AI/rocm-systems
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Refactor Barrier_all and Sync_all APIs to use default context (#159)

Browse Source 
* Refactor `Barrier_all` and `Sync_all` to use default context

- Removed context-specific implementations of barrier_all and sync_all
- Added barrier_all and sync_all to the default context implementation
- Updated functional tests to use the default context for barrier_all and sync_all

* Update `Barrier_all` and `Sync_all` API usage in documentation

* Update `CHANGELOG`

---------

Co-authored-by: Yiltan <ytemucin@amd.com>
This commit is contained in:
Avinash Kethineedi
2025-06-17 11:16:18 -05:00
committed by GitHub
parent 551603829c
commit bf48bcabf2
10 changed files with 115 additions and 130 deletions
Download Patch File Download Diff File Expand all files Collapse all files
docs/api/coll.rst
+9 -9
View File
@@ -11,16 +11,16 @@ Collective routines
ROCSHMEM_BARRIER_ALL
--------------------
.. cpp:function:: __device__ void rocshmem_ctx_barrier_all(rocshmem_ctx_t ctx)
.. cpp:function:: __device__ void rocshmem_ctx_barrier_all_wave(rocshmem_ctx_t ctx)
.. cpp:function:: __device__ void rocshmem_ctx_barrier_all_wg(rocshmem_ctx_t ctx)
.. cpp:function:: __device__ void rocshmem_barrier_all()
.. cpp:function:: __device__ void rocshmem_barrier_all_wave()
.. cpp:function:: __device__ void rocshmem_barrier_all_wg()
:param ctx: Context with which to perform this operation.
:returns: None.
**Description:**
This routine performs a collective barrier across all PEs in the system.
The caller is blocked until the barrier is resolved and all updates local and remote are completed.
These APIs should be called from only one thread/wavefront/workgroup within the grid to avoid undefined behavior.
ROCSHMEM_BARRIER
----------------
@@ -58,15 +58,15 @@ ensure the completion of remote memory updates issued via OpenSHMEM routines.
ROCSHMEM_SYNC_ALL
-----------------
.. cpp:function:: __device__ void rocshmem_ctx_sync_all(rocshmem_ctx_t ctx)
.. cpp:function:: __device__ void rocshmem_ctx_sync_all_wave(rocshmem_ctx_t ctx)
.. cpp:function:: __device__ void rocshmem_ctx_sync_all_wg(rocshmem_ctx_t ctx)
.. cpp:function:: __device__ void rocshmem_sync_all()
.. cpp:function:: __device__ void rocshmem_sync_all_wave()
.. cpp:function:: __device__ void rocshmem_sync_all_wg()
:param ctx: Context with which to perform this operation.
:returns: None.
**Description:**
This routine behaves the same as ``rocshmem_team_sync_wg`` when called on the world team.
These routines behaves the same way as ``rocshmem_team_sync_*`` when called on the world team.
These APIs should be called from only one thread/wavefront/workgroup within the grid to avoid undefined behavior.
ROSHMEM_ALLTOALL
----------------