2016-04-16 Basile Starynkevitch <basile@starynkevitch.net>

	{{merging with even more of GCC 6, using subversion 1.9
	svn merge -r230196:230700  ^/trunk
	}}


git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/branches/melt-branch@235061 138bc75d-0d04-0410-961f-82ee72b054a4

5 files changed, 241 insertions, 142 deletions

diff --git a/liboffloadmic/ChangeLog b/liboffloadmic/ChangeLog
index 93419bdfaaf..fb9fa851512 100644
--- a/liboffloadmic/ChangeLog
+++ b/liboffloadmic/ChangeLog
@@ -1,3 +1,56 @@
+2015-11-19  Ilya Verbin  <ilya.verbin@intel.com>
+
+	* plugin/libgomp-plugin-intelmic.cpp (struct TargetImageDesc): New.
+	(ImgDescMap): New typedef.
+	(image_descriptors): New static var.
+	(init): Allocate image_descriptors.
+	(offload): Remove vars2 argument.  Pass NULL to __offload_offload1
+	instead of vars2.
+	(unregister_main_image): New static function.
+	(register_main_image): Call unregister_main_image at exit.
+	(GOMP_OFFLOAD_init_device): Print device number, fix offload args.
+	(GOMP_OFFLOAD_fini_device): Likewise.
+	(get_target_table): Remove vd1g and vd2g, don't pass them to offload.
+	(offload_image): Remove declaration of the struct TargetImage.
+	Free table.  Insert new descriptor into image_descriptors.
+	(GOMP_OFFLOAD_unload_image): Call __offload_unregister_image, free
+	the corresponding descriptor, and remove it from address_table and
+	image_descriptors.
+	(GOMP_OFFLOAD_alloc): Print device number, remove vd1g.
+	(GOMP_OFFLOAD_free): Likewise.
+	(GOMP_OFFLOAD_host2dev): Print device number, remove vd1g and vd2g.
+	(GOMP_OFFLOAD_dev2host): Likewise.
+	(GOMP_OFFLOAD_run): Print device number, remove vd1g.
+	* plugin/offload_target_main.cpp (__offload_target_table_p1): Remove
+	vd2, don't pass it to __offload_target_enter.
+	(__offload_target_table_p2): Likewise.
+	(__offload_target_alloc): Likewise.
+	(__offload_target_free): Likewise.
+	(__offload_target_host2tgt_p1): Likewise.
+	(__offload_target_host2tgt_p2): Likewise.
+	(__offload_target_tgt2host_p1): Likewise.
+	(__offload_target_tgt2host_p2): Likewise.
+	(__offload_target_run): Likewise.
+	(__offload_target_tgt2tgt): Remove vd1g, don't pass it to
+	__offload_target_enter.
+
+2015-11-14  Ilya Verbin  <ilya.verbin@intel.com>
+
+	* runtime/offload_host.cpp (task_completion_callback): New
+	variable.
+	(offload_proxy_task_completed_ooo): Call task_completion_callback.
+	(__offload_register_task_callback): New function.
+	* runtime/offload_host.h (__offload_register_task_callback): New
+	declaration.
+	* plugin/libgomp-plugin-intelmic.cpp (offload): Add async_data
+	argument, handle async offloading.
+	(register_main_image): Call register_main_image.
+	(GOMP_OFFLOAD_init_device, get_target_table, GOMP_OFFLOAD_alloc,
+	GOMP_OFFLOAD_free, GOMP_OFFLOAD_host2dev, GOMP_OFFLOAD_dev2host,
+	GOMP_OFFLOAD_dev2dev): Adjust offload callers.
+	(GOMP_OFFLOAD_async_run): New function.
+	(GOMP_OFFLOAD_run): Implement using GOMP_OFFLOAD_async_run.
+
 2015-10-26  Ilya Verbin  <ilya.verbin@intel.com>
 	    Aleksander Ivanushenko  <aleksander.ivanushenko@intel.com>
 
diff --git a/liboffloadmic/plugin/libgomp-plugin-intelmic.cpp b/liboffloadmic/plugin/libgomp-plugin-intelmic.cpp
index 26ac6fed982..f8c172562ae 100644
--- a/liboffloadmic/plugin/libgomp-plugin-intelmic.cpp
+++ b/liboffloadmic/plugin/libgomp-plugin-intelmic.cpp
@@ -65,6 +65,17 @@ typedef std::vector<AddrVect> DevAddrVect;
 /* Addresses for all images and all devices.  */
 typedef std::map<const void *, DevAddrVect> ImgDevAddrMap;
 
+/* Image descriptor needed by __offload_[un]register_image.  */
+struct TargetImageDesc {
+  int64_t size;
+  /* 10 characters is enough for max int value.  */
+  char name[sizeof ("lib0000000000.so")];
+  char data[];
+};
+
+/* Image descriptors, indexed by a pointer obtained from libgomp.  */
+typedef std::map<const void *, TargetImageDesc *> ImgDescMap;
+
 
