markalle

I like sndrcv() for the data copying, and I wish an analogue of ompi_datatype_add() was part of the MPI standard, that's a nice API option to have. I think it...

This feels like it could be a can of worms involving what fp_ind means and how it should be updated, it's used a lot of places. Many uses of fp_ind...

Is anyone able to pass the whitespace checker first try? Some of its corrections are super questionable I think. LIke in the below "bad" code the opening "{" is very...

Ah, I didn't know how to run the git pre-commit hook. Re-pushing and looking at CI results for every indentation fix was making it hard to pass, but being able...

Yeah, this is a tough area, and my 'fix' is definitely not comprehensive. I also wrote this code a long time ago so I'm not remembering why I did some...

Fixes https://github.com/open-mpi/ompi/issues/8918 not well tested yet. So far just tested on Mac. Needs tested somewhere bigendian, and somewhere that __float128 is defined since I only just added that path and...

I wasn't saying your code is wrong for big-endian machines, I'm saying it only converts data that's in the local endianness, it can't be used directly on big-endian data on...

I didn't go too far with performance, but those memcpy() were awfully expensive in at least some cases, so I put an alignment check and conditionally avoid the memcpy now,...

Thanks, that's a good argument for that format then. So I just repushed to put the padding back where that header had it.

Repushed. I used the new code conversions where @dalcinl added f64 as well as f80, but renamed the functions to `f80_to_f128()` etc. The usage from the OMPI level is in...