问题
In MIPS the R-format is encoded like this
opcode (6) | rs (5) | rt (5) | rd (5) | shamt (5) | funct (6)
The shamt field is used only in shift or rotate instructions. In other cases it's always 0. So why couldn't they use it for other more useful things?
IMHO it'll be much better to use that field to store the register number that stores the carry/borrow in add or sub instructions. For example
addu $Rdest, $Rcarry, $Rsrc1, $Rsrc2
$Rcarry will only be set to 0 or 1 depending on the carry bit. That'll make it far more useful for multi-precision arithmetic. Architectures with a flag register can do a double-word addition in 2 instructions whereas MIPS needs 3 because the carry needs to be calculated in a separate instruction. Multiplications will also benefit from that because both the high and low of the output can be specified in the instruction itself instead of wasting several opcodes for the move from/to HI/LO registers
Similarly, the rs field is fixed at 0 for shift instructions. Why couldn't they use it for things like support for multi-word shifting like x86's shrd/shld? Or simply use those 5 bits in rs/shamt as a sub-opcode field. That'd vastly increase the opcode space. With that bigger space probably MIPSr6 wouldn't need a breaking change by removing some instructions to leave space for future expansion
来源：https://stackoverflow.com/questions/55597730/why-didnt-mips-designers-use-the-5-remaining-bits-in-r-format-instructions-for