You copied the Doc URL to your clipboard.

BFloat16 floating-point number format

Arm® Compiler [ALPHA] supports the floating-point __bf16 type.

Note

This topic includes descriptions of [ALPHA] features. See Support level definitions.

BFloat16 is a floating-point format that occupies 16 bits. It is [ALPHA] supported by Armv8.2 and later Application profile architectures.

Note

The __bf16 type is a storage format only type, and it can only be used by intrinsics. An error is raised if arithmetic operations in C or C++ expressions are performed using the __bf16 type.

BFloat16 floating-point format

Arm Compiler uses the BFloat16 binary floating-point format which is a truncated form of the IEEE 754 standard.

Figure B6-2 BFloat16 floating-point format


Where:

   S (bit[15]):     Sign bit
   E (bits[14:7]):  Biased exponent
   T (bits[6:0]):   Fraction

Note

See the Arm® C Language Extensions for more information.