Change Details:
- Prefetch header in GRO path. This reduces napi_frags_skb time from 9% to 5%.
- Changed the configurable limit of RxQ depth to 16384 (was 2048).
- bnad_rx_unmap_q elements are cachealigned.
Signed-off-by: Rasesh Mody <rmody@brocade.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
unmap_q = rcb->unmap_q;
bnad = rcb->bnad;
+
+ /* prefetch header */
+ prefetch(page_address(unmap_q->unmap[sop_ci].page) +
+ unmap_q->unmap[sop_ci].page_offset);
+
for (vec = 1, ci = sop_ci; vec <= nvecs; vec++) {
unmap = &unmap_q->unmap[ci];
BNA_QE_INDX_INC(ci, rcb->q_depth);
#define BNAD_IOCETH_TIMEOUT 10000
#define BNAD_MIN_Q_DEPTH 512
-#define BNAD_MAX_RXQ_DEPTH 2048
+#define BNAD_MAX_RXQ_DEPTH 16384
#define BNAD_MAX_TXQ_DEPTH 2048
#define BNAD_JUMBO_MTU 9000
struct bnad_rx_unmap {
struct page *page;
- u32 page_offset;
struct sk_buff *skb;
struct bnad_rx_vector vector;
+ u32 page_offset;
};
enum bnad_rxbuf_type {
int alloc_order;
u32 map_size;
enum bnad_rxbuf_type type;
- struct bnad_rx_unmap unmap[0];
+ struct bnad_rx_unmap unmap[0] ____cacheline_aligned;
};
#define BNAD_PCI_DEV_IS_CAT2(_bnad) \