			use_avx2 = true;
+
+		PMD_DRV_LOG(DEBUG, "Using %sVector Tx (port %d).",
+			    use_avx2 ? "avx2 " : "",
 			    dev->data->port_id);
-		dev->tx_pkt_burst = iavf_xmit_pkts;
-		dev->tx_pkt_prepare = iavf_prep_pkts;
+		dev->tx_pkt_burst = use_avx2 ?
+				    iavf_xmit_pkts_vec_avx2 :
+				    iavf_xmit_pkts_vec;
+		dev->tx_pkt_prepare = NULL;
+
+		return;
 	}
+#endif
+
+	PMD_DRV_LOG(DEBUG, "Using Basic Tx callback (port=%d).",
+		    dev->data->port_id);
+	dev->tx_pkt_burst = iavf_xmit_pkts;
+	dev->tx_pkt_prepare = iavf_prep_pkts;
 }
 
 void
@@ -1923,38 +1925,3 @@ iavf_dev_tx_desc_status(void *tx_queue, uint16_t offset)
 	return RTE_ETH_TX_DESC_FULL;
 }
 
-__rte_weak uint16_t
-iavf_recv_pkts_vec(__rte_unused void *rx_queue,
-		  __rte_unused struct rte_mbuf **rx_pkts,
-		  __rte_unused uint16_t nb_pkts)
-{
-	return 0;
-}
-
-__rte_weak uint16_t
-iavf_recv_scattered_pkts_vec(__rte_unused void *rx_queue,
-			    __rte_unused struct rte_mbuf **rx_pkts,
-			    __rte_unused uint16_t nb_pkts)
-{
-	return 0;
-}
-
-__rte_weak uint16_t
-iavf_xmit_fixed_burst_vec(__rte_unused void *tx_queue,
-			 __rte_unused struct rte_mbuf **tx_pkts,
-			 __rte_unused uint16_t nb_pkts)
-{
-	return 0;
-}
-
-__rte_weak int
-iavf_rxq_vec_setup(__rte_unused struct iavf_rx_queue *rxq)
-{
-	return -1;
-}
-
-__rte_weak int
-iavf_txq_vec_setup(__rte_unused struct iavf_tx_queue *txq)
-{
-	return -1;
-}
diff --git a/drivers/net/iavf/iavf_rxtx.h b/drivers/net/iavf/iavf_rxtx.h
index c86720bda..225a0c4c4 100644
--- a/drivers/net/iavf/iavf_rxtx.h
+++ b/drivers/net/iavf/iavf_rxtx.h
@@ -19,6 +19,7 @@
 /* used for Vector PMD */
 #define IAVF_VPMD_RX_MAX_BURST    32
 #define IAVF_VPMD_TX_MAX_BURST    32
+#define IAVF_RXQ_REARM_THRESH     32
 #define IAVF_VPMD_DESCS_PER_LOOP  4
 #define IAVF_VPMD_TX_MAX_FREE_BUF 64
 
@@ -200,6 +201,17 @@ uint16_t iavf_recv_scattered_pkts_vec(void *rx_queue,
 				     uint16_t nb_pkts);
 uint16_t iavf_xmit_fixed_burst_vec(void *tx_queue, struct rte_mbuf **tx_pkts,
 				  uint16_t nb_pkts);
+uint16_t iavf_recv_pkts_vec_avx2(void *rx_queue, struct rte_mbuf **rx_pkts,
+				 uint16_t nb_pkts);
+uint16_t iavf_recv_scattered_pkts_vec_avx2(void *rx_queue,
+					   struct rte_mbuf **rx_pkts,
+					   uint16_t nb_pkts);
+uint16_t iavf_xmit_pkts_vec(void *tx_queue, struct rte_mbuf **tx_pkts,
+			    uint16_t nb_pkts);
+uint16_t iavf_xmit_pkts_vec_avx2(void *tx_queue, struct rte_mbuf **tx_pkts,
+				 uint16_t nb_pkts);
+int iavf_rx_vec_dev_check(struct rte_eth_dev *dev);
+int iavf_tx_vec_dev_check(struct rte_eth_dev *dev);
 int iavf_rxq_vec_setup(struct iavf_rx_queue *rxq);
 int iavf_txq_vec_setup(struct iavf_tx_queue *txq);
 
