diff --git a/opal/mca/common/cuda/common_cuda.c b/opal/mca/common/cuda/common_cuda.c
index 56183043c26..14169678bb8 100644
--- a/opal/mca/common/cuda/common_cuda.c
+++ b/opal/mca/common/cuda/common_cuda.c
@@ -232,7 +232,6 @@ static void cuda_dump_memhandle(int, void *, char *) __opal_attribute_unused__ ;
 int mca_common_cuda_stage_one_init(void)
 {
     int retval, i, j;
-    int advise_support = 1;
     char *cudalibs[] = {"libcuda.so.1", "libcuda.dylib", NULL};
     char *searchpaths[] = {"", "/usr/lib64", NULL};
     char **errmsgs = NULL;
@@ -337,12 +336,7 @@ int mca_common_cuda_stage_one_init(void)
     }
 
     if (!OPAL_HAVE_DL_SUPPORT) {
-        if (OPAL_ERR_NOT_SUPPORTED == retval) {
-            opal_show_help("help-mpi-common-cuda.txt", "dlopen disabled", true);
-        } else {
-            opal_show_help("help-mpi-common-cuda.txt", "unknown ltdl error", true,
-                           "opal_lt_dlinit", retval, opal_lt_dlerror());
-        }
+        opal_show_help("help-mpi-common-cuda.txt", "dlopen disabled", true);
         return 1;
     }
 
