This patch adds the basic NEON Rx path to the iavf driver. It does not include scatter or flex varieties.
Tested on N1SDP platform with Intel XL710 NIC and 40G connection. Tested with a single core and testpmd rxonly mode. Saw no significant performance difference between scalar and Arm vPMD paths using this test in iavf and saw the same results when comparing scalar and Arm vPMD path in i40e. Signed-off-by: Kathleen Capella <kathleen.cape...@arm.com> Reviewed-by: Ruifeng Wang <ruifeng.w...@arm.com> --- drivers/net/iavf/iavf_rxtx.c | 14 + drivers/net/iavf/iavf_rxtx_vec_neon.c | 415 ++++++++++++++++++++++++++ drivers/net/iavf/meson.build | 2 + 3 files changed, 431 insertions(+) create mode 100644 drivers/net/iavf/iavf_rxtx_vec_neon.c diff --git a/drivers/net/iavf/iavf_rxtx.c b/drivers/net/iavf/iavf_rxtx.c index 14d4dbe967..109ba756f8 100644 --- a/drivers/net/iavf/iavf_rxtx.c +++ b/drivers/net/iavf/iavf_rxtx.c @@ -3059,7 +3059,21 @@ iavf_set_rx_function(struct rte_eth_dev *dev) return; } +#elif defined RTE_ARCH_ARM + int check_ret; + check_ret = iavf_rx_vec_dev_check(dev); + if (check_ret >= 0 && + rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_128) { + PMD_DRV_LOG(DEBUG, "Using a Vector Rx callback (port=%d).", + dev->data->port_id); + for (i = 0; i < dev->data->nb_rx_queues; i++) { + rxq = dev->data->rx_queues[i]; + (void)iavf_rxq_vec_setup(rxq); + } + dev->rx_pkt_burst = iavf_recv_pkts_vec; + return; + } #endif if (dev->data->scattered_rx) { PMD_DRV_LOG(DEBUG, "Using a Scattered Rx callback (port=%d).", diff --git a/drivers/net/iavf/iavf_rxtx_vec_neon.c b/drivers/net/iavf/iavf_rxtx_vec_neon.c new file mode 100644 index 0000000000..83825aa427 --- /dev/null +++ b/drivers/net/iavf/iavf_rxtx_vec_neon.c @@ -0,0 +1,415 @@ +/* SPDX-License-Identifier: BSD-3-Clause + * Copyright(c) 2022 Intel Corporation + * Copyright(c) 2022 Arm Limited + */ + +#include <stdint.h> +#include <ethdev_driver.h> +#include <rte_malloc.h> +#include <rte_vect.h> + +#include "iavf.h" +#include "iavf_rxtx.h" +#include "iavf_rxtx_vec_common.h" + +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 **rxep = &rxq->sw_ring[rxq->rxrearm_start]; + struct rte_mbuf *mb0, *mb1; + uint64x2_t dma_addr0, dma_addr1; + uint64x2_t zero = vdupq_n_u64(0); + uint64_t paddr; + + rxdp = rxq->rx_ring + rxq->rxrearm_start; + + /* Pull 'n' more MBUFs into the software ring */ + if (unlikely(rte_mempool_get_bulk(rxq->mp, + (void *)rxep, + IAVF_RXQ_REARM_THRESH) < 0)) { + if (rxq->rxrearm_nb + IAVF_RXQ_REARM_THRESH >= + rxq->nb_rx_desc) { + for (i = 0; i < IAVF_VPMD_DESCS_PER_LOOP; i++) { + rxep[i] = &rxq->fake_mbuf; + vst1q_u64((uint64_t *)&rxdp[i].read, zero); + } + } + rte_eth_devices[rxq->port_id].data->rx_mbuf_alloc_failed += + IAVF_RXQ_REARM_THRESH; + return; + } + + /* Initialize the mbufs in vector, process 2 mbufs in one loop */ + for (i = 0; i < IAVF_RXQ_REARM_THRESH; i += 2, rxep += 2) { + mb0 = rxep[0]; + mb1 = rxep[1]; + + paddr = mb0->buf_iova + RTE_PKTMBUF_HEADROOM; + dma_addr0 = vdupq_n_u64(paddr); + + /* flush desc with pa dma_addr */ + vst1q_u64((uint64_t *)&rxdp++->read, dma_addr0); + + paddr = mb1->buf_iova + RTE_PKTMBUF_HEADROOM; + dma_addr1 = vdupq_n_u64(paddr); + vst1q_u64((uint64_t *)&rxdp++->read, dma_addr1); + } + + 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)); + + rte_io_wmb(); + /* Update the tail pointer on the NIC */ + IAVF_PCI_REG_WRITE_RELAXED(rxq->qrx_tail, rx_id); +} + +static inline void +desc_to_olflags_v(struct iavf_rx_queue *rxq, volatile union iavf_rx_desc *rxdp, + uint64x2_t descs[4], struct rte_mbuf **rx_pkts) +{ + RTE_SET_USED(rxdp); + + uint32x4_t vlan0, vlan1, rss, l3_l4e; + const uint64x2_t mbuf_init = {rxq->mbuf_initializer, 0}; + uint64x2_t rearm0, rearm1, rearm2, rearm3; + + /* 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 uint32x4_t rss_vlan_msk = { + 0x1c03804, 0x1c03804, 0x1c03804, 0x1c03804}; + + const uint32x4_t cksum_mask = { + RTE_MBUF_F_RX_IP_CKSUM_GOOD | RTE_MBUF_F_RX_IP_CKSUM_BAD | + RTE_MBUF_F_RX_L4_CKSUM_GOOD | RTE_MBUF_F_RX_L4_CKSUM_BAD | + RTE_MBUF_F_RX_OUTER_IP_CKSUM_BAD, + RTE_MBUF_F_RX_IP_CKSUM_GOOD | RTE_MBUF_F_RX_IP_CKSUM_BAD | + RTE_MBUF_F_RX_L4_CKSUM_GOOD | RTE_MBUF_F_RX_L4_CKSUM_BAD | + RTE_MBUF_F_RX_OUTER_IP_CKSUM_BAD, + RTE_MBUF_F_RX_IP_CKSUM_GOOD | RTE_MBUF_F_RX_IP_CKSUM_BAD | + RTE_MBUF_F_RX_L4_CKSUM_GOOD | RTE_MBUF_F_RX_L4_CKSUM_BAD | + RTE_MBUF_F_RX_OUTER_IP_CKSUM_BAD, + RTE_MBUF_F_RX_IP_CKSUM_GOOD | RTE_MBUF_F_RX_IP_CKSUM_BAD | + RTE_MBUF_F_RX_L4_CKSUM_GOOD | RTE_MBUF_F_RX_L4_CKSUM_BAD | + RTE_MBUF_F_RX_OUTER_IP_CKSUM_BAD}; + + /* map rss and vlan type to rss hash and vlan flag */ + const uint8x16_t vlan_flags = { + 0, 0, 0, 0, + RTE_MBUF_F_RX_VLAN | RTE_MBUF_F_RX_VLAN_STRIPPED, 0, 0, 0, + 0, 0, 0, 0, + 0, 0, 0, 0}; + + const uint8x16_t rss_flags = { + 0, RTE_MBUF_F_RX_FDIR, 0, 0, + 0, 0, RTE_MBUF_F_RX_RSS_HASH, RTE_MBUF_F_RX_RSS_HASH | RTE_MBUF_F_RX_FDIR, + 0, 0, 0, 0, + 0, 0, 0, 0}; + + const uint8x16_t l3_l4e_flags = { + (RTE_MBUF_F_RX_IP_CKSUM_GOOD | RTE_MBUF_F_RX_L4_CKSUM_GOOD) >> 1, + RTE_MBUF_F_RX_IP_CKSUM_BAD >> 1, + (RTE_MBUF_F_RX_IP_CKSUM_GOOD | RTE_MBUF_F_RX_L4_CKSUM_BAD) >> 1, + (RTE_MBUF_F_RX_L4_CKSUM_BAD | RTE_MBUF_F_RX_IP_CKSUM_BAD) >> 1, + (RTE_MBUF_F_RX_IP_CKSUM_GOOD | RTE_MBUF_F_RX_OUTER_IP_CKSUM_BAD) >> 1, + (RTE_MBUF_F_RX_OUTER_IP_CKSUM_BAD | RTE_MBUF_F_RX_IP_CKSUM_BAD) >> 1, + (RTE_MBUF_F_RX_IP_CKSUM_GOOD | RTE_MBUF_F_RX_OUTER_IP_CKSUM_BAD | + RTE_MBUF_F_RX_L4_CKSUM_BAD) >> 1, + (RTE_MBUF_F_RX_OUTER_IP_CKSUM_BAD | RTE_MBUF_F_RX_L4_CKSUM_BAD | + RTE_MBUF_F_RX_IP_CKSUM_BAD) >> 1, + 0, 0, 0, 0, 0, 0, 0, 0}; + + vlan0 = vzipq_u32(vreinterpretq_u32_u64(descs[0]), + vreinterpretq_u32_u64(descs[2])).val[1]; + vlan1 = vzipq_u32(vreinterpretq_u32_u64(descs[1]), + vreinterpretq_u32_u64(descs[3])).val[1]; + vlan0 = vzipq_u32(vlan0, vlan1).val[0]; + + vlan1 = vandq_u32(vlan0, rss_vlan_msk); + vlan0 = vreinterpretq_u32_u8(vqtbl1q_u8(vlan_flags, + vreinterpretq_u8_u32(vlan1))); + + const uint32x4_t desc_fltstat = vshrq_n_u32(vlan1, 11); + rss = vreinterpretq_u32_u8(vqtbl1q_u8(rss_flags, + vreinterpretq_u8_u32(desc_fltstat))); + + l3_l4e = vshrq_n_u32(vlan1, 22); + l3_l4e = vreinterpretq_u32_u8(vqtbl1q_u8(l3_l4e_flags, + vreinterpretq_u8_u32(l3_l4e))); + /* then we shift left 1 bit */ + l3_l4e = vshlq_n_u32(l3_l4e, 1); + /* we need to mask out the redundant bits */ + l3_l4e = vandq_u32(l3_l4e, cksum_mask); + + vlan0 = vorrq_u32(vlan0, rss); + vlan0 = vorrq_u32(vlan0, l3_l4e); + + rearm0 = vsetq_lane_u64(vgetq_lane_u32(vlan0, 0), mbuf_init, 1); + rearm1 = vsetq_lane_u64(vgetq_lane_u32(vlan0, 1), mbuf_init, 1); + rearm2 = vsetq_lane_u64(vgetq_lane_u32(vlan0, 2), mbuf_init, 1); + rearm3 = vsetq_lane_u64(vgetq_lane_u32(vlan0, 3), mbuf_init, 1); + + vst1q_u64((uint64_t *)&rx_pkts[0]->rearm_data, rearm0); + vst1q_u64((uint64_t *)&rx_pkts[1]->rearm_data, rearm1); + vst1q_u64((uint64_t *)&rx_pkts[2]->rearm_data, rearm2); + vst1q_u64((uint64_t *)&rx_pkts[3]->rearm_data, rearm3); +} + +#define PKTLEN_SHIFT 10 +#define IAVF_UINT16_BIT (CHAR_BIT * sizeof(uint16_t)) + +static inline void +desc_to_ptype_v(uint64x2_t descs[4], struct rte_mbuf **__rte_restrict rx_pkts, + uint32_t *__rte_restrict ptype_tbl) +{ + int i; + uint8_t ptype; + uint8x16_t tmp; + + for (i = 0; i < 4; i++) { + tmp = vreinterpretq_u8_u64(vshrq_n_u64(descs[i], 30)); + ptype = vgetq_lane_u8(tmp, 8); + rx_pkts[i]->packet_type = ptype_tbl[ptype]; + } +} + +/** + * vPMD raw receive routine, only accept(nb_pkts >= IAVF_VPMD_DESCS_PER_LOOP) + * + * Notice: + * - nb_pkts < IAVF_VPMD_DESCS_PER_LOOP, just return no packet + * - floor align nb_pkts to a IAVF_VPMD_DESCS_PER_LOOP power-of-two + */ +static inline uint16_t +_recv_raw_pkts_vec(struct iavf_rx_queue *__rte_restrict rxq, + struct rte_mbuf **__rte_restrict rx_pkts, + uint16_t nb_pkts, uint8_t *split_packet) +{ + RTE_SET_USED(split_packet); + + volatile union iavf_rx_desc *rxdp; + struct rte_mbuf **sw_ring; + uint16_t nb_pkts_recd; + int pos; + uint32_t *ptype_tbl = rxq->vsi->adapter->ptype_tbl; + + /* mask to shuffle from desc. to mbuf */ + uint8x16_t shuf_msk = { + 0xFF, 0xFF, /* pkt_type set as unknown */ + 0xFF, 0xFF, /* pkt_type set as unknown */ + 14, 15, /* octet 15~14, low 16 bits pkt_len */ + 0xFF, 0xFF, /* skip high 16 bits pkt_len, zero out */ + 14, 15, /* octet 15~14, 16 bits data_len */ + 2, 3, /* octet 2~3, low 16 bits vlan_macip */ + 4, 5, 6, 7 /* octet 4~7, 32bits rss */ + }; + + uint16x8_t crc_adjust = { + 0, 0, /* ignore pkt_type field */ + rxq->crc_len, /* sub crc on pkt_len */ + 0, /* ignore high-16bits of pkt_len */ + rxq->crc_len, /* sub crc on data_len */ + 0, 0, 0 /* ignore non-length fields */ + }; + /* nb_pkts has to be floor-aligned to IAVF_VPMD_DESCS_PER_LOOP */ + nb_pkts = RTE_ALIGN_FLOOR(nb_pkts, IAVF_VPMD_DESCS_PER_LOOP); + + rxdp = rxq->rx_ring + rxq->rx_tail; + + rte_prefetch_non_temporal(rxdp); + + /* 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; + + /* 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 += IAVF_VPMD_DESCS_PER_LOOP, + rxdp += IAVF_VPMD_DESCS_PER_LOOP) { + uint64x2_t descs[IAVF_VPMD_DESCS_PER_LOOP]; + uint8x16_t pkt_mb1, pkt_mb2, pkt_mb3, pkt_mb4; + uint16x8x2_t sterr_tmp1, sterr_tmp2; + uint64x2_t mbp1, mbp2; + uint16x8_t staterr; + uint16x8_t tmp; + uint64_t stat; + + int32x4_t len_shl = {0, 0, 0, PKTLEN_SHIFT}; + + /* A.1 load desc[3-0] */ + descs[3] = vld1q_u64((uint64_t *)(rxdp + 3)); + descs[2] = vld1q_u64((uint64_t *)(rxdp + 2)); + descs[1] = vld1q_u64((uint64_t *)(rxdp + 1)); + descs[0] = vld1q_u64((uint64_t *)(rxdp)); + + /* Use acquire fence to order loads of descriptor qwords */ + rte_atomic_thread_fence(__ATOMIC_ACQUIRE); + /* A.2 reload qword0 to make it ordered after qword1 load */ + descs[3] = vld1q_lane_u64((uint64_t *)(rxdp + 3), descs[3], 0); + descs[2] = vld1q_lane_u64((uint64_t *)(rxdp + 2), descs[2], 0); + descs[1] = vld1q_lane_u64((uint64_t *)(rxdp + 1), descs[1], 0); + descs[0] = vld1q_lane_u64((uint64_t *)(rxdp), descs[0], 0); + + /* B.1 load 4 mbuf point */ + mbp1 = vld1q_u64((uint64_t *)&sw_ring[pos]); + mbp2 = vld1q_u64((uint64_t *)&sw_ring[pos + 2]); + + /* B.2 copy 4 mbuf point into rx_pkts */ + vst1q_u64((uint64_t *)&rx_pkts[pos], mbp1); + vst1q_u64((uint64_t *)&rx_pkts[pos + 2], mbp2); + + /* pkts shift the pktlen field to be 16-bit aligned*/ + uint32x4_t len3 = vshlq_u32(vreinterpretq_u32_u64(descs[3]), + len_shl); + descs[3] = vreinterpretq_u64_u16(vsetq_lane_u16 + (vgetq_lane_u16(vreinterpretq_u16_u32(len3), 7), + vreinterpretq_u16_u64(descs[3]), + 7)); + uint32x4_t len2 = vshlq_u32(vreinterpretq_u32_u64(descs[2]), + len_shl); + descs[2] = vreinterpretq_u64_u16(vsetq_lane_u16 + (vgetq_lane_u16(vreinterpretq_u16_u32(len2), 7), + vreinterpretq_u16_u64(descs[2]), + 7)); + uint32x4_t len1 = vshlq_u32(vreinterpretq_u32_u64(descs[1]), + len_shl); + descs[1] = vreinterpretq_u64_u16(vsetq_lane_u16 + (vgetq_lane_u16(vreinterpretq_u16_u32(len1), 7), + vreinterpretq_u16_u64(descs[1]), + 7)); + uint32x4_t len0 = vshlq_u32(vreinterpretq_u32_u64(descs[0]), + len_shl); + descs[0] = vreinterpretq_u64_u16(vsetq_lane_u16 + (vgetq_lane_u16(vreinterpretq_u16_u32(len0), 7), + vreinterpretq_u16_u64(descs[0]), + 7)); + desc_to_olflags_v(rxq, rxdp, descs, &rx_pkts[pos]); + + /* D.1 pkts convert format from desc to pktmbuf */ + pkt_mb4 = vqtbl1q_u8(vreinterpretq_u8_u64(descs[3]), shuf_msk); + pkt_mb3 = vqtbl1q_u8(vreinterpretq_u8_u64(descs[2]), shuf_msk); + pkt_mb2 = vqtbl1q_u8(vreinterpretq_u8_u64(descs[1]), shuf_msk); + pkt_mb1 = vqtbl1q_u8(vreinterpretq_u8_u64(descs[0]), shuf_msk); + + /* D.2 pkts set in_port/nb_seg and remove crc */ + tmp = vsubq_u16(vreinterpretq_u16_u8(pkt_mb4), crc_adjust); + pkt_mb4 = vreinterpretq_u8_u16(tmp); + tmp = vsubq_u16(vreinterpretq_u16_u8(pkt_mb3), crc_adjust); + pkt_mb3 = vreinterpretq_u8_u16(tmp); + tmp = vsubq_u16(vreinterpretq_u16_u8(pkt_mb2), crc_adjust); + pkt_mb2 = vreinterpretq_u8_u16(tmp); + tmp = vsubq_u16(vreinterpretq_u16_u8(pkt_mb1), crc_adjust); + pkt_mb1 = vreinterpretq_u8_u16(tmp); + + /* D.3 copy final data to rx_pkts */ + vst1q_u8((void *)&rx_pkts[pos + 3]->rx_descriptor_fields1, + pkt_mb4); + vst1q_u8((void *)&rx_pkts[pos + 2]->rx_descriptor_fields1, + pkt_mb3); + vst1q_u8((void *)&rx_pkts[pos + 1]->rx_descriptor_fields1, + pkt_mb2); + vst1q_u8((void *)&rx_pkts[pos]->rx_descriptor_fields1, + pkt_mb1); + + desc_to_ptype_v(descs, &rx_pkts[pos], ptype_tbl); + + if (likely(pos + IAVF_VPMD_DESCS_PER_LOOP < nb_pkts)) + rte_prefetch_non_temporal(rxdp + IAVF_VPMD_DESCS_PER_LOOP); + + /* C.1 4=>2 filter staterr info only */ + sterr_tmp2 = vzipq_u16(vreinterpretq_u16_u64(descs[1]), + vreinterpretq_u16_u64(descs[3])); + sterr_tmp1 = vzipq_u16(vreinterpretq_u16_u64(descs[0]), + vreinterpretq_u16_u64(descs[2])); + + /* C.2 get 4 pkts staterr value */ + staterr = vzipq_u16(sterr_tmp1.val[1], + sterr_tmp2.val[1]).val[0]; + + staterr = vshlq_n_u16(staterr, IAVF_UINT16_BIT - 1); + staterr = vreinterpretq_u16_s16( + vshrq_n_s16(vreinterpretq_s16_u16(staterr), + IAVF_UINT16_BIT - 1)); + stat = ~vgetq_lane_u64(vreinterpretq_u64_u16(staterr), 0); + + /* C.4 calc available number of desc */ + if (unlikely(stat == 0)) { + nb_pkts_recd += IAVF_VPMD_DESCS_PER_LOOP; + } else { + nb_pkts_recd += __builtin_ctzl(stat) / IAVF_UINT16_BIT; + 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 < IAVF_VPMD_DESCS_PER_LOOP, just return no packet + * - nb_pkts > IAVF_VPMD_RX_BURST, only scan IAVF_VPMD_RX_BURST + * numbers of DD bits + */ +uint16_t +iavf_recv_pkts_vec(void *__rte_restrict rx_queue, + struct rte_mbuf **__rte_restrict rx_pkts, uint16_t nb_pkts) +{ + return _recv_raw_pkts_vec(rx_queue, rx_pkts, nb_pkts, NULL); +} + +static void __rte_cold +iavf_rx_queue_release_mbufs_neon(struct iavf_rx_queue *rxq) +{ + _iavf_rx_queue_release_mbufs_vec(rxq); +} + +static const struct iavf_rxq_ops neon_vec_rxq_ops = { + .release_mbufs = iavf_rx_queue_release_mbufs_neon, +}; + +int __rte_cold +iavf_rxq_vec_setup(struct iavf_rx_queue *rxq) +{ + rxq->ops = &neon_vec_rxq_ops; + return iavf_rxq_vec_setup_default(rxq); +} + +int __rte_cold +iavf_rx_vec_dev_check(struct rte_eth_dev *dev) +{ + return iavf_rx_vec_dev_check_default(dev); +} diff --git a/drivers/net/iavf/meson.build b/drivers/net/iavf/meson.build index 5eb230f687..d32d91885a 100644 --- a/drivers/net/iavf/meson.build +++ b/drivers/net/iavf/meson.build @@ -65,6 +65,8 @@ if arch_subdir == 'x86' c_args: avx512_args) objs += iavf_avx512_lib.extract_objects('iavf_rxtx_vec_avx512.c') endif +elif arch_subdir == 'arm' + sources += files('iavf_rxtx_vec_neon.c') endif headers = files('rte_pmd_iavf.h') -- 2.17.1