Date: Mon, 18 Feb 2013 13:28:39 -0800 From: Adrian Chadd <adrian@freebsd.org> To: Mukunda Haveri <mukunda@pointred.co> Cc: freebsd-wireless@freebsd.org, freebsd-mips@freebsd.org Subject: Re: WiFi TDMA AR7161, results and moving forward Message-ID: <CAJ-Vmo=eRboDNE-5JhQWq8zvsPnGt%2B7-KjZGBcw8RdQmK2q5jw@mail.gmail.com> In-Reply-To: <CAK3PU_C9AW3xO1mofwT1iH_bgrBDYk3zJaLMJCbiRN-jakm%2BFA@mail.gmail.com> References: <CAK3PU_C9AW3xO1mofwT1iH_bgrBDYk3zJaLMJCbiRN-jakm%2BFA@mail.gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On 18 February 2013 03:31, Mukunda Haveri <mukunda@pointred.co> wrote: > > Thanks to Adrian's WiFi scripts, we were able to get the TDMA working on the > Compex-AR7161 board. The results were surprising; we are able to do, close > to 100 mbps one way iperf tests and 40 mbps bidirectional Iperf in non-TDMA > mode. We were able to achieve this, only after disabling all the debug > options in the kernel. Porting the U-Boot to the Compex-boards did take lot > of effort, but not the "FreeBsd". Many thanks to all the "scientists" who > made this possible. Nice! Which wireless cards are you using? > Moving forward, it is observed that the TDMA throughput peaks at 9 mbps and > refuses to move beyond. After reducing the slot duration to 1 ms, the > throughput increased to around 12 mbps. I was expecting the TDMA to yield a > better throughput because of collision-less scheme. I would like to > understand if our observation is expected or if there's some inherent > limitation within the TDMA controller. > > It will be good to have some feedback from TDMers with similar experience or > better. Right now the TDMA code doesn't implement MCS rates or TX aggregation. Thus you're not going to get 11n like throughput. The first thing to implement is allowing for MCS rates (non-aggregation) and make sure all the packet duration calculations are being done "right". After that, we need to implement delayed blockack support in net80211 and the ath driver. That requires the stack to support handling BA requests/responses and ath(4) to mark all frame descriptors in a delayed-BA TID to be no-ack. Once that's done, we can tie it all together to make it work over TDMA. :-) Thanks, Adrian
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CAJ-Vmo=eRboDNE-5JhQWq8zvsPnGt%2B7-KjZGBcw8RdQmK2q5jw>