diff --git a/drivers/net/iavf/iavf_rxtx_vec_avx2.c b/drivers/net/iavf/iavf_rxtx_vec_avx2.c
new file mode 100644
index 000000000..f0c00be56
--- /dev/null
+++ b/drivers/net/iavf/iavf_rxtx_vec_avx2.c
@@ -0,0 +1,867 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2019 Intel Corporation
+ */
+
+#include "base/iavf_prototype.h"
+#include "iavf_rxtx_vec_common.h"
+
+#include <x86intrin.h>
+
+#ifndef __INTEL_COMPILER
+#pragma GCC diagnostic ignored "-Wcast-qual"
+#endif
+
+static inline void
+iavf_rxq_rearm(struct iavf_rx_queue *rxq)
+{
+	int i;
+	uint16_t rx_id;
+	volatile union iavf_rx_desc *rxdp;
+	struct rte_mbuf **rxp = &rxq->sw_ring[rxq->rxrearm_start];
+
+	rxdp = rxq->rx_ring + rxq->rxrearm_start;
+
+	/* Pull 'n' more MBUFs into the software ring */
+	if (rte_mempool_get_bulk(rxq->mp,
+				 (void *)rxp,
+				 IAVF_RXQ_REARM_THRESH) < 0) {
+		if (rxq->rxrearm_nb + IAVF_RXQ_REARM_THRESH >=
+		    rxq->nb_rx_desc) {
+			__m128i dma_addr0;
+
+			dma_addr0 = _mm_setzero_si128();
+			for (i = 0; i < IAVF_VPMD_DESCS_PER_LOOP; i++) {
+				rxp[i] = &rxq->fake_mbuf;
+				_mm_store_si128((__m128i *)&rxdp[i].read,
+						dma_addr0);
+			}
+		}
+		rte_eth_devices[rxq->port_id].data->rx_mbuf_alloc_failed +=
+			IAVF_RXQ_REARM_THRESH;
+		return;
+	}
+
+#ifndef RTE_LIBRTE_IAVF_16BYTE_RX_DESC
+	struct rte_mbuf *mb0, *mb1;
+	__m128i dma_addr0, dma_addr1;
+	__m128i hdr_room = _mm_set_epi64x(RTE_PKTMBUF_HEADROOM,
+			RTE_PKTMBUF_HEADROOM);
+	/* Initialize the mbufs in vector, process 2 mbufs in one loop */
+	for (i = 0; i < IAVF_RXQ_REARM_THRESH; i += 2, rxp += 2) {
+		__m128i vaddr0, vaddr1;
+
+		mb0 = rxp[0];
+		mb1 = rxp[1];
+
+		/* load buf_addr(lo 64bit) and buf_physaddr(hi 64bit) */
+		RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, buf_physaddr) !=
+				offsetof(struct rte_mbuf, buf_addr) + 8);
+		vaddr0 = _mm_loadu_si128((__m128i *)&mb0->buf_addr);
+		vaddr1 = _mm_loadu_si128((__m128i *)&mb1->buf_addr);
+
+		/* convert pa to dma_addr hdr/data */
+		dma_addr0 = _mm_unpackhi_epi64(vaddr0, vaddr0);
+		dma_addr1 = _mm_unpackhi_epi64(vaddr1, vaddr1);
+
+		/* add headroom to pa values */
+		dma_addr0 = _mm_add_epi64(dma_addr0, hdr_room);
+		dma_addr1 = _mm_add_epi64(dma_addr1, hdr_room);
+
+		/* flush desc with pa dma_addr */
+		_mm_store_si128((__m128i *)&rxdp++->read, dma_addr0);
+		_mm_store_si128((__m128i *)&rxdp++->read, dma_addr1);
+	}
+#else
+	struct rte_mbuf *mb0, *mb1, *mb2, *mb3;
+	__m256i dma_addr0_1, dma_addr2_3;
+	__m256i hdr_room = _mm256_set1_epi64x(RTE_PKTMBUF_HEADROOM);
+	/* Initialize the mbufs in vector, process 4 mbufs in one loop */
+	for (i = 0; i < IAVF_RXQ_REARM_THRESH;
+			i += 4, rxp += 4, rxdp += 4) {
+		__m128i vaddr0, vaddr1, vaddr2, vaddr3;
+		__m256i vaddr0_1, vaddr2_3;
+
+		mb0 = rxp[0];
+		mb1 = rxp[1];
+		mb2 = rxp[2];
+		mb3 = rxp[3];
+
+		/* load buf_addr(lo 64bit) and buf_physaddr(hi 64bit) */
+		RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, buf_physaddr) !=
+				offsetof(struct rte_mbuf, buf_addr) + 8);
+		vaddr0 = _mm_loadu_si128((__m128i *)&mb0->buf_addr);
+		vaddr1 = _mm_loadu_si128((__m128i *)&mb1->buf_addr);
+		vaddr2 = _mm_loadu_si128((__m128i *)&mb2->buf_addr);
+		vaddr3 = _mm_loadu_si128((__m128i *)&mb3->buf_addr);
+
+		/**
+		 * merge 0 & 1, by casting 0 to 256-bit and inserting 1
+		 * into the high lanes. Similarly for 2 & 3
+		 */
+		vaddr0_1 =
+			_mm256_inserti128_si256(_mm256_castsi128_si256(vaddr0),
+						vaddr1, 1);
+		vaddr2_3 =
+			_mm256_inserti128_si256(_mm256_castsi128_si256(vaddr2),
+						vaddr3, 1);
+
+		/* convert pa to dma_addr hdr/data */
+		dma_addr0_1 = _mm256_unpackhi_epi64(vaddr0_1, vaddr0_1);
+		dma_addr2_3 = _mm256_unpackhi_epi64(vaddr2_3, vaddr2_3);
+
+		/* add headroom to pa values */
+		dma_addr0_1 = _mm256_add_epi64(dma_addr0_1, hdr_room);
+		dma_addr2_3 = _mm256_add_epi64(dma_addr2_3, hdr_room);
+
+		/* flush desc with pa dma_addr */
+		_mm256_store_si256((__m256i *)&rxdp->read, dma_addr0_1);
+		_mm256_store_si256((__m256i *)&(rxdp + 2)->read, dma_addr2_3);
+	}
+
+#endif
+
+	rxq->rxrearm_start += IAVF_RXQ_REARM_THRESH;
+	if (rxq->rxrearm_start >= rxq->nb_rx_desc)
+		rxq->rxrearm_start = 0;
+
+	rxq->rxrearm_nb -= IAVF_RXQ_REARM_THRESH;
+
+	rx_id = (uint16_t)((rxq->rxrearm_start == 0) ?
+			     (rxq->nb_rx_desc - 1) : (rxq->rxrearm_start - 1));
+
+	/* Update the tail pointer on the NIC */
+	IAVF_PCI_REG_WRITE(rxq->qrx_tail, rx_id);
+}
+
+#define PKTLEN_SHIFT     10
+
+static inline uint16_t
+_iavf_recv_raw_pkts_vec_avx2(struct iavf_rx_queue *rxq,
+			     struct rte_mbuf **rx_pkts,
+			     uint16_t nb_pkts, uint8_t *split_packet)
+{
+#define IAVF_DESCS_PER_LOOP_AVX 8
+
+	/* const uint32_t *ptype_tbl = rxq->vsi->adapter->ptype_tbl; */
+	static const uint32_t type_table[UINT8_MAX + 1] __rte_cache_aligned = {
+		/* [0] reserved */
+		[1] = RTE_PTYPE_L2_ETHER,
+		/* [2] - [21] reserved */
+		[22] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
+			RTE_PTYPE_L4_FRAG,
+		[23] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
+			RTE_PTYPE_L4_NONFRAG,
+		[24] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
+			RTE_PTYPE_L4_UDP,
+		/* [25] reserved */
+		[26] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
+			RTE_PTYPE_L4_TCP,
+		[27] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
+			RTE_PTYPE_L4_SCTP,
+		[28] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
+			RTE_PTYPE_L4_ICMP,
+		/* All others reserved */
+	};
+	const __m256i mbuf_init = _mm256_set_epi64x(0, 0,
+			0, rxq->mbuf_initializer);
+	/* struct iavf_rx_entry *sw_ring = &rxq->sw_ring[rxq->rx_tail]; */
+	struct rte_mbuf **sw_ring = &rxq->sw_ring[rxq->rx_tail];
+	volatile union iavf_rx_desc *rxdp = rxq->rx_ring + rxq->rx_tail;
+	const int avx_aligned = ((rxq->rx_tail & 1) == 0);
+
+	rte_prefetch0(rxdp);
+
+	/* nb_pkts has to be floor-aligned to IAVF_DESCS_PER_LOOP_AVX */
+	nb_pkts = RTE_ALIGN_FLOOR(nb_pkts, IAVF_DESCS_PER_LOOP_AVX);
+
+	/* See if we need to rearm the RX queue - gives the prefetch a bit
+	 * of time to act
+	 */
+	if (rxq->rxrearm_nb > IAVF_RXQ_REARM_THRESH)
+		iavf_rxq_rearm(rxq);
+
+	/* Before we start moving massive data around, check to see if
+	 * there is actually a packet available
+	 */
+	if (!(rxdp->wb.qword1.status_error_len &
