In the process of converting the assembly from AT&T syntax to Intel
syntax, bjorn3 made some mistakes. Firstly, `A+B(%reg)` is translated to
`[reg + A + B]`, and not `[reg * B + A]`. I can see how this was
confused because x86 also has an addressing mode that takes up to 4
arguments, where `A(%r1,%r2,B)` is translated to `[r1 + A + r2 * B]`
with the constraints that `B` must be a power of two <= 8, and `A` (if
present), must be an immediate. In this case, the `72+8` and such is
constfolded by the assembler into `80`.
This commit and the previous AT&T syntax version produce the exact same
assembly output now.
Connor-GH
aecec88fb6