Changes for getting GPU details from C funtion

Author: punAhuja

java/cuvs-java/src/main/java/com/nvidia/cuvs/common/Util.java

@@ -35,23 +35,25 @@
 
 public class Util {
 
-  /**
-   * Returns the number of GPUs connected to the system using CuVSResources.
-   *
-   * @param resources The CuVSResources object managing native resources.
-   * @return Number of GPUs connected, or -1 if an error occurred.
-   */
-  public static int getNumberOfGPUs(CuVSResources resources) {
-    try {
-      MethodHandle getNumberOfGPUsHandle = resources.linker.downcallHandle(
-          resources.libcuvsNativeLibrary.find("get_number_of_gpus")
-              .orElseThrow(() -> new IllegalStateException("get_number_of_gpus not found in library")),
-          FunctionDescriptor.of(ValueLayout.JAVA_INT));
-
-      return (int) getNumberOfGPUsHandle.invokeExact();
+  public static String getGpuDetails(CuVSResources resources, int maxGpus, int maxDetailLength) {
+    try (Arena arena = Arena.ofConfined()) {
+      MemorySegment detailSegment = arena.allocate(maxGpus * maxDetailLength);
+      MethodHandle getGpuDetailsHandle = resources.linker.downcallHandle(
+          resources.libcuvsNativeLibrary.find("get_gpu_details")
+              .orElseThrow(() -> new IllegalStateException("get_gpu_details not found in library")),
+          FunctionDescriptor.of(ValueLayout.JAVA_INT, ValueLayout.ADDRESS, ValueLayout.JAVA_INT, ValueLayout.JAVA_INT));
+
+      int gpuCount = (int) getGpuDetailsHandle.invoke(detailSegment, maxGpus, maxDetailLength);
+      if (gpuCount < 0) {
+        throw new RuntimeException("Failed to retrieve GPU details");
+      }
+
+      // Convert MemorySegment to String
+      String details = new String(detailSegment.toArray(ValueLayout.JAVA_BYTE), 0, gpuCount * maxDetailLength);
+      return details.trim();
     } catch (Throwable e) {
-      System.err.println("Failed to invoke get_number_of_gpus: " + e.getMessage());
-      return -1; // Return -1 to indicate an error
+      System.err.println("Error invoking get_gpu_details: " + e.getMessage());
+      throw new RuntimeException("Failed to invoke get_gpu_details", e);
     }
   }
 

java/cuvs-java/src/test/java/com/nvidia/cuvs/CagraRandomizedTest.java

@@ -10,7 +10,6 @@
 import org.slf4j.LoggerFactory;
 
 import com.carrotsearch.randomizedtesting.RandomizedContext;
-import com.nvidia.cuvs.common.Util;
 
 public class CagraRandomizedTest extends LuceneTestCase {
   private Random random;
@@ -56,21 +55,12 @@ public void testResultsTopKWithRandomValues() throws Throwable {
     }
 
     log.info("Queries:");
-    for (float[] query : queries) {
+    for (float[] query : queries) {  
       log.info(java.util.Arrays.toString(query));
     }
 
     CuVSResources resources = new CuVSResources();
 
-    int gpuCount = Util.getNumberOfGPUs(resources);
-    if (gpuCount == -1) {
-      log.info("Failed to detect GPUs.");
-    } else if (gpuCount == 0) {
-      log.info("No GPUs detected.");
-    } else {
-      log.info("Number of GPUs detected: {}", gpuCount);
-    }
-
     CagraIndexParams indexParams = new CagraIndexParams.Builder(resources).build();
 
     CagraIndex index = new CagraIndex.Builder(resources).withDataset(dataset).withIndexParams(indexParams).build();
@@ -87,4 +77,3 @@ public void testResultsTopKWithRandomValues() throws Throwable {
   }
 
 }
-

java/cuvs-java/src/test/java/com/nvidia/cuvs/common/TestUtil.java

@@ -0,0 +1,24 @@
+package com.nvidia.cuvs.common;
+
+import static org.junit.Assert.assertTrue;
+
+import org.junit.Test;
+
+import com.nvidia.cuvs.CuVSResources;
+
+public class TestUtil {
+
+  @Test
+  public void testGpuDetails() throws Throwable {
+    try {
+      CuVSResources resources = new CuVSResources();
+      String details = Util.getGpuDetails(resources, 10, 256);
+      System.out.println("GPU Details: " + details);
+      assertTrue("GPU details should not be empty", !details.isEmpty());
+    } catch (RuntimeException e) {
+      e.printStackTrace();
+      throw new AssertionError("Test failed due to an exception: " + e.getMessage());
+    }
+  }
+
+}

java/internal/src/cuvs_java.c

@@ -20,6 +20,8 @@
 #include <cuda_runtime.h>
 #include <stdio.h>
 #include <stdlib.h>
+#include <cuda_runtime.h>
+#include <string.h>
 
 cuvsResources_t create_resource(int *returnValue) {
   cuvsResources_t cuvsResources;
@@ -104,3 +106,37 @@ int get_number_of_gpus() {
     }   
     return deviceCount;
 }
+
+int get_gpu_details(char *details, int max_gpus, int max_detail_length) {
+    int deviceCount = 0;
+    cudaError_t err = cudaGetDeviceCount(&deviceCount);
+
+    if (err != cudaSuccess || deviceCount == 0) {
+        fprintf(stderr, "cudaGetDeviceCount failed or no GPUs found: %s\n", cudaGetErrorString(err));
+        return -1;
+    }
+
+    for (int i = 0; i < deviceCount && i < max_gpus; i++) {
+        struct cudaDeviceProp deviceProp;
+        err = cudaGetDeviceProperties(&deviceProp, i);
+        if (err != cudaSuccess) {
+            snprintf(&details[i * max_detail_length], max_detail_length, 
+                "Error fetching properties for device %d", i);
+            continue;
+        }
+
+        size_t freeMem = 0, totalMem = 0;
+        cudaSetDevice(i);
+        err = cudaMemGetInfo(&freeMem, &totalMem);
+        if (err != cudaSuccess) {
+            snprintf(&details[i * max_detail_length], max_detail_length, 
+                "%s | Memory info unavailable", deviceProp.name);
+            continue;
+        }
+
+        snprintf(&details[i * max_detail_length], max_detail_length, 
+            "%s | Total: %zuMB | Free: %zuMB", deviceProp.name, totalMem / (1024 * 1024), freeMem / (1024 * 1024));
+    }
+
+    return deviceCount;
+}