* [PATCH 1/4] e1000e: save skb counts in TX to avoid cache misses
@ 2010-04-24 0:25 Tom Herbert
2010-04-24 6:30 ` David Miller
` (2 more replies)
0 siblings, 3 replies; 4+ messages in thread
From: Tom Herbert @ 2010-04-24 0:25 UTC (permalink / raw)
To: davem, netdev
In e1000_tx_map, precompute number of segements and bytecounts which
are derived from fields in skb; these are stored in buffer_info. When
cleaning tx in e1000_clean_tx_irq use the values in the associated
buffer_info for statistics counting, this eliminates cache misses
on skb fields.
Signed-off-by: Tom Herbert <therbert@google.com>
---
diff --git a/drivers/net/e1000e/e1000.h b/drivers/net/e1000e/e1000.h
index 12648a1..d6da75b 100644
--- a/drivers/net/e1000e/e1000.h
+++ b/drivers/net/e1000e/e1000.h
@@ -188,6 +188,8 @@ struct e1000_buffer {
unsigned long time_stamp;
u16 length;
u16 next_to_watch;
+ unsigned int segs;
+ unsigned int bytecount;
u16 mapped_as_page;
};
/* Rx */
diff --git a/drivers/net/e1000e/netdev.c b/drivers/net/e1000e/netdev.c
index 5f70c43..4f5034a 100644
--- a/drivers/net/e1000e/netdev.c
+++ b/drivers/net/e1000e/netdev.c
@@ -646,14 +635,8 @@ static bool e1000_clean_tx_irq(struct e1000_adapter *adapter)
cleaned = (i == eop);
if (cleaned) {
- struct sk_buff *skb = buffer_info->skb;
- unsigned int segs, bytecount;
- segs = skb_shinfo(skb)->gso_segs ?: 1;
- /* multiply data chunks by size of headers */
- bytecount = ((segs - 1) * skb_headlen(skb)) +
- skb->len;
- total_tx_packets += segs;
- total_tx_bytes += bytecount;
+ total_tx_packets += buffer_info->segs;
+ total_tx_bytes += buffer_info->bytecount;
}
e1000_put_txbuf(adapter, buffer_info);
@@ -3906,7 +3889,7 @@ static int e1000_tx_map(struct e1000_adapter *adapter,
struct e1000_buffer *buffer_info;
unsigned int len = skb_headlen(skb);
unsigned int offset = 0, size, count = 0, i;
- unsigned int f;
+ unsigned int f, bytecount, segs;
i = tx_ring->next_to_use;
@@ -3965,7 +3948,13 @@ static int e1000_tx_map(struct e1000_adapter *adapter,
}
}
+ segs = skb_shinfo(skb)->gso_segs ?: 1;
+ /* multiply data chunks by size of headers */
+ bytecount = ((segs - 1) * skb_headlen(skb)) + skb->len;
+
tx_ring->buffer_info[i].skb = skb;
+ tx_ring->buffer_info[i].segs = segs;
+ tx_ring->buffer_info[i].bytecount = bytecount;
tx_ring->buffer_info[first].next_to_watch = i;
return count;
^ permalink raw reply related [flat|nested] 4+ messages in thread* Re: [PATCH 1/4] e1000e: save skb counts in TX to avoid cache misses
2010-04-24 0:25 [PATCH 1/4] e1000e: save skb counts in TX to avoid cache misses Tom Herbert
@ 2010-04-24 6:30 ` David Miller
2010-04-26 15:55 ` Alexander Duyck
2010-04-26 22:01 ` Jeff Kirsher
2 siblings, 0 replies; 4+ messages in thread
From: David Miller @ 2010-04-24 6:30 UTC (permalink / raw)
To: therbert
Cc: netdev, jeffrey.t.kirsher, jesse.brandeburg, bruce.w.allan,
alexander.h.duyck, peter.p.waskiewicz.jr, john.ronciak
Intel folks, please integrate Tom's patches.
Tom, please always CC: the developers listed in MAINTAINERS
for the driver you are submitting patches to.
Thanks.
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [PATCH 1/4] e1000e: save skb counts in TX to avoid cache misses
2010-04-24 0:25 [PATCH 1/4] e1000e: save skb counts in TX to avoid cache misses Tom Herbert
2010-04-24 6:30 ` David Miller
@ 2010-04-26 15:55 ` Alexander Duyck
2010-04-26 22:01 ` Jeff Kirsher
2 siblings, 0 replies; 4+ messages in thread
From: Alexander Duyck @ 2010-04-26 15:55 UTC (permalink / raw)
To: Tom Herbert; +Cc: davem@davemloft.net, netdev@vger.kernel.org
Tom Herbert wrote:
> In e1000_tx_map, precompute number of segements and bytecounts which
> are derived from fields in skb; these are stored in buffer_info. When
> cleaning tx in e1000_clean_tx_irq use the values in the associated
> buffer_info for statistics counting, this eliminates cache misses
> on skb fields.
