Added cuda logs for the engine, which shows device info and memory usage

src/detect.c

@@ -2829,19 +2829,46 @@ int SigGroupBuild (DetectEngineCtx *de_ctx) {
     SigAddressPrepareStage2(de_ctx);
 
 #ifdef __SC_CUDA_SUPPORT__
+    unsigned int cuda_total = 0;
+    unsigned int cuda_free_before_alloc = 0;
     /* we register a module that would require cuda handler service.  This
      * module would hold the context for all the patterns in the rules */
     de_ctx->cuda_rc_mod_handle = SCCudaHlRegisterModule("SC_RULES_CONTENT_B2G_CUDA");
+    if (de_ctx->mpm_matcher == MPM_B2G_CUDA) {
+        CUcontext dummy_context;
+        if (SCCudaHlGetCudaContext(&dummy_context,
+                                   de_ctx->cuda_rc_mod_handle) == -1) {
+            SCLogError(SC_ERR_B2G_CUDA_ERROR, "Error getting a cuda context for the "
+                       "module SC_RULES_CONTENT_B2G_CUDA");
+        }
+        if (SCCudaMemGetInfo(&cuda_free_before_alloc, &cuda_total) == 0) {
+            SCLogInfo("Total Memory available in the CUDA context used for mpm "
+                      "with b2g: %.2f MB", cuda_total/(1024.0 * 1024.0));
+            SCLogInfo("Free Memory available in the CUDA context used for b2g "
+                      "mpm before any allocation is made on the GPU for the "
+                      "context: %.2f MB", cuda_free_before_alloc/(1024.0 * 1024.0));
+        }
+    }
+
 #endif
 
     SigAddressPrepareStage3(de_ctx);
 
 #ifdef __SC_CUDA_SUPPORT__
+    unsigned int cuda_free_after_alloc = 0;
     /* if a user has selected some other mpm algo other than b2g_cuda, inspite of
      * enabling cuda support, then no cuda contexts or cuda vars would be created.
      * Pop the cuda context, only on confirming that the MPM algo selected is the
      * CUDA mpm algo */
     if (de_ctx->mpm_matcher == MPM_B2G_CUDA) {
+        if (SCCudaMemGetInfo(&cuda_free_after_alloc, &cuda_total) == 0) {
+            SCLogInfo("Free Memory available in the CUDA context used for b2g mpm "
+                      "after allocation is made on the GPU for the context: %.2f MB",
+                      cuda_free_after_alloc/(1024.0 * 1024.0));
+            SCLogInfo("Total memory consumed by the CUDA context for the b2g mpm: "
+                      "%.2f MB", (cuda_free_before_alloc/(1024.0 * 1024.0)) -
+                      (cuda_free_after_alloc/(1024.0 * 1024.0)));
+        }
         /* the AddressPrepareStage3 actually handles the creation of device
          * pointers on the gpu.  The cuda context that stage3 used would still be
          * attached to this host thread.  We need to pop this cuda context so that

src/suricata.c

@@ -390,11 +390,6 @@ int main(int argc, char **argv)
     /* Initialize the configuration module. */
     ConfInit();
 
-#ifdef __SC_CUDA_SUPPORT__
-    /* Init the CUDA environment */
-    SCCudaInitCudaEnvironment();
-#endif
-
     struct option long_opts[] = {
         {"dump-config", 0, &dump_config, 1},
         {"pfring-int",  required_argument, 0, 0},
@@ -563,6 +558,11 @@ int main(int argc, char **argv)
     SCLogInfo("This is %s version %s", PROG_NAME, PROG_VER);
     UtilCpuPrintSummary();
 
+#ifdef __SC_CUDA_SUPPORT__
+    /* Init the CUDA environment */
+    SCCudaInitCudaEnvironment();
+#endif
+
     if (!CheckValidDaemonModes(daemon, run_mode)) {
         exit(EXIT_FAILURE);
     }

src/util-cuda.c

@@ -3944,6 +3944,43 @@ void SCCudaPrintDeviceList(SCCudaDevices *devices)
     return;
 }
 
+/**
+ * \brief Prints some basic information for the default device(the first devie)
+ *        we will be using on this cuda platform for use by our engine.  This
+ *        function is basically to be used to print some minimal information to
+ *        the user at engine startup.
+ *
+ * \param devices Pointer to a SCCudaDevices instance that holds information on
+ *                the devices.
+ */
+void SCCudaPrintBasicDeviceInfo(SCCudaDevices *devices)
+{
+    int i = 0;
+
+    if (devices == NULL) {
+        SCLogError(SC_ERR_CUDA_ERROR, "CUDA environment not initialized.  "
+                   "Please initialized the CUDA environment by calling "
+                   "SCCudaInitCudaEnvironment() before making any calls "
+                   "to the CUDA API.");
+        return;
+    }
+
+    SCLogInfo("Printing graphics card device details used by our engine");
+    SCLogInfo("No of devices:  %d", devices->count);
+
+    for (i = 0; i < devices->count && i <= 1; i++) {
+        SCLogInfo("Device Name: %s", devices->devices[i]->name);
+        SCLogInfo("Device Major Revision: %d", devices->devices[i]->major_rev);
+        SCLogInfo("Device Minor Revision: %d", devices->devices[i]->minor_rev);
+        SCLogInfo("Device Multiprocessor Count: %d",
+                   devices->devices[i]->attr_multiprocessor_count);
+        SCLogInfo("Device Clock Rate: %d", devices->devices[i]->attr_clock_rate);
+        SCLogInfo("Device Clock Frequency: %d", devices->devices[i]->prop.clockRate);
+    }
+
+    return;
+}
+
 /**
  * \brief Gets the device list, for the CUDA platform environment initialized by
  *        the engine.
@@ -4013,6 +4050,8 @@ int SCCudaInitCudaEnvironment(void)
         goto error;
     }
 
+    SCCudaPrintBasicDeviceInfo(devices);
+
     return 0;
 
  error:

src/util-cuda.h

@@ -154,6 +154,7 @@ int SCCudaCtxSynchronize(void);
 int SCCudaDriverGetVersion(int *);
 
 void SCCudaPrintDeviceList(SCCudaDevices *);
+void SCCudaPrintBasicDeviceInfo(SCCudaDevices *);
 SCCudaDevices *SCCudaGetDeviceList(void);
 
 int SCCudaInitCudaEnvironment(void);