Raul Akhmetshin comments

Results 47 comments of


                                            Raul Akhmetshin

AZP: Add L40G test

/azp run

Feature Request: ability to print all ucx_perftest inputs and outputs in a structured, machine-readable format (like JSON)

Different network devices can be used for different data transfer operations. `mlx5_0` is a name of IB network device. This device is used in your case. This information can be...

Feature Request: ability to print all ucx_perftest inputs and outputs in a structured, machine-readable format (like JSON)

You can use `ucx_info` tool. E.g. `ucx_info -d | grep Device | sort --unique` will show you all devices that can be used by UCX. You can also use `ibstat...

UCT/CUDA_IPC: use rcache instead of pgtable

> @brminich / @ivankochin / @rakhmets Can you help with restarting the failed tests? I don't expect the failures to be related to the PR. > > @brminich when time...

UCP/PROTOV2: Fix AM multi-eager protocols

Should we revert 0e92c1d8fa749daa4c1b9ee4be6c807f4688eb75 in this PR to confirm the fix?

GTest failures when running on the L40G GPU

Please find the following conclusions based on CI logs. The failed tests are tests containing call to `nvmlInit`. The tests fails because they are aborted by watchdog at the end...

Attempts communication over unroutable IP interface

Hi @G-Ragghianti, We merged PR https://github.com/openucx/ucx/pull/9475 that should fix the issue. We also pushed the changes to v1.16.x branch (https://github.com/openucx/ucx/pull/9487).

UCP/RMA: Do not flush UCT EP if UCT iface is not active.

/azp run

Error compiling for v1.14 or 1.15

The first compilation error was fixed by https://github.com/openucx/ucx/pull/8804

Error compiling for v1.14 or 1.15

Could you please check if this patch https://github.com/openucx/ucx/pull/9795 fixes the second issue?