Mikhail Brinskiy

## What Add support for host memory pipeline thru cuda staging buffers. This gives the best performance on systems with multi-node NVLINK.

## What disable gdrcopy to avoid dmabuf failures, like: ``` [ RUN ] shm_ib/test_ucp_am_nbx_eager_memtype.basic/19 [1724858288.255042] [swx-rdmz-ucx-gpu-02:70262:0] ib_md.c:293 UCX ERROR ibv_reg_dmabuf_mr(address=0x7f3245200000, length=65536, access=0xf) failed: Invalid argument [1724858288.255086] [swx-rdmz-ucx-gpu-02:70262:0] ucp_mm.c:70 UCX ERROR...

## What Enable using cuda staging buffers for pipeline protocols by default. ## Why ? Using ppln protocols with cuda staging buffers provides better performance for the following cases: -...

AZP: Use different ports in one wire-compat job - v1.17.x

2

## What port #10071

## What Fix MNNVL reachability check