eal: introduce macro for always inline

[dpdk.git] / examples / l3fwd / l3fwd_lpm_sse.h

diff --git a/examples/l3fwd/l3fwd_lpm_sse.h b/examples/l3fwd/l3fwd_lpm_sse.h
index d64d6d2..fa1b902 100644 (file)
--- a/examples/l3fwd/l3fwd_lpm_sse.h
+++ b/examples/l3fwd/l3fwd_lpm_sse.h
@@ -36,12 +36,11 @@
 
 #include "l3fwd_sse.h"
 
-static inline __attribute__((always_inline)) uint16_t
+static __rte_always_inline uint16_t
 lpm_get_dst_port(const struct lcore_conf *qconf, struct rte_mbuf *pkt,
                uint8_t portid)
 {
-       uint32_t next_hop_ipv4;
-       uint8_t next_hop_ipv6;
+       uint32_t next_hop;
        struct ipv6_hdr *ipv6_hdr;
        struct ipv4_hdr *ipv4_hdr;
        struct ether_hdr *eth_hdr;
@@ -51,9 +50,11 @@ lpm_get_dst_port(const struct lcore_conf *qconf, struct rte_mbuf *pkt,
                eth_hdr = rte_pktmbuf_mtod(pkt, struct ether_hdr *);
                ipv4_hdr = (struct ipv4_hdr *)(eth_hdr + 1);
 
-               return (uint16_t) ((rte_lpm_lookup(qconf->ipv4_lookup_struct,
-                               rte_be_to_cpu_32(ipv4_hdr->dst_addr), &next_hop_ipv4) == 0) ?
-                                               next_hop_ipv4 : portid);
+               return (uint16_t) (
+                       (rte_lpm_lookup(qconf->ipv4_lookup_struct,
+                                       rte_be_to_cpu_32(ipv4_hdr->dst_addr),
+                                       &next_hop) == 0) ?
+                                               next_hop : portid);
 
        } else if (RTE_ETH_IS_IPV6_HDR(pkt->packet_type)) {
 
@@ -61,8 +62,8 @@ lpm_get_dst_port(const struct lcore_conf *qconf, struct rte_mbuf *pkt,
                ipv6_hdr = (struct ipv6_hdr *)(eth_hdr + 1);
 
                return (uint16_t) ((rte_lpm6_lookup(qconf->ipv6_lookup_struct,
-                               ipv6_hdr->dst_addr, &next_hop_ipv6) == 0)
-                               ? next_hop_ipv6 : portid);
+                               ipv6_hdr->dst_addr, &next_hop) == 0)
+                               ? next_hop : portid);
 
        }
 
@@ -74,18 +75,17 @@ lpm_get_dst_port(const struct lcore_conf *qconf, struct rte_mbuf *pkt,
  * precalculated. If packet is ipv6 dst_addr is taken directly from packet
  * header and dst_ipv4 value is not used.
  */
-static inline __attribute__((always_inline)) uint16_t
+static __rte_always_inline uint16_t
 lpm_get_dst_port_with_ipv4(const struct lcore_conf *qconf, struct rte_mbuf *pkt,
        uint32_t dst_ipv4, uint8_t portid)
 {
-       uint32_t next_hop_ipv4;
-       uint8_t next_hop_ipv6;
+       uint32_t next_hop;
        struct ipv6_hdr *ipv6_hdr;
        struct ether_hdr *eth_hdr;
 
        if (RTE_ETH_IS_IPV4_HDR(pkt->packet_type)) {
                return (uint16_t) ((rte_lpm_lookup(qconf->ipv4_lookup_struct, dst_ipv4,
-                       &next_hop_ipv4) == 0) ? next_hop_ipv4 : portid);
+                       &next_hop) == 0) ? next_hop : portid);
 
        } else if (RTE_ETH_IS_IPV6_HDR(pkt->packet_type)) {
 
@@ -93,8 +93,8 @@ lpm_get_dst_port_with_ipv4(const struct lcore_conf *qconf, struct rte_mbuf *pkt,
                ipv6_hdr = (struct ipv6_hdr *)(eth_hdr + 1);
 
                return (uint16_t) ((rte_lpm6_lookup(qconf->ipv6_lookup_struct,
-                               ipv6_hdr->dst_addr, &next_hop_ipv6) == 0)
-                               ? next_hop_ipv6 : portid);
+                               ipv6_hdr->dst_addr, &next_hop) == 0)
+                               ? next_hop : portid);
 
        }
 
@@ -145,9 +145,9 @@ static inline void
 processx4_step2(const struct lcore_conf *qconf,
                __m128i dip,
                uint32_t ipv4_flag,
-               uint32_t portid,
+               uint8_t portid,
                struct rte_mbuf *pkt[FWDSTEP],
-               uint32_t dprt[FWDSTEP])
+               uint16_t dprt[FWDSTEP])
 {
        rte_xmm_t dst;
        const  __m128i bswap_mask = _mm_set_epi8(12, 13, 14, 15, 8, 9, 10, 11,
@@ -158,7 +158,11 @@ processx4_step2(const struct lcore_conf *qconf,
 
        /* if all 4 packets are IPV4. */
        if (likely(ipv4_flag)) {
-               rte_lpm_lookupx4(qconf->ipv4_lookup_struct, dip, dprt, portid);
+               rte_lpm_lookupx4(qconf->ipv4_lookup_struct, dip, dst.u32,
+                       portid);
+               /* get rid of unused upper 16 bit for each dport. */
+               dst.x = _mm_packs_epi32(dst.x, dst.x);
+               *(uint64_t *)dprt = dst.u64[0];
        } else {
                dst.x = dip;
                dprt[0] = lpm_get_dst_port_with_ipv4(qconf, pkt[0], dst.u32[0], portid);
@@ -177,7 +181,7 @@ l3fwd_lpm_send_packets(int nb_rx, struct rte_mbuf **pkts_burst,
                        uint8_t portid, struct lcore_conf *qconf)
 {
        int32_t j;
-       uint32_t dst_port[MAX_PKT_BURST];
+       uint16_t dst_port[MAX_PKT_BURST];
        __m128i dip[MAX_PKT_BURST / FWDSTEP];
        uint32_t ipv4_flag[MAX_PKT_BURST / FWDSTEP];
        const int32_t k = RTE_ALIGN_FLOOR(nb_rx, FWDSTEP);
@@ -195,9 +199,11 @@ l3fwd_lpm_send_packets(int nb_rx, struct rte_mbuf **pkts_burst,
        case 3:
                dst_port[j] = lpm_get_dst_port(qconf, pkts_burst[j], portid);
                j++;
+               /* fall-through */
        case 2:
                dst_port[j] = lpm_get_dst_port(qconf, pkts_burst[j], portid);
                j++;
+               /* fall-through */
        case 1:
                dst_port[j] = lpm_get_dst_port(qconf, pkts_burst[j], portid);
                j++;