flac dev - Apr 2015 - "keep qlp coeff precision such that only 32-bit math is required"

Martijn van Beurden wrote:
> Yes, indeed. I removed the 17-bits part because I just matched
> the code in evaluate_lpc_subframe_ with the process_subframe_
> code. It appears it only makes sense that those two pieces code
> are the same.


A bit of history:

1) The conditional "if(subframe_bps <= 16)" was added to
evaluate_lpc_subframe_()
in the commit
http://git.xiph.org/?p=flac.git;a=commitdiff;h=20ac2c1cac298b8c0666e191ae913f6481e79879


2) The conditional "if(subframe_bps <= 16)" was added to
process_subframe_()
in the commit
http://git.xiph.org/?p=flac.git;a=commitdiff;h=c9c0d130c562044e641e2cab09894213c5de4d88


3) The conditional in process_subframe_() was changed to "if(subframe_bps
<= 17)"
in the commit
http://git.xiph.org/?p=flac.git;a=commitdiff;h=df59845d04a3c43e3caea5d9838d08b9b17ba572

So the question is: why the condition in evaluate_lpc_subframe_() wasn't
changed in this commit.
Maybe there were some reasons, but maybe Josh forgot about it.

2015-04-19 22:36 GMT+02:00 lvqcl <lvqcl.mail at gmail.com>:
> So the question is: why the condition in evaluate_lpc_subframe_()
wasn't
> changed in this commit.
> Maybe there were some reasons, but maybe Josh forgot about it.
>
Or maybe the question is: why is this code in evaluate_lpc_subframe anyway,
i.e, why is this code duplicated? If process_subframe_ sets the
qlp_precision for evaluate_lpc_subframe, why should the latter ignore that?

We can only speculate about this, but I think this code has been duplicated
by accident, and therefore it wasn't changed as the code shouldn't have
been there in the first place.

P.S.: sorry for the duplicate lvqcl
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
http://lists.xiph.org/pipermail/flac-dev/attachments/20150420/77220c36/attachment.htm

Martijn van Beurden wrote:
> Or maybe the question is: why is this code in evaluate_lpc_subframe anyway,
> i.e, why is this code duplicated? If process_subframe_ sets the
> qlp_precision for evaluate_lpc_subframe, why should the latter ignore that?
>
> We can only speculate about this, but I think this code has been duplicated
> by accident, and therefore it wasn't changed as the code shouldn't
have
> been there in the first place.
Well, the code in process_subframe_() calculates MAX value of
qlp_coeff_precision,
while the code in evaluate_lpc_subframe_() adjusts CURRENT qlp_coeff_precision.

It's possible to move the logic from evaluate_lpc_subframe_() into
process_subframe_(), see the attached patch, but I'm not sure that the
code becomes much clearer after it.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: tmp1.patch
Type: application/octet-stream
Size: 1518 bytes
Desc: not available
Url :
http://lists.xiph.org/pipermail/flac-dev/attachments/20150420/741a6064/attachment.obj