[MPlayer-dev-eng] [BUG] surround audio plugin broken

[Arpi]

Hi,

> > also it can add a given bias value to the pcm values (while doint the imdct)
> > helping the float2int process (there are integeronly tricks to convert an
> > ieee float in a given range to integer) 
> 
> Do you know which file?

resample_c.c::convert()

but the trick is easier to understand from mp3lib's decod386.c, it' seven
explained there:

#define WRITE_SAMPLE(samples,sum,clip) { \
  union { double dtemp; int itemp[2]; } u; int v; \
  u.dtemp = ((((65536.0 * 65536.0 * 16)+(65536.0 * 0.5))* 65536.0)) + (sum);\
  v = u.itemp[MANTISSA_OFFSET] - 0x80000000; \
  if( v > 32767) { *(samples) = 0x7fff; (clip)++; } \
  else if( v < -32768) { *(samples) = -0x8000; (clip)++; } \
  else { *(samples) = v; } \
}

 * Here's how it works:
 * ((((65536.0 * 65536.0 * 16)+(65536.0 * 0.5))* 65536.0)) is
 * 0x0010000080000000LL in hex.  It computes 0x0010000080000000LL + sum
 * as a double IEEE fp value and extracts the low-order 32-bits from the
 * IEEE fp representation stored in memory.  The 2^56 bit in the constant
 * is intended to force the bits of "sum" into the least significant bits
 * of the double mantissa.  After an integer substraction of 0x80000000
 * we have the original double value "sum" converted to an 32-bit int value.


A'rpi / Astral & ESP-team

--
Developer of MPlayer, the Movie Player for Linux - http://www.MPlayerHQ.hu