net/i40e: make vector driver filenames consistent
authorJianbo Liu <jianbo.liu@linaro.org>
Fri, 14 Oct 2016 04:00:03 +0000 (09:30 +0530)
committerBruce Richardson <bruce.richardson@intel.com>
Wed, 26 Oct 2016 17:38:17 +0000 (19:38 +0200)
To be consistent with the naming for ARM NEON implementation,
i40e_rxtx_vec.c is renamed to i40e_rxtx_vec_sse.c.

Signed-off-by: Jianbo Liu <jianbo.liu@linaro.org>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
drivers/net/i40e/Makefile
drivers/net/i40e/i40e_rxtx_vec.c [deleted file]
drivers/net/i40e/i40e_rxtx_vec_sse.c [new file with mode: 0644]

index 9e92b38..13085fb 100644 (file)
@@ -100,7 +100,7 @@ SRCS-$(CONFIG_RTE_LIBRTE_I40E_PMD) += i40e_rxtx.c
 ifeq ($(CONFIG_RTE_ARCH_ARM64),y)
 SRCS-$(CONFIG_RTE_LIBRTE_I40E_INC_VECTOR) += i40e_rxtx_vec_neon.c
 else
-SRCS-$(CONFIG_RTE_LIBRTE_I40E_INC_VECTOR) += i40e_rxtx_vec.c
+SRCS-$(CONFIG_RTE_LIBRTE_I40E_INC_VECTOR) += i40e_rxtx_vec_sse.c
 endif
 SRCS-$(CONFIG_RTE_LIBRTE_I40E_PMD) += i40e_ethdev_vf.c
 SRCS-$(CONFIG_RTE_LIBRTE_I40E_PMD) += i40e_pf.c
@@ -108,7 +108,7 @@ SRCS-$(CONFIG_RTE_LIBRTE_I40E_PMD) += i40e_fdir.c
 
 # vector PMD driver needs SSE4.1 support
 ifeq ($(findstring RTE_MACHINE_CPUFLAG_SSE4_1,$(CFLAGS)),)
-CFLAGS_i40e_rxtx_vec.o += -msse4.1
+CFLAGS_i40e_rxtx_vec_sse.o += -msse4.1
 endif
 
 
diff --git a/drivers/net/i40e/i40e_rxtx_vec.c b/drivers/net/i40e/i40e_rxtx_vec.c
deleted file mode 100644 (file)
index 3607312..0000000
+++ /dev/null
@@ -1,609 +0,0 @@
-/*-
- *   BSD LICENSE
- *
- *   Copyright(c) 2010-2015 Intel Corporation. All rights reserved.
- *   All rights reserved.
- *
- *   Redistribution and use in source and binary forms, with or without
- *   modification, are permitted provided that the following conditions
- *   are met:
- *
- *     * Redistributions of source code must retain the above copyright
- *       notice, this list of conditions and the following disclaimer.
- *     * Redistributions in binary form must reproduce the above copyright
- *       notice, this list of conditions and the following disclaimer in
- *       the documentation and/or other materials provided with the
- *       distribution.
- *     * Neither the name of Intel Corporation nor the names of its
- *       contributors may be used to endorse or promote products derived
- *       from this software without specific prior written permission.
- *
- *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
- *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
- *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
- *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
- *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
- *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
- *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
- *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
- *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
- *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
- *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
- */
-
-#include <stdint.h>
-#include <rte_ethdev.h>
-#include <rte_malloc.h>
-
-#include "base/i40e_prototype.h"
-#include "base/i40e_type.h"
-#include "i40e_ethdev.h"
-#include "i40e_rxtx.h"
-#include "i40e_rxtx_vec_common.h"
-
-#include <tmmintrin.h>
-
-#ifndef __INTEL_COMPILER
-#pragma GCC diagnostic ignored "-Wcast-qual"
-#endif
-
-static inline void
-i40e_rxq_rearm(struct i40e_rx_queue *rxq)
-{
-       int i;
-       uint16_t rx_id;
-       volatile union i40e_rx_desc *rxdp;
-       struct i40e_rx_entry *rxep = &rxq->sw_ring[rxq->rxrearm_start];
-       struct rte_mbuf *mb0, *mb1;
-       __m128i hdr_room = _mm_set_epi64x(RTE_PKTMBUF_HEADROOM,
-                       RTE_PKTMBUF_HEADROOM);
-       __m128i dma_addr0, dma_addr1;
-
-       rxdp = rxq->rx_ring + rxq->rxrearm_start;
-
-       /* Pull 'n' more MBUFs into the software ring */
-       if (rte_mempool_get_bulk(rxq->mp,
-                                (void *)rxep,
-                                RTE_I40E_RXQ_REARM_THRESH) < 0) {
-               if (rxq->rxrearm_nb + RTE_I40E_RXQ_REARM_THRESH >=
-                   rxq->nb_rx_desc) {
-                       dma_addr0 = _mm_setzero_si128();
-                       for (i = 0; i < RTE_I40E_DESCS_PER_LOOP; i++) {
-                               rxep[i].mbuf = &rxq->fake_mbuf;
-                               _mm_store_si128((__m128i *)&rxdp[i].read,
-                                               dma_addr0);
-                       }
-               }
-               rte_eth_devices[rxq->port_id].data->rx_mbuf_alloc_failed +=
-                       RTE_I40E_RXQ_REARM_THRESH;
-               return;
-       }
-
-       /* Initialize the mbufs in vector, process 2 mbufs in one loop */
-       for (i = 0; i < RTE_I40E_RXQ_REARM_THRESH; i += 2, rxep += 2) {
-               __m128i vaddr0, vaddr1;
-               uintptr_t p0, p1;
-
-               mb0 = rxep[0].mbuf;
-               mb1 = rxep[1].mbuf;
-
-                /* Flush mbuf with pkt template.
-                * Data to be rearmed is 6 bytes long.
-                * Though, RX will overwrite ol_flags that are coming next
-                * anyway. So overwrite whole 8 bytes with one load:
-                * 6 bytes of rearm_data plus first 2 bytes of ol_flags.
-                */
-               p0 = (uintptr_t)&mb0->rearm_data;
-               *(uint64_t *)p0 = rxq->mbuf_initializer;
-               p1 = (uintptr_t)&mb1->rearm_data;
-               *(uint64_t *)p1 = rxq->mbuf_initializer;
-
-               /* load buf_addr(lo 64bit) and buf_physaddr(hi 64bit) */
-               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);
-       }
-
-       rxq->rxrearm_start += RTE_I40E_RXQ_REARM_THRESH;
-       if (rxq->rxrearm_start >= rxq->nb_rx_desc)
-               rxq->rxrearm_start = 0;
-
-       rxq->rxrearm_nb -= RTE_I40E_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 */
-       I40E_PCI_REG_WRITE(rxq->qrx_tail, rx_id);
-}
-
-/* Handling the offload flags (olflags) field takes computation
- * time when receiving packets. Therefore we provide a flag to disable
- * the processing of the olflags field when they are not needed. This
- * gives improved performance, at the cost of losing the offload info
- * in the received packet
- */
-#ifdef RTE_LIBRTE_I40E_RX_OLFLAGS_ENABLE
-
-static inline void
-desc_to_olflags_v(__m128i descs[4], struct rte_mbuf **rx_pkts)
-{
-       __m128i vlan0, vlan1, rss, l3_l4e;
-
-       /* mask everything except RSS, flow director and VLAN flags
-        * bit2 is for VLAN tag, bit11 for flow director indication
-        * bit13:12 for RSS indication.