@@ -350,94 +344,55 @@ int mca_common_cuda_stage_one_init(void)
      * that works.  If it does, all is good.  If not, print out all
      * the messages about why things failed.  This code was careful
      * to try and save away all error messages if the loading ultimately
-     * failed to help with debugging.  
+     * failed to help with debugging.
+     *
      * NOTE: On the first loop we just utilize the default loading
      * paths from the system.  For the second loop, set /usr/lib64 to
      * the search path and try again.  This is done to handle the case
-     * where we have both 32 and 64 bit libcuda.so libraries installed.
-     * Even when running in 64-bit mode, the /usr/lib directory
-     * is searched first and we may find a 32-bit libcuda.so.1 library.
-     * Loading of this library will fail as libtool does not handle having
-     * the wrong ABI in the search path (unlike ld or ld.so).  Note that
-     * we only set this search path after the original search.  This is
-     * so that LD_LIBRARY_PATH and run path settings are respected.
-     * Setting this search path overrides them (rather then being appended). */
-    if (advise_support) {
-        if (0 != (retval = opal_lt_dladvise_global(&advise))) {
-            opal_show_help("help-mpi-common-cuda.txt", "unknown ltdl error", true,
-                           "opal_lt_dladvise_global", retval, opal_lt_dlerror());
-            opal_lt_dladvise_destroy(&advise);
-            return 1;
+     * where we have both 32 and 64 bit libcuda.so libraries
+     * installed.  Even when running in 64-bit mode, the /usr/lib
+     * directory is searched first and we may find a 32-bit
+     * libcuda.so.1 library.  Loading of this library will fail as the
+     * OPAL DL framework does not handle having the wrong ABI in the
+     * search path (unlike ld or ld.so).  Note that we only set this
+     * search path after the original search.  This is so that
+     * LD_LIBRARY_PATH and run path settings are respected.  Setting
+     * this search path overrides them (rather then being
+     * appended). */
+    j = 0;
+    while (searchpaths[j] != NULL) {
+        /* Set explicit search path if entry is not empty string */
+        if (strcmp("", searchpaths[j])) {
+            opal_lt_dlsetsearchpath(searchpaths[j]);
         }
-        j = 0;
-        while (searchpaths[j] != NULL) {
-            /* Set explicit search path if entry is not empty string */
-            if (strcmp("", searchpaths[j])) {
-                opal_lt_dlsetsearchpath(searchpaths[j]);
-            }
-            i = 0;
-            while (cudalibs[i] != NULL) {
-                const char *str;
-                libcuda_handle = opal_lt_dlopenadvise(cudalibs[i], advise);
-                if (NULL == libcuda_handle) {
-                    str = opal_lt_dlerror();
-                    if (NULL != str) {
-                        opal_argv_append(&errsize, &errmsgs, str);
-                    } else {
-                        opal_argv_append(&errsize, &errmsgs, "lt_dlerror() returned NULL.");
-                    }
-                    opal_output_verbose(10, mca_common_cuda_output,
-                                        "CUDA: Library open error: %s",
-                                        errmsgs[errsize-1]);
+        i = 0;
+        while (cudalibs[i] != NULL) {
+            const char *str;
+            retval = opal_dl_open(cudalibs[i], true, false,
+                                  &libcuda_handle, &str);
+            if (OPAL_SUCCESS != retval || NULL == libcuda_handle) {
+                if (NULL != str) {
+                    opal_argv_append(&errsize, &errmsgs, str);
                 } else {
-                    opal_output_verbose(10, mca_common_cuda_output,
-                                        "CUDA: Library successfully opened %s",
-                                        cudalibs[i]);
-                    stage_one_init_passed = true;
-                    break;
+                    opal_argv_append(&errsize, &errmsgs,
+                                     "opal_dl_open() returned NULL.");
                 }
-                i++;
+                opal_output_verbose(10, mca_common_cuda_output,
+                                    "CUDA: Library open error: %s",
+                                    errmsgs[errsize-1]);
+            } else {
+                opal_output_verbose(10, mca_common_cuda_output,
+                                    "CUDA: Library successfully opened %s",
+                                    cudalibs[i]);
+                stage_one_init_passed = true;
+                break;
             }
-            if (true == stage_one_init_passed) break; /* Break out of outer loop */
-            j++;
+            i++;
         }
-        opal_lt_dladvise_destroy(&advise);
-    } else {
-        j = 0;
-        /* No lt_dladvise support.  This should rarely happen. */
-        while (searchpaths[j] != NULL) {
-            /* Set explicit search path if entry is not empty string */
-            if (strcmp("", searchpaths[j])) {
-                opal_lt_dlsetsearchpath(searchpaths[j]);
-            }
-            i = 0;
-            while (cudalibs[i] != NULL) {
-                const char *str;
-                libcuda_handle = opal_lt_dlopen(cudalibs[i]);
-                if (NULL == libcuda_handle) {
-                    str = opal_lt_dlerror();
-                    if (NULL != str) {
-                        opal_argv_append(&errsize, &errmsgs, str);
-                    } else {
-                        opal_argv_append(&errsize, &errmsgs, "lt_dlerror() returned NULL.");
-                    }
-
-                    opal_output_verbose(10, mca_common_cuda_output,
-                                        "CUDA: Library open error: %s",
-                                        errmsgs[errsize-1]);
-
-                } else {
-                    opal_output_verbose(10, mca_common_cuda_output,
-                                        "CUDA: Library successfully opened %s",
-                                        cudalibs[i]);
-                    stage_one_init_passed = true;
-                    break;
-                }
-                i++;
-            }
-            if (true == stage_one_init_passed) break; /* Break out of outer loop */
-            j++;
+        if (true == stage_one_init_passed) {
+            break; /* Break out of outer loop */
         }
+        j++;
     }
 
     if (true != stage_one_init_passed) {
diff --git a/opal/mca/common/cuda/help-mpi-common-cuda.txt b/opal/mca/common/cuda/help-mpi-common-cuda.txt
index f91826dee40..b4b5b96b0dc 100644
--- a/opal/mca/common/cuda/help-mpi-common-cuda.txt
+++ b/opal/mca/common/cuda/help-mpi-common-cuda.txt
@@ -1,6 +1,7 @@
 # -*- text -*-
 #
 # Copyright (c) 2011-2014 NVIDIA.  All rights reserved.
+# Copyright (c) 2015 Cisco Systems, Inc.  All rights reserved.
 # $COPYRIGHT$
 # 
 # Additional copyrights may follow
@@ -158,14 +159,6 @@ Open MPI was compiled without dynamic library support (e.g., with the
 
 If you need CUDA support, reconfigure Open MPI with dynamic library support enabled.
 #
-[unknown ltdl error]
-While attempting to load the supporting libcuda.so library, an error
-occurred.  This really should rarely happen.  Please notify the Open
-MPI developers. 
-   Function:     %s
-   Return Value: %d
-   Error string: %s
-#
 [dlopen failed]
 The library attempted to open the following supporting CUDA libraries, 
 but each of them failed.  CUDA-aware support is disabled.