jamal writes:
> I will test with a newer piece of hardware and one of the older ones i
> have (both Xeons) - perhaps this weekend.
> Robert may have some results perhaps on this driver, Robert?
> It would also be nice for the intel folks to post their full results
> somewhere.
I agree with you Jamal. Prefetch addresses the latency problem we suffer
from but to me it seems risky too. Benchmarks are very HW and SW dependent.
I think Intel had an interesting application note for P4/XEON describing
pitfalls etc. How much prefetching can we add? When does it get wrong or
to aggressive? Who'll be hurt? How can it be tuned?
> On the copybreak change
> -----------------------
>
> Lets defer this since i already said a lot on prefetch ;->
> But someone like Robert with results could shed more light.
Anyway below is short test I've run comparing the vanilla e1000
2.6.14 driver with e1000 6.2.15 (having copybreak and prefetch)
Dual Opteron 1.6 GHz with two dual e1000 82546GB both CPU:s used.
Buses on 100 MHz resp 133 MHz. eth0->eth1 is on CPU0 and eth2->eth3
on CPU1. Table show aggregated performance with 64 byte pkts.
In this combination of hardware and in this forwarding test
copybreak is bad but prefetching helps.
e1000 vanilla 1150 kpps
e1000 6.2.15 1084
e1000 6.2.15 copybreak disabled 1216
e1000 6.2.15 copybreak disabled and no rx prefetch 1097
Jesse showed me some numbers with s dramatic improvment in his HW/SW setup.
Cheers
--ro
Input 756 + 711 kpps. Vanilla.
Kernel Interface table
Iface MTU Met RX-OK RX-ERR RX-DRP RX-OVR TX-OK TX-ERR TX-DRP TX-OVR Flags
eth0 1500 0 7530922 2469078 2469078 2469078 5 0 0 0 BRU
eth1 1500 0 1 0 0 0 7530501 0 0 0 BRU
eth2 1500 0 8186038 1813962 1813962 1813962 3 0 0 0 BRU
eth3 1500 0 1 0 0 0 8185553 0 0 0 BRU
0072e9ab 00000000 0000620f 00000000 00000000 00000000 00000000 00000000 00000000
007ce8b7 00000000 00006a96 00000000 00000000 00000000 00000000 00000000 00000000
e1000 6.2.15 730+690 input
Kernel Interface table
Iface MTU Met RX-OK RX-ERR RX-DRP RX-OVR TX-OK TX-ERR TX-DRP TX-OVR Flags
eth0 1500 0 7738174 0 2262190 0 245 0 0 0 BRU
eth1 1500 0 1 0 0 0 7737389 0 0 0 BRU
eth2 1500 0 7546265 0 2453735 0 3 0 0 0 BRU
eth3 1500 0 1 0 0 0 7545846 0 0 0 BRU
007612f3 00000000 000064c0 00000000 00000000 00000000 00000000 00000000 00000000
00732622 00000000 00006242 00000000 00000000 00000000 00000000 00000000 00000000
e1000 6.2.15 782+747 input copybreak disabled
Kernel Interface table
Iface MTU Met RX-OK RX-ERR RX-DRP RX-OVR TX-OK TX-ERR TX-DRP TX-OVR Flags
eth0 1500 0 7530164 0 2470117 0 166 0 0 0 BRU
eth1 1500 0 1 0 0 0 7528073 0 0 0 BRU
eth2 1500 0 8416010 0 1583990 0 3 0 0 0 BRU
eth3 1500 0 1 0 0 0 8414874 0 0 0 BRU
0072e6c2 00000000 0000620a 00000000 00000000 00000000 00000000 00000000 00000000
00806b57 00000000 00006d94 00000000 00000000 00000000 00000000 00000000 00000000
e1000 6.2.15 711+681 input copybreak disabled and no rx prefetch
Kernel Interface table
Iface MTU Met RX-OK RX-ERR RX-DRP RX-OVR TX-OK TX-ERR TX-DRP TX-OVR Flags
eth0 1500 0 7465662 0 2534338 0 5 0 0 0 BRU
eth1 1500 0 1 0 0 0 7464878 0 0 0 BRU
eth2 1500 0 8337205 0 1662795 0 3 0 0 0 BRU
eth3 1500 0 1 0 0 0 8336784 0 0 0 BRU
0071eabf 00000000 00006135 00000000 00000000 00000000 00000000 00000000 00000000
007f3736 00000000 00006c8e 00000000 00000000 00000000 00000000 00000000 00000000
-
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
