Fix AMDGPU synchronize in tests & update doc (#628)

* Add CUDA-aware MPI all-to-all tests and fix typo. * Update ROCQueue init to latest syntax
JuliaParallel · Sep 17, 2022 · 4cd7118 · 4cd7118
1 parent 6ef9d6b
commit 4cd7118
Show file tree

Hide file tree

Showing 2 changed files with 7 additions and 2 deletions.
diff --git a/docs/src/usage.md b/docs/src/usage.md
@@ -74,7 +74,12 @@ If your MPI implementation has been compiled with CUDA support, then `CUDA.CuArr
 [CUDA.jl](https://github.com/JuliaGPU/CUDA.jl) package) can be passed directly as
 send and receive buffers for point-to-point and collective operations (they may also work with one-sided operations, but these are not often supported).
 
-If using Open MPI, the status of CUDA support can be checked via the
+Successfully running the [alltoall\_test\_cuda.jl](https://gist.github.com/luraess/0063e90cb08eb2208b7fe204bbd90ed2) 
+should confirm your MPI implementation to have the CUDA support enabled. Moreover, successfully running the 
+[alltoall\_test\_cuda\_multigpu.jl](https://gist.github.com/luraess/ed93cc09ba04fe16f63b4219c1811566) should confirm 
+your CUDA-aware MPI implementation to use multiple Nvidia GPUs (one GPU per rank).
+
+If using OpenMPI, the status of CUDA support can be checked via the
 [`MPI.has_cuda()`](@ref) function.
 
 ## ROCm-aware MPI support

diff --git a/test/common.jl b/test/common.jl
@@ -10,7 +10,7 @@ elseif get(ENV,"JULIA_MPI_TEST_ARRAYTYPE","") == "ROCArray"
     ArrayType = AMDGPU.ROCArray
     function synchronize()
         # TODO: AMDGPU synchronization story is complicated. HSA does not provide a consistent notion of global queues. We need a mechanism for all GPUArrays.jl provided kernels to be synchronized.
-        queue = AMDGPU.get_default_queue()
+        queue = AMDGPU.ROCQueue()
         barrier = AMDGPU.barrier_and!(queue, AMDGPU.active_kernels(queue))
         AMDGPU.HIP.hipDeviceSynchronize() # Sync all HIP kernels e.g. BLAS. N.B. this is blocking Julia progress
         wait(barrier)