Function "LOD" with "MANT_SIZE > 14"?

dicecco1 / fpga_cpfp

HLS Custom-Precision Floating-Point Library

MIT License

13 stars 4 forks source link

Function "LOD" with "MANT_SIZE > 14"? #1

Open Oxygen-Chu opened 5 years ago

Oxygen-Chu commented 5 years ago

I see the the function "LOD" can only be treated with "MANT_SIZE < 15". So how about Single-Precision with "MANT_SIZE = 23", and even Double-Precision with "MANT_SIZE = 53"?

dicecco1 commented 5 years ago

This library was specifically for low-precision floating-point operations. For single-precision or double-precision it'd be much better to use the built in support

Oxygen-Chu commented 5 years ago

Yes, but CPFP has a specific advantage over built-in float/double that the CPFP is portable, and not dependent upon Xilinx FP Core. Will it be possible for you to have an updating for CPFP to support real arbitary precision floating-point synthesis?

Best Regards : -) Oxygen

dicecco1 commented 5 years ago

I'm not working on this project anymore unfortunately but it shouldn't be too much extra effort to augment for larger mantissa sizes