Fast Matrix Transposition

This repository contains functions with different approaches to Transposition of Matrix and performance tests of them.

We sequentially implement :

1. Naive matrix transposition (single-thread)
   
2. Parallel naive matrix transposition (multi-threads)
   
3. SSE matrix transposition (single-thread)
   
4. SSE Block matrix transposition (single-thread)
   
5. and fastest variant SSE Parallel Block matrix transposition (multi-threads)
   

Results of performance tests:

All tests were performed on "Intel(R) Core(TM) i7-8650U CPU @ 1.90GHz" with 16 GB of DDR4 RAM
As can be seen significant difference apears starting with matrix about 2000x2000

SSE Block matrix transposition approach faster than any other single-thread approaches
SSE Parallel Block matrix transposition fastest at all

How-to-start tests:

To start performance test build and run following:

FastMatrixTransposition [matrix_size] [block_size] [number_of_threads] [number_of_tests performance tests for each approach]

Will outputed average times for each approah

Example of output:

8000,194659.156250,97701.046875,140561.281250,86731.703125,62631.093750

Format of output:

[matrix_size],[naive approach], [parallel naive approach], [SSE matrix transposition], [Block transposition], [Block SSE parallel transposition]

All times in nano seconds