Merge pull request #47 from williamfgc/ci-refactoring

Refactor CI and address AMD sync issue
JuliaORNL · Feb 28, 2024 · 2deecc2 · 2deecc2
2 parents ba1e820 + dd1343f
commit 2deecc2
Show file tree

Hide file tree

Showing 5 changed files with 8 additions and 19 deletions.
diff --git a/.github/workflows/ci-cpu.yaml b/.github/workflows/ci-cpu.yaml
@@ -24,7 +24,7 @@ jobs:
         run: julia --project -e 'using Pkg; Pkg.instantiate();'
 
       - name: Test Threads
-        run: julia -t 2 --project -e 'using Pkg; Pkg.test()'
+        run: julia -t 4 --project -e 'using Pkg; Pkg.test()'
 
   macos:
     runs-on: macos-latest

diff --git a/.github/workflows/ci-gpu-AMD.yaml b/.github/workflows/ci-gpu-AMD.yaml
@@ -73,7 +73,7 @@ jobs:
 
       - name: Instantiate
         if: steps.check.outputs.triggered == 'true'
-        run: | 
+        run: |
           source /etc/profile.d/lmod.sh
           module load julia/1.9.1
           module load rocm/6.0.0
@@ -88,7 +88,7 @@ jobs:
           source /etc/profile.d/lmod.sh
           module load julia/1.9.1
           module load rocm/6.0.0
-          julia --project -e 'using Pkg; Pkg.test(;test_args=["perf"])'
+          julia --project -e 'using Pkg; Pkg.test()'
 
       - name: Report PR status
         if: always() && steps.check.outputs.triggered == 'true'

diff --git a/.github/workflows/ci-gpu-NVIDIA.yaml b/.github/workflows/ci-gpu-NVIDIA.yaml
@@ -81,7 +81,7 @@ jobs:
 
       - name: Test
         if: steps.check.outputs.triggered == 'true'
-        run: julia --project -e 'using Pkg; Pkg.test(;test_args=["perf"])'
+        run: julia --project -e 'using Pkg; Pkg.test()'
 
       - name: Report PR status
         if: always() && steps.check.outputs.triggered == 'true'

diff --git a/ext/JACCAMDGPU/JACCAMDGPU.jl b/ext/JACCAMDGPU/JACCAMDGPU.jl
@@ -7,7 +7,7 @@ function JACC.parallel_for(N::I, f::F, x...) where {I<:Integer,F<:Function}
   threads = min(N, numThreads)
   blocks = ceil(Int, N / threads)
   @roc groupsize = threads gridsize = threads * blocks _parallel_for_amdgpu(f, x...)
-  AMDGPU.synchronize()
+  # AMDGPU.synchronize()
 end
 
 function JACC.parallel_for((M, N)::Tuple{I,I}, f::F, x...) where {I<:Integer,F<:Function}
@@ -17,7 +17,7 @@ function JACC.parallel_for((M, N)::Tuple{I,I}, f::F, x...) where {I<:Integer,F<:
   Mblocks = ceil(Int, M / Mthreads)
   Nblocks = ceil(Int, N / Nthreads)
   @roc groupsize = (Mthreads, Nthreads) gridsize = (Mblocks * Mthreads, Nblocks * Nthreads) _parallel_for_amdgpu_MN(f, x...)
-  AMDGPU.synchronize()
+  # AMDGPU.synchronize()
 end
 
 function JACC.parallel_reduce(N::I, f::F, x...) where {I<:Integer,F<:Function}

diff --git a/test/runtests.jl b/test/runtests.jl
@@ -3,20 +3,12 @@ import JACC
 const backend = JACC.JACCPreferences.backend
 
 @static if backend == "cuda"
-    println("CUDA backend loaded")
+    @show "CUDA backend loaded"
     include("tests_cuda.jl")
-    if "perf" in ARGS
-        println("Running performance tests")
-        include("tests_cuda_perf.jl")
-    end
 
 elseif backend == "amdgpu"
     @show "AMDGPU backend loaded"
     include("tests_amdgpu.jl")
-    if "perf" in ARGS
-        println("Running performance tests")
-        include("tests_amdgpu_perf.jl")
-    end
 
 elseif backend == "oneapi"
     @show "OneAPI backend loaded"
@@ -25,8 +17,5 @@ elseif backend == "oneapi"
 elseif backend == "threads"
     @show "Threads backend loaded"
     include("tests_threads.jl")
-    if "perf" in ARGS
-        println("Running performance tests")
-        include("tests_threads_perf.jl")
-    end
+
 end