+			rte_cpu_to_le_32(1 << IAVF_RX_DESC_STATUS_DD_SHIFT)))
+		return 0;
+
+	/* constants used in processing loop */
+	const __m256i crc_adjust =
+		_mm256_set_epi16
+			(/* first descriptor */
+			 0, 0, 0,       /* ignore non-length fields */
+			 -rxq->crc_len, /* sub crc on data_len */
+			 0,             /* ignore high-16bits of pkt_len */
+			 -rxq->crc_len, /* sub crc on pkt_len */
+			 0, 0,          /* ignore pkt_type field */
+			 /* second descriptor */
+			 0, 0, 0,       /* ignore non-length fields */
+			 -rxq->crc_len, /* sub crc on data_len */
+			 0,             /* ignore high-16bits of pkt_len */
+			 -rxq->crc_len, /* sub crc on pkt_len */
+			 0, 0           /* ignore pkt_type field */
+			);
+
+	/* 8 packets DD mask, LSB in each 32-bit value */
+	const __m256i dd_check = _mm256_set1_epi32(1);
+
+	/* 8 packets EOP mask, second-LSB in each 32-bit value */
+	const __m256i eop_check = _mm256_slli_epi32(dd_check,
+			IAVF_RX_DESC_STATUS_EOF_SHIFT);
+
+	/* mask to shuffle from desc. to mbuf (2 descriptors)*/
+	const __m256i shuf_msk =
+		_mm256_set_epi8
+			(/* first descriptor */
+			 7, 6, 5, 4,  /* octet 4~7, 32bits rss */
+			 3, 2,        /* octet 2~3, low 16 bits vlan_macip */
+			 15, 14,      /* octet 15~14, 16 bits data_len */
+			 0xFF, 0xFF,  /* skip high 16 bits pkt_len, zero out */
+			 15, 14,      /* octet 15~14, low 16 bits pkt_len */
+			 0xFF, 0xFF,  /* pkt_type set as unknown */
+			 0xFF, 0xFF,  /*pkt_type set as unknown */
+			 /* second descriptor */
+			 7, 6, 5, 4,  /* octet 4~7, 32bits rss */
+			 3, 2,        /* octet 2~3, low 16 bits vlan_macip */
+			 15, 14,      /* octet 15~14, 16 bits data_len */
+			 0xFF, 0xFF,  /* skip high 16 bits pkt_len, zero out */
+			 15, 14,      /* octet 15~14, low 16 bits pkt_len */
+			 0xFF, 0xFF,  /* pkt_type set as unknown */
+			 0xFF, 0xFF   /*pkt_type set as unknown */
+			);
+	/**
+	 * compile-time check the above crc and shuffle layout is correct.
+	 * NOTE: the first field (lowest address) is given last in set_epi
+	 * calls above.
+	 */
+	RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, pkt_len) !=
+			offsetof(struct rte_mbuf, rx_descriptor_fields1) + 4);
+	RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, data_len) !=
+			offsetof(struct rte_mbuf, rx_descriptor_fields1) + 8);
+	RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, vlan_tci) !=
+			offsetof(struct rte_mbuf, rx_descriptor_fields1) + 10);
+	RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, hash) !=
+			offsetof(struct rte_mbuf, rx_descriptor_fields1) + 12);
+
+	/* Status/Error flag masks */
+	/**
+	 * mask everything except RSS, flow director and VLAN flags
+	 * bit2 is for VLAN tag, bit11 for flow director indication
+	 * bit13:12 for RSS indication. Bits 3-5 of error
+	 * field (bits 22-24) are for IP/L4 checksum errors
+	 */
+	const __m256i flags_mask =
+		 _mm256_set1_epi32((1 << 2) | (1 << 11) |
+				   (3 << 12) | (7 << 22));
+	/**
+	 * data to be shuffled by result of flag mask. If VLAN bit is set,
+	 * (bit 2), then position 4 in this array will be used in the
+	 * destination
+	 */
+	const __m256i vlan_flags_shuf =
+		_mm256_set_epi32(0, 0, PKT_RX_VLAN | PKT_RX_VLAN_STRIPPED, 0,
+				 0, 0, PKT_RX_VLAN | PKT_RX_VLAN_STRIPPED, 0);
+	/**
+	 * data to be shuffled by result of flag mask, shifted down 11.
+	 * If RSS/FDIR bits are set, shuffle moves appropriate flags in
+	 * place.
+	 */
+	const __m256i rss_flags_shuf =
+		_mm256_set_epi8(0, 0, 0, 0, 0, 0, 0, 0,
+				PKT_RX_RSS_HASH | PKT_RX_FDIR, PKT_RX_RSS_HASH,
+				0, 0, 0, 0, PKT_RX_FDIR, 0,/* end up 128-bits */
+				0, 0, 0, 0, 0, 0, 0, 0,
+				PKT_RX_RSS_HASH | PKT_RX_FDIR, PKT_RX_RSS_HASH,
+				0, 0, 0, 0, PKT_RX_FDIR, 0);
+
+	/**
+	 * data to be shuffled by the result of the flags mask shifted by 22
+	 * bits.  This gives use the l3_l4 flags.
+	 */
+	const __m256i l3_l4_flags_shuf = _mm256_set_epi8(0, 0, 0, 0, 0, 0, 0, 0,
+			/* shift right 1 bit to make sure it not exceed 255 */
+			(PKT_RX_EIP_CKSUM_BAD | PKT_RX_L4_CKSUM_BAD |
+			 PKT_RX_IP_CKSUM_BAD) >> 1,
+			(PKT_RX_IP_CKSUM_GOOD | PKT_RX_EIP_CKSUM_BAD |
+			 PKT_RX_L4_CKSUM_BAD) >> 1,
+			(PKT_RX_EIP_CKSUM_BAD | PKT_RX_IP_CKSUM_BAD) >> 1,
+			(PKT_RX_IP_CKSUM_GOOD | PKT_RX_EIP_CKSUM_BAD) >> 1,
+			(PKT_RX_L4_CKSUM_BAD | PKT_RX_IP_CKSUM_BAD) >> 1,
+			(PKT_RX_IP_CKSUM_GOOD | PKT_RX_L4_CKSUM_BAD) >> 1,
+			PKT_RX_IP_CKSUM_BAD >> 1,
+			(PKT_RX_IP_CKSUM_GOOD | PKT_RX_L4_CKSUM_GOOD) >> 1,
+			/* second 128-bits */
+			0, 0, 0, 0, 0, 0, 0, 0,
+			(PKT_RX_EIP_CKSUM_BAD | PKT_RX_L4_CKSUM_BAD |
+			 PKT_RX_IP_CKSUM_BAD) >> 1,
+			(PKT_RX_IP_CKSUM_GOOD | PKT_RX_EIP_CKSUM_BAD |
+			 PKT_RX_L4_CKSUM_BAD) >> 1,
+			(PKT_RX_EIP_CKSUM_BAD | PKT_RX_IP_CKSUM_BAD) >> 1,
+			(PKT_RX_IP_CKSUM_GOOD | PKT_RX_EIP_CKSUM_BAD) >> 1,
+			(PKT_RX_L4_CKSUM_BAD | PKT_RX_IP_CKSUM_BAD) >> 1,
+			(PKT_RX_IP_CKSUM_GOOD | PKT_RX_L4_CKSUM_BAD) >> 1,
+			PKT_RX_IP_CKSUM_BAD >> 1,
+			(PKT_RX_IP_CKSUM_GOOD | PKT_RX_L4_CKSUM_GOOD) >> 1);
+
+	const __m256i cksum_mask =
+		 _mm256_set1_epi32(PKT_RX_IP_CKSUM_GOOD | PKT_RX_IP_CKSUM_BAD |
+				   PKT_RX_L4_CKSUM_GOOD | PKT_RX_L4_CKSUM_BAD |
+				   PKT_RX_EIP_CKSUM_BAD);
+
+	RTE_SET_USED(avx_aligned); /* for 32B descriptors we don't use this */
+
+	uint16_t i, received;
+
+	for (i = 0, received = 0; i < nb_pkts;
+	     i += IAVF_DESCS_PER_LOOP_AVX,
+	     rxdp += IAVF_DESCS_PER_LOOP_AVX) {
+		/* step 1, copy over 8 mbuf pointers to rx_pkts array */
+		_mm256_storeu_si256((void *)&rx_pkts[i],
+				    _mm256_loadu_si256((void *)&sw_ring[i]));
+#ifdef RTE_ARCH_X86_64
+		_mm256_storeu_si256
+			((void *)&rx_pkts[i + 4],
+			 _mm256_loadu_si256((void *)&sw_ring[i + 4]));
+#endif
+
+		__m256i raw_desc0_1, raw_desc2_3, raw_desc4_5, raw_desc6_7;
+#ifdef RTE_LIBRTE_IAVF_16BYTE_RX_DESC
+		/* for AVX we need alignment otherwise loads are not atomic */
+		if (avx_aligned) {
+			/* load in descriptors, 2 at a time, in reverse order */
+			raw_desc6_7 = _mm256_load_si256((void *)(rxdp + 6));
+			rte_compiler_barrier();
+			raw_desc4_5 = _mm256_load_si256((void *)(rxdp + 4));
+			rte_compiler_barrier();
+			raw_desc2_3 = _mm256_load_si256((void *)(rxdp + 2));
+			rte_compiler_barrier();
