Turns out the C version of the cubic resampler is just slightly faster than
even the SSE3 version of the FIR4 resampler. This is likely due to not using a
64KB random-access lookup table along with unaligned loads, both offseting the
gains from SSE.
The context state properties are less likely to change compared to the listener
state, and future changes may prefer more infrequent updates to the context
state.
Note that this puts the MetersPerUnit in as a context state, even though it's
handled through the listener functions. Considering the infrequency that it's
updated at (generally set just once for the context's lifetime), it makes more
sense to put it there than with the more frequently updated listener
properties. The aforementioned future changes would also prefer MetersPerUnit
to not be updated unnecessarily.
This improves the transition width, allowing more of the higher frequencies
remain audible. It would be preferrable to have an upper limit of 32 points
instead of 48, to reduce the overall table size and the CPU cost for down-
sampling.
Rather than storing individual pointers to filter, scale delta, phase delta,
and scale phase delta entries, per phase index, the new table layout makes it
trivial to access the per-phase filter and delta entries given the base offset
and coefficient count.
This is just for the output limiter right now, but in the future can be used
for the compressor EFX effect. The parameters are also hardcoded, but can be
made configurable after 1.18.
The previous value couldn't actually be expressed as a float and got rounded up
to the next whole number value, leaving the potential for an overrun in the
squared sum.
This is a bit more efficient than calling the normal HRTF mixing function
twice, and helps solve the problem of the values generated from convolution not
being consistent with the new HRIR.
This is intended to do conversions for interleaved samples, and supports
changing from one DevFmtType to another as well as resampling. It does not
handle remixing channels.
The mixer is more optimized to use the resampling functions directly. However,
this should prove useful for recording with certain backends that won't do the
conversion themselves.
This improves fading between HRIRs as sources pan around. In particular, it
improves the issue with individual coefficients having various rounding errors
in the stepping values, as well as issues with interpolating delay values.
It does this by doing two mixing passes for each source. First using the last
coefficients that fade to silence, and then again using the new coefficients
that fade from silence. When added together, it creates a linear fade from one
to the other. Additionally, the gain is applied separately so the individual
coefficients don't step with rounding errors. Although this does increase CPU
cost since it's doing two mixes per source, each mix is a bit cheaper now since
the stepping is simplified to a single gain value, and the overall quality is
improved.
NFC filters currently only work when rendering to ambisonic buffers, which
includes HQ rendering and ambisonic output. There are two new config options:
'decoder/nfc' (default on) enables or disables use of NFC filters globally, and
'decoder/nfc-ref-delay' (default 0) specifies the reference delay parameter for
NFC-HOA rendering with ambisonic output (a value of 0 disables NFC).
Currently, NFC filters rely on having an appropriate value set for
AL_METERS_PER_UNIT to get the correct scaling. HQ rendering uses the averaged
speaker distances as a control/reference, and currently doesn't correct for
individual speaker distances (if the speakers are all equidistant, this is
fine, otherwise per-speaker correction should be done as well).
