Commit 1a860388 authored by Rocky Bernstein's avatar Rocky Bernstein

Doc calculation typo.

parent 5dab5a0d
......@@ -2,7 +2,7 @@
* Common pixel/chroma manipulation routines.
*****************************************************************************
* Copyright (C) 2003, 2004 VideoLAN
* $Id: pixmap.c,v 1.1 2004/01/30 13:17:12 rocky Exp $
* $Id: pixmap.c,v 1.2 2004/01/30 13:23:08 rocky Exp $
*
* Author: Rocky Bernstein
*
......@@ -149,14 +149,16 @@ find_cmap_rgb8_nearest(const vout_thread_t *p_vout, const ogt_yuvt_t *p_yuv,
/* Interval range calculations to show that we don't overflow the
word sizes below. pixels component values start out 8
bits. When we subtract two components we get 9 bits, then
square to 10 bits. Next we scale by 6 we get 16 bits. XXX_COEF
all fit into 5 bits, so when we multiply we should have 21 bits
maximum. So computations can be done using 32-bit
precision. However before storing back distance components we
scale back down by 12 bits making the precision 9 bits.
The squared distance is the sum of three of the 9-bit numbers
described above. This then uses 21-bits and also fits in a
square to 10 bits. Next we scale by 6 to give 16
bits. XXX_COEF all fit into 5 bits, so when we multiply we
should have 21 bits maximum. So computations can be done using
32-bit precision. However before storing back distance
components we scale back down by 12 bits making the precision 9
bits. (This checks out since it is basically the range of the
square of the initial 8-bit value.)
The squared distance is the sum of three of the 9-bit components
described above. This then uses 27-bits and also fits in a
32-bit word.
*/
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment