libfixmath.net

• origin : https://code.google.com/archive/p/libfixmath/

fix16

using fix16_t = System.Int32;

public struct fix16
{
  private fix16_t value;
}

• 32-bit numbers, fix16_t, which have 16 bit integer part and 16 bit fractional part.
• max : 32767.999_985
• min : -32768.0
  • The minimum value is also used to represent fix16_overflow for overflow detection, so for some operations it cannot be determined whether it overflowed or the result was the smallest possible value. In practice this does not matter much.
• The smallest unit (machine precision) of the datatype is 1/65536 = 0.000_015

wiki

• https://code.google.com/archive/p/libfixmath/wikis/FunctionReference.wiki

Define Option

• I already removed FIXMATH_NO_64BIT, FIXMATH_OPTIMIZE_8BIT from origin source.

Accuracy

    FIXMATH_NO_ROUNDING: Disable rounding. Results may round randomly up or down, i.e. their accuracy is +-1. Runs slightly faster.
    FIXMATH_NO_OVERFLOW: Disable overflow detection and saturating arithmetic support. Overflowing computations will give garbage results. Runs slightly faster.

Platform

    FIXMATH_NO_CACHE: Do not use cache for exp etc. function results. Uses less RAM, runs slightly slower.

Algorithms

    FIXMATH_SIN_LUT: This uses a look-up table generated by the fixsingen tool from svn, it is faster on some devices but the lookup table takes up ~200KiB (205376 bytes) in memory and can be slower dependant on usage.
    FIXMATH_FAST_SIN: This enables a faster but less accurate approximation of the sin function, for code where accuracy isn't as important such as games logic or graphics rendering then this is often a worthwhile tradeoff.

why.

When I try to test deterministic logic on game, I tried to find float library. and I found some. but I decided to customize from original libfixmath.

• https://github.com/asik/FixedMath.Net

• https://github.com/sungiant/abacus

• I don't need 64bit float.

• float library slower than float. so as possible as I choose more faster logic.