+			raw_desc0_1 = _mm256_load_si256((void *)(rxdp + 0));
+		} else
+#endif
+		{
+			const __m128i raw_desc7 =
+				_mm_load_si128((void *)(rxdp + 7));
+			rte_compiler_barrier();
+			const __m128i raw_desc6 =
+				_mm_load_si128((void *)(rxdp + 6));
+			rte_compiler_barrier();
+			const __m128i raw_desc5 =
+				_mm_load_si128((void *)(rxdp + 5));
+			rte_compiler_barrier();
+			const __m128i raw_desc4 =
+				_mm_load_si128((void *)(rxdp + 4));
+			rte_compiler_barrier();
+			const __m128i raw_desc3 =
+				_mm_load_si128((void *)(rxdp + 3));
+			rte_compiler_barrier();
+			const __m128i raw_desc2 =
+				_mm_load_si128((void *)(rxdp + 2));
+			rte_compiler_barrier();
+			const __m128i raw_desc1 =
+				_mm_load_si128((void *)(rxdp + 1));
+			rte_compiler_barrier();
+			const __m128i raw_desc0 =
+				_mm_load_si128((void *)(rxdp + 0));
+
+			raw_desc6_7 =
+				_mm256_inserti128_si256
+					(_mm256_castsi128_si256(raw_desc6),
+					 raw_desc7, 1);
+			raw_desc4_5 =
+				_mm256_inserti128_si256
+					(_mm256_castsi128_si256(raw_desc4),
+					 raw_desc5, 1);
+			raw_desc2_3 =
+				_mm256_inserti128_si256
+					(_mm256_castsi128_si256(raw_desc2),
+					 raw_desc3, 1);
+			raw_desc0_1 =
+				_mm256_inserti128_si256
+					(_mm256_castsi128_si256(raw_desc0),
+					 raw_desc1, 1);
+		}
+
+		if (split_packet) {
+			int j;
+
+			for (j = 0; j < IAVF_DESCS_PER_LOOP_AVX; j++)
+				rte_mbuf_prefetch_part2(rx_pkts[i + j]);
+		}
+
+		/**
+		 * convert descriptors 4-7 into mbufs, adjusting length and
+		 * re-arranging fields. Then write into the mbuf
+		 */
+		const __m256i len6_7 = _mm256_slli_epi32(raw_desc6_7,
+							 PKTLEN_SHIFT);
+		const __m256i len4_5 = _mm256_slli_epi32(raw_desc4_5,
+							 PKTLEN_SHIFT);
+		const __m256i desc6_7 = _mm256_blend_epi16(raw_desc6_7,
+							   len6_7, 0x80);
+		const __m256i desc4_5 = _mm256_blend_epi16(raw_desc4_5,
+							   len4_5, 0x80);
+		__m256i mb6_7 = _mm256_shuffle_epi8(desc6_7, shuf_msk);
+		__m256i mb4_5 = _mm256_shuffle_epi8(desc4_5, shuf_msk);
+
+		mb6_7 = _mm256_add_epi16(mb6_7, crc_adjust);
+		mb4_5 = _mm256_add_epi16(mb4_5, crc_adjust);
+		/**
+		 * to get packet types, shift 64-bit values down 30 bits
+		 * and so ptype is in lower 8-bits in each
+		 */
+		const __m256i ptypes6_7 = _mm256_srli_epi64(desc6_7, 30);
+		const __m256i ptypes4_5 = _mm256_srli_epi64(desc4_5, 30);
+		const uint8_t ptype7 = _mm256_extract_epi8(ptypes6_7, 24);
+		const uint8_t ptype6 = _mm256_extract_epi8(ptypes6_7, 8);
+		const uint8_t ptype5 = _mm256_extract_epi8(ptypes4_5, 24);
+		const uint8_t ptype4 = _mm256_extract_epi8(ptypes4_5, 8);
+
+		mb6_7 = _mm256_insert_epi32(mb6_7, type_table[ptype7], 4);
+		mb6_7 = _mm256_insert_epi32(mb6_7, type_table[ptype6], 0);
+		mb4_5 = _mm256_insert_epi32(mb4_5, type_table[ptype5], 4);
+		mb4_5 = _mm256_insert_epi32(mb4_5, type_table[ptype4], 0);
+		/* merge the status bits into one register */
+		const __m256i status4_7 = _mm256_unpackhi_epi32(desc6_7,
+				desc4_5);
+
+		/**
+		 * convert descriptors 0-3 into mbufs, adjusting length and
+		 * re-arranging fields. Then write into the mbuf
+		 */
+		const __m256i len2_3 = _mm256_slli_epi32(raw_desc2_3,
+							 PKTLEN_SHIFT);
+		const __m256i len0_1 = _mm256_slli_epi32(raw_desc0_1,
+							 PKTLEN_SHIFT);
+		const __m256i desc2_3 = _mm256_blend_epi16(raw_desc2_3,
+							   len2_3, 0x80);
+		const __m256i desc0_1 = _mm256_blend_epi16(raw_desc0_1,
+							   len0_1, 0x80);
+		__m256i mb2_3 = _mm256_shuffle_epi8(desc2_3, shuf_msk);
+		__m256i mb0_1 = _mm256_shuffle_epi8(desc0_1, shuf_msk);
+
+		mb2_3 = _mm256_add_epi16(mb2_3, crc_adjust);
+		mb0_1 = _mm256_add_epi16(mb0_1, crc_adjust);
+		/* get the packet types */
+		const __m256i ptypes2_3 = _mm256_srli_epi64(desc2_3, 30);
+		const __m256i ptypes0_1 = _mm256_srli_epi64(desc0_1, 30);
+		const uint8_t ptype3 = _mm256_extract_epi8(ptypes2_3, 24);
+		const uint8_t ptype2 = _mm256_extract_epi8(ptypes2_3, 8);
+		const uint8_t ptype1 = _mm256_extract_epi8(ptypes0_1, 24);
+		const uint8_t ptype0 = _mm256_extract_epi8(ptypes0_1, 8);
+
+		mb2_3 = _mm256_insert_epi32(mb2_3, type_table[ptype3], 4);
+		mb2_3 = _mm256_insert_epi32(mb2_3, type_table[ptype2], 0);
+		mb0_1 = _mm256_insert_epi32(mb0_1, type_table[ptype1], 4);
+		mb0_1 = _mm256_insert_epi32(mb0_1, type_table[ptype0], 0);
+		/* merge the status bits into one register */
+		const __m256i status0_3 = _mm256_unpackhi_epi32(desc2_3,
+								desc0_1);
+
+		/**
+		 * take the two sets of status bits and merge to one
+		 * After merge, the packets status flags are in the
+		 * order (hi->lo): [1, 3, 5, 7, 0, 2, 4, 6]
+		 */
+		__m256i status0_7 = _mm256_unpacklo_epi64(status4_7,
+							  status0_3);
+
+		/* now do flag manipulation */
+
+		/* get only flag/error bits we want */
+		const __m256i flag_bits =
+			_mm256_and_si256(status0_7, flags_mask);
+		/* set vlan and rss flags */
+		const __m256i vlan_flags =
+			_mm256_shuffle_epi8(vlan_flags_shuf, flag_bits);
+		const __m256i rss_flags =
+			_mm256_shuffle_epi8(rss_flags_shuf,
+					    _mm256_srli_epi32(flag_bits, 11));
+		/**
+		 * l3_l4_error flags, shuffle, then shift to correct adjustment
+		 * of flags in flags_shuf, and finally mask out extra bits
+		 */
+		__m256i l3_l4_flags = _mm256_shuffle_epi8(l3_l4_flags_shuf,
+				_mm256_srli_epi32(flag_bits, 22));
+		l3_l4_flags = _mm256_slli_epi32(l3_l4_flags, 1);
+		l3_l4_flags = _mm256_and_si256(l3_l4_flags, cksum_mask);
+
+		/* merge flags */
+		const __m256i mbuf_flags = _mm256_or_si256(l3_l4_flags,
+				_mm256_or_si256(rss_flags, vlan_flags));
+		/**
+		 * At this point, we have the 8 sets of flags in the low 16-bits
+		 * of each 32-bit value in vlan0.
+		 * We want to extract these, and merge them with the mbuf init
+		 * data so we can do a single write to the mbuf to set the flags
+		 * and all the other initialization fields. Extracting the
+		 * appropriate flags means that we have to do a shift and blend
+		 * for each mbuf before we do the write. However, we can also
+		 * add in the previously computed rx_descriptor fields to
+		 * make a single 256-bit write per mbuf
+		 */
+		/* check the structure matches expectations */
+		RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, ol_flags) !=
