Implementation of a library of double precision operations in pure GLSL 1.30 for GPU using bit twiddling operations and integer math.
This library is the translation of SoftFloat by John R. Hauser in GLSL in order to address GPUs instead of CPUs.
-
The first goal is to be able to compute FP64 on GPUs without FP64 hardware support.
-
A streach goal of this project is to integrate it into Mesa and create GL_ARB_gpu_shader_fp64.
For more details, see my presentation at XDC'16, my website or simply email me.
This project began with the Google Summer of Code 2016.
Elie Tournier