Improve prefetching in handle_mirror

Some of the prefetching code was commented out, which was wrong. The
prefetching is back in.

Change-Id: Id7e7f09bd62faa7b9faf3d677962a99ba2a71080
Signed-off-by: Luc Provoost <luc.provoost@intel.com>

diff --git a/VNFs/DPPD-PROX/handle_mirror.c b/VNFs/DPPD-PROX/handle_mirror.c
index 8e6eb6d..53e341e 100644 (file)
--- a/VNFs/DPPD-PROX/handle_mirror.c
+++ b/VNFs/DPPD-PROX/handle_mirror.c
@@ -61,9 +61,9 @@ static int handle_mirror_bulk(struct task_base *tbase, struct rte_mbuf **mbufs,
        rte_memcpy(mbufs2, mbufs, sizeof(mbufs[0]) * n_pkts);
        /* prefetch for optimization */
        prox_rte_ether_hdr * hdr[MAX_PKT_BURST];
-       //for (uint16_t i = 0; i < n_pkts; ++i) {
-       //      PREFETCH0(mbufs2[i]);
-       //}
+       for (uint16_t j = 0; j < n_pkts; ++j) {
+               PREFETCH0(mbufs2[j]);
+       }
        for (uint16_t j = 0; j < n_pkts; ++j) {
                hdr[j] = rte_pktmbuf_mtod(mbufs2[j], prox_rte_ether_hdr *);
                PREFETCH0(hdr[j]);