Greetings,

I am interested in using this for a research project for which I build suffix arrays and inverse suffix arrays. I have so far been using the original divsufsort and although it seems to have some openmp capability built in, I don't seem to be able to get different performance turning openmp on and off (i.e. both compiler options give the same performance result).

My project is already built in VC++ and I was wondering if I could use your parallel version of divsufsort to improve my results. Do you have any notes or pointers for building your code in VC++ (2013 and above)? Would it be enough to swap some of the original divsufsort code with your version to get better results?

Sincerely,
Faizal

Hi Julian, and thanks again for this nice resource.

Apparently I run into a bug for short inputs (less than hundred characters).
To reproduce it you can feed the following file https://www.cs.helsinki.fi/u/dvalenzu/tmp/parallel_divsufsort/problematic_text.txt
to the timing example https://github.com/jlabeit/parallel-divsufsort/blob/master/examples/timing.cpp
which in my workstation gets stuck (laptop with 4 cores).

Best Regards,
Daniel

Hi,

I found another case, now with a larger instance ( ~50MB), which gives a segmentation fault.
The steps to reproduce it:

• Compile the code using OpenMP [I don't know if the behavior is the same with the other flags as I have used only the OpenMP compulation]
• Download the input from https://www.cs.helsinki.fi/u/dvalenzu/tmp/parallel_divsufsort/genome_prob_2.txt.7z and d decompress it, e.g. 7z .
• Runing the example code from https://github.com/jlabeit/parallel-divsufsort/blob/master/examples/timing.cpp triggers the segmentation fault in machines that can run with 16 threads (or more).
  The same execution in a computer that can run up to 4 threads goes without problem.

Let me know if you are able to reproduce the problem or if you need additional details,
Best,
Daniel