[flac-dev] "keep qlp coeff precision such that only 32-bit math is required"

[lvqcl]

Erik de Castro Lopo wrote:

> There should be some indication of why in the git history.

http://git.xiph.org/?p=flac.git;a=commitdiff;h=27846708fe6271e5e3965a4bbad99baa1ca24c49

Now I remember a discussion about a bug in -p switch: the old code substracts
lpc_order instead of FLAC__bitmath_ilog2(lpc_order), and this commit fixes this.

It seems that the logic in process_subframe_() and in evaluate_lpc_subframe_()
is the same, so the constants in the conditions should be the same: either
both equal to 16, or both equal to 17.



I built two flac executables, the 1st is from current git head, the 2nd - with
my patch applied. Then I took a 16-bit WAV file and compressed it with both
encoders.

Standard compression levels (-0...-8): no difference (the resulting files are bit-identical).
-p option: no difference (the resulting files are bit-identical).
-q N option (N = 13...15): the 2nd encoder produces slightly smaller files.
-l N --lax option (N = 16...31): again, the 2nd encoder produces slightly smaller files.

About decoding speed: the difference isn't very big.
Except for "-l N --lax" option and 32-bit decoder: the files created by the 2nd
encoder have noticeably faster decompression speed.

So it seems that my patch is slightly beneficial for those who use
non-standard encoding presets. Especially when they use non-Subset
encoding options.