>
> Signed-off-by: Tom Herbert <therbert@google.com>
> ---
> diff --git a/drivers/net/e1000e/e1000.h b/drivers/net/e1000e/e1000.h
> index 12648a1..d6da75b 100644
> --- a/drivers/net/e1000e/e1000.h
> +++ b/drivers/net/e1000e/e1000.h
> @@ -188,6 +188,8 @@ struct e1000_buffer {
> unsigned long time_stamp;
> u16 length;
> u16 next_to_watch;
> + unsigned int segs;
> + unsigned int bytecount;
> u16 mapped_as_page;
> };
> /* Rx */
Does segs need to be a full int here? I think you can probably get away
with either an unsigned short or just define it as a u16, then combine
it with mapped_as_page to cut down on the overall size.
Thanks,
Alex
^ permalink raw reply [flat|nested] 4+ messages in thread* Re: [PATCH 1/4] e1000e: save skb counts in TX to avoid cache misses
2010-04-24 0:25 [PATCH 1/4] e1000e: save skb counts in TX to avoid cache misses Tom Herbert
2010-04-24 6:30 ` David Miller
2010-04-26 15:55 ` Alexander Duyck
@ 2010-04-26 22:01 ` Jeff Kirsher
2 siblings, 0 replies; 4+ messages in thread
From: Jeff Kirsher @ 2010-04-26 22:01 UTC (permalink / raw)
To: Tom Herbert; +Cc: davem, netdev
On Fri, Apr 23, 2010 at 17:25, Tom Herbert <therbert@google.com> wrote:
> In e1000_tx_map, precompute number of segements and bytecounts which
> are derived from fields in skb; these are stored in buffer_info. When
> cleaning tx in e1000_clean_tx_irq use the values in the associated
> buffer_info for statistics counting, this eliminates cache misses
> on skb fields.
>
> Signed-off-by: Tom Herbert <therbert@google.com>
> ---
> diff --git a/drivers/net/e1000e/e1000.h b/drivers/net/e1000e/e1000.h
> index 12648a1..d6da75b 100644
> --- a/drivers/net/e1000e/e1000.h
> +++ b/drivers/net/e1000e/e1000.h
> @@ -188,6 +188,8 @@ struct e1000_buffer {
> unsigned long time_stamp;
> u16 length;
> u16 next_to_watch;
> + unsigned int segs;
> + unsigned int bytecount;
> u16 mapped_as_page;
> };
> /* Rx */
> diff --git a/drivers/net/e1000e/netdev.c b/drivers/net/e1000e/netdev.c
> index 5f70c43..4f5034a 100644
> --- a/drivers/net/e1000e/netdev.c
> +++ b/drivers/net/e1000e/netdev.c
> @@ -646,14 +635,8 @@ static bool e1000_clean_tx_irq(struct e1000_adapter *adapter)
> cleaned = (i == eop);
>
> if (cleaned) {
> - struct sk_buff *skb = buffer_info->skb;
> - unsigned int segs, bytecount;
> - segs = skb_shinfo(skb)->gso_segs ?: 1;
> - /* multiply data chunks by size of headers */
> - bytecount = ((segs - 1) * skb_headlen(skb)) +
> - skb->len;
> - total_tx_packets += segs;
> - total_tx_bytes += bytecount;
> + total_tx_packets += buffer_info->segs;
> + total_tx_bytes += buffer_info->bytecount;
> }
>
> e1000_put_txbuf(adapter, buffer_info);
> @@ -3906,7 +3889,7 @@ static int e1000_tx_map(struct e1000_adapter *adapter,
> struct e1000_buffer *buffer_info;
> unsigned int len = skb_headlen(skb);
> unsigned int offset = 0, size, count = 0, i;
> - unsigned int f;
> + unsigned int f, bytecount, segs;
>
> i = tx_ring->next_to_use;
>
> @@ -3965,7 +3948,13 @@ static int e1000_tx_map(struct e1000_adapter *adapter,
> }
> }
>
> + segs = skb_shinfo(skb)->gso_segs ?: 1;
> + /* multiply data chunks by size of headers */
> + bytecount = ((segs - 1) * skb_headlen(skb)) + skb->len;
> +
> tx_ring->buffer_info[i].skb = skb;
> + tx_ring->buffer_info[i].segs = segs;
> + tx_ring->buffer_info[i].bytecount = bytecount;
> tx_ring->buffer_info[first].next_to_watch = i;
>
> return count;
> --
>
I have added your patches to my queue of e1000e patches, thanks.
--
Cheers,
Jeff
^ permalink raw reply [flat|nested] 4+ messages in thread
end of thread, other threads:[~2010-04-26 22:01 UTC | newest]
Thread overview: 4+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2010-04-24 0:25 [PATCH 1/4] e1000e: save skb counts in TX to avoid cache misses Tom Herbert
2010-04-24 6:30 ` David Miller
2010-04-26 15:55 ` Alexander Duyck
2010-04-26 22:01 ` Jeff Kirsher
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).