A tricky problem exists where most of the core kernel code should not be using FPU instructions on most arches but some application code may want it.
Idea: Have a MODULE_FLAG variable that can be set per module that opts in for fpu codegen (if the arch supports it). This would allow for particular applications to use the fpu without others.
May be problematic on some arches because of ABI issues however, so might be really difficult to mix fpu and no-fpu code in those cases.
A tricky problem exists where most of the core kernel code should not be using FPU instructions on most arches but some application code may want it.
Idea: Have a MODULE_FLAG variable that can be set per module that opts in for fpu codegen (if the arch supports it). This would allow for particular applications to use the fpu without others.
May be problematic on some arches because of ABI issues however, so might be really difficult to mix fpu and no-fpu code in those cases.