Mpbenchmark is an application benchmark for evaluating the efficiency of multiprocessor support for parallel processing. Mpbenchmark was a revised version of "JetBench", written in many languagues such as C(OpenMP), Ada, C# and Java. This repository contains mpbenchmark writtein in mordern C++. This is a command line application used to analyse the real time performance of a multicore platform. The application is designed to be platform independent by avoiding target specific libraries and hardware counters and timers.
The original source code for mpbenchmark delivered optimal performance with implementations in Ada and C, attributed to their nature as compiled languages. As anticipated, Java and C# implementations yielded lower performance metrics. Given that the original codebase was dated, and faced challenges inherent to C—such as manual memory management and a lack of object-oriented programming features—it became imperative to transition to modern C++. With advancements in C++, especially the advent of std::thread introduced in C++11, memory management concerns have been significantly alleviated. It's intriguing to envisage how a modern compiled language like C++ compares in performance against the original C implementation.
Moreover, an alternative version is also produced to leverage SIMD instructions(the application will determine whether to use AVX2 or NEON depending on the CPU). The SIMD optimised version perfoms over 50% faster than the previous solutions in C(OpenMP) and Ada, please see the benchmark time plot below.
Measurements were taken using 16 threads using the following system:
CPU: Amd Ryzen 7 5800H
RAM: 16GB
OS: Ubuntu Linux 22.04.3 LTS
The C++ solution especially the optimised SIMD version darastically outperforms all the solutions.
Linux:
1- First make sure to install CMake and the fmt C++ library. CMake and fmt can be installed on Debian/Ubuntu distribution like shown below:
1.1 - Intall CMake using sudo apt install cmake. Then verify the intallation cmake --version.
1.2 - Install fmt library using the command sudo apt install libfmt-dev.
2- Clone this repository.
2.1 - To use the SIMD optimised version(recommended). Checkout to the WFD branch, run git checkout WFD.
2.2 - To use the regular C++ version without SIMD stay on this branch.
Now create a "build" folder inside "C++_BenchMark". You can do this by running the commands:
cd C++_BenchMark
mkdir build && cd build
3- Now run cmake and build the project. You can use these commands:
cmake ..
cmake --build .
4- Once the build is successfull then run the executable. You can supply it with the engine type(1, 2 or 3) and specify the threads to use as the command line arguments. You can also leave this blank to use the default parameters. Run this for the default configuration:
./mpbenchmark
Run this to use engine 1 and 12 threads:
./mpbenchmark 1 12
Windows(MinGW):
Here we make use of the MinGW compiler for windows.
1- Firstly install cmake and the fmt library C++ library. fmt library can be installed via vcpkg. Installing fmt is more tricky on Windows so I will share these steps.
2- Installing fmt :
2.1 - Clone the git repo git clone https://github.com/microsoft/vcpkg.git.
2.2 - Navigate inside the folder cd vcpkg. Then run the .bat file: .\bootstrap-vcpkg.bat.
2.3 - Now install the static library for MinGW. ./vcpkg install fmt:x64-mingw-static. Dynamic library can also be used but I used static for this example.
3- Once fmt is installed. Navigate inside the mpbenchmark folder that contains the cloned repo. Create a new folder for building mkdir build. Now we need to run CMake using the MinGW compiler, specify where to look for the vcpkg toolchain and specify the triplet here for MinGW compiler:
cmake -G "MinGW Makefiles" -B [build directory] -S [source directory] -DCMAKE_TOOLCHAIN_FILE=[path to vcpkg]\scripts\buildsystems\vcpkg.cmake" -DVCPKG_TARGET_TRIPLET=x64-mingw-static
Example for my pc would be:
cmake -G "MinGW Makefiles" -B "C:\Users\user1\Desktop\mpbenchmark\build" -S "C:\Users\user1\Desktop\mpbenchmark" -DCMAKE_TOOLCHAIN_FILE="C:\Users\user1\Documents\vcpkg\scripts\buildsystems\vcpkg.cmake" -DVCPKG_TARGET_TRIPLET=x64-mingw-static
4- Now we build the project. cmake --build "C:\Users\user1\Desktop\mpbenchmark\build"
5- Once built, find and execute the executable produced in the build folder.