If one is able to create custom sections in an ELF file then is it feasible/possible to dump machine code of different CPU arch into the same ELF file in different section and at runtime map the correct .text section based on the current machine ? Would this not allow for universal binary distribution based on the ELF format ? I get the downside of this is that the executable size would be huge, but from a concept point/experimentation point of view is the idea possible ?

如果能够在ELF文件中创建自定义节，那么是否可以/可能将不同CPU Arch的机器代码转储到不同节中的同一ELF文件中，并在运行时基于当前计算机映射正确的.Text节？这会不会允许基于ELF格式的通用二进制分发？我知道这样做的缺点是可执行文件的大小将是巨大的，但从概念的角度/实验的角度来看，这个想法可能吗？