Setting things up for using parallel STL C++17

[Scott Kostyshak]

As part of C++17, now the STL can optionally use parallelization. And in fact it's so easy that even I can do it.

On Ubuntu 22.04, I just had to install the package 'libtbb-dev'. Then, we need to add -ltbb to the link command. Then, we add the following include in whichever source file we want to use parallelization:

  #include <execution>

Then, adapting the STL algorithm call to run in parallel is trivial. Here is a helpful post in case you are curious. It also gives a simple test program if you want to try it out: https://stackoverflow.com/questions/51031060/are-c17-parallel-algorithms-implemented-already/55989883#55989883

Of course, the hard part is ensuring the relevant code is thread-safe.

I think it would be nice to set things up so that we could someday optionally allow to use such parallelization in LyX. It can be controlled at run-time; for example, we could have an LFUN that could toggle it on/off. We would not need to expose anything to the user, but it would be nice to set things up so that we can test it and then someday in a few years maybe flip a switch to expose things to the user.

Does anyone know of any candidates in LyX's code for using such parallelization? The best candidates are simple for loops where the iterations are independent, and where the for-loop takes long enough time that the user notices a lag. I'm going to send an email for a particular use case, but it is actually not a good candidate since the iterations are not independent and the underlying code is not thread-safe.

Scott
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 488 bytes
Desc: not available
URL: <http://lists.lyx.org/pipermail/lyx-devel/attachments/20220829/a707667a/attachment.asc>