So instead of having common optimized kernels for AMD, Intel, all ARM firms, all FPGAs and all ASICS. Each member is reinventing the wheel in it's own backend?

So instead of having common optimized kernels for AMD, Intel, all ARM firms, all FPGAs and all ASICS. Each member is reinventing the wheel in it's own backend? Not so surprising ^^