netdev - Re: [RFC] myri10ge: small rx

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <4D8A0FE1.1040500@myri.com>
Date:	Wed, 23 Mar 2011 11:21:05 -0400
From:	Andrew Gallatin <gallatin@...i.com>
To:	Stanislaw Gruszka <sgruszka@...hat.com>
CC:	netdev@...r.kernel.org, Brice Goglin <brice@...i.com>
Subject: Re: [RFC] myri10ge: small rx_done refactoring

On 03/23/11 08:52, Stanislaw Gruszka wrote:

> on my i686 system, what should also make myri10ge_clean_rx_done()
> being faster.

I tested this on my very old, very weak dual-core athlon64 systems.
These machines can barely achieve 10Gb/s using a 1500b MTU with LRO.
Running 35 60 second netperf tests into the machines with the stock
driver, and again with this patch applied, I see a tiny bandwidth
increase (1.4Mb/s on average) which is statistically significant
( p <  0.001). There is no statistically significant CPU load
reduction.

> +	if (len<= mgp->small_bytes) {
> +		rx =&ss->rx_small;
> +		bytes = mgp->small_bytes;
> +	} else {
> +		rx =&ss->rx_big,

Small nit:  the "," above should be a ";"

Between the small bandwidth increase, and the code size reduction,
I'm very appreciative of this patch.

Thank you,

Drew
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html