Hi
Dne Mon, 25 Jan 2010 19:25:26 +0200
Nanakos Chrysostomos <debian_mpg@wired-net.gr> napsal(a):
On 25/1/2010 16:10, Michal Čihař wrote:
Your patch:
++ //tempsample = mad_f_mul(*left_ch++, options.volume);
++ tempsample = (mad_fixed_t) ((*left_ch++ *
(double)options.volume)/MAD_F_ONE);
Seems obviously wrong, because one of key features of mpg321 is
that it
uses just fixed-point calculation, while you introduce floating
point
arithmetic by this patch.
Can you please tell me where mpg321 introduces floating point
arithmetic, especially with this patch??
Converting to double, doing calculation in double and converting back
to integer will do that.
This patch sends mpg321 to the
previous state. Please take a look at 0.2.10.3 , 0.2.10.6. It has
nothing to do with the decoding of the mp3 file. For the audio
dithering
responsible are the functions from libmad which uses fixed-point
calculation as you can see, audio dithering function does also. This
patch will avoid compilation problems for MIPS and MIPSEL
architectures
[0],[1] introduced by the use of mad_f_mul function. Please take a
look
to /usr/include/mad.h. There is no difference by the introduced
patch.
Package upload again to mentors.d.n.
If the bug is in mad, then please fix it there. Or at least file a bug
report with appropriate severity.
--
Michal Čihař | http://cihar.com | http://blog.cihar.com