-        */
-       const __m128i rss_vlan_msk = _mm_set_epi32(
-                       0x1c03804, 0x1c03804, 0x1c03804, 0x1c03804);
-
-       /* map rss and vlan type to rss hash and vlan flag */
-       const __m128i vlan_flags = _mm_set_epi8(0, 0, 0, 0,
-                       0, 0, 0, 0,
-                       0, 0, 0, PKT_RX_VLAN_PKT | PKT_RX_VLAN_STRIPPED,
-                       0, 0, 0, 0);
-
-       const __m128i rss_flags = _mm_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);
-
-       const __m128i l3_l4e_flags = _mm_set_epi8(0, 0, 0, 0, 0, 0, 0, 0,
-                       PKT_RX_EIP_CKSUM_BAD | PKT_RX_L4_CKSUM_BAD | PKT_RX_IP_CKSUM_BAD,
-                       PKT_RX_EIP_CKSUM_BAD | PKT_RX_L4_CKSUM_BAD,
-                       PKT_RX_EIP_CKSUM_BAD | PKT_RX_IP_CKSUM_BAD,
-                       PKT_RX_EIP_CKSUM_BAD,
-                       PKT_RX_L4_CKSUM_BAD | PKT_RX_IP_CKSUM_BAD,
-                       PKT_RX_L4_CKSUM_BAD,
-                       PKT_RX_IP_CKSUM_BAD,
-                       0);
-
-       vlan0 = _mm_unpackhi_epi32(descs[0], descs[1]);
-       vlan1 = _mm_unpackhi_epi32(descs[2], descs[3]);
-       vlan0 = _mm_unpacklo_epi64(vlan0, vlan1);
-
-       vlan1 = _mm_and_si128(vlan0, rss_vlan_msk);
-       vlan0 = _mm_shuffle_epi8(vlan_flags, vlan1);
-
-       rss = _mm_srli_epi32(vlan1, 11);
-       rss = _mm_shuffle_epi8(rss_flags, rss);
-
-       l3_l4e = _mm_srli_epi32(vlan1, 22);
-       l3_l4e = _mm_shuffle_epi8(l3_l4e_flags, l3_l4e);
-
-       vlan0 = _mm_or_si128(vlan0, rss);
-       vlan0 = _mm_or_si128(vlan0, l3_l4e);
-
-       rx_pkts[0]->ol_flags = _mm_extract_epi16(vlan0, 0);
-       rx_pkts[1]->ol_flags = _mm_extract_epi16(vlan0, 2);
-       rx_pkts[2]->ol_flags = _mm_extract_epi16(vlan0, 4);
-       rx_pkts[3]->ol_flags = _mm_extract_epi16(vlan0, 6);
-}
-#else
-#define desc_to_olflags_v(desc, rx_pkts) do {} while (0)
-#endif
-
-#define PKTLEN_SHIFT     10
-
-static inline void
-desc_to_ptype_v(__m128i descs[4], struct rte_mbuf **rx_pkts)
-{
-       __m128i ptype0 = _mm_unpackhi_epi64(descs[0], descs[1]);
-       __m128i ptype1 = _mm_unpackhi_epi64(descs[2], descs[3]);
-
-       ptype0 = _mm_srli_epi64(ptype0, 30);
-       ptype1 = _mm_srli_epi64(ptype1, 30);
-
-       rx_pkts[0]->packet_type = i40e_rxd_pkt_type_mapping(_mm_extract_epi8(ptype0, 0));
-       rx_pkts[1]->packet_type = i40e_rxd_pkt_type_mapping(_mm_extract_epi8(ptype0, 8));
-       rx_pkts[2]->packet_type = i40e_rxd_pkt_type_mapping(_mm_extract_epi8(ptype1, 0));
-       rx_pkts[3]->packet_type = i40e_rxd_pkt_type_mapping(_mm_extract_epi8(ptype1, 8));
-}
-
- /*
- * Notice:
- * - nb_pkts < RTE_I40E_DESCS_PER_LOOP, just return no packet
- * - nb_pkts > RTE_I40E_VPMD_RX_BURST, only scan RTE_I40E_VPMD_RX_BURST
- *   numbers of DD bits
- */
-static inline uint16_t
-_recv_raw_pkts_vec(struct i40e_rx_queue *rxq, struct rte_mbuf **rx_pkts,
-                  uint16_t nb_pkts, uint8_t *split_packet)
-{
-       volatile union i40e_rx_desc *rxdp;
-       struct i40e_rx_entry *sw_ring;
-       uint16_t nb_pkts_recd;
-       int pos;
-       uint64_t var;
-       __m128i shuf_msk;
-
-       __m128i crc_adjust = _mm_set_epi16(
-                               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 */
-                       );
-       __m128i dd_check, eop_check;
-
-       /* nb_pkts shall be less equal than RTE_I40E_MAX_RX_BURST */
-       nb_pkts = RTE_MIN(nb_pkts, RTE_I40E_MAX_RX_BURST);
-
-       /* nb_pkts has to be floor-aligned to RTE_I40E_DESCS_PER_LOOP */
-       nb_pkts = RTE_ALIGN_FLOOR(nb_pkts, RTE_I40E_DESCS_PER_LOOP);
-
-       /* Just the act of getting into the function from the application is
-        * going to cost about 7 cycles
-        */
-       rxdp = rxq->rx_ring + rxq->rx_tail;
-
-       rte_prefetch0(rxdp);
-
-       /* See if we need to rearm the RX queue - gives the prefetch a bit
-        * of time to act
-        */
-       if (rxq->rxrearm_nb > RTE_I40E_RXQ_REARM_THRESH)
-               i40e_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 << I40E_RX_DESC_STATUS_DD_SHIFT)))
-               return 0;
-
-       /* 4 packets DD mask */
-       dd_check = _mm_set_epi64x(0x0000000100000001LL, 0x0000000100000001LL);
-
-       /* 4 packets EOP mask */
-       eop_check = _mm_set_epi64x(0x0000000200000002LL, 0x0000000200000002LL);
-
-       /* mask to shuffle from desc. to mbuf */
-       shuf_msk = _mm_set_epi8(
-               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 */
-               );
-
-       /* Cache is empty -> need to scan the buffer rings, but first move
-        * the next 'n' mbufs into the cache
-        */
-       sw_ring = &rxq->sw_ring[rxq->rx_tail];
-
-       /* A. load 4 packet in one loop
-        * [A*. mask out 4 unused dirty field in desc]
-        * B. copy 4 mbuf point from swring to rx_pkts
-        * C. calc the number of DD bits among the 4 packets
-        * [C*. extract the end-of-packet bit, if requested]
-        * D. fill info. from desc to mbuf
-        */
-
-       for (pos = 0, nb_pkts_recd = 0; pos < nb_pkts;
-                       pos += RTE_I40E_DESCS_PER_LOOP,
-                       rxdp += RTE_I40E_DESCS_PER_LOOP) {
-               __m128i descs[RTE_I40E_DESCS_PER_LOOP];
