Compiled code statistics

[Fiwo735]

Low priority, but at some point in the future we should add a way to measure compiled code (by students' compiler) statistics, like binary size / # of machine instructions, simulated cycles, actual compilation time. These would allow for a better reasoning about any of the implemented "optimisations extensions", once we put more focus on extensions in our grading.

That could be implemented behind a CLI flag (--analyse_binaries?)

[LH-and-FPGA]

I remember that Spike can obtain the number of cycles a program uses during simulation. I think this should be a relatively good method, because sometimes having more lines of assembly code does not necessarily mean worse performance. If you want, I can create a testing script for that?

[Fiwo735]

If your group has already created a way of measuring some statistics (e.g. # of cycles) and would like to share that, feel free to open a PR. Otherwise, don't worry about that, it's not something that will see any use this year, so this can be addressed by you or someone else in the future.

[Fiwo735]

#67 introduced benchmarking mode with --benchmarking for benchmark programs in benchmark/. What's left is to document this mode of operation and provide more benchmark programs.