+				 offsetof(struct rte_mbuf, rearm_data) + 8);
+		RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, rearm_data) !=
+				 RTE_ALIGN(offsetof(struct rte_mbuf,
+						    rearm_data),
+					   16));
+		/* build up data and do writes */
+		__m256i rearm0, rearm1, rearm2, rearm3, rearm4, rearm5,
+			rearm6, rearm7;
+		rearm6 = _mm256_blend_epi32(mbuf_init,
+					    _mm256_slli_si256(mbuf_flags, 8),
+					    0x04);
+		rearm4 = _mm256_blend_epi32(mbuf_init,
+					    _mm256_slli_si256(mbuf_flags, 4),
+					    0x04);
+		rearm2 = _mm256_blend_epi32(mbuf_init, mbuf_flags, 0x04);
+		rearm0 = _mm256_blend_epi32(mbuf_init,
+					    _mm256_srli_si256(mbuf_flags, 4),
+					    0x04);
+		/* permute to add in the rx_descriptor e.g. rss fields */
+		rearm6 = _mm256_permute2f128_si256(rearm6, mb6_7, 0x20);
+		rearm4 = _mm256_permute2f128_si256(rearm4, mb4_5, 0x20);
+		rearm2 = _mm256_permute2f128_si256(rearm2, mb2_3, 0x20);
+		rearm0 = _mm256_permute2f128_si256(rearm0, mb0_1, 0x20);
+		/* write to mbuf */
+		_mm256_storeu_si256((__m256i *)&rx_pkts[i + 6]->rearm_data,
+				    rearm6);
+		_mm256_storeu_si256((__m256i *)&rx_pkts[i + 4]->rearm_data,
+				    rearm4);
+		_mm256_storeu_si256((__m256i *)&rx_pkts[i + 2]->rearm_data,
+				    rearm2);
+		_mm256_storeu_si256((__m256i *)&rx_pkts[i + 0]->rearm_data,
+				    rearm0);
+
+		/* repeat for the odd mbufs */
+		const __m256i odd_flags =
+			_mm256_castsi128_si256
+				(_mm256_extracti128_si256(mbuf_flags, 1));
+		rearm7 = _mm256_blend_epi32(mbuf_init,
+					    _mm256_slli_si256(odd_flags, 8),
+					    0x04);
+		rearm5 = _mm256_blend_epi32(mbuf_init,
+					    _mm256_slli_si256(odd_flags, 4),
+					    0x04);
+		rearm3 = _mm256_blend_epi32(mbuf_init, odd_flags, 0x04);
+		rearm1 = _mm256_blend_epi32(mbuf_init,
+					    _mm256_srli_si256(odd_flags, 4),
+					    0x04);
+		/* since odd mbufs are already in hi 128-bits use blend */
+		rearm7 = _mm256_blend_epi32(rearm7, mb6_7, 0xF0);
+		rearm5 = _mm256_blend_epi32(rearm5, mb4_5, 0xF0);
+		rearm3 = _mm256_blend_epi32(rearm3, mb2_3, 0xF0);
+		rearm1 = _mm256_blend_epi32(rearm1, mb0_1, 0xF0);
+		/* again write to mbufs */
+		_mm256_storeu_si256((__m256i *)&rx_pkts[i + 7]->rearm_data,
+				    rearm7);
+		_mm256_storeu_si256((__m256i *)&rx_pkts[i + 5]->rearm_data,
+				    rearm5);
+		_mm256_storeu_si256((__m256i *)&rx_pkts[i + 3]->rearm_data,
+				    rearm3);
+		_mm256_storeu_si256((__m256i *)&rx_pkts[i + 1]->rearm_data,
+				    rearm1);
+
+		/* extract and record EOP bit */
+		if (split_packet) {
+			const __m128i eop_mask =
+				_mm_set1_epi16(1 << IAVF_RX_DESC_STATUS_EOF_SHIFT);
+			const __m256i eop_bits256 = _mm256_and_si256(status0_7,
+								     eop_check);
+			/* pack status bits into a single 128-bit register */
+			const __m128i eop_bits =
+				_mm_packus_epi32
+					(_mm256_castsi256_si128(eop_bits256),
+					 _mm256_extractf128_si256(eop_bits256,
+								  1));
+			/**
+			 * flip bits, and mask out the EOP bit, which is now
+			 * a split-packet bit i.e. !EOP, rather than EOP one.
+			 */
+			__m128i split_bits = _mm_andnot_si128(eop_bits,
+					eop_mask);
+			/**
+			 * eop bits are out of order, so we need to shuffle them
+			 * back into order again. In doing so, only use low 8
+			 * bits, which acts like another pack instruction
+			 * The original order is (hi->lo): 1,3,5,7,0,2,4,6
+			 * [Since we use epi8, the 16-bit positions are
+			 * multiplied by 2 in the eop_shuffle value.]
+			 */
+			__m128i eop_shuffle =
+				_mm_set_epi8(/* zero hi 64b */
+					     0xFF, 0xFF, 0xFF, 0xFF,
+					     0xFF, 0xFF, 0xFF, 0xFF,
+					     /* move values to lo 64b */
+					     8, 0, 10, 2,
+					     12, 4, 14, 6);
+			split_bits = _mm_shuffle_epi8(split_bits, eop_shuffle);
+			*(uint64_t *)split_packet =
+				_mm_cvtsi128_si64(split_bits);
+			split_packet += IAVF_DESCS_PER_LOOP_AVX;
+		}
+
+		/* perform dd_check */
+		status0_7 = _mm256_and_si256(status0_7, dd_check);
+		status0_7 = _mm256_packs_epi32(status0_7,
+					       _mm256_setzero_si256());
+
+		uint64_t burst = __builtin_popcountll
+					(_mm_cvtsi128_si64
+						(_mm256_extracti128_si256
+							(status0_7, 1)));
+		burst += __builtin_popcountll
+				(_mm_cvtsi128_si64
+					(_mm256_castsi256_si128(status0_7)));
+		received += burst;
+		if (burst != IAVF_DESCS_PER_LOOP_AVX)
+			break;
+	}
+
+	/* update tail pointers */
+	rxq->rx_tail += received;
+	rxq->rx_tail &= (rxq->nb_rx_desc - 1);
+	if ((rxq->rx_tail & 1) == 1 && received > 1) { /* keep avx2 aligned */
+		rxq->rx_tail--;
+		received--;
+	}
+	rxq->rxrearm_nb += received;
+	return received;
+}
+
+/**
+ * Notice:
+ * - nb_pkts < IAVF_DESCS_PER_LOOP, just return no packet
+ */
+uint16_t
+iavf_recv_pkts_vec_avx2(void *rx_queue, struct rte_mbuf **rx_pkts,
+			uint16_t nb_pkts)
+{
+	return _iavf_recv_raw_pkts_vec_avx2(rx_queue, rx_pkts, nb_pkts, NULL);
+}
+
+/**
+ * vPMD receive routine that reassembles single burst of 32 scattered packets
+ * Notice:
+ * - nb_pkts < IAVF_DESCS_PER_LOOP, just return no packet
+ */
+static uint16_t
+iavf_recv_scattered_burst_vec_avx2(void *rx_queue, struct rte_mbuf **rx_pkts,
+				   uint16_t nb_pkts)
+{
+	struct iavf_rx_queue *rxq = rx_queue;
+	uint8_t split_flags[IAVF_VPMD_RX_MAX_BURST] = {0};
+
+	/* get some new buffers */
+	uint16_t nb_bufs = _iavf_recv_raw_pkts_vec_avx2(rxq, rx_pkts, nb_pkts,
+						       split_flags);
+	if (nb_bufs == 0)
+		return 0;
+
+	/* happy day case, full burst + no packets to be joined */
+	const uint64_t *split_fl64 = (uint64_t *)split_flags;
+
+	if (!rxq->pkt_first_seg &&
+	    split_fl64[0] == 0 && split_fl64[1] == 0 &&
+	    split_fl64[2] == 0 && split_fl64[3] == 0)
+		return nb_bufs;
+
+	/* reassemble any packets that need reassembly*/
+	unsigned int i = 0;
+
+	if (!rxq->pkt_first_seg) {
+		/* find the first split flag, and only reassemble then*/
+		while (i < nb_bufs && !split_flags[i])
+			i++;
+		if (i == nb_bufs)
+			return nb_bufs;
+		rxq->pkt_first_seg = rx_pkts[i];
+	}
+	return i + reassemble_packets(rxq, &rx_pkts[i], nb_bufs - i,
+					     &split_flags[i]);
+}
+
+/**
+ * vPMD receive routine that reassembles scattered packets.
+ * Main receive routine that can handle arbitrary burst sizes
+ * Notice:
