mscclpp

mirror of https://github.com/microsoft/mscclpp.git synced 2026-03-26 01:57:46 +00:00

Files

Changho Hwang 105239fc6c Use GpuIpcMem for NVLS connections (#719 )

* Now `NvlsConnection` internally reuses `GpuIpcMem` for multicast
memory handling.
* Removed unnecessary barriers from `connectNvlsCollective()` (CUDA API
handles this automatically).
* Updated `GpuIpcMem::map()` and `GpuIpcMem::mapMulticast()` to return a
shared pointer with custom deleter for unmapping, which prevents misuse
of raw pointers and reduces states to be stored in the `GpuIpcMem`
instance.
* Now for `RuntimeIpc` type handles, for consistency with other types,
`cudaIpcOpenMemHandle` will be called in `GpuIpcMem::map()` instead of
the ctor of `GpuIpcMem`.

---------

Co-authored-by: Binyang Li <binyli@microsoft.com>
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
Co-authored-by: Copilot <198982749+Copilot@users.noreply.github.com>
Co-authored-by: Binyang2014 <9415966+Binyang2014@users.noreply.github.com>

2026-01-15 13:16:04 +08:00

csrc

Use GpuIpcMem for NVLS connections (#719 )

2026-01-15 13:16:04 +08:00

mscclpp

Add handle cache for AMD platform (#698 )

2025-12-21 18:39:12 -08:00

mscclpp_benchmark

Fix Python bindings and tests (#690 )

2025-11-21 12:53:12 -08:00

test

Add handle cache for AMD platform (#698 )