Re: [dpdk-dev] [PATCH v3] vhost: read last used index once

[Maxime Coquelin <maxime.coquelin@redhat.com>]

On 4/7/21 11:55 AM, Balazs Nemeth wrote:
> Instead of calculating the address of a packed descriptor based on the
> vq->desc_packed and vq->last_used_idx every time, store that base
> address in desc_base. On arm, this saves 176 bytes in code size of
> function in which vhost_flush_enqueue_batch_packed gets inlined.
> 
> Signed-off-by: Balazs Nemeth <bnemeth@redhat.com>
> ---
>  lib/librte_vhost/virtio_net.c | 13 ++++++++-----
>  1 file changed, 8 insertions(+), 5 deletions(-)
> 
> diff --git a/lib/librte_vhost/virtio_net.c b/lib/librte_vhost/virtio_net.c
> index 179c57b46..f091384a6 100644
> --- a/lib/librte_vhost/virtio_net.c
> +++ b/lib/librte_vhost/virtio_net.c
> @@ -217,6 +217,8 @@ vhost_flush_enqueue_batch_packed(struct virtio_net *dev,
>  {
>  	uint16_t i;
>  	uint16_t flags;
> +	uint16_t last_used_idx = vq->last_used_idx;
> +	struct vring_packed_desc *desc_base = &vq->desc_packed[last_used_idx];
>  
>  	if (vq->shadow_used_idx) {
>  		do_data_copy_enqueue(dev, vq);
> @@ -226,16 +228,17 @@ vhost_flush_enqueue_batch_packed(struct virtio_net *dev,
>  	flags = PACKED_DESC_ENQUEUE_USED_FLAG(vq->used_wrap_counter);
>  
>  	vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE) {
> -		vq->desc_packed[vq->last_used_idx + i].id = ids[i];
> -		vq->desc_packed[vq->last_used_idx + i].len = lens[i];
> +		desc_base[i].id = ids[i];
> +		desc_base[i].len = lens[i];
>  	}
>  
>  	rte_atomic_thread_fence(__ATOMIC_RELEASE);
>  
> -	vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE)
> -		vq->desc_packed[vq->last_used_idx + i].flags = flags;
> +	vhost_for_each_try_unroll(i, 0, PACKED_BATCH_SIZE) {
> +		desc_base[i].flags = flags;
> +	}
>  
> -	vhost_log_cache_used_vring(dev, vq, vq->last_used_idx *
> +	vhost_log_cache_used_vring(dev, vq, last_used_idx *
>  				   sizeof(struct vring_packed_desc),
>  				   sizeof(struct vring_packed_desc) *
>  				   PACKED_BATCH_SIZE);
> 

Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>

Thanks,
Maxime