+ * - nb_pkts < IAVF_DESCS_PER_LOOP, just return no packet
+ */
+uint16_t
+iavf_recv_scattered_pkts_vec_avx2(void *rx_queue, struct rte_mbuf **rx_pkts,
+				  uint16_t nb_pkts)
+{
+	uint16_t retval = 0;
+
+	while (nb_pkts > IAVF_VPMD_RX_MAX_BURST) {
+		uint16_t burst = iavf_recv_scattered_burst_vec_avx2(rx_queue,
+				rx_pkts + retval, IAVF_VPMD_RX_MAX_BURST);
+		retval += burst;
+		nb_pkts -= burst;
+		if (burst < IAVF_VPMD_RX_MAX_BURST)
+			return retval;
+	}
+	return retval + iavf_recv_scattered_burst_vec_avx2(rx_queue,
+				rx_pkts + retval, nb_pkts);
+}
+
+static inline void
+iavf_vtx1(volatile struct iavf_tx_desc *txdp,
+	  struct rte_mbuf *pkt, uint64_t flags)
+{
+	uint64_t high_qw =
+		(IAVF_TX_DESC_DTYPE_DATA |
+		 ((uint64_t)flags  << IAVF_TXD_QW1_CMD_SHIFT) |
+		 ((uint64_t)pkt->data_len << IAVF_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
+iavf_vtx(volatile struct iavf_tx_desc *txdp,
+	 struct rte_mbuf **pkt, uint16_t nb_pkts,  uint64_t flags)
+{
+	const uint64_t hi_qw_tmpl = (IAVF_TX_DESC_DTYPE_DATA |
+			((uint64_t)flags  << IAVF_TXD_QW1_CMD_SHIFT));
+
+	/* if unaligned on 32-bit boundary, do one to align */
+	if (((uintptr_t)txdp & 0x1F) != 0 && nb_pkts != 0) {
+		iavf_vtx1(txdp, *pkt, flags);
+		nb_pkts--, txdp++, pkt++;
+	}
+
+	/* do two at a time while possible, in bursts */
+	for (; nb_pkts > 3; txdp += 4, pkt += 4, nb_pkts -= 4) {
+		uint64_t hi_qw3 =
+			hi_qw_tmpl |
+			((uint64_t)pkt[3]->data_len <<
+			 IAVF_TXD_QW1_TX_BUF_SZ_SHIFT);
+		uint64_t hi_qw2 =
+			hi_qw_tmpl |
+			((uint64_t)pkt[2]->data_len <<
+			 IAVF_TXD_QW1_TX_BUF_SZ_SHIFT);
+		uint64_t hi_qw1 =
+			hi_qw_tmpl |
+			((uint64_t)pkt[1]->data_len <<
+			 IAVF_TXD_QW1_TX_BUF_SZ_SHIFT);
+		uint64_t hi_qw0 =
+			hi_qw_tmpl |
+			((uint64_t)pkt[0]->data_len <<
+			 IAVF_TXD_QW1_TX_BUF_SZ_SHIFT);
+
+		__m256i desc2_3 =
+			_mm256_set_epi64x
+				(hi_qw3,
+				 pkt[3]->buf_physaddr + pkt[3]->data_off,
+				 hi_qw2,
+				 pkt[2]->buf_physaddr + pkt[2]->data_off);
+		__m256i desc0_1 =
+			_mm256_set_epi64x
+				(hi_qw1,
+				 pkt[1]->buf_physaddr + pkt[1]->data_off,
+				 hi_qw0,
+				 pkt[0]->buf_physaddr + pkt[0]->data_off);
+		_mm256_store_si256((void *)(txdp + 2), desc2_3);
+		_mm256_store_si256((void *)txdp, desc0_1);
+	}
+
+	/* do any last ones */
+	while (nb_pkts) {
+		iavf_vtx1(txdp, *pkt, flags);
+		txdp++, pkt++, nb_pkts--;
+	}
+}
+
+static inline uint16_t
+iavf_xmit_fixed_burst_vec_avx2(void *tx_queue, struct rte_mbuf **tx_pkts,
+			       uint16_t nb_pkts)
+{
+	struct iavf_tx_queue *txq = (struct iavf_tx_queue *)tx_queue;
+	volatile struct iavf_tx_desc *txdp;
+	struct iavf_tx_entry *txep;
+	uint16_t n, nb_commit, tx_id;
+	uint64_t flags = IAVF_TX_DESC_CMD_EOP;
+	uint64_t rs = IAVF_TX_DESC_CMD_RS | IAVF_TX_DESC_CMD_EOP;
+
+	/* cross rx_thresh boundary is not allowed */
+	nb_pkts = RTE_MIN(nb_pkts, txq->rs_thresh);
+
+	if (txq->nb_free < txq->free_thresh)
+		iavf_tx_free_bufs(txq);
+
+	nb_commit = nb_pkts = (uint16_t)RTE_MIN(txq->nb_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_free = (uint16_t)(txq->nb_free - nb_pkts);
+
+	n = (uint16_t)(txq->nb_tx_desc - tx_id);
+	if (nb_commit >= n) {
+		tx_backlog_entry(txep, tx_pkts, n);
+
+		iavf_vtx(txdp, tx_pkts, n - 1, flags);
+		tx_pkts += (n - 1);
+		txdp += (n - 1);
+
+		iavf_vtx1(txdp, *tx_pkts++, rs);
+
+		nb_commit = (uint16_t)(nb_commit - n);
+
+		tx_id = 0;
+		txq->next_rs = (uint16_t)(txq->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);
+
+	iavf_vtx(txdp, tx_pkts, nb_commit, flags);
+
+	tx_id = (uint16_t)(tx_id + nb_commit);
+	if (tx_id > txq->next_rs) {
+		txq->tx_ring[txq->next_rs].cmd_type_offset_bsz |=
+			rte_cpu_to_le_64(((uint64_t)IAVF_TX_DESC_CMD_RS) <<
+					 IAVF_TXD_QW1_CMD_SHIFT);
+		txq->next_rs =
+			(uint16_t)(txq->next_rs + txq->rs_thresh);
+	}
+
+	txq->tx_tail = tx_id;
+
+	IAVF_PCI_REG_WRITE(txq->qtx_tail, txq->tx_tail);
+
+	return nb_pkts;
+}
+
+uint16_t
+iavf_xmit_pkts_vec_avx2(void *tx_queue, struct rte_mbuf **tx_pkts,
+			uint16_t nb_pkts)
+{
+	uint16_t nb_tx = 0;
+	struct iavf_tx_queue *txq = (struct iavf_tx_queue *)tx_queue;
+
+	while (nb_pkts) {
+		uint16_t ret, num;
+
+		num = (uint16_t)RTE_MIN(nb_pkts, txq->rs_thresh);
+		ret = iavf_xmit_fixed_burst_vec_avx2(tx_queue, &tx_pkts[nb_tx],
+						     num);
+		nb_tx += ret;
+		nb_pkts -= ret;
+		if (ret < num)
+			break;
+	}
+
+	return nb_tx;
+}
diff --git a/drivers/net/iavf/iavf_rxtx_vec_common.h b/drivers/net/iavf/iavf_rxtx_vec_common.h
index db509d71f..a6ba22758 100644
--- a/drivers/net/iavf/iavf_rxtx_vec_common.h
+++ b/drivers/net/iavf/iavf_rxtx_vec_common.h
@@ -207,4 +207,69 @@ iavf_rxq_vec_setup_default(struct iavf_rx_queue *rxq)
 	rxq->mbuf_initializer = *(uint64_t *)p;
 	return 0;
 }
+
+static inline int
+iavf_rx_vec_queue_default(struct iavf_rx_queue *rxq)
+{
+	if (!rxq)
+		return -1;
+
+	if (!rte_is_power_of_2(rxq->nb_rx_desc))
+		return -1;
+
+	if (rxq->rx_free_thresh < IAVF_VPMD_RX_MAX_BURST)
+		return -1;
+
+	if (rxq->nb_rx_desc % rxq->rx_free_thresh)
+		return -1;
+
+	return 0;
+}
+
+static inline int
+iavf_tx_vec_queue_default(struct iavf_tx_queue *txq)
+{
+	if (!txq)
+		return -1;
+
+	if (txq->offloads & IAVF_NO_VECTOR_FLAGS)
+		return -1;
+
+	if (txq->rs_thresh < IAVF_VPMD_TX_MAX_BURST ||
+	    txq->rs_thresh > IAVF_VPMD_TX_MAX_FREE_BUF)
+		return -1;
+
+	return 0;
+}
+
+static inline int
+iavf_rx_vec_dev_check_default(struct rte_eth_dev *dev)
+{
+	int i;
+	struct iavf_rx_queue *rxq;
+
+	for (i = 0; i < dev->data->nb_rx_queues; i++) {
+		rxq = dev->data->rx_queues[i];
+		if (iavf_rx_vec_queue_default(rxq))
+			return -1;
+	}
+
+	return 0;
+}
+
+static inline int
+iavf_tx_vec_dev_check_default(struct rte_eth_dev *dev)
+{
+	int i;
+	struct iavf_tx_queue *txq;
+
+	for (i = 0; i < dev->data->nb_tx_queues; i++) {
+		txq = dev->data->tx_queues[i];
+		if (iavf_tx_vec_queue_default(txq))
+			return -1;
+	}
+
+	return 0;
+}
+
 #endif
diff --git a/drivers/net/iavf/iavf_rxtx_vec_sse.c b/drivers/net/iavf/iavf_rxtx_vec_sse.c
index cc71f23a5..2b16dc1b5 100644
--- a/drivers/net/iavf/iavf_rxtx_vec_sse.c
+++ b/drivers/net/iavf/iavf_rxtx_vec_sse.c
@@ -622,6 +622,27 @@ iavf_xmit_fixed_burst_vec(void *tx_queue, struct rte_mbuf **tx_pkts,
 	return nb_pkts;
 }
 
