Berkeley SoftFloat-3e is an IEEE conformant software floating-point library. It is written in C89, and supports 16, 32, 64, 80, and 128bit floating-point formats.

It has been choosen to implement 64bit `long double` on the Ti84CE, using the following functions:
Arithmetic:
* f64_add
* f64_sub
* f64_mul
* f64_div
* f64_rem
* f64_mulAdd (Fused-multiply add)
* f64_sqrt
Conversion:
* f64_to_f32
* f64_to_i32
* f64_to_i64
* i64_to_f64
* ui64_to_f64
* f64_roundToInt

The SoftFloat library is configured to use the `8086-SSE` rules for NaN propagation, but this can customized, and possibly disabled to reduce binary size.

Every unused source file from `SoftFloat-3e/source` has been excluded to speed up compile times.

The following files are used for implementing `float fmaf(float, float, float)`, but are not needed for `long double`:
* SoftFloat-3e/source/f32_mulAdd.c
* SoftFloat-3e/source/s_mulAddF32.c
* SoftFloat-3e/source/8086-SSE/s_propagateNaNF32UI.c

The following files may or may not be needed at -O0, or when `SOFTFLOAT_BUILTIN_CLZ` is not defined in `platform.h`:
* SoftFloat-3e/source/s_countLeadingZeros8.c
* SoftFloat-3e/source/s_countLeadingZeros16.c
* SoftFloat-3e/source/s_countLeadingZeros32.c
* SoftFloat-3e/source/s_countLeadingZeros64.c