-               __m128i pkt_mb1, pkt_mb2, pkt_mb3, pkt_mb4;
-               __m128i zero, staterr, sterr_tmp1, sterr_tmp2;
-               __m128i mbp1, mbp2; /* two mbuf pointer in one XMM reg. */
-
-               /* B.1 load 1 mbuf point */
-               mbp1 = _mm_loadu_si128((__m128i *)&sw_ring[pos]);
-               /* Read desc statuses backwards to avoid race condition */
-               /* A.1 load 4 pkts desc */
-               descs[3] = _mm_loadu_si128((__m128i *)(rxdp + 3));
-
-               /* B.2 copy 2 mbuf point into rx_pkts  */
-               _mm_storeu_si128((__m128i *)&rx_pkts[pos], mbp1);
-
-               /* B.1 load 1 mbuf point */
-               mbp2 = _mm_loadu_si128((__m128i *)&sw_ring[pos+2]);
-
-               descs[2] = _mm_loadu_si128((__m128i *)(rxdp + 2));
-               /* B.1 load 2 mbuf point */
-               descs[1] = _mm_loadu_si128((__m128i *)(rxdp + 1));
-               descs[0] = _mm_loadu_si128((__m128i *)(rxdp));
-
-               /* B.2 copy 2 mbuf point into rx_pkts  */
-               _mm_storeu_si128((__m128i *)&rx_pkts[pos+2], mbp2);
-
-               if (split_packet) {
-                       rte_mbuf_prefetch_part2(rx_pkts[pos]);
-                       rte_mbuf_prefetch_part2(rx_pkts[pos + 1]);
-                       rte_mbuf_prefetch_part2(rx_pkts[pos + 2]);
-                       rte_mbuf_prefetch_part2(rx_pkts[pos + 3]);
-               }
-
-               /* avoid compiler reorder optimization */
-               rte_compiler_barrier();
-
-               /* pkt 3,4 shift the pktlen field to be 16-bit aligned*/
-               const __m128i len3 = _mm_slli_epi32(descs[3], PKTLEN_SHIFT);
-               const __m128i len2 = _mm_slli_epi32(descs[2], PKTLEN_SHIFT);
-
-               /* merge the now-aligned packet length fields back in */
-               descs[3] = _mm_blend_epi16(descs[3], len3, 0x80);
-               descs[2] = _mm_blend_epi16(descs[2], len2, 0x80);
-
-               /* D.1 pkt 3,4 convert format from desc to pktmbuf */
-               pkt_mb4 = _mm_shuffle_epi8(descs[3], shuf_msk);
-               pkt_mb3 = _mm_shuffle_epi8(descs[2], shuf_msk);
-
-               /* C.1 4=>2 filter staterr info only */
-               sterr_tmp2 = _mm_unpackhi_epi32(descs[3], descs[2]);
-               /* C.1 4=>2 filter staterr info only */
-               sterr_tmp1 = _mm_unpackhi_epi32(descs[1], descs[0]);
-
-               desc_to_olflags_v(descs, &rx_pkts[pos]);
-
-               /* D.2 pkt 3,4 set in_port/nb_seg and remove crc */
-               pkt_mb4 = _mm_add_epi16(pkt_mb4, crc_adjust);
-               pkt_mb3 = _mm_add_epi16(pkt_mb3, crc_adjust);
-
-               /* pkt 1,2 shift the pktlen field to be 16-bit aligned*/
-               const __m128i len1 = _mm_slli_epi32(descs[1], PKTLEN_SHIFT);
-               const __m128i len0 = _mm_slli_epi32(descs[0], PKTLEN_SHIFT);
-
-               /* merge the now-aligned packet length fields back in */
-               descs[1] = _mm_blend_epi16(descs[1], len1, 0x80);
-               descs[0] = _mm_blend_epi16(descs[0], len0, 0x80);
-
-               /* D.1 pkt 1,2 convert format from desc to pktmbuf */
-               pkt_mb2 = _mm_shuffle_epi8(descs[1], shuf_msk);
-               pkt_mb1 = _mm_shuffle_epi8(descs[0], shuf_msk);
-
-               /* C.2 get 4 pkts staterr value  */
-               zero = _mm_xor_si128(dd_check, dd_check);
-               staterr = _mm_unpacklo_epi32(sterr_tmp1, sterr_tmp2);
-
-               /* D.3 copy final 3,4 data to rx_pkts */
-               _mm_storeu_si128((void *)&rx_pkts[pos+3]->rx_descriptor_fields1,
-                                pkt_mb4);
-               _mm_storeu_si128((void *)&rx_pkts[pos+2]->rx_descriptor_fields1,
-                                pkt_mb3);
-
-               /* D.2 pkt 1,2 set in_port/nb_seg and remove crc */
-               pkt_mb2 = _mm_add_epi16(pkt_mb2, crc_adjust);
-               pkt_mb1 = _mm_add_epi16(pkt_mb1, crc_adjust);
-
-               /* C* extract and record EOP bit */
-               if (split_packet) {
-                       __m128i eop_shuf_mask = _mm_set_epi8(
-                                       0xFF, 0xFF, 0xFF, 0xFF,
-                                       0xFF, 0xFF, 0xFF, 0xFF,
-                                       0xFF, 0xFF, 0xFF, 0xFF,
-                                       0x04, 0x0C, 0x00, 0x08
-                                       );
-
-                       /* and with mask to extract bits, flipping 1-0 */
-                       __m128i eop_bits = _mm_andnot_si128(staterr, eop_check);
-                       /* the staterr values are not in order, as the count
-                        * count of dd bits doesn't care. However, for end of
-                        * packet tracking, we do care, so shuffle. This also
-                        * compresses the 32-bit values to 8-bit
-                        */
-                       eop_bits = _mm_shuffle_epi8(eop_bits, eop_shuf_mask);
-                       /* store the resulting 32-bit value */
-                       *(int *)split_packet = _mm_cvtsi128_si32(eop_bits);
-                       split_packet += RTE_I40E_DESCS_PER_LOOP;
-
