Skip to content

nir/inline_uniforms: allow bcsel/fcsel-based inlining conditionally

if the resulting value from a csel op comes from an intrinsic, we can optimize out the alu to potentially avoid intensive instructions, e.g., a case where a csel chooses either the result of a load_ubo or a constant value that can be optimized down to the constant value through inlining

Merge request reports