Add basic BF16 routines (#45) * Improve bf16_to_f32 performance Signed-off-by: Andrew Waterman <andrew@sifive.com> * Add basic BF16 routines These are implemented so as to make them more-or-less obviously correct (with the exception of the FMA, which is a bit more subtle, because of double rounding). As such, they are implemented by widening into formats with existing native support, then rounding back to BF16. The hope is that optimized, functionally equivalent variants will eventually supplant these. Signed-off-by: Andrew Waterman <andrew@sifive.com> --------- Signed-off-by: Andrew Waterman <andrew@sifive.com>
John R. Hauser
2018 January 20
Berkeley SoftFloat is a software implementation of binary floating-point that conforms to the IEEE Standard for Floating-Point Arithmetic. SoftFloat is distributed in the form of C source code. Building the SoftFloat sources generates a library file (typically softfloat.a or libsoftfloat.a) containing the floating-point subroutines.
The SoftFloat package is documented in the following files in the doc subdirectory:
Other files in the package comprise the source code for SoftFloat.