Lisandro Dalcin

@nariba-fj Any update? Has Fujitsu implemented any new macros in `mpi.h`?

@markalle I'll insist with a point I made before. I believe my code is overall correct for BE arches, but what I'm not sure about is the helper union layout...

> Is there a reason you initially coded it as above? I followed the layout in `ieee754.h` from glibc. Look for `union ieee854_long_double` in that header [[link]](https://github.com/bminor/glibc/blob/master/sysdeps/ieee754/ieee754.h#L136) However, I found...

@markalle I believe we missed a few important points in our original discussions. * I have a Raspberry Pi at home (armv7l). In there, `long double` is the same as...

@markalle I've updated my [gist](https://gist.github.com/dalcinl/05cccf7b11cdf169a750485f67b499b7) with float64 float128 conversion routines. Please note I also changed the union names used in the implementation.

> I never did find a big endian machine using 80-bit floats to test on @markalle float80 is an Intel format, and Intel processors are little-endian. So I doubt you...

Perhaps it is time to revisit this `memory` type. I added it to easy supporting Python 2 and 3. The class is a striped-down version of the built-in `memoryview` type,...

> * Maybe we could store an `is_cpu_accessible` flag in the [`internal`](https://docs.python.org/3/c-api/buffer.html#c.Py_buffer.internal) field of the buffer to check if we should disable accessing GPUs? That's an excellent idea! > *...

@drew-parsons `AF_UNIX` is not available on Windows https://github.com/mpi4py/mpi4py/runs/7763801705?check_suite_focus=true

> I guess that's obvious :) The case wouldn't be in the way though, it just wouldn't occur on Windows installations. What do you mean? The case did happen, look...