* [PATCH] vhost-user: optimize stats counters performance
@ 2024-08-02 14:32 Morten Brørup
2024-08-06 8:24 ` Chenbo Xia
` (2 more replies)
0 siblings, 3 replies; 4+ messages in thread
From: Morten Brørup @ 2024-08-02 14:32 UTC (permalink / raw)
To: maxime.coquelin, chenbox; +Cc: dev, Morten Brørup
Optimized the performance of updating the statistics counters by reducing
the number of branches.
Ordered the packet size comparisons according to the probability with
typical internet traffic mix.
Signed-off-by: Morten Brørup <mb@smartsharesystems.com>
---
lib/vhost/virtio_net.c | 40 ++++++++++++++--------------------------
1 file changed, 14 insertions(+), 26 deletions(-)
diff --git a/lib/vhost/virtio_net.c b/lib/vhost/virtio_net.c
index 370402d849..25a495df56 100644
--- a/lib/vhost/virtio_net.c
+++ b/lib/vhost/virtio_net.c
@@ -53,7 +53,7 @@ is_valid_virt_queue_idx(uint32_t idx, int is_tx, uint32_t nr_vring)
}
static inline void
-vhost_queue_stats_update(struct virtio_net *dev, struct vhost_virtqueue *vq,
+vhost_queue_stats_update(const struct virtio_net *dev, struct vhost_virtqueue *vq,
struct rte_mbuf **pkts, uint16_t count)
__rte_shared_locks_required(&vq->access_lock)
{
@@ -64,37 +64,25 @@ vhost_queue_stats_update(struct virtio_net *dev, struct vhost_virtqueue *vq,
return;
for (i = 0; i < count; i++) {
- struct rte_ether_addr *ea;
- struct rte_mbuf *pkt = pkts[i];
+ const struct rte_ether_addr *ea;
+ const struct rte_mbuf *pkt = pkts[i];
uint32_t pkt_len = rte_pktmbuf_pkt_len(pkt);
stats->packets++;
stats->bytes += pkt_len;
- if (pkt_len == 64) {
- stats->size_bins[1]++;
- } else if (pkt_len > 64 && pkt_len < 1024) {
- uint32_t bin;
-
- /* count zeros, and offset into correct bin */
- bin = (sizeof(pkt_len) * 8) - rte_clz32(pkt_len) - 5;
- stats->size_bins[bin]++;
- } else {
- if (pkt_len < 64)
- stats->size_bins[0]++;
- else if (pkt_len < 1519)
- stats->size_bins[6]++;
- else
- stats->size_bins[7]++;
- }
+ if (pkt_len >= 1024)
+ stats->size_bins[6 + (pkt_len > 1518)]++;
+ else if (pkt_len <= 64)
+ stats->size_bins[pkt_len >> 6]++;
+ else
+ stats->size_bins[32UL - rte_clz32(pkt_len) - 5]++;
- ea = rte_pktmbuf_mtod(pkt, struct rte_ether_addr *);
- if (rte_is_multicast_ether_addr(ea)) {
- if (rte_is_broadcast_ether_addr(ea))
- stats->broadcast++;
- else
- stats->multicast++;
- }
+ ea = rte_pktmbuf_mtod(pkt, const struct rte_ether_addr *);
+ RTE_BUILD_BUG_ON(offsetof(struct virtqueue_stats, broadcast) !=
+ offsetof(struct virtqueue_stats, multicast) + sizeof(uint64_t));
+ if (unlikely(rte_is_multicast_ether_addr(ea)))
+ (&stats->multicast)[rte_is_broadcast_ether_addr(ea)]++;
}
}
--
2.43.0
^ permalink raw reply [flat|nested] 4+ messages in thread* Re: [PATCH] vhost-user: optimize stats counters performance
2024-08-02 14:32 [PATCH] vhost-user: optimize stats counters performance Morten Brørup
@ 2024-08-06 8:24 ` Chenbo Xia
2024-09-10 14:59 ` Maxime Coquelin
2024-09-19 12:05 ` Maxime Coquelin
2 siblings, 0 replies; 4+ messages in thread
From: Chenbo Xia @ 2024-08-06 8:24 UTC (permalink / raw)
To: Morten Brørup; +Cc: Maxime Coquelin, dev
> On Aug 2, 2024, at 22:32, Morten Brørup <mb@smartsharesystems.com> wrote:
>
> External email: Use caution opening links or attachments
>
>
> Optimized the performance of updating the statistics counters by reducing
> the number of branches.
>
> Ordered the packet size comparisons according to the probability with
> typical internet traffic mix.