-                       /* zero-out next pointers */
-                       rx_pkts[pos]->next = NULL;
-                       rx_pkts[pos + 1]->next = NULL;
-                       rx_pkts[pos + 2]->next = NULL;
-                       rx_pkts[pos + 3]->next = NULL;
-               }
-
-               /* C.3 calc available number of desc */
-               staterr = _mm_and_si128(staterr, dd_check);
-               staterr = _mm_packs_epi32(staterr, zero);
-
-               /* D.3 copy final 1,2 data to rx_pkts */
-               _mm_storeu_si128((void *)&rx_pkts[pos+1]->rx_descriptor_fields1,
-                                pkt_mb2);
-               _mm_storeu_si128((void *)&rx_pkts[pos]->rx_descriptor_fields1,
-                                pkt_mb1);
-               desc_to_ptype_v(descs, &rx_pkts[pos]);
-               /* C.4 calc avaialbe number of desc */
-               var = __builtin_popcountll(_mm_cvtsi128_si64(staterr));
-               nb_pkts_recd += var;
-               if (likely(var != RTE_I40E_DESCS_PER_LOOP))
-                       break;
-       }
-
-       /* Update our internal tail pointer */
-       rxq->rx_tail = (uint16_t)(rxq->rx_tail + nb_pkts_recd);
-       rxq->rx_tail = (uint16_t)(rxq->rx_tail & (rxq->nb_rx_desc - 1));
-       rxq->rxrearm_nb = (uint16_t)(rxq->rxrearm_nb + nb_pkts_recd);
-
-       return nb_pkts_recd;
-}
-
- /*
- * Notice:
- * - nb_pkts < RTE_I40E_DESCS_PER_LOOP, just return no packet
- * - nb_pkts > RTE_I40E_VPMD_RX_BURST, only scan RTE_I40E_VPMD_RX_BURST
- *   numbers of DD bits
- */
-uint16_t
-i40e_recv_pkts_vec(void *rx_queue, struct rte_mbuf **rx_pkts,
-                  uint16_t nb_pkts)
-{
-       return _recv_raw_pkts_vec(rx_queue, rx_pkts, nb_pkts, NULL);
-}
-
- /* vPMD receive routine that reassembles scattered packets
- * Notice:
- * - nb_pkts < RTE_I40E_DESCS_PER_LOOP, just return no packet
- * - nb_pkts > RTE_I40E_VPMD_RX_BURST, only scan RTE_I40E_VPMD_RX_BURST
- *   numbers of DD bits
- */
-uint16_t
-i40e_recv_scattered_pkts_vec(void *rx_queue, struct rte_mbuf **rx_pkts,
-                            uint16_t nb_pkts)
-{
-
-       struct i40e_rx_queue *rxq = rx_queue;
-       uint8_t split_flags[RTE_I40E_VPMD_RX_BURST] = {0};
-
-       /* get some new buffers */
-       uint16_t nb_bufs = _recv_raw_pkts_vec(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 == NULL &&
-                       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 i = 0;
-
-       if (rxq->pkt_first_seg == NULL) {
-               /* find the first split flag, and only reassemble then*/
-               while (i < nb_bufs && !split_flags[i])
-                       i++;
-               if (i == nb_bufs)
-                       return nb_bufs;
-       }
-       return i + reassemble_packets(rxq, &rx_pkts[i], nb_bufs - i,
-               &split_flags[i]);
-}
-
-static inline void
-vtx1(volatile struct i40e_tx_desc *txdp,
-               struct rte_mbuf *pkt, uint64_t flags)
-{
-       uint64_t high_qw = (I40E_TX_DESC_DTYPE_DATA |
-                       ((uint64_t)flags  << I40E_TXD_QW1_CMD_SHIFT) |
-                       ((uint64_t)pkt->data_len << I40E_TXD_QW1_TX_BUF_SZ_SHIFT));
-
-       __m128i descriptor = _mm_set_epi64x(high_qw,
-                               pkt->buf_physaddr + pkt->data_off);
-       _mm_store_si128((__m128i *)txdp, descriptor);
-}
-
-static inline void
-vtx(volatile struct i40e_tx_desc *txdp,
-               struct rte_mbuf **pkt, uint16_t nb_pkts,  uint64_t flags)
-{
-       int i;
-
-       for (i = 0; i < nb_pkts; ++i, ++txdp, ++pkt)
-               vtx1(txdp, *pkt, flags);
-}
-
-uint16_t
-i40e_xmit_pkts_vec(void *tx_queue, struct rte_mbuf **tx_pkts,
-                  uint16_t nb_pkts)
-{
-       struct i40e_tx_queue *txq = (struct i40e_tx_queue *)tx_queue;
-       volatile struct i40e_tx_desc *txdp;
-       struct i40e_tx_entry *txep;
-       uint16_t n, nb_commit, tx_id;
-       uint64_t flags = I40E_TD_CMD;
-       uint64_t rs = I40E_TX_DESC_CMD_RS | I40E_TD_CMD;
-       int i;
-
-       /* cross rx_thresh boundary is not allowed */
-       nb_pkts = RTE_MIN(nb_pkts, txq->tx_rs_thresh);
-
-       if (txq->nb_tx_free < txq->tx_free_thresh)
-               i40e_tx_free_bufs(txq);
-
-       nb_commit = nb_pkts = (uint16_t)RTE_MIN(txq->nb_tx_free, nb_pkts);
-       if (unlikely(nb_pkts == 0))
-               return 0;
-
-       tx_id = txq->tx_tail;
-       txdp = &txq->tx_ring[tx_id];
-       txep = &txq->sw_ring[tx_id];
-
-       txq->nb_tx_free = (uint16_t)(txq->nb_tx_free - nb_pkts);
-
-       n = (uint16_t)(txq->nb_tx_desc - tx_id);
-       if (nb_commit >= n) {
-               tx_backlog_entry(txep, tx_pkts, n);
-
-               for (i = 0; i < n - 1; ++i, ++tx_pkts, ++txdp)
-                       vtx1(txdp, *tx_pkts, flags);
-
-               vtx1(txdp, *tx_pkts++, rs);
-
-               nb_commit = (uint16_t)(nb_commit - n);
-
-               tx_id = 0;
-               txq->tx_next_rs = (uint16_t)(txq->tx_rs_thresh - 1);
-
-               /* avoid reach the end of ring */
-               txdp = &txq->tx_ring[tx_id];
-               txep = &txq->sw_ring[tx_id];
-       }
-
-       tx_backlog_entry(txep, tx_pkts, nb_commit);
-
-       vtx(txdp, tx_pkts, nb_commit, flags);
-
