TODO:

1. Check for denormal and pseudodenormal operands in ALL instructions

2. Unmasked underflow should correct the result by magic number
   for all operations, including float32 and float64

3. The instructions still implemented using old FPU library:
	FSIN, FCOS, FSINCOS, FPTAN, FPATAN, FYL2XP1, F2XM1, FYL2X

4. Set SW_C1 according to PRECISION_UP or PRECISION_DOWN conditions

5. Mode float_tininess_before_rounding is deprecated, remove from 
   softfloat code