On Tue, 2008-12-09 at 20:50 +0100, Henning Rogge wrote:
> On Monday 08 December 2008 20:43:35 you wrote:
> > And on your second patch:
> > > +static u16 nl80211_calculate_bitrate(struct rate_info *rate)
> > > +{
> > > + int modulation, streams, bitrate;
> > > +
> > > + if (!(rate->flags & RATE_INFO_FLAGS_MCS))
> > > + return rate->legacy;
> > > +
> > > + modulation = rate->mcs & 7;
> > > + streams = rate->mcs >> 3;
>
> > I don't think this gives correct results for MCS rates 32 through 76,
> > does it? It should at least not calculate anything then. If we did it in
> > userspace then programs could also actually calculate the correct rate
> > as a float, rather than this approximation. Not that it'll matter, I
> > guess. Doesn't anybody know how to do the actual calculation?
> It had a small type (corrected in the new patches), but I think the values are
> good. I have attached a test programm to this mail that outputs the bandwith
> values calculated by the function.
No, that's the thing, for 32 through 76 the calculated values are
totally wrong, that isn't a logical extension of the scheme, MCS 73 for
20 MHz for example is:
four streams, using 64-qam, 64-qam, 16-qam, qpsk with a coding rate of
3/4 for a net bitrate of 195.0 MBits
johannes
Attachment:
signature.asc
Description: This is a digitally signed message part
