gpu-compute,tests: Move GPU tests to testlib (#1270)

A new host tag `gcn_gpu` has been added. This allows for selection of
those GPU tests which depend upon the gcn-gpu docker image to run.

In addition to this, the square GPU tests has been moved to the CI
tests. This ensures some GPU code is compiled and run on every PR.

.github/workflows/ci-tests.yaml

@@ -221,6 +221,32 @@ jobs:
                   path: tests/testing-results
                   retention-days: 30
 
+    gpu-tests:
+        runs-on: [self-hosted, linux, x64]
+        container: ghcr.io/gem5/gcn-gpu:latest
+        timeout-minutes: 180
+        needs: [pre-commit, check-for-change-id]
+        steps:
+            - uses: actions/checkout@v4
+
+            # Build the VEGA_X86/gem5.opt binary.
+            - name: Build VEGA_X86/gem5.opt
+              run: scons build/VEGA_X86/gem5.opt -j`nproc`
+
+            # Run the GPU tests.
+            - name: Run Testlib GPU Tests
+              working-directory: ${{ github.workspace }}/tests
+              run: ./main.py run  --skip-build -vvv -t $(nproc) --host gcn_gpu gem5/gpu
+
+            # Upload the tests/testing-results directory as an artifact.
+            - name: Upload results
+              if: success() || failure()
+              uses: actions/upload-artifact@v4
+              with:
+                  name: ci-tests-run-${{ github.run_number }}-attempt-${{ github.run_attempt }}-gpu-status-${{ steps.run-tests.outcome }}-output
+                  path: tests/testing-results
+                  retention-days: 30
+
     ci-tests:
     # It is 'testlib-quick' which needs to pass for the pull request to be
     # merged. This job is a dummy job that depends on all the other jobs.
@@ -232,5 +258,6 @@ jobs:
             - unittests-all-opt
             - pre-commit
             - check-for-change-id
+            - gpu-tests
         steps:
             - run: echo "This job's status is ${{ job.status }}."

.github/workflows/daily-tests.yaml

@@ -80,6 +80,30 @@ jobs:
                   retention-days: 7
             - run: echo "This job's status is ${{ job.status }}."
 
+    gpu-tests:
+        runs-on: [self-hosted, linux, x64]
+        container: ghcr.io/gem5/gcn-gpu:latest
+        timeout-minutes: 300
+
+        steps:
+            - uses: actions/checkout@v4
+              with:
+                  ref: develop
+            - name: Build VEGA_X86/gem5.opt
+              run: scons build/VEGA_X86/gem5.opt -j`nproc`
+
+            - name: Run Testlib GPU Tests
+              working-directory: ${{ github.workspace }}/tests
+              run: ./main.py run --length=long --skip-build -vvv -t $(nproc) --host gcn_gpu
+
+            - name: Upload results
+              if: success() || failure()
+              uses: actions/upload-artifact@v4
+              with:
+                  name: gpu_COMMIT.${{github.sha}}_RUN.${{github.run_id}}_ATTEMPT.${{github.run_attempt}}
+                  path: tests/testing-results
+                  retention-days: 7
+
   # This runs the SST-gem5 integration compilation and tests it with
   # ext/sst/sst/example.py.
     sst-test:
@@ -124,30 +148,6 @@ jobs:
             - name: Continue gem5 within SystemC test
               run: LD_LIBRARY_PATH=build/ARM/:/opt/systemc/lib-linux64/ ./util/systemc/gem5_within_systemc/gem5.opt.sc m5out/config.ini
 
-  # Runs the gem5 Nighyly GPU tests.
-    gpu-tests:
-        runs-on: [self-hosted, linux, x64]
-        container: ghcr.io/gem5/gcn-gpu:latest
-        timeout-minutes: 720 # 12 hours
-
-        steps:
-            - uses: actions/checkout@v4
-            - name: Compile build/VEGA_X86/gem5.opt
-              run: scons build/VEGA_X86/gem5.opt -j $(nproc)
-            - name: Get Square test-prog from gem5-resources
-              run: build/VEGA_X86/gem5.opt util/obtain-resource.py square-gpu-test -p square
-            - name: Run Square test with VEGA_X86/gem5.opt (SE mode)
-              run: |
-                  mkdir -p tests/testing-results
-                  ./build/VEGA_X86/gem5.opt configs/example/apu_se.py --reg-alloc-policy=dynamic -n3 -c square
-            - name: Get allSyncPrims-1kernel from gem5-resources
-              run: build/VEGA_X86/gem5.opt util/obtain-resource.py allSyncPrims-1kernel -p allSyncPrims-1kernel
-            - name: Run allSyncPrims-1kernel sleepMutex test with VEGA_X86/gem5.opt (SE mode)
-              run: ./build/VEGA_X86/gem5.opt configs/example/apu_se.py --reg-alloc-policy=dynamic -n3 -c allSyncPrims-1kernel --options="sleepMutex 10 16
-                  4"
-            - name: Run allSyncPrims-1kernel lfTreeBarrUsing test with VEGA_X86/gem5.opt (SE mode)
-              run: ./build/VEGA_X86/gem5.opt configs/example/apu_se.py --reg-alloc-policy=dynamic -n3 -c allSyncPrims-1kernel --options="lfTreeBarrUniq
-                  10 16 4"
     daily-tests:
         # The dummy job is used to indicate whether the daily tests have
         # passed or not. This can be used as status check for pull requests.

