Switch example tools to use VBR and 960 sample frames by default on the basis that if the user doesn't have any particular requirements that they probably want this.

Minor change in the VBR behavior to hot-start with some internal state
parameters which were observed to be quite consistent across bitrates,
framesizes, and content. This also prevents it from completely burning
the reserve capacity on the first frame if its a short.

Also switch some maximum frame sizes to match the OPUS draft maximums.

Also making use of alloc_trim_analysis() again because the effect of
inter-channel correlation on the bitstream is really in terms of absolute
number of bits/samples.

The idea here is that it's better to fold a higher band -- even if it was
coded less accurately -- than a lower band that may have a different
temporal structure.

Also fixed a fixed-point breakage.

Previous limit was effectively 120.

Both ended up causing more harm than good (e.g. violating energy conservation)
and provided little benefit. This also saves ~3 kB code size on x86-86.

John Ridges authored 14 years ago and Jean-Marc Valin committed 14 years ago

I've been compiling CELT for 64-bit Windows and it all seems to work
fine (with the occasional benign warning message) except for one place,
and that is in "float_cast.h" where there is some inline assembly
language that gets complied for Windows. Since the Microsoft 64-bit
compiler won't allow inline assembly language (and *still* doesn't have
"lrintf").

Still need to make it use the right channel in stereo.