v.24.11New Feature
New BFloat16 Data Type Introduced for 16-bit Floating Point Numbers
A new data type, BFloat16, represents 16-bit floating point numbers with 8-bit exponent, sign, and 7-bit mantissa. This closes #44206. This closes #49937. #64712 (Alexey Milovidov).Why it matters
This feature provides an efficient way to store and process floating point data with reduced precision and memory footprint compared to standard 32-bit floats, which is valuable for machine learning workloads and other applications where approximate calculations are sufficient.How to use it
Users can declare columns or variables using theBFloat16 data type in their table schemas or queries, e.g., CREATE TABLE example (col BFloat16). Standard floating point operations and functions support this type transparently.