-       tx_id = (uint16_t)(tx_id + nb_commit);
-       if (tx_id > txq->tx_next_rs) {
-               txq->tx_ring[txq->tx_next_rs].cmd_type_offset_bsz |=
-                       rte_cpu_to_le_64(((uint64_t)I40E_TX_DESC_CMD_RS) <<
-                                               I40E_TXD_QW1_CMD_SHIFT);
-               txq->tx_next_rs =
-                       (uint16_t)(txq->tx_next_rs + txq->tx_rs_thresh);
-       }
-
-       txq->tx_tail = tx_id;
-
-       I40E_PCI_REG_WRITE(txq->qtx_tail, txq->tx_tail);
-
-       return nb_pkts;
-}
-
-void __attribute__((cold))
-i40e_rx_queue_release_mbufs_vec(struct i40e_rx_queue *rxq)
-{
-       _i40e_rx_queue_release_mbufs_vec(rxq);
-}
-
-int __attribute__((cold))
-i40e_rxq_vec_setup(struct i40e_rx_queue *rxq)
-{
-       return i40e_rxq_vec_setup_default(rxq);
-}
-
-int __attribute__((cold))
-i40e_txq_vec_setup(struct i40e_tx_queue __rte_unused *txq)
-{
-       return 0;
-}
-
-int __attribute__((cold))
-i40e_rx_vec_dev_conf_condition_check(struct rte_eth_dev *dev)
-{
-#ifndef RTE_LIBRTE_IEEE1588
-       /* need SSE4.1 support */
-       if (!rte_cpu_get_flag_enabled(RTE_CPUFLAG_SSE4_1))
-               return -1;
-#endif
-
-       return i40e_rx_vec_dev_conf_condition_check_default(dev);
-}
diff --git a/drivers/net/i40e/i40e_rxtx_vec_sse.c b/drivers/net/i40e/i40e_rxtx_vec_sse.c
new file mode 100644 (file)
index 0000000..3607312
--- /dev/null
@@ -0,0 +1,609 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) 2010-2015 Intel Corporation. All rights reserved.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Intel Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <stdint.h>
+#include <rte_ethdev.h>
+#include <rte_malloc.h>
+
+#include "base/i40e_prototype.h"
+#include "base/i40e_type.h"
+#include "i40e_ethdev.h"
+#include "i40e_rxtx.h"
+#include "i40e_rxtx_vec_common.h"
+
+#include <tmmintrin.h>
+
+#ifndef __INTEL_COMPILER
+#pragma GCC diagnostic ignored "-Wcast-qual"
+#endif
+
+static inline void
+i40e_rxq_rearm(struct i40e_rx_queue *rxq)
+{
+       int i;
+       uint16_t rx_id;
+       volatile union i40e_rx_desc *rxdp;
+       struct i40e_rx_entry *rxep = &rxq->sw_ring[rxq->rxrearm_start];
+       struct rte_mbuf *mb0, *mb1;
+       __m128i hdr_room = _mm_set_epi64x(RTE_PKTMBUF_HEADROOM,
+                       RTE_PKTMBUF_HEADROOM);
+       __m128i dma_addr0, dma_addr1;
+
+       rxdp = rxq->rx_ring + rxq->rxrearm_start;
+
+       /* Pull 'n' more MBUFs into the software ring */
+       if (rte_mempool_get_bulk(rxq->mp,
+                                (void *)rxep,
+                                RTE_I40E_RXQ_REARM_THRESH) < 0) {
+               if (rxq->rxrearm_nb + RTE_I40E_RXQ_REARM_THRESH >=
+                   rxq->nb_rx_desc) {
+                       dma_addr0 = _mm_setzero_si128();
+                       for (i = 0; i < RTE_I40E_DESCS_PER_LOOP; i++) {
+                               rxep[i].mbuf = &rxq->fake_mbuf;
+                               _mm_store_si128((__m128i *)&rxdp[i].read,
+                                               dma_addr0);
+                       }
+               }
+               rte_eth_devices[rxq->port_id].data->rx_mbuf_alloc_failed +=
+                       RTE_I40E_RXQ_REARM_THRESH;
+               return;
+       }
+
+       /* Initialize the mbufs in vector, process 2 mbufs in one loop */
+       for (i = 0; i < RTE_I40E_RXQ_REARM_THRESH; i += 2, rxep += 2) {
+               __m128i vaddr0, vaddr1;
+               uintptr_t p0, p1;
+
+               mb0 = rxep[0].mbuf;
+               mb1 = rxep[1].mbuf;
+
+                /* Flush mbuf with pkt template.
+                * Data to be rearmed is 6 bytes long.
+                * Though, RX will overwrite ol_flags that are coming next
+                * anyway. So overwrite whole 8 bytes with one load:
+                * 6 bytes of rearm_data plus first 2 bytes of ol_flags.
+                */
+               p0 = (uintptr_t)&mb0->rearm_data;
+               *(uint64_t *)p0 = rxq->mbuf_initializer;
+               p1 = (uintptr_t)&mb1->rearm_data;
+               *(uint64_t *)p1 = rxq->mbuf_initializer;
+
+               /* load buf_addr(lo 64bit) and buf_physaddr(hi 64bit) */
+               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);
+       }
+
+       rxq->rxrearm_start += RTE_I40E_RXQ_REARM_THRESH;
+       if (rxq->rxrearm_start >= rxq->nb_rx_desc)
+               rxq->rxrearm_start = 0;
+
+       rxq->rxrearm_nb -= RTE_I40E_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 */
+       I40E_PCI_REG_WRITE(rxq->qrx_tail, rx_id);
+}
+
+/* Handling the offload flags (olflags) field takes computation
+ * time when receiving packets. Therefore we provide a flag to disable
+ * the processing of the olflags field when they are not needed. This
+ * gives improved performance, at the cost of losing the offload info
+ * in the received packet
+ */
+#ifdef RTE_LIBRTE_I40E_RX_OLFLAGS_ENABLE
+
+static inline void
+desc_to_olflags_v(__m128i descs[4], struct rte_mbuf **rx_pkts)
+{
+       __m128i vlan0, vlan1, rss, l3_l4e;
+
+       /* mask everything except RSS, flow director and VLAN flags
+        * bit2 is for VLAN tag, bit11 for flow director indication
+        * bit13:12 for RSS indication.
