libSoftFloat 1.0 Released, Still Working Towards Emulated FP64 Support For GPUs

Last week marked the release of libSoftFloat 1.0, the library working to implement double-precision operations in pure GLSL 1.30 via bit twiddling operations and integer math. This is the most hopeful effort yet for getting OpenGL FP64 support exposed for older GPUs that lack native support.

LibSoftFloat started as a Google Summer of Code 2016 project under the X.Org Foundation umbrella for providing a "soft" FP64 implementation for older GPUs, such as the Radeon HD 5000/6000 series that with R600 Gallium3D don't currently expose OpenGL 4.0 due to lacking FP64 support while the proprietary drivers have long provided such emulated support.

A Phoronix reader pointed out that in the past week libSoftFloat 1.0 was released via the GitHub project site. It's still part of the project's goals to integrate it into Mesa for exposing ARB_gpu_shader_fp64 for graphics processors that otherwise wouldn't have the support, but it's not quite there yet. See that GitHub project site for more details on the current status.