Running time: operands in memory versus registers

Hi, I'm trying to compare the running time for operands in memory versus operands in registers. The specs are below:

- number of instructions: n
- 2-addresses per instruction
- 10 times longer to access memory versus register
- each instruction has at least 1 memory reference due to fetch operation in fetch/execute cycle

This is my what I did---please correct me if I'm wrong:

Instruction with operands in registers:
10 units of measurement for memory reference in fetch operation + 1 unit for register 1 + 1 unit for register 2 = 12 units

Instruction with operans in memory:
10 units (fetch operation) + 10 units for memory address 1 + 10 units for memory address 2 = 30 units

30/12 = 2.5

2.5 times longer for operands in memory versus in registers

Thanks!

In your spec, you have specified 2 addresses per instruction.

When a fetch operation for the instruction happens, the entire instruction including the operands and opcode is got into the instruction register and then decoded and executed..

so what do u mean when you say the operands can be in memory/registers?