Date: Tue, 16 Apr 2013 21:44:06 +1000 (EST) From: Bruce Evans <brde@optusnet.com.au> To: Sepherosa Ziehau <sepherosa@gmail.com> Cc: YongHyeon PYUN <pyunyh@gmail.com>, Sean Bruno <sean_bruno@yahoo.com>, bde <bde@FreeBSD.org>, Bruce Evans <brde@optusnet.com.au>, "freebsd-net@freebsd.org" <freebsd-net@FreeBSD.org> Subject: Re: bge(4) sysctl tuneables -- a blast from the past. Message-ID: <20130416213235.O1783@besplex.bde.org> In-Reply-To: <CAMOc5cy%2BG4QaV5dcPCt%2ByJE%2BmbKWApQEJ%2BAxKaZHeUukYUmrHw@mail.gmail.com> References: <1365781568.1418.1.camel@localhost> <20130413200512.G1165@besplex.bde.org> <1366065356.1350.7.camel@localhost> <20130416052500.GA1428@michelle.cdnetworks.com> <20130416162150.X1106@besplex.bde.org> <CAMOc5cy%2BG4QaV5dcPCt%2ByJE%2BmbKWApQEJ%2BAxKaZHeUukYUmrHw@mail.gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Tue, 16 Apr 2013, Sepherosa Ziehau wrote: > On Tue, Apr 16, 2013 at 3:14 PM, Bruce Evans <brde@optusnet.com.au> wrote: > >> bge_rx_coal_ticks == 0 && bge_rx_max_coal_bds == 0 might work accidentally >> if there are enough tx interrupts. There is also the DEVICE_POLLING >> mistake. >> In polling mode, these parameters of course have no effect (Polling mode >> disables interrupts, and the coal parameters have no effect when interrupts >> are disabled). > > As far as I have tested, coalesce BDs and ticks also control how often > status block is updated, so it does affect polling(4) This might explain why polling worked even worse than expected for most things. However, I got reduced latency using it (from ~55 usec ping latency to ~30 usec). 30 usec would be impossible with the default parameters (normal is an average of at least half of bge_rx_coal_ticks). Maybe I had the parameters reduced to bge_rx_coal_bds = 1 when I found 30 usec (then bge_rx_coal_ticks doesn't matter). This is probably what polling should use. Bruce
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20130416213235.O1783>