a+b + C, C is an unsigned 32-bit constant.

Total available ALUs Modern CPUs have 4 ALUs, but we use only 3 because random math executes together with other main loop code.

Available ALUs for MUL Modern CPUs typically have only 1 ALU which can do multiplications.

a*b.

Never generate more than 70 instructions (final RET instruction doesn't count here).

Always generate at least 60 instructions.

finish execution.

rotate left "a" by "b & 31" bits.

rotate right "a" by "b & 31" bits.

a-b.

Generate code with minimal theoretical latency = 45 cycles, which is equivalent to 15 multiplications.

a^b.