View on GitHub

Graphitti

A project to facilitate construction of high-performance neural simulations

Array Performance Testing

This file tests the efficency of C++ arrays, Vectors, and Valarrays. Each is tested in four areas: Initialization, Filling the array with zeroes, Sequential access of each element in the array, and Random access of each element in the array. It also tests different methods such as Vector [] with .at() for accessing elements.

How to Use

Compile the program by running (replace the X with 0-4 for different compiler optimization levels)

g++ -OX -o arraytest arraySpeedTest.cpp

Then run the test using

./arraytest

Results

When used properly, Array, Vector, and Valarray have very similar performance. Built-in compiler optimization makes a huge difference in speed. If you know the size of your data beforehand, then reserve the size of the array before using it. Avoid functions that perform bounds checking.

Sample Output

Initialization

| | Array | Vector | Valarray | |—————————|————-|————-|————-| | No optimization | 43989 µs | 182074 µs | 58962 µs | | O1 | 23012 µs | 42519 µs | 24857 µs | | O2 | 20268 µs | 37518 µs | 21361 µs |

Access and Sum

| | Array | Vector [] | Vector .at() | Valarray | |—————————|————-|————-|————–|————-| | No optimization | 22617 µs | 30256 µs | 105913 µs | 30225 µs | | O1 | 6809 µs | 6338 µs | 6639 µs | 6544 µs | | O2 | 6906 µs | 6638 µs | 6449 µs | 6586 µs |

Random Access and Sum

| | Array | Vector [] | Vector .at() | Valarray | |—————————|————-|————-|————–|————-| | No optimization | 105570 µs | 142558 µs | 289157 µs | 143044 µs | | O1 | 73207 µs | 73429 µs | 73406 µs | 73156 µs | | O2 | 73473 µs | 73400 µs | 73767 µs | 75460 µs |