The objective of this lab is to have some hands-on experimentation on the subject of performance as discussed within chapter 1 of Patterson and Hennessy.
The course webpage for labs can be found at staff.fnwi.uva.nl/e.h.steffens
We will use the SIM-PL logic simulator to simulate 3 different architectures. It should already be installed on your system if not you can Download the SIM-PL program from the course webpage and install it .
Notice the clock used in the simulation is unitless. Actually time is not important in running these simulations.. When a time ivalue s requird in the answer take an acceptable unit (microseconds or nanoseconds) and give the clocktime or frequency used .
Experiments Performance measurement
The assignment is to compare the performance statistics of three different architectures using the performances measures stated in the first chapter of the book.
In the archive there are three different architectures:
- Single Cycle
To measure the performance of the different architectures there are three different programs called:
You task is to measure and/or calculate the following statistics for every program and architecture.
- The number of instructions that have been executed for every architecture.
- the CPI
- the cpu time
Explain what architecture performed the best for every executable and explain why.
A program counter is present in the architecture. Can you explain what the program counter function is?
What do you notice about the behavior when running a simulation?
In the programs there are different instructions. Explain for the following instructions what do they do and how the are used.
Now change the first program to subtract two numbers, do not change the initialization of the variables can you explain what happens.
Finally explain why the clock rate of a cpu is not a good benchmark for the performance of a cpu.
Hand in a report( pdf fprmat) with your results. For details see the course webpage at staff.fnwi.uva.nl/e.h.steffens