Change CUDA install validation test to support both v11.4 and v11.x > 11.4 on ARM (#4305)

Starting with CUDA Toolkit v11.5, the samples where the deviceQuery source code is installed
was moved out of the CUDA install package and into its own github repository. The structure
of the sample directy tree also changed thus requiring different paths for compiling and
executing the deviceQuery utility.

Co-authored-by: Luca Carrogu <[email protected]>

Author: davprat

cli/src/pcluster/resources/imagebuilder/parallelcluster_validate.yaml

@@ -167,6 +167,36 @@ phases:
               VERSION=$(echo ${PATTERN} | tr -d '\n' | cut -d = -f 2 | xargs)
               echo ${VERSION}
 
+      - name: CudaSamplesSrcDir
+        action: ExecuteBash
+        inputs:
+          commands:
+            - |
+              set -v
+              cuda_ver="{{ validate.CudaVersion.outputs.stdout }}"
+              if [ ${cuda_ver} \> '11.4' ]; then
+                PATTERN=$(grep -F "default['cluster']['nvidia']['cuda_sample_version']" {{ CookbookDefaultFile }})
+                VERSION=$(echo ${PATTERN} | tr -d '\n' | cut -d = -f 2 | xargs)
+                echo cuda-samples-${VERSION}/Samples
+              else
+                echo cuda-${cuda_ver}/samples
+              fi
+
+      - name: CudaSamplesBinDir
+        action: ExecuteBash
+        inputs:
+          commands:
+            - |
+              set -v
+              cuda_ver="{{ validate.CudaVersion.outputs.stdout }}"
+              if [ ${cuda_ver} \> '11.4' ]; then
+                PATTERN=$(grep -F "default['cluster']['nvidia']['cuda_sample_version']" {{ CookbookDefaultFile }})
+                VERSION=$(echo ${PATTERN} | tr -d '\n' | cut -d = -f 2 | xargs)
+                echo cuda-samples-${VERSION}/bin
+              else
+                echo cuda-${cuda_ver}/samples/bin
+              fi
+
       - name: ArmPLVersion
         action: ExecuteBash
         inputs:
@@ -280,7 +310,7 @@ phases:
                   apt list --installed | grep "nvidia-fabric.*manager" | grep "${nvidia_driver_version}" || exit 1
                   apt-mark showhold | grep "nvidia-fabric.*manager" || exit 1
                 fi
-                echo "Fabric Manager match Nvidia driver and version is locked"              
+                echo "Fabric Manager match Nvidia driver and version is locked"
               fi
 
               echo "Testing CUDA installation with nvcc"
@@ -294,11 +324,11 @@ phases:
               echo "Testing CUDA with deviceQuery..."
               if [ {{ validate.OperatingSystemArchitecture.outputs.stdout }} != 'arm64' ]; then
                 /usr/local/cuda-${cuda_ver}/extras/demo_suite/deviceQuery | grep -o "Result = PASS"
-                [[ $? -ne 0 ]] && echo "CUDA deviceQuery test failed" && exit 1              
+                [[ $? -ne 0 ]] && echo "CUDA deviceQuery test failed" && exit 1
               else
-                 cd /usr/local/cuda-${cuda_ver}/samples/1_Utilities/deviceQuery
+                 cd /usr/local/{{ validate.CudaSamplesSrcDir.outputs.stdout }}/1_Utilities/deviceQuery
                  make
-                 /usr/local/cuda-${cuda_ver}/samples/bin/sbsa/linux/release/deviceQuery | grep -o "Result = PASS"
+                 /usr/local/{{ validate.CudaSamplesBinDir.outputs.stdout }}/sbsa/linux/release/deviceQuery | grep -o "Result = PASS"
                  [[ $? -ne 0 ]] && echo "CUDA deviceQuery test failed" && exit 1
               fi
               echo "CUDA deviceQuery test passed"