+        */
+       const __m128i rss_vlan_msk = _mm_set_epi32(
+                       0x1c03804, 0x1c03804, 0x1c03804, 0x1c03804);
+
+       /* map rss and vlan type to rss hash and vlan flag */
+       const __m128i vlan_flags = _mm_set_epi8(0, 0, 0, 0,
+                       0, 0, 0, 0,
+                       0, 0, 0, PKT_RX_VLAN_PKT | PKT_RX_VLAN_STRIPPED,
+                       0, 0, 0, 0);
+
+       const __m128i rss_flags = _mm_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);
+
+       const __m128i l3_l4e_flags = _mm_set_epi8(0, 0, 0, 0, 0, 0, 0, 0,
+                       PKT_RX_EIP_CKSUM_BAD | PKT_RX_L4_CKSUM_BAD | PKT_RX_IP_CKSUM_BAD,
+                       PKT_RX_EIP_CKSUM_BAD | PKT_RX_L4_CKSUM_BAD,
+                       PKT_RX_EIP_CKSUM_BAD | PKT_RX_IP_CKSUM_BAD,
+                       PKT_RX_EIP_CKSUM_BAD,
+                       PKT_RX_L4_CKSUM_BAD | PKT_RX_IP_CKSUM_BAD,
+                       PKT_RX_L4_CKSUM_BAD,
+                       PKT_RX_IP_CKSUM_BAD,
+                       0);
+
+       vlan0 = _mm_unpackhi_epi32(descs[0], descs[1]);
+       vlan1 = _mm_unpackhi_epi32(descs[2], descs[3]);
+       vlan0 = _mm_unpacklo_epi64(vlan0, vlan1);
+
+       vlan1 = _mm_and_si128(vlan0, rss_vlan_msk);
+       vlan0 = _mm_shuffle_epi8(vlan_flags, vlan1);
+
+       rss = _mm_srli_epi32(vlan1, 11);
+       rss = _mm_shuffle_epi8(rss_flags, rss);
+
+       l3_l4e = _mm_srli_epi32(vlan1, 22);
+       l3_l4e = _mm_shuffle_epi8(l3_l4e_flags, l3_l4e);
+
+       vlan0 = _mm_or_si128(vlan0, rss);
+       vlan0 = _mm_or_si128(vlan0, l3_l4e);
+
+       rx_pkts[0]->ol_flags = _mm_extract_epi16(vlan0, 0);
+       rx_pkts[1]->ol_flags = _mm_extract_epi16(vlan0, 2);
+       rx_pkts[2]->ol_flags = _mm_extract_epi16(vlan0, 4);
+       rx_pkts[3]->ol_flags = _mm_extract_epi16(vlan0, 6);
+}
+#else
+#define desc_to_olflags_v(desc, rx_pkts) do {} while (0)
+#endif
+
+#define PKTLEN_SHIFT     10
+
+static inline void
+desc_to_ptype_v(__m128i descs[4], struct rte_mbuf **rx_pkts)
+{
+       __m128i ptype0 = _mm_unpackhi_epi64(descs[0], descs[1]);
+       __m128i ptype1 = _mm_unpackhi_epi64(descs[2], descs[3]);
+
+       ptype0 = _mm_srli_epi64(ptype0, 30);
+       ptype1 = _mm_srli_epi64(ptype1, 30);
+
+       rx_pkts[0]->packet_type = i40e_rxd_pkt_type_mapping(_mm_extract_epi8(ptype0, 0));
+       rx_pkts[1]->packet_type = i40e_rxd_pkt_type_mapping(_mm_extract_epi8(ptype0, 8));
+       rx_pkts[2]->packet_type = i40e_rxd_pkt_type_mapping(_mm_extract_epi8(ptype1, 0));
+       rx_pkts[3]->packet_type = i40e_rxd_pkt_type_mapping(_mm_extract_epi8(ptype1, 8));
+}
+
+ /*
+ * Notice:
+ * - nb_pkts < RTE_I40E_DESCS_PER_LOOP, just return no packet
+ * - nb_pkts > RTE_I40E_VPMD_RX_BURST, only scan RTE_I40E_VPMD_RX_BURST
+ *   numbers of DD bits
+ */
+static inline uint16_t
+_recv_raw_pkts_vec(struct i40e_rx_queue *rxq, struct rte_mbuf **rx_pkts,
+                  uint16_t nb_pkts, uint8_t *split_packet)
+{
+       volatile union i40e_rx_desc *rxdp;
+       struct i40e_rx_entry *sw_ring;
+       uint16_t nb_pkts_recd;
+       int pos;
+       uint64_t var;
+       __m128i shuf_msk;
+
+       __m128i crc_adjust = _mm_set_epi16(
+                               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 */
+                       );
+       __m128i dd_check, eop_check;
+
+       /* nb_pkts shall be less equal than RTE_I40E_MAX_RX_BURST */
+       nb_pkts = RTE_MIN(nb_pkts, RTE_I40E_MAX_RX_BURST);
+
+       /* nb_pkts has to be floor-aligned to RTE_I40E_DESCS_PER_LOOP */
+       nb_pkts = RTE_ALIGN_FLOOR(nb_pkts, RTE_I40E_DESCS_PER_LOOP);
+
+       /* Just the act of getting into the function from the application is
+        * going to cost about 7 cycles
+        */
+       rxdp = rxq->rx_ring + rxq->rx_tail;
+
+       rte_prefetch0(rxdp);
+
+       /* See if we need to rearm the RX queue - gives the prefetch a bit
+        * of time to act
+        */
+       if (rxq->rxrearm_nb > RTE_I40E_RXQ_REARM_THRESH)
+               i40e_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 << I40E_RX_DESC_STATUS_DD_SHIFT)))
+               return 0;
+
+       /* 4 packets DD mask */
+       dd_check = _mm_set_epi64x(0x0000000100000001LL, 0x0000000100000001LL);
+
+       /* 4 packets EOP mask */
+       eop_check = _mm_set_epi64x(0x0000000200000002LL, 0x0000000200000002LL);
+
+       /* mask to shuffle from desc. to mbuf */
+       shuf_msk = _mm_set_epi8(
+               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 */
+               );
+
+       /* Cache is empty -> need to scan the buffer rings, but first move
+        * the next 'n' mbufs into the cache
+        */
+       sw_ring = &rxq->sw_ring[rxq->rx_tail];
+
+       /* A. load 4 packet in one loop
+        * [A*. mask out 4 unused dirty field in desc]
+        * B. copy 4 mbuf point from swring to rx_pkts
+        * C. calc the number of DD bits among the 4 packets
+        * [C*. extract the end-of-packet bit, if requested]
+        * D. fill info. from desc to mbuf
+        */
+
+       for (pos = 0, nb_pkts_recd = 0; pos < nb_pkts;
+                       pos += RTE_I40E_DESCS_PER_LOOP,
+                       rxdp += RTE_I40E_DESCS_PER_LOOP) {
+               __m128i descs[RTE_I40E_DESCS_PER_LOOP];
+               __m128i pkt_mb1, pkt_mb2, pkt_mb3, pkt_mb4;
+               __m128i zero, staterr, sterr_tmp1, sterr_tmp2;
+               __m128i mbp1, mbp2; /* two mbuf pointer in one XMM reg. */
+
+               /* B.1 load 1 mbuf point */
+               mbp1 = _mm_loadu_si128((__m128i *)&sw_ring[pos]);
+               /* Read desc statuses backwards to avoid race condition */
