[pulseaudio-discuss] [PATCHv3 0/3] resampler: refactor calc_map_table()

[Stefan Huber]

Hi Tanu,

thanks for the fast replies!


> I'd be happy with either of these two alternatives: only update
> calc_map_table() without a separate patch, or update resampler.c in
> its entirety with a separate patch.

The PATCHv3 1/3 contains the renaming of pa_bool_t to pa_bool and
TRUE/true, FALSE/false, respectively.

PATCHv3 2/3 is basically the unaltered PATCHv2 1/2, but rebased onto
PATCHv3 1/3.

PATCHv3 3/3 adds your comments:

> > + result of 1..6) factors should be multiplied by 0.7 (in the case
> > + of S:Center) and 0.8 (in the case of S:LFE). If C-front is only
> 
> On that last quoted line, the "If" doesn't seem to belong there.

Fixed.

> > +  if (!ic_connected[ic] && on_left(r->i_cm.map[ic])) {
> >        m->map_table_f[oc][ic] = .1f / (float) ic_unconnected_left;
> > +  }
> 
> No braces for single-line ifs, please.

Fixed for multiple occasions in calc_map_table().


PATCHv3 3B/3 is an alternative patch to PATCHv3 3/3. Instead of
carefully trying to maintain the normalization-property of each row for
each step, this patch simply normalizes the row in a post-processing
step. As we do not require to perform normalizations after each step,
we can add some further code simplifications.

After some time of thinking I tend to prefer PATCH 3B over PATCH 3 as it
is a bit simpler and cleaner. What do you think?


Best,
Stefan


Stefan Huber (3):
  resampler: replace pa_bool_t by bool
  resampler: refactor calc_map_table()
  resampler: generate normalized rows in calc_map_table()

 src/pulsecore/resampler.c |  459 +++++++++++++++++++++------------------------
 1 file changed, 218 insertions(+), 241 deletions(-)

-- 
1.7.9.5