Pieter Noordhuis

Thanks for the issue report! Is this consistently reproducible for you? These tests are passing for me. It is possible this is a timing issue, e.g. a destruction time problem...

Hi @skypjack, thanks for creating the issue! First of all, thanks for creating uvw. It's the top library that came up for using libuv from C++ and it's been a...

Apologies for the lack of action on this issue, @skypjack. I haven't gotten around to working on this. In the mean time, PyTorch (where Gloo is predominantly used) is in...

Thanks for the suggestion. Since you're using `find_package(Gloo)` I assume this is a problem in the generated CMake module files?

Thanks for the kind words! I don't have access to 40G IB, only 100G IB. So the result wouldn't be an apples to apples comparison.

@bkovalev These benchmarks have nothing to do with any outside framework. They are native gloo algorithm benchmarks. Also see the `gloo/benchmark` directory for more info. You can build the benchmark...

This is not intentional; it has not been needed so far so nobody noticed it missing :) I found [this TODO](https://github.com/facebookincubator/gloo/blob/825d3ecbb5a97353c3bd03f00fbd19cb52c211dd/gloo/allreduce_builder.cc#L119) that I added myself about this very problem. Could...

Do you have more info about the setup, any environment information, stack traces, etc? To my knowledge we have never tried running this with soft RoCE.

Hi there! Thanks for reporting the issue. I think this is expected given the number of elements you're using in your test in combination with the number of machines. The...

The TF bump is happening in #1460.