After you optimize edit benches/bench_io_opt1.rs and add back the "Medium file" and "Large file" benchmarks. They should pass in reasonable time thanks to your improvements.
An included build script generates a necessary module (src/data_paths.rs) and some test files. If things aren't working right try cargo clean then cargo build.
MappedFile.hpp from the C++ original is not yet included, but you should be able to get huge speedups without it.
The rest of this course deals with low-level CPU and memory issues. You find those with perf and the pmu-tools/ repo (which I believe wraps perf). This lab is different. Here is a suggested approach. This is also an approach that should work for a real world program. The binary we are optimizing was built by cargo bench and lives in target/release/deps/.
First you have to discover whether we are spending most of the time in our code ("%usr") or waiting on a syscall ("%system"), meaning the kernel is doing something on our behalf:
pidstat -e ./bench_io_opt1-<hash> --bench --profile-time 5
This is an I/O optimization exercise, so we expect "%system" to be high. This means there's no point doing low-level CPU cycle counting or worrying about processor caches. Our program is basically idle, waiting for syscalls to return.
Second find out which syscalls we are blocked on. The -D 1 says wait 1ms after startup before recording. This allows us to ignore process launch syscalls for a much cleaner output.
sudo perf trace -D 1 -s ./bench_io_opt1-<hash> --bench --profile-time 5
There will be a single syscall which is causing all our problems. Hopefully it is the one you were expecting.
Finally (and optionally) find out where the syscalls came from. In this case you already know, but it's worth the exercise.
sudo perf trace record --call-graph fp -D 1 -s ./bench_io_opt1-<hash> --bench --profile-time 5
sudo perf report
The solution and bench1 functions were most likely inlined, so the calls to read will be reported from bench_io_opt1::main, which isn't especially useful here. Now make it faster!