16 bits hex range for mips assembly instruction beq and bne

0
votes

I have implemented beq instruction for mips assembly language. As my understanding for beq instruction (beq $s, $t, i), i can take on integer values or hex values. I have establish a bound for 16 bits integer value. I was wondering what is the bound for 16 bits hex values, so when i is too large (or too small?) it would output error before executing it. Following is beq instruction in binary.

Branch On Equal
beq $s, $t, i
0001 00ss ssst tttt iiii iiii iiii iiii

I tried (i > 0xffff) but it seems not cover all the cases. What should i do here? Thanks.

2
assemblyhexmips
what exactly do you mean by "it seems not cover all the cases"? – TomT
It does not pass the tests given by my professor.I'm pretty sure it's related to the range of those 16 bits hex. – Jing
It's unclear to me exactly what you're trying to do. Are you writing a MIPS emulator? Or a MIPS assembler? Either way, the offset is signed, so the range is -32768..+32767, which then will be multiplied by 4 before being added to PC. – Michael
What Michael's comment means, it's +/- 32k words. zero = no offset (= nop), 0xffff = -1 = 1 word backwards (branch to itself), ... – turboscrew

2 Answers

0
votes

It should be in between 0 and 2^16 - 1 for hex(0 <= i <= 65535);
It should be in between -2^15 and 2^15 - 1 for int(-32768<=i<=32767)

0
votes

When you need to conditionally branch farther than +- 2^15 instructions (byte offset = two's complement 16 bits << 2), you can conditionally jump over a non-conditional jump that can reach the target.

beq $s, $t,  target      # short jump, approx +- 2^17 bytes

vs. j uses pseudo-direct addressing, keeping the top 4 bits of PC and replacing the rest with a 26-bit << 2 immediate. This lets you reach any target address in the same aligned 1/16th of address space as your code.

bne $s, $t,  .nojump
j   target               # anywhere with the same top 4 bits
.nojump:

vs. putting the address in a register and using jr to reach any 32-bit address:

bne $s, $t,  .nojump
lui  $t0,      %hi(target)
ori  $t0, $t0, %lo(target)      # this is what  la $t0, target does.  It might use addui, but whatever
jr   $t0
.nojump: