Float64 Support For Intel's Vulkan Driver Is Almost Here
While it took a long time for Intel's Mesa driver to begin supporting the ARB_gpu_shader_fp64 extension for double-precision floating-point data types in shaders, fortunately it looks like Intel should soon land the Float64 support in their Vulkan driver soon.
The latest patches from Igalia this morning are their revised set of 25 patches for supporting the Float64 capability in the Intel Vulkan driver. The shaderFloat64 capability signifies 64-bit floats (doubles) support within the shader code.
Adding Float64 support to the Intel Vulkan driver required changes to NIR as well as their SPIR-V code. The Float64 support for the Intel ANV Vulkan driver is being enabled currently for Intel's Broadwell "Gen8" hardware and newer, with Mesa Git not yet supporting ARB_gpu_shader_fp64 yet for Haswell.
These new Float64 patches for the Intel Vulkan driver can be found via this patch series.
The latest patches from Igalia this morning are their revised set of 25 patches for supporting the Float64 capability in the Intel Vulkan driver. The shaderFloat64 capability signifies 64-bit floats (doubles) support within the shader code.
Adding Float64 support to the Intel Vulkan driver required changes to NIR as well as their SPIR-V code. The Float64 support for the Intel ANV Vulkan driver is being enabled currently for Intel's Broadwell "Gen8" hardware and newer, with Mesa Git not yet supporting ARB_gpu_shader_fp64 yet for Haswell.
These new Float64 patches for the Intel Vulkan driver can be found via this patch series.
5 Comments