mikucionisaau / number_series

Extended Exercise 1

Number Series, Wrapper and Benchmark

In this assignment, you have to implement a data type (class) number_series, to store a number series (just a sequence of integers).

0. Check that the initial project works: select All CTest target and run, all tests should pass. Ask for help if something does not work, see if you can work with main.cpp instead, try running just main.

1. Implement number_series class

   • Uncomment one test at a time in number_series_test.cpp, recompile and observe errors and test failures.
   • Implement missing functionality (solve TODO: comments) in number_series.hpp until the test passes.
   • Use std::vector<int> to model data.
   • Maintain the minimum and the maximum values of a number series as members of your class (so that you do not have to compute them when needed).
   • Implement the necessary constructors/destructors/assignment operators. Implement them only if the ones generated by default are not good. Study Section 17.6 or Item 17 in EMC++ to learn about what is generated by the compiler.
   • Your class should have a static member factory function make_random that returns a random number series of a desired length. See uniform_int_distribution example how to generate random numbers.
   • Implement operator+ and operator+= to add two number series element-wise. Decide yourself what to do if the number series have different lengths.
   • Implement operator< to compare the amplitudes of two number series. The amplitude is the difference between the maximum and the minimum values.

2. Solve TODOs in number_series_bm.cpp benchmark and run number_series_bm target in Release profile:

   • Fill a vector of 100'000 random number series, each with 100 elements.
   • Add a random number series to each of number series in the vector.
   • Sort the vector according to the amplitudes of number series.
   • Record how much time it takes to sort (remember to write down the timings in comments).

3. Create a wrapper class, that has just one private data member: std::unique_ptr<number_series> and add the same interface as for number_series. Think about the copy/move constructors/assignment operators, that you may need to provide or rely on the default ones. On the client side in number_series_bm.cpp, demonstrate that you can now use essentially the same code as in step 2.

   • What do you expect regarding the performance of sorting objects of this wrapper class when compared to sorting the original number_series objects?
   • Measure the performance and see whether your expectations were confirmed.

4. Add int dummy[100]; as an extra data member of number_series.

   • Rerun number_series_bm again.
   • How did the performance change? Why?

In your code, pay attention to safety and performance:

1. Use const whenever it makes sense (arguments, member functions).
2. Use pass-by-reference for efficiency whenever possible.
3. Avoid using plain pointers.
4. Use Release profile (without Debug info overhead) when benchmarking. To add Release profile in CLion visit Settings > Build, Execution and Deployment > CMake > Profiles and click +, it should create Release profile automatically (no further tweaking should be needed).

Unit Tests

number_series_test.cpp includes unit tests to guide you through the implementation.

Benchmarks

sort_bm.cpp includes a sorting example demonstrating how to use Google Benchmark library. number_series_bm.cpp your code for benchmarking the number_series and number_series_wrap.

Libraries included

• doctest: unit test framework.
• Google Benchmark: microbenchmark support library.

CMake should fetch those libraries automatically during cmake reload, and the libraries should build automatically when building the project.