For certain types of filters where the intermediate sum of coefficients
can go above the fixed-point equivalent of 1.0 in the middle of a filter,
the sum of a 31-bit calculation can overflow in both directions and can
thus not be represented in a 32-bit signed or unsigned integer. To work
around this, we subtract 0x40000000 from a signed integer base, so that
we're halfway signed/unsigned, which makes it fit even if it overflows.
After the filter finishes, we add the scaled bias back after a shift.
We use the same trick for 16-bit bpc YUV output routines.
Signed-off-by: Mans Rullgard <mans@mansr.com>
for (i = 0; i < (dstW >> 1); i++) {
int j;
- int Y1 = 0;
- int Y2 = 0;
+ int Y1 = -0x40000000;
+ int Y2 = -0x40000000;
int U = -128 << 23; // 19
int V = -128 << 23;
int R, G, B;
// 8bit: 12+15=27; 16-bit: 12+19=31
Y1 >>= 14; // 10
+ Y1 += 0x10000;
Y2 >>= 14;
+ Y2 += 0x10000;
U >>= 14;
V >>= 14;