mcjtag / fpx

A Parameterizable Floating Point Operations (Verilog)

A parameterizable floating point operations (written in Verilog)

Implemented operations:

1. Multiplication
2. Addition
3. Conversion (real-to-integer)
4. Conversion (integer-to-real)
5. Absolute Value
6. Comparison
7. Complex Multiplication
8. Complex Addition
9. Complex Conjugation

Parameters

• NUM_W - Number Width
• EXP_W - Exponent Width
• MAN_W - Mantissa Width
• TRIGOUT - Output Trigger (1 - enable, 0 - disable)

IEEE754 settings

• FP16 : NUM_W = 16, EXP_W = 5, MAN_W = 10 (Half-precision)
• FP32 : NUM_W = 32, EXP_W = 8, MAN_W = 23 (Single-precision)
• FP64 : NUM_W = 64, EXP_W = 11, MAN_W = 52 (Double-precision)
• FP128: NUM_W = 128, EXP_W = 15, MAN_W = 112 (Quadruple-precision)
• FP256: NUM_W = 256, EXP_W = 19, MAN_W = 236 (Octuple-precision)

Note: It is able to use any custom values for Exponent and Mantissa Width.