Byte ordering during machine code generation

Is there a way to specify the byte ordering in a way that’s neither big nor little endian during machine code generation?
For example, if B3B2B1B0 is a 32-bit instruction with B3 the most significant byte, is there a way to write it as B1 B0 B3 B2 ?

The ARM backend does something similar to what you’re describing for Thumb code; see ARMMCCodeEmitter::encodeInstruction.