Hello Maarten Brock,
Request to Add support for half precision float (16 bits ) for 8051 in SDCC. Since for speed and code size optimization it is very important.
Thank you
Please do not assign tickets to developers. That you can is considered a flaw in the sourceforge system by us. It is up to the developers themselves to pick up tickets.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
The MCS-51 floating point library that ships with SDCC can be compiled to use a reduced precision mode. This mode uses 32 bit floats, but shrinks the mantissa for internal calculations, thereby speeding them up.
If this is not sufficient, please specify which type of half precision floating point you want.
Is it IEEE 754 binary16, is it bfloat16 or is it something else, entirely?
Should anybody ever feel like adding this kind of functionality, this would obviously matter.
👍
1
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Please do not assign tickets to developers. That you can is considered a flaw in the sourceforge system by us. It is up to the developers themselves to pick up tickets.
Noted
The MCS-51 floating point library that ships with SDCC can be compiled to use a reduced precision mode. This mode uses 32 bit floats, but shrinks the mantissa for internal calculations, thereby speeding them up.
If this is not sufficient, please specify which type of half precision floating point you want.
Is it IEEE 754 binary16, is it bfloat16 or is it something else, entirely?
Should anybody ever feel like adding this kind of functionality, this would obviously matter.
ideally type should be parametrizable (standard vs fast)
IEEE 754 binary16 I believe it the best
https://github.com/artyom-beilis/float16