half_float
half_float copied to clipboard
acgessler
C++ implementation of a 16 bit floating-point type mimicking most of the IEEE 754 behaviour. Compatible with the half data type used as texture format by OpenGl/Direct3D.
Although your project has a copyright header and disclaimer on README file, it is missing the license file. Could you please add it?
Hello, are you aware of the other half implementation for x86_64? https://github.com/melowntech/half If so, could you compare them, please? I'll be experimenting with using `float16_t` for a neural net library...
Hey couple of questions, 1. Is this faster than float arithmetic? Have you run some benchmarks? 2. How fast or slow would this be compared to just casting to float...
24 bit precision floating point is also not supported natively by the CPU. The code may be generalized using templates to create a custom precision floating point. e.g. ``` template...
