Influential linux settings on socket behaviour and performance

Question

I'am looking forward to improve my application speed/throughput, sending packets from a UDP socket in C.

I want to tune the performance of my system in order get some extra Mb/s.

I have already looked into :

/proc/sys/net/core

/proc/sys/net/ipv4

/proc/sys/net/ipv6

From those, I know that wmem_max helps me increase the write size of my socket buffer. I'am wondering what else I can change, keeping in mind that my goal is to increase the performance of sending packets:

ls /proc/sys/net/core/

busy_poll                     fb_tunnels_only_for_init_net  netdev_budget                 rmem_default                  wmem_default
busy_read                     flow_limit_cpu_bitmap         netdev_budget_usecs           rmem_max                      wmem_max
default_qdisc                 flow_limit_table_len          netdev_max_backlog            rps_sock_flow_entries         xfrm_acq_expires
dev_weight                    max_skb_frags                 netdev_rss_key                somaxconn                     xfrm_aevent_etime
dev_weight_rx_bias            message_burst                 netdev_tstamp_prequeue        tstamp_allow_data             xfrm_aevent_rseqth
dev_weight_tx_bias            message_cost                  optmem_max                    warnings                      xfrm_larval_drop