> On Thu, 2014-07-24 at 09:49 +0200, Peter Meerwald wrote:
> > for example, the conversion function for
> > convert_from_float32ne(PA_SAMPLE_S16LE) can also be used for
> > convert_to_s16ne(PA_SAMPLE_FLOAT32LE)
> >
> > Signed-off-by: Peter Meerwald <pmeerw at pmeerw.net>
> > ---
> > src/pulsecore/sconv_neon.c | 2 ++
> > src/pulsecore/sconv_sse.c | 3 ++-
> > 2 files changed, 4 insertions(+), 1 deletion(-)
> >
> > diff --git a/src/pulsecore/sconv_neon.c b/src/pulsecore/sconv_neon.c
> > index 6c2a2b3..8e6a23c 100644
> > --- a/src/pulsecore/sconv_neon.c
> > +++ b/src/pulsecore/sconv_neon.c
> > @@ -89,4 +89,6 @@ void pa_convert_func_init_neon(pa_cpu_arm_flag_t flags) {
> > pa_log_info("Initialising ARM NEON optimized conversions.");
> > pa_set_convert_from_float32ne_function(PA_SAMPLE_S16LE, (pa_convert_func_t) pa_sconv_s16le_from_f32ne_neon);
> > pa_set_convert_to_float32ne_function(PA_SAMPLE_S16LE, (pa_convert_func_t) pa_sconv_s16le_to_f32ne_neon);
> > + pa_set_convert_from_s16ne_function(PA_SAMPLE_FLOAT32LE, (pa_convert_func_t) pa_sconv_s16le_to_f32ne_neon);
> > + pa_set_convert_to_s16ne_function(PA_SAMPLE_FLOAT32LE, (pa_convert_func_t) pa_sconv_s16le_from_f32ne_neon);
>
> Can the NE/LE mismatch cause problems? I suppose it's fine on x86,
> because NE means always LE, but isn't it possible that NE on ARM can
> mean BE, in which case the optimized functions can't be used?
ARM in principal is bi-endian, I don't think BE no ARM is used much (at
least I don't have a test system :)
so I suggest to check for endianness and apply the optimizations for LE
only
regards, p.
--
Peter Meerwald
+43-664-2444418 (mobile)
