Commit graph

3 commits

Author SHA1 Message Date
Seungha Yang
6fa405c1e2 cudaipcclient: Protect IPC handle import/close with global lock
Protect import/close with signle lock to avoid importing a IPC handle
while it's being closed by another cudaipcsrc from other thread.
Also fixing cuda context leak

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/5230>
2023-08-25 10:06:58 +00:00
L. E. Segovia
b1a5707fcb windows: Fix mutexes leaking into the exports table
Translation unit-local variables must be marked static on Windows,
otherwise they're made available to the whole binary.

See:

https://learn.microsoft.com/en-us/cpp/cpp/program-and-linkage-cpp?view=msvc-170#external-vs-internal-linkage

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/5185>
2023-08-16 08:17:05 +00:00
Seungha Yang
7b6023d9cf nvcodec: Add support for CUDA IPC
Adding cudaipc{src,sink} element for CUDA IPC support.

Implementation note:
* For the communication between end points, Win32 named-pipe
and unix domain socket will be used on Windows and Linux respectively.

* cudaipcsink behaves as a server, and all GPU resources will be owned by
the server process and exported for other processes, then cudaipcsrc
(client) will import each exported handle.

* User can select IPC mode via "ipc-mode" property of cudaipcsink.
There are two IPC mode, one is "legacy" which uses legacy CUDA IPC
method and the other is "mmap" which uses CUDA virtual memory API
with OS's resource handle sharing method such as DuplicateHandle()
on Windows. The "mmap" mode might be better than "legacy" in terms
of stability since it relies on OS's resource management but
it would consume more GPU memory than "legacy" mode.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/4510>
2023-08-14 13:41:01 +00:00