+               /* A.1 load 4 pkts desc */
+               descs[3] = _mm_loadu_si128((__m128i *)(rxdp + 3));
+
+               /* B.2 copy 2 mbuf point into rx_pkts  */
+               _mm_storeu_si128((__m128i *)&rx_pkts[pos], mbp1);
+
+               /* B.1 load 1 mbuf point */
+               mbp2 = _mm_loadu_si128((__m128i *)&sw_ring[pos+2]);
+
+               descs[2] = _mm_loadu_si128((__m128i *)(rxdp + 2));
+               /* B.1 load 2 mbuf point */
+               descs[1] = _mm_loadu_si128((__m128i *)(rxdp + 1));
+               descs[0] = _mm_loadu_si128((__m128i *)(rxdp));
+
+               /* B.2 copy 2 mbuf point into rx_pkts  */
+               _mm_storeu_si128((__m128i *)&rx_pkts[pos+2], mbp2);
+
+               if (split_packet) {
+                       rte_mbuf_prefetch_part2(rx_pkts[pos]);
+                       rte_mbuf_prefetch_part2(rx_pkts[pos + 1]);
+                       rte_mbuf_prefetch_part2(rx_pkts[pos + 2]);
+                       rte_mbuf_prefetch_part2(rx_pkts[pos + 3]);
+               }
+
+               /* avoid compiler reorder optimization */
+               rte_compiler_barrier();
+
+               /* pkt 3,4 shift the pktlen field to be 16-bit aligned*/
+               const __m128i len3 = _mm_slli_epi32(descs[3], PKTLEN_SHIFT);
+               const __m128i len2 = _mm_slli_epi32(descs[2], PKTLEN_SHIFT);
+
+               /* merge the now-aligned packet length fields back in */
+               descs[3] = _mm_blend_epi16(descs[3], len3, 0x80);
+               descs[2] = _mm_blend_epi16(descs[2], len2, 0x80);
+
+               /* D.1 pkt 3,4 convert format from desc to pktmbuf */
+               pkt_mb4 = _mm_shuffle_epi8(descs[3], shuf_msk);
+               pkt_mb3 = _mm_shuffle_epi8(descs[2], shuf_msk);
+
+               /* C.1 4=>2 filter staterr info only */
+               sterr_tmp2 = _mm_unpackhi_epi32(descs[3], descs[2]);
+               /* C.1 4=>2 filter staterr info only */
+               sterr_tmp1 = _mm_unpackhi_epi32(descs[1], descs[0]);
+
+               desc_to_olflags_v(descs, &rx_pkts[pos]);
+
+               /* D.2 pkt 3,4 set in_port/nb_seg and remove crc */
+               pkt_mb4 = _mm_add_epi16(pkt_mb4, crc_adjust);
+               pkt_mb3 = _mm_add_epi16(pkt_mb3, crc_adjust);
+
+               /* pkt 1,2 shift the pktlen field to be 16-bit aligned*/
+               const __m128i len1 = _mm_slli_epi32(descs[1], PKTLEN_SHIFT);
+               const __m128i len0 = _mm_slli_epi32(descs[0], PKTLEN_SHIFT);
+
+               /* merge the now-aligned packet length fields back in */
+               descs[1] = _mm_blend_epi16(descs[1], len1, 0x80);
+               descs[0] = _mm_blend_epi16(descs[0], len0, 0x80);
+
+               /* D.1 pkt 1,2 convert format from desc to pktmbuf */
+               pkt_mb2 = _mm_shuffle_epi8(descs[1], shuf_msk);
+               pkt_mb1 = _mm_shuffle_epi8(descs[0], shuf_msk);
+
+               /* C.2 get 4 pkts staterr value  */
+               zero = _mm_xor_si128(dd_check, dd_check);
+               staterr = _mm_unpacklo_epi32(sterr_tmp1, sterr_tmp2);
+
+               /* D.3 copy final 3,4 data to rx_pkts */
+               _mm_storeu_si128((void *)&rx_pkts[pos+3]->rx_descriptor_fields1,
+                                pkt_mb4);
+               _mm_storeu_si128((void *)&rx_pkts[pos+2]->rx_descriptor_fields1,
+                                pkt_mb3);
+
+               /* D.2 pkt 1,2 set in_port/nb_seg and remove crc */
+               pkt_mb2 = _mm_add_epi16(pkt_mb2, crc_adjust);
+               pkt_mb1 = _mm_add_epi16(pkt_mb1, crc_adjust);
+
+               /* C* extract and record EOP bit */
+               if (split_packet) {
+                       __m128i eop_shuf_mask = _mm_set_epi8(
+                                       0xFF, 0xFF, 0xFF, 0xFF,
+                                       0xFF, 0xFF, 0xFF, 0xFF,
+                                       0xFF, 0xFF, 0xFF, 0xFF,
+                                       0x04, 0x0C, 0x00, 0x08
+                                       );
+
+                       /* and with mask to extract bits, flipping 1-0 */
+                       __m128i eop_bits = _mm_andnot_si128(staterr, eop_check);
+                       /* the staterr values are not in order, as the count
+                        * count of dd bits doesn't care. However, for end of
+                        * packet tracking, we do care, so shuffle. This also
+                        * compresses the 32-bit values to 8-bit
+                        */
+                       eop_bits = _mm_shuffle_epi8(eop_bits, eop_shuf_mask);
+                       /* store the resulting 32-bit value */
+                       *(int *)split_packet = _mm_cvtsi128_si32(eop_bits);
+                       split_packet += RTE_I40E_DESCS_PER_LOOP;
+
+                       /* zero-out next pointers */
+                       rx_pkts[pos]->next = NULL;
+                       rx_pkts[pos + 1]->next = NULL;
+                       rx_pkts[pos + 2]->next = NULL;
+                       rx_pkts[pos + 3]->next = NULL;
+               }
+
+               /* C.3 calc available number of desc */
+               staterr = _mm_and_si128(staterr, dd_check);
+               staterr = _mm_packs_epi32(staterr, zero);
+
+               /* D.3 copy final 1,2 data to rx_pkts */
+               _mm_storeu_si128((void *)&rx_pkts[pos+1]->rx_descriptor_fields1,
+                                pkt_mb2);
+               _mm_storeu_si128((void *)&rx_pkts[pos]->rx_descriptor_fields1,
+                                pkt_mb1);
+               desc_to_ptype_v(descs, &rx_pkts[pos]);
