davidlt/kernel-ark
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity
d62fda082c
kernel-ark/drivers
History
Eric Dumazet d62fda082c bnx2: bnx2_tx_int() optimizations 
When using bnx2 in a high transmit load, bnx2_tx_int() cost is pretty high.

There are two reasons.

One is an expensive call to bnx2_get_hw_tx_cons(bnapi) for each freed skb

One is cpu stalls when accessing skb_is_gso(skb) / skb_shinfo(skb)->nr_frags
because of two cache line misses.
(One to get skb->end/head to compute skb_shinfo(skb),
 one to get is_gso/nr_frags)

This patch :

1) avoids calling bnx2_get_hw_tx_cons(bnapi) too many times.

2) makes bnx2_start_xmit() cache is_gso & nr_frags into sw_tx_bd descriptor.
   This uses a litle bit more ram (256 longs per device on x86), but helps a lot.

3) uses a prefetch(&skb->end) to speedup dev_kfree_skb(), bringing
  cache line that will be needed in skb_release_data()

result is 5 % bandwidth increase in benchmarks, involving UDP or TCP receive
 & transmits, when a cpu is dedicated to ksoftirqd for bnx2.

bnx2_tx_int going from 3.33 % cpu to 0.5 % cpu in oprofile

Note : skb_dma_unmap() still very expensive but this is for another patch,
not related to bnx2 (2.9 % of cpu, while it does nothing on x86_32)

Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-05-17 20:47:44 -07:00
..
accessibility
acpi Revert "ACPI battery: fix async boot oops" 2009-04-11 12:45:20 -07:00
amba
ata ata: fix obviously wrong comment 2009-04-13 04:22:04 -04:00
atm dma-mapping: replace all DMA_32BIT_MASK macro with DMA_BIT_MASK(32) 2009-04-07 08:31:11 -07:00
auxdisplay auxdisplay: remove PARPORT dependency 2009-04-01 08:59:17 -07:00
base dma-mapping: replace all DMA_24BIT_MASK macro with DMA_BIT_MASK(24) 2009-04-07 08:31:12 -07:00
block xsysace: Fix dereferencing of cf_id after hd_driveid removal 2009-04-08 14:13:04 +02:00
bluetooth
cdrom
char esp: fix section mismatch warning 2009-04-07 08:31:03 -07:00
clocksource clocksource: sh_cmt: use remove_irq() and remove clockevent workaround 2009-04-02 11:18:48 +09:00
connector
cpufreq
cpuidle
crypto dma-mapping: replace all DMA_32BIT_MASK macro with DMA_BIT_MASK(32) 2009-04-07 08:31:11 -07:00
dca
dio
dma dma-mapping: replace all DMA_32BIT_MASK macro with DMA_BIT_MASK(32) 2009-04-07 08:31:11 -07:00
edac edac: AMD8111 driver Kconfig & Makefile 2009-04-02 19:05:04 -07:00
eisa
firewire
firmware dma-mapping: replace all DMA_32BIT_MASK macro with DMA_BIT_MASK(32) 2009-04-07 08:31:11 -07:00
gpio Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jikos/trivial 2009-04-03 15:24:35 -07:00
gpu Merge branch 'linus' into release 2009-04-05 02:14:15 -04:00
hid Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jikos/hid 2009-04-03 15:25:44 -07:00
hwmon Merge branch 'hwmon-for-linus' of git://jdelvare.pck.nerim.net/jdelvare-2.6 2009-04-07 08:44:43 -07:00
i2c i2c: Let new-style drivers implement attach_adapter 2009-04-13 17:02:14 +02:00
ide at91_ide: turn on PIO 6 support 2009-04-08 14:13:04 +02:00
idle dma-mapping: replace all DMA_64BIT_MASK macro with DMA_BIT_MASK(64) 2009-04-07 08:31:10 -07:00
ieee1394 dma-mapping: replace all DMA_32BIT_MASK macro with DMA_BIT_MASK(32) 2009-04-07 08:31:11 -07:00
infiniband Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland/infiniband 2009-04-09 16:42:26 -07:00
input Merge branch 'next' into for-linus 2009-04-08 00:00:33 -07:00
isdn Documentation/isdn/INTERFACE.CAPI 2009-04-27 05:37:39 -07:00
leds leds: just ignore invalid GPIOs in leds-gpio 2009-04-08 14:13:48 +01:00
lguest
macintosh
mca
md dm kcopyd: fix callback race 2009-04-09 00:27:17 +01:00
media Merge git://git.infradead.org/~dwmw2/firmware-2.6 2009-04-10 12:01:22 -07:00
memstick dma-mapping: replace all DMA_32BIT_MASK macro with DMA_BIT_MASK(32) 2009-04-07 08:31:11 -07:00
message dma-mapping: replace all DMA_32BIT_MASK macro with DMA_BIT_MASK(32) 2009-04-07 08:31:11 -07:00
mfd mfd: fix da903x warning 2009-04-05 00:32:25 +02:00
misc dma-mapping: replace all DMA_32BIT_MASK macro with DMA_BIT_MASK(32) 2009-04-07 08:31:11 -07:00
mmc mmc_spi: support for non-byte-aligned cards 2009-04-08 20:45:51 +02:00
mtd Merge branch 'linux-next' of git://git.infradead.org/ubi-2.6 2009-04-06 15:00:00 -07:00
net bnx2: bnx2_tx_int() optimizations 2009-05-17 20:47:44 -07:00
nubus
of openfirmware: Add OF phylib support code 2009-04-27 02:53:47 -07:00
oprofile Merge branch 'tracing-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip 2009-04-05 11:04:19 -07:00
parisc parisc: drivers: fix warnings 2009-04-02 00:40:41 +00:00
parport parport: Use the PCI IRQ if offered 2009-04-07 08:44:06 -07:00
pci pci: update 82576 sr-iov quirk with latest device IDs 2009-05-06 15:33:44 -07:00
pcmcia [ARM] 5442/1: pxa/cm-x255: fix reverse RDY gpios in PCMCIA driver 2009-04-02 11:02:32 +01:00
platform rfkill: remove user_claim stuff 2009-04-22 16:54:27 -04:00
pnp dma-mapping: replace all DMA_24BIT_MASK macro with DMA_BIT_MASK(24) 2009-04-07 08:31:12 -07:00
power Merge git://git.infradead.org/battery-2.6 2009-04-08 17:45:02 -07:00
ps3
rapidio dma-mapping: replace all DMA_32BIT_MASK macro with DMA_BIT_MASK(32) 2009-04-07 08:31:11 -07:00
regulator regulator: twl4030 VAUX3 supports 3.0V 2009-03-31 11:29:54 +01:00
rtc Merge git://git.kernel.org/pub/scm/linux/kernel/git/lethal/sh-2.6 2009-04-05 11:15:54 -07:00
s390 workqueue: add to_delayed_work() helper function 2009-04-02 19:04:50 -07:00
sbus
scsi PM/Hibernate: Wait for SCSI devices scan to complete during resume 2009-04-13 11:37:07 -07:00
serial 3c589_cs: add cis(firmware) of 3Com multifunction pcmcia card 2009-05-01 19:21:26 -07:00
sh sh: intc: set_irq_wake() support 2009-04-02 11:19:57 +09:00
sn dma-mapping: replace all DMA_64BIT_MASK macro with DMA_BIT_MASK(64) 2009-04-07 08:31:10 -07:00
spi spi_imx: set spi_master.dma_alignment = 4 2009-04-07 08:31:07 -07:00
ssb
staging serqt: initial clean up pass for tty side 2009-04-07 08:44:05 -07:00
tc
telephony
thermal
uio
usb cxacru: Fix negative dB output 2009-04-27 03:22:15 -07:00
uwb dma-mapping: replace all DMA_32BIT_MASK macro with DMA_BIT_MASK(32) 2009-04-07 08:31:11 -07:00
video fbdev: i.MX31: fix panning, error handling, clean up 2009-04-07 08:31:09 -07:00
virtio
w1 mfd: remove DS1WM clock handling 2009-04-05 00:32:22 +02:00
watchdog [WATCHDOG] omap_wdt.c: move probe function to .devinit.text 2009-04-08 06:21:57 +00:00
xen Merge git://git.kernel.org/pub/scm/linux/kernel/git/rusty/linux-2.6-cpumask 2009-04-05 10:33:07 -07:00
zorro
Kconfig
Makefile