>
> Signed-off-by: Morten Brørup <mb@smartsharesystems.com>
> ---
> lib/vhost/virtio_net.c | 40 ++++++++++++++--------------------------
> 1 file changed, 14 insertions(+), 26 deletions(-)
>
> diff --git a/lib/vhost/virtio_net.c b/lib/vhost/virtio_net.c
> index 370402d849..25a495df56 100644
> --- a/lib/vhost/virtio_net.c
> +++ b/lib/vhost/virtio_net.c
> @@ -53,7 +53,7 @@ is_valid_virt_queue_idx(uint32_t idx, int is_tx, uint32_t nr_vring)
> }
>
> static inline void
> -vhost_queue_stats_update(struct virtio_net *dev, struct vhost_virtqueue *vq,
> +vhost_queue_stats_update(const struct virtio_net *dev, struct vhost_virtqueue *vq,
> struct rte_mbuf **pkts, uint16_t count)
> __rte_shared_locks_required(&vq->access_lock)
> {
> @@ -64,37 +64,25 @@ vhost_queue_stats_update(struct virtio_net *dev, struct vhost_virtqueue *vq,
> return;
>
> for (i = 0; i < count; i++) {
> - struct rte_ether_addr *ea;
> - struct rte_mbuf *pkt = pkts[i];
> + const struct rte_ether_addr *ea;
> + const struct rte_mbuf *pkt = pkts[i];
> uint32_t pkt_len = rte_pktmbuf_pkt_len(pkt);
>
> stats->packets++;
> stats->bytes += pkt_len;
>
> - if (pkt_len == 64) {
> - stats->size_bins[1]++;
> - } else if (pkt_len > 64 && pkt_len < 1024) {
> - uint32_t bin;
> -
> - /* count zeros, and offset into correct bin */
> - bin = (sizeof(pkt_len) * 8) - rte_clz32(pkt_len) - 5;
> - stats->size_bins[bin]++;
> - } else {
> - if (pkt_len < 64)
> - stats->size_bins[0]++;
> - else if (pkt_len < 1519)
> - stats->size_bins[6]++;
> - else
> - stats->size_bins[7]++;
> - }
> + if (pkt_len >= 1024)
> + stats->size_bins[6 + (pkt_len > 1518)]++;
> + else if (pkt_len <= 64)
> + stats->size_bins[pkt_len >> 6]++;
> + else
> + stats->size_bins[32UL - rte_clz32(pkt_len) - 5]++;
>
> - ea = rte_pktmbuf_mtod(pkt, struct rte_ether_addr *);
> - if (rte_is_multicast_ether_addr(ea)) {
> - if (rte_is_broadcast_ether_addr(ea))
> - stats->broadcast++;
> - else
> - stats->multicast++;
> - }
> + ea = rte_pktmbuf_mtod(pkt, const struct rte_ether_addr *);
> + RTE_BUILD_BUG_ON(offsetof(struct virtqueue_stats, broadcast) !=
> + offsetof(struct virtqueue_stats, multicast) + sizeof(uint64_t));
> + if (unlikely(rte_is_multicast_ether_addr(ea)))
> + (&stats->multicast)[rte_is_broadcast_ether_addr(ea)]++;
> }
> }
>
> --
> 2.43.0
>
Reviewed-by: Chenbo Xia <chenbox@nvidia.com>
^ permalink raw reply [flat|nested] 4+ messages in thread* Re: [PATCH] vhost-user: optimize stats counters performance
2024-08-02 14:32 [PATCH] vhost-user: optimize stats counters performance Morten Brørup
2024-08-06 8:24 ` Chenbo Xia
@ 2024-09-10 14:59 ` Maxime Coquelin
2024-09-19 12:05 ` Maxime Coquelin
2 siblings, 0 replies; 4+ messages in thread
From: Maxime Coquelin @ 2024-09-10 14:59 UTC (permalink / raw)
To: Morten Brørup, chenbox; +Cc: dev
On 8/2/24 16:32, Morten Brørup wrote:
> Optimized the performance of updating the statistics counters by reducing
> the number of branches.
>
> Ordered the packet size comparisons according to the probability with
> typical internet traffic mix.
>
> Signed-off-by: Morten Brørup <mb@smartsharesystems.com>
> ---
> lib/vhost/virtio_net.c | 40 ++++++++++++++--------------------------
> 1 file changed, 14 insertions(+), 26 deletions(-)
>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
Thanks,
Maxime
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [PATCH] vhost-user: optimize stats counters performance
2024-08-02 14:32 [PATCH] vhost-user: optimize stats counters performance Morten Brørup
2024-08-06 8:24 ` Chenbo Xia
2024-09-10 14:59 ` Maxime Coquelin
@ 2024-09-19 12:05 ` Maxime Coquelin
2 siblings, 0 replies; 4+ messages in thread
From: Maxime Coquelin @ 2024-09-19 12:05 UTC (permalink / raw)
To: Morten Brørup, chenbox; +Cc: dev
On 8/2/24 16:32, Morten Brørup wrote:
> Optimized the performance of updating the statistics counters by reducing
> the number of branches.
>
> Ordered the packet size comparisons according to the probability with
> typical internet traffic mix.
>
> Signed-off-by: Morten Brørup <mb@smartsharesystems.com>
> ---
> lib/vhost/virtio_net.c | 40 ++++++++++++++--------------------------
> 1 file changed, 14 insertions(+), 26 deletions(-)
>
Applied to next-virtio/for-next-net.
Thanks,
Maxime
^ permalink raw reply [flat|nested] 4+ messages in thread
end of thread, other threads:[~2024-09-19 12:05 UTC | newest]
Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2024-08-02 14:32 [PATCH] vhost-user: optimize stats counters performance Morten Brørup
2024-08-06 8:24 ` Chenbo Xia
2024-09-10 14:59 ` Maxime Coquelin
2024-09-19 12:05 ` Maxime Coquelin
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).