+               /* C.4 calc avaialbe number of desc */
+               var = __builtin_popcountll(_mm_cvtsi128_si64(staterr));
+               nb_pkts_recd += var;
+               if (likely(var != RTE_I40E_DESCS_PER_LOOP))
+                       break;
+       }
+
+       /* Update our internal tail pointer */
+       rxq->rx_tail = (uint16_t)(rxq->rx_tail + nb_pkts_recd);
+       rxq->rx_tail = (uint16_t)(rxq->rx_tail & (rxq->nb_rx_desc - 1));
+       rxq->rxrearm_nb = (uint16_t)(rxq->rxrearm_nb + nb_pkts_recd);
+
+       return nb_pkts_recd;
+}
+
+ /*
+ * Notice:
+ * - nb_pkts < RTE_I40E_DESCS_PER_LOOP, just return no packet
+ * - nb_pkts > RTE_I40E_VPMD_RX_BURST, only scan RTE_I40E_VPMD_RX_BURST
+ *   numbers of DD bits
+ */
+uint16_t
+i40e_recv_pkts_vec(void *rx_queue, struct rte_mbuf **rx_pkts,
+                  uint16_t nb_pkts)
+{
+       return _recv_raw_pkts_vec(rx_queue, rx_pkts, nb_pkts, NULL);
+}
+
+ /* vPMD receive routine that reassembles scattered packets
+ * Notice:
+ * - nb_pkts < RTE_I40E_DESCS_PER_LOOP, just return no packet
+ * - nb_pkts > RTE_I40E_VPMD_RX_BURST, only scan RTE_I40E_VPMD_RX_BURST
+ *   numbers of DD bits
+ */
+uint16_t
+i40e_recv_scattered_pkts_vec(void *rx_queue, struct rte_mbuf **rx_pkts,
+                            uint16_t nb_pkts)
+{
+
+       struct i40e_rx_queue *rxq = rx_queue;
+       uint8_t split_flags[RTE_I40E_VPMD_RX_BURST] = {0};
+
+       /* get some new buffers */
+       uint16_t nb_bufs = _recv_raw_pkts_vec(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 == NULL &&
+                       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 i = 0;
+
+       if (rxq->pkt_first_seg == NULL) {
+               /* find the first split flag, and only reassemble then*/
+               while (i < nb_bufs && !split_flags[i])
+                       i++;
+               if (i == nb_bufs)
+                       return nb_bufs;
+       }
+       return i + reassemble_packets(rxq, &rx_pkts[i], nb_bufs - i,
+               &split_flags[i]);
+}
+
+static inline void
+vtx1(volatile struct i40e_tx_desc *txdp,
+               struct rte_mbuf *pkt, uint64_t flags)
+{
+       uint64_t high_qw = (I40E_TX_DESC_DTYPE_DATA |
+                       ((uint64_t)flags  << I40E_TXD_QW1_CMD_SHIFT) |
+                       ((uint64_t)pkt->data_len << I40E_TXD_QW1_TX_BUF_SZ_SHIFT));
+
+       __m128i descriptor = _mm_set_epi64x(high_qw,
+                               pkt->buf_physaddr + pkt->data_off);
+       _mm_store_si128((__m128i *)txdp, descriptor);
+}
+
+static inline void
+vtx(volatile struct i40e_tx_desc *txdp,
+               struct rte_mbuf **pkt, uint16_t nb_pkts,  uint64_t flags)
+{
+       int i;
+
+       for (i = 0; i < nb_pkts; ++i, ++txdp, ++pkt)
+               vtx1(txdp, *pkt, flags);
+}
+
+uint16_t
+i40e_xmit_pkts_vec(void *tx_queue, struct rte_mbuf **tx_pkts,
+                  uint16_t nb_pkts)
+{
+       struct i40e_tx_queue *txq = (struct i40e_tx_queue *)tx_queue;
+       volatile struct i40e_tx_desc *txdp;
+       struct i40e_tx_entry *txep;
+       uint16_t n, nb_commit, tx_id;
+       uint64_t flags = I40E_TD_CMD;
+       uint64_t rs = I40E_TX_DESC_CMD_RS | I40E_TD_CMD;
+       int i;
+
+       /* cross rx_thresh boundary is not allowed */
+       nb_pkts = RTE_MIN(nb_pkts, txq->tx_rs_thresh);
+
+       if (txq->nb_tx_free < txq->tx_free_thresh)
+               i40e_tx_free_bufs(txq);
+
+       nb_commit = nb_pkts = (uint16_t)RTE_MIN(txq->nb_tx_free, nb_pkts);
+       if (unlikely(nb_pkts == 0))
+               return 0;
+
+       tx_id = txq->tx_tail;
+       txdp = &txq->tx_ring[tx_id];
+       txep = &txq->sw_ring[tx_id];
+
+       txq->nb_tx_free = (uint16_t)(txq->nb_tx_free - nb_pkts);
+
+       n = (uint16_t)(txq->nb_tx_desc - tx_id);
+       if (nb_commit >= n) {
+               tx_backlog_entry(txep, tx_pkts, n);
+
+               for (i = 0; i < n - 1; ++i, ++tx_pkts, ++txdp)
+                       vtx1(txdp, *tx_pkts, flags);
+
+               vtx1(txdp, *tx_pkts++, rs);
+
+               nb_commit = (uint16_t)(nb_commit - n);
+
+               tx_id = 0;
+               txq->tx_next_rs = (uint16_t)(txq->tx_rs_thresh - 1);
+
+               /* avoid reach the end of ring */
+               txdp = &txq->tx_ring[tx_id];
+               txep = &txq->sw_ring[tx_id];
+       }
+
+       tx_backlog_entry(txep, tx_pkts, nb_commit);
+
+       vtx(txdp, tx_pkts, nb_commit, flags);
+
+       tx_id = (uint16_t)(tx_id + nb_commit);
+       if (tx_id > txq->tx_next_rs) {
+               txq->tx_ring[txq->tx_next_rs].cmd_type_offset_bsz |=
+                       rte_cpu_to_le_64(((uint64_t)I40E_TX_DESC_CMD_RS) <<
+                                               I40E_TXD_QW1_CMD_SHIFT);
+               txq->tx_next_rs =
+                       (uint16_t)(txq->tx_next_rs + txq->tx_rs_thresh);
+       }
+
+       txq->tx_tail = tx_id;
+
+       I40E_PCI_REG_WRITE(txq->qtx_tail, txq->tx_tail);
+
+       return nb_pkts;
+}
+
+void __attribute__((cold))
+i40e_rx_queue_release_mbufs_vec(struct i40e_rx_queue *rxq)
+{
+       _i40e_rx_queue_release_mbufs_vec(rxq);
+}
+
+int __attribute__((cold))
+i40e_rxq_vec_setup(struct i40e_rx_queue *rxq)
+{
+       return i40e_rxq_vec_setup_default(rxq);
+}
+
+int __attribute__((cold))
+i40e_txq_vec_setup(struct i40e_tx_queue __rte_unused *txq)
+{
+       return 0;
+}
+
+int __attribute__((cold))
+i40e_rx_vec_dev_conf_condition_check(struct rte_eth_dev *dev)
+{
+#ifndef RTE_LIBRTE_IEEE1588
+       /* need SSE4.1 support */
+       if (!rte_cpu_get_flag_enabled(RTE_CPUFLAG_SSE4_1))
+               return -1;
+#endif
+
+       return i40e_rx_vec_dev_conf_condition_check_default(dev);
+}