First version with working meson builder configuration, with tests!

[AtilaSaraiva]

Hello again! I made the working version, I still have some things to do, but all the tests are working. The most troublesome thing is that I had to write a small python script (for portability) to copy the *.cl files to another directory in memory, and then create the *.cl.o files from there with ld, each at their own folder, which is deleted afterwards. This is not ideal but is working.
I'd much rather have the focal library to be able to handle any symbol name.

Either way I will add how to use it on the docs and I thinking about using a simple shell script instead of python, just need to study it a little bit.

[AtilaSaraiva]

Oh I closed by accident haha

[codecov-commenter]

Codecov Report

Merging #15 (98277f4) into master (fee5d7a) will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff           @@
##           master      #15   +/-   ##
=======================================
  Coverage   71.89%   71.89%           
=======================================
  Files          10       10           
  Lines        1327     1327           
=======================================
  Hits          954      954           
  Misses        373      373           

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update fee5d7a...98277f4. Read the comment docs.

[LKedward]

Many thanks @AtilaSaraiva , looks good to me! Yes you are right about the *.cl object files, it is a bit awkward for the build process. I agree it would be better for Focal to be able to handle any symbol name; this might be possible, I'll have a think about it.
I think I might simplify the tests anyway so that the kernel sources are included in the Fortran files as strings since the kernels are quite simple and only a few tests actually use them anyway.

Is this PR ready to merge now or is there more work to do on it? Thanks again!

[AtilaSaraiva]

It is my pleasure! I'd like to check for mistakes before the merge, I'll do it asap.
I've been thinking about the problem with .cl files and I stumbled upon the clBlas repo. They write their kernels on normal c files, and use a python program to convert to a header file with a string variable containing the kernel, and compile the kernel and extract its binary, or at least this is what I managed to understand. We could do something like that!

Additionally, I'd like to ask a question, is it possible to use clBlas or other OpenCL libraries from fortran using focal? I thought a simple interface would do it but I'm not so sure, gotta test.

Either way, I still have to update the docs to reflect the meson builder addition. I'll try to create some time this week to do that. It is in another repository right?

[LKedward]

Awesome, sounds good! Yes that seems like a possible alternative.
It should be possible to use other OpenCL libraries from Focal, but I haven't tried this yet. I'd quite like to try clBlas with Focal.
Yes the documentation is in https://github.com/lkedward/focal-docs.
Cheers, no rush - let me know when you're ready to merge.

[AtilaSaraiva]

@LKedward pls do forgive me for my absence, the last few months have been crazy. I added support for generating pkg-config configurations for the library, making it easier for people with different build systems to easily import focal. This will also help me when I ultimately package this for Nix.

Now the meson configuration is looking good IMHO!

[AtilaSaraiva]

Oh I remembered, the .mod files need a special script to be installed, gonna work on that.