From fb2f9825cd581ca843d527ac7e2ef420dbdb31f5 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ludovic=20R=C3=A4ss?= <61313342+luraess@users.noreply.github.com> Date: Thu, 29 Jun 2023 14:34:32 +0200 Subject: [PATCH 1/3] Update Project.toml --- Project.toml | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/Project.toml b/Project.toml index 4539f0b8a..b38b1585c 100644 --- a/Project.toml +++ b/Project.toml @@ -19,7 +19,7 @@ Serialization = "9e88b42a-f829-5b0c-bbe9-9e923198166b" Sockets = "6462fe0b-24de-5631-8697-dd941f90decc" [compat] -AMDGPU = "0.3, 0.4" +AMDGPU = "0.3, 0.4, 0.5" CUDA = "3, 4" DocStringExtensions = "0.8, 0.9" MPIPreferences = "0.1.6" From 161b1d0906925a2ebef7db39d9b00619276a2f2a Mon Sep 17 00:00:00 2001 From: Simon Byrne Date: Sun, 9 Jul 2023 13:34:20 -0700 Subject: [PATCH 2/3] bump version --- Project.toml | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/Project.toml b/Project.toml index b38b1585c..ac6eb981f 100644 --- a/Project.toml +++ b/Project.toml @@ -1,7 +1,7 @@ name = "MPI" uuid = "da04e1cc-30fd-572f-bb4f-1f8673147195" authors = [] -version = "0.20.11" +version = "0.20.12" [deps] Distributed = "8ba89e20-285c-5b6f-9357-94700520ee1b" From 1367d68f06a8d4714eaa0e7d45c677d66304ffa1 Mon Sep 17 00:00:00 2001 From: Ludovic Raess Date: Sat, 22 Jul 2023 12:40:41 +0300 Subject: [PATCH 3/3] UPdate AMDGPU sync --- test/common.jl | 8 +------- 1 file changed, 1 insertion(+), 7 deletions(-) diff --git a/test/common.jl b/test/common.jl index 20b959991..75c87beec 100644 --- a/test/common.jl +++ b/test/common.jl @@ -8,13 +8,7 @@ if get(ENV,"JULIA_MPI_TEST_ARRAYTYPE","") == "CuArray" elseif get(ENV,"JULIA_MPI_TEST_ARRAYTYPE","") == "ROCArray" import AMDGPU 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.default_queue() - 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) - end + synchronize() = AMDGPU.HIP.synchronize() else ArrayType = Array synchronize() = nothing