gloo

gloo copied to clipboard

The `notify_send_ready` and `notify_recv_ready` messages used in the tcp backend (and future uv backend, see #195) need better documentation. The protocol how these are sequenced as well.

pietern

Installed gloo looks for libgloo in build dir

2

When importing Gloo (via find_package(Gloo) ), it tells me it was not able to find libglood.a in ${BUILD_DIR}. The file is correctly installed in ${INSTALL_DIR}. This is on Linux x86_64....

xsacha

This is what we do in PyTorch upstream today but it would be good to move the functionality into Gloo. This would be a new type of context that wraps...

pietern

The comments mention it is usable for any `#nodes == c * base ^ x`, for any `c >= 1`, `base >= 2`, and `x >= 1`, but in reality...

pietern

Reduce and Allreduce ops apply sanity check to enforce non-empty inputs [[here](https://github.com/facebookincubator/gloo/blob/master/gloo/allreduce.cc#L95)]. Allgather returns error code 8 on empty inputs. Does it make sense to support empty inputs in these...

mrshenli

test failure: connection reset by peer

2

``` [ RUN ] Transport/TransportMultiProcTest.UnboundIoErrors/44 terminate called after throwing an instance of 'gloo::IoException' what(): [/home/lumin/Debian/gloo.pkg/gloo/gloo/transport/tcp/pair.cc:374] writev [::1]:54172: Connection reset by peer /home/lumin/Debian/gloo.pkg/gloo/gloo/test/transport_test.cc:200: Failure Value of: WIFEXITED(result) Actual: false Expected: true...

cdluminate

This is used in a few places in `gloo/transport/**`.

pietern

The shutdown code may be a bit too aggressive here. We should have a stress test for the termination scenario where we loop on (context creation, barrier, context destruction) and...

pietern

Cannot run with SoftRoCE

2

Hello! It seems fine to run Gloo with RoCE, but it seems to be stuck with SoftRoCE. It should just run out of box but it looks like it cannot...

Luo-Liang

benchmark --verify error

2

Hi! I'm testing the benchmark program. When I use the --verify flag, I am getting some complaints. what(): [enforce fail at /home/ubuntu/gloo/gloo/benchmark/main.cc:91] T(offset + expected) == input[i]. 2.4e+07 vs 2.4e+07....

Luo-Liang