.github/workflows/weekly-tests.yaml

@@ -8,73 +8,6 @@ on:
     workflow_dispatch:
 
 jobs:
-    build-gcn-gpu-gem5:
-        runs-on: [self-hosted, linux, x64]
-        container: ghcr.io/gem5/gcn-gpu:latest
-        steps:
-            - uses: actions/checkout@v4
-            - name: Build gem5
-              run: scons build/VEGA_X86/gem5.opt -j $(nproc) --ignore-style
-            - uses: actions/upload-artifact@v4
-              with:
-                  name: weekly-test-${{ github.run_number }}-attempt-${{ github.run_attempt }}-gem5-build-vega
-                  path: build/VEGA_X86/gem5.opt
-                  retention-days: 5
-            - run: echo "This job's status is ${{ job.status }}."
-
-    LULESH-tests:
-        runs-on: [self-hosted, linux, x64]
-        container: ghcr.io/gem5/gcn-gpu:latest
-        needs: build-gcn-gpu-gem5
-        timeout-minutes: 480 # 8 hours
-        steps:
-            - uses: actions/checkout@v4
-
-            - name: Download build/VEGA_X86/gem5.opt
-              uses: actions/download-artifact@v4
-              with:
-                  name: weekly-test-${{ github.run_number }}-attempt-${{ github.run_attempt }}-gem5-build-vega
-                  path: build/VEGA_X86
-              # `download-artifact` does not preserve permissions so we need to set
-              # them again.
-            - run: chmod u+x build/VEGA_X86/gem5.opt
-
-            - name: Obtain LULESH
-              # Obtains the latest LULESH compatible with this version of gem5 via
-              # gem5 Resources.
-              run: build/VEGA_X86/gem5.opt util/obtain-resource.py lulesh -p lulesh
-
-            - name: Run LULUESH tests
-              working-directory: ${{ github.workspace }}
-              run: |
-                  build/VEGA_X86/gem5.opt configs/example/apu_se.py -n3 --mem-size=8GB --reg-alloc-policy=dynamic --dgpu --gfx-version=gfx900 -c \
-                  lulesh --options="0.01 2"
-
-    HACC-tests:
-        runs-on: [self-hosted, linux, x64]
-        container: ghcr.io/gem5/gcn-gpu:latest
-        needs: build-gcn-gpu-gem5
-        timeout-minutes: 120 # 2 hours
-        steps:
-            - uses: actions/checkout@v4
-            - uses: actions/download-artifact@v4
-              with:
-                  name: weekly-test-${{ github.run_number }}-attempt-${{ github.run_attempt }}-gem5-build-vega
-                  path: build/VEGA_X86
-            - run: chmod u+x build/VEGA_X86/gem5.opt
-            - name: make hip directory
-              run: mkdir hip
-            - name: Compile m5ops and x86
-              working-directory: ${{ github.workspace }}/util/m5
-              run: |
-                  export TERM=xterm-256color
-                  scons build/x86/out/m5
-            - name: Download tests
-              run: build/VEGA_X86/gem5.opt util/obtain-resource.py hacc-force-tree -p hip/ForceTreeTest
-            - name: Run HACC tests
-              run: |
-                  build/VEGA_X86/gem5.opt configs/example/apu_se.py -n3 --reg-alloc-policy=dynamic --benchmark-root=hip -c ForceTreeTest --options="0.5 0.1 64 0.1 1 N 12 rcb"
-
     build-gem5:
         runs-on: [self-hosted, linux, x64]
         container: ghcr.io/gem5/ubuntu-24.04_all-dependencies:latest
@@ -127,6 +60,30 @@ jobs:
                   retention-days: 7
             - run: echo "This job's status is ${{ job.status }}."
 
+    gpu-tests:
+        runs-on: [self-hosted, linux, x64]
+        container: ghcr.io/gem5/gcn-gpu:latest
+        timeout-minutes: 300
+
+        steps:
+            - uses: actions/checkout@v4
+              with:
+                  ref: develop
+            - name: Build VEGA_X86/gem5.opt
+              run: scons build/VEGA_X86/gem5.opt -j`nproc`
+
+            - name: Run Testlib GPU Tests
+              working-directory: ${{ github.workspace }}/tests
+              run: ./main.py run --length=very-long --skip-build -vvv -t $(nproc) --host gcn_gpu
+
+            - name: Upload results
+              if: success() || failure()
+              uses: actions/upload-artifact@v4.0.0
+              with:
+                  name: gpu_COMMIT.${{github.sha}}_RUN.${{github.run_id}}_ATTEMPT.${{github.run_attempt}}
+                  path: tests/testing-results
+                  retention-days: 7
+
     dramsys-tests:
         runs-on: [self-hosted, linux, x64]
         container: ghcr.io/gem5/ubuntu-22.04_all-dependencies:latest
@@ -158,7 +115,6 @@ jobs:
         needs:
             - testlib-very-long-tests
             - dramsys-tests
-            - LULESH-tests
-            - HACC-tests
+            - gpu-tests
         steps:
             - run: echo "This weekly tests have passed."