i found this link from some forum, so i thought the source code must have been useful to some: font2openvg.cpp.txt
The question is, why divide 4096 to convert a fixed to float?
float convFTFixed( const FT_Pos &x )
{
return (float)x / 4096.0f;
}
Asked
Active
Viewed 71 times
1 Answers
0
The type FT_Pos is an integer type intended to be $2^{12}$ times the real value being represented (thus the precision is a bit better than three decimal places, and the absolute precision is invariant as the number grows).
With today's hardware, this is really not a useful thing to do, but in the past, when memory might have been tight or double precision operations hugely expensive, it is conceivable that representing potentially large numbers in this way would have let you avoid using doubles and yet still get more than 23 bits of precision.
Mark Fischler
- 42,297
-
you meant 2^12, so if i have 16.6, should i do x/64? – noypi Apr 20 '17 at 00:39
-
The representation of 16.6 (decimal) is 0x00010948 in that fixed point format. Necessarily, this is not exact. – Mark Fischler Apr 20 '17 at 22:42
-
sorry, i meant 26:6, 6 bits for fraction. – noypi Apr 21 '17 at 05:40
-
it's ok now, i understand how this works, thanks for your help. – noypi Apr 21 '17 at 05:40