The other strict-ness
< has a softer sibling. c.le.s — compare, less-or-equal — tests <=, and it's a different instruction from c.lt.s, not a flag trick: whether the boundary value itself passes is baked into the mnemonic. Here's fits(x), which tests x <= 4.0f:
0: lui at, 0x4080 # 4.0f
4: mtc1 at, ft0
8: or v0, zero, zero
c: c.le.s fa0, ft0 # flag = (x <= 4.0f)
10: nop
14: bc1f 0x20
18: nop
1c: addiu v0, zero, 1
20: jr ra
24: nop
Argument first, constant second, .le instead of .lt — otherwise the boolean skeleton you already own. So the reading rule for the pair is one table:
c.lt.s → strict < (or a mirrored >)c.le.s → <= (or a mirrored >=)
And yes — the mirror applies here too. There's no greater-or-equal compare either, so a >= b compiles as "b <= a": a c.le.s with the operands facing the other way. Strictness comes from the mnemonic, direction from the operand order. Two glances and you've recovered the exact C comparison, boundary behavior included.
That precision matters more with floats than you might expect — hp >= maxHp and hp > maxHp are different game logic on the boundary frame, and the assembly never lets the two blur.
The target is a >= between the two arguments. Mirror, soften, done.
Your task
Write func_8034b4bc to reproduce the target assembly.