+uint16_t
+iavf_xmit_pkts_vec(void *tx_queue, struct rte_mbuf **tx_pkts,
+		   uint16_t nb_pkts)
+{
+	uint16_t nb_tx = 0;
+	struct iavf_tx_queue *txq = (struct iavf_tx_queue *)tx_queue;
+
+	while (nb_pkts) {
+		uint16_t ret, num;
+
+		num = (uint16_t)RTE_MIN(nb_pkts, txq->rs_thresh);
+		ret = iavf_xmit_fixed_burst_vec(tx_queue, &tx_pkts[nb_tx], num);
+		nb_tx += ret;
+		nb_pkts -= ret;
+		if (ret < num)
+			break;
+	}
+
+	return nb_tx;
+}
+
 static void __attribute__((cold))
 iavf_rx_queue_release_mbufs_sse(struct iavf_rx_queue *rxq)
 {
@@ -655,3 +676,15 @@ iavf_rxq_vec_setup(struct iavf_rx_queue *rxq)
 	rxq->ops = &sse_vec_rxq_ops;
 	return iavf_rxq_vec_setup_default(rxq);
 }
+
+int __attribute__((cold))
+iavf_rx_vec_dev_check(struct rte_eth_dev *dev)
+{
+	return iavf_rx_vec_dev_check_default(dev);
+}
+
+int __attribute__((cold))
+iavf_tx_vec_dev_check(struct rte_eth_dev *dev)
+{
+	return iavf_tx_vec_dev_check_default(dev);
+}
diff --git a/drivers/net/iavf/meson.build b/drivers/net/iavf/meson.build
index e5a2f5553..ce820573b 100644
--- a/drivers/net/iavf/meson.build
+++ b/drivers/net/iavf/meson.build
@@ -15,6 +15,22 @@ sources = files(
 )
 
 if arch_subdir == 'x86'
-	dpdk_conf.set('RTE_LIBRTE_IAVF_INC_VECTOR', 1)
 	sources += files('iavf_rxtx_vec_sse.c')
+
+	# compile AVX2 version if either:
+	# a. we have AVX supported in minimum instruction set baseline
+	# b. it's not minimum instruction set, but supported by compiler
+	if dpdk_conf.has('RTE_MACHINE_CPUFLAG_AVX2')
+		cflags += ['-DCC_AVX2_SUPPORT']
+		sources += files('iavf_rxtx_vec_avx2.c')
+	elif cc.has_argument('-mavx2')
+		cflags += ['-DCC_AVX2_SUPPORT']
+		iavf_avx2_lib = static_library('iavf_avx2_lib',
+				'iavf_rxtx_vec_avx2.c',
+				dependencies: [static_rte_ethdev,
+					static_rte_kvargs, static_rte_hash],
+				include_directories: includes,
+				c_args: [cflags, '-mavx2'])
+		objs += iavf_avx2_lib.extract_objects('iavf_rxtx_vec_avx2.c')
+	endif
 endif
-- 
2.17.1


^ permalink raw reply	[flat|nested] 11+ messages in thread

* Re: [dpdk-dev] [PATCH v5] net/iavf: enable AVX2 for iavf
  2019-10-16  8:14 ` [dpdk-dev] [PATCH v5] " Leyi Rong
@ 2019-10-20  1:12   ` Ye Xiaolong
  0 siblings, 0 replies; 11+ messages in thread
From: Ye Xiaolong @ 2019-10-20  1:12 UTC (permalink / raw)
  To: Leyi Rong; +Cc: wenzhuo.lu, qi.z.zhang, dev

On 10/16, Leyi Rong wrote:
>This patch enables AVX data path for iavf PMD.
>
>Signed-off-by: Leyi Rong <leyi.rong@intel.com>
>
>---
>v5:
>- Eliminates CONFIG_RTE_LIBRTE_IAVF_INC_VECTOR and __rte_weak.
>
>v4:
>- Romoves redundant IAVF_NO_VECTOR_FLAGS define.
>
>v3:
>- Adds release note.
>
>v2:
>- Fixes build error on aarch64.
>---
> config/common_base                      |   1 -
> doc/guides/rel_notes/release_19_11.rst  |   4 +
> drivers/net/iavf/Makefile               |  23 +-
> drivers/net/iavf/iavf_ethdev.c          |   5 -
> drivers/net/iavf/iavf_rxtx.c            | 141 ++--
> drivers/net/iavf/iavf_rxtx.h            |  12 +
> drivers/net/iavf/iavf_rxtx_vec_avx2.c   | 867 ++++++++++++++++++++++++
> drivers/net/iavf/iavf_rxtx_vec_common.h |  65 ++
> drivers/net/iavf/iavf_rxtx_vec_sse.c    |  33 +
> drivers/net/iavf/meson.build            |  18 +-
> 10 files changed, 1074 insertions(+), 95 deletions(-)
> create mode 100644 drivers/net/iavf/iavf_rxtx_vec_avx2.c
>

Applied to dpdk-next-net-intel.

Thanks,
Xiaolong

^ permalink raw reply	[flat|nested] 11+ messages in thread

end of thread, other threads:[~2019-10-20  1:15 UTC | newest]

Thread overview: 11+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2019-09-04 10:06 [dpdk-dev] [PATCH] net/iavf: enable AVX2 for iavf Leyi Rong
2019-09-04 13:29 ` Aaron Conole
2019-09-06  3:28 ` [dpdk-dev] [PATCH v2] " Leyi Rong
2019-09-19  8:44 ` [dpdk-dev] [PATCH v3] " Leyi Rong
2019-09-19 12:55   ` Zhang, Qi Z
2019-09-23  6:02 ` [dpdk-dev] [PATCH v4] " Leyi Rong
2019-09-23  6:49   ` Zhang, Qi Z
2019-09-30  9:47   ` Ye Xiaolong
2019-09-30 13:08   ` Ferruh Yigit
2019-10-16  8:14 ` [dpdk-dev] [PATCH v5] " Leyi Rong
2019-10-20  1:12   ` Ye Xiaolong

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).