cuda: add minimal CUDA program + cleanup

gpu/cuda/.gitignore

@@ -3,3 +3,4 @@ hello-world.ptx
 threads
 inc
 array-add
+minimal

gpu/cuda/Makefile

@@ -1,3 +1,5 @@
+minimal: src/minimal.cu
+	nvcc -o $@ $<
 
 hello-world: src/hello-world.cu
 	nvcc -lnppc -o $@ $<
@@ -17,4 +19,4 @@ array-add: src/array-add.cu
 
 .PHONY: clean
 clean:
-	${RM} hello-world threads inc hello-world.ptx
+	${RM} hello-world threads inc hello-world.ptx minimal

gpu/cuda/cuda-env.sh

@@ -1,8 +1,8 @@
-VERSION=12.2
+VERSION=12.4
 export PATH=/usr/local/cuda-${VERSION}/bin:$PATH                                    
 export LD_LIBRARY_PATH=/usr/local/cuda-${VERSION}/lib64:$LD_LIBRARY_PATH 
 
 # Source Spack so that we can switch to use [email protected]
-. ~/work/linux/spack/share/spack/setup-env.sh
-spack load [email protected]
-gcc --version
+#. ~/work/linux/spack/share/spack/setup-env.sh
+#spack load [email protected]
+#gcc --version

gpu/cuda/src/hello-world.cu

@@ -7,9 +7,46 @@ __global__ void helloWorld() {
 }
 
 int main() {
+    cudaError_t cudaStatus = cudaGetLastError();
+    // Print CUDA version
+    int runtimeVersion = 0;
+    cudaStatus = cudaRuntimeGetVersion(&runtimeVersion);
+    if (cudaStatus != cudaSuccess) {
+        fprintf(stderr, "cudaRuntimeGetVersion failed: %s\n", cudaGetErrorString(cudaStatus));
+        return 1;
+    }
+    printf("CUDA Runtime version: %d\n", runtimeVersion);
+
+    // Check for CUDA device
+    int deviceCount = 0;
+    cudaStatus = cudaGetDeviceCount(&deviceCount);
+    if (cudaStatus != cudaSuccess) {
+        fprintf(stderr, "cudaGetDeviceCount failed: %s\n", cudaGetErrorString(cudaStatus));
+        return 1;
+    }
+    printf("Number of CUDA devices: %d\n", deviceCount);
+
+    if (deviceCount == 0) {
+        fprintf(stderr, "No CUDA-capable devices found\n");
+        return 1;
+    }
+
+
+
+    // Print device information
+    /*
+    cudaDeviceProp prop;
+    cudaStatus = cudaGetDeviceProperties(&prop, 0);
+    if (cudaStatus != cudaSuccess) {
+        fprintf(stderr, "cudaGetDeviceProperties failed: %s\n", cudaGetErrorString(cudaStatus));
+        return 1;
+    }
+    printf("Device name: %s\n", prop.name);
+    printf("Compute capability: %d.%d\n", prop.major, prop.minor);
+    */
+
     helloWorld<<<1, 1>>>();
 
-    cudaError_t cudaStatus = cudaGetLastError();
     if (cudaStatus != cudaSuccess) {
         fprintf(stderr, "helloWorld launch failed: %s\n", cudaGetErrorString(cudaStatus));
         return 1;

gpu/cuda/src/minimal.cu

@@ -0,0 +1,32 @@
+#include <stdio.h>
+
+__global__ void myKernel() {}
+
+int main() {
+    cudaError_t err = cudaSuccess;
+    int count = 0;
+
+    err = cudaGetDeviceCount(&count);
+    if (err != cudaSuccess) {
+        printf("cudaGetDeviceCount failed: %s\n", cudaGetErrorString(err));
+        return 1;
+    }
+
+    printf("CUDA device count: %d\n", count);
+
+    myKernel<<<1,1>>>();
+    err = cudaGetLastError();
+    if (err != cudaSuccess) {
+        printf("Kernel launch failed: %s\n", cudaGetErrorString(err));
+        return 1;
+    }
+
+    err = cudaDeviceSynchronize();
+    if (err != cudaSuccess) {
+        printf("cudaDeviceSynchronize failed: %s\n", cudaGetErrorString(err));
+        return 1;
+    }
+
+    printf("CUDA program ran successfully\n");
+    return 0;
+}