 /* Total number of available devices.  */
 static int num_devices;
@@ -76,6 +87,9 @@ static int num_images;
    second key is number of device.  Contains a vector of pointer pairs.  */
 static ImgDevAddrMap *address_table;
 
+/* Descriptors of all images, registered in liboffloadmic.  */
+static ImgDescMap *image_descriptors;
+
 /* Thread-safe registration of the main image.  */
 static pthread_once_t main_image_is_registered = PTHREAD_ONCE_INIT;
 
@@ -156,6 +170,7 @@ init (void)
 
 out:
   address_table = new ImgDevAddrMap;
+  image_descriptors = new ImgDescMap;
   num_devices = _Offload_number_of_devices ();
 }
 
@@ -192,11 +207,22 @@ GOMP_OFFLOAD_get_num_devices (void)
 
 static void
 offload (const char *file, uint64_t line, int device, const char *name,
-	 int num_vars, VarDesc *vars, VarDesc2 *vars2)
+	 int num_vars, VarDesc *vars, const void **async_data)
 {
   OFFLOAD ofld = __offload_target_acquire1 (&device, file, line);
   if (ofld)
-    __offload_offload1 (ofld, name, 0, num_vars, vars, vars2, 0, NULL, NULL);
+    {
+      if (async_data == NULL)
+	__offload_offload1 (ofld, name, 0, num_vars, vars, NULL, 0, NULL, NULL);
+      else
+	{
+	  OffloadFlags flags;
+	  flags.flags = 0;
+	  flags.bits.omp_async = 1;
+	  __offload_offload3 (ofld, name, 0, num_vars, vars, NULL, 0, NULL,
+			      async_data, 0, NULL, flags, NULL);
+	}
+    }
   else
     {
       fprintf (stderr, "%s:%d: Offload target acquire failed\n", file, line);
@@ -205,9 +231,27 @@ offload (const char *file, uint64_t line, int device, const char *name,
 }
 
 static void
+unregister_main_image ()
+{
+  __offload_unregister_image (&main_target_image);
+}
+
+static void
 register_main_image ()
 {
+  /* Do not check the return value, because old versions of liboffloadmic did
+     not have return values.  */
   __offload_register_image (&main_target_image);
+
+  /* liboffloadmic will call GOMP_PLUGIN_target_task_completion when
+     asynchronous task on target is completed.  */
+  __offload_register_task_callback (GOMP_PLUGIN_target_task_completion);
+
+  if (atexit (unregister_main_image) != 0)
+    {
+      fprintf (stderr, "%s: atexit failed\n", __FILE__);
+      exit (1);
+    }
 }
 
 /* liboffloadmic loads and runs offload_target_main on all available devices
@@ -215,16 +259,16 @@ register_main_image ()
 extern "C" void
 GOMP_OFFLOAD_init_device (int device)
 {
-  TRACE ("");
+  TRACE ("(device = %d)", device);
   pthread_once (&main_image_is_registered, register_main_image);
-  offload (__FILE__, __LINE__, device, "__offload_target_init_proc", 0,
-	   NULL, NULL);
+  offload (__FILE__, __LINE__, device, "__offload_target_init_proc", 0, NULL,
+	   NULL);
 }
 
 extern "C" void
 GOMP_OFFLOAD_fini_device (int device)
 {
-  TRACE ("");
+  TRACE ("(device = %d)", device);
   /* Unreachable for GOMP_OFFLOAD_CAP_OPENMP_400.  */
   abort ();
 }
@@ -237,10 +281,9 @@ get_target_table (int device, int &num_funcs, int &num_vars, void **&table)
   vd1[0].size = sizeof (num_funcs);
   vd1[1].ptr = &num_vars;
   vd1[1].size = sizeof (num_vars);
-  VarDesc2 vd1g[2] = { { "num_funcs", 0 }, { "num_vars", 0 } };
 
-  offload (__FILE__, __LINE__, device, "__offload_target_table_p1", 2,
-	   vd1, vd1g);
+  offload (__FILE__, __LINE__, device, "__offload_target_table_p1", 2, vd1,
+	   NULL);
 
   int table_size = num_funcs + 2 * num_vars;
   if (table_size > 0)
@@ -251,10 +294,9 @@ get_target_table (int device, int &num_funcs, int &num_vars, void **&table)
       vd2 = vd_tgt2host;
       vd2.ptr = table;
       vd2.size = table_size * sizeof (void *);
-      VarDesc2 vd2g = { "table", 0 };
 
-      offload (__FILE__, __LINE__, device, "__offload_target_table_p2", 1,
-	       &vd2, &vd2g);
+      offload (__FILE__, __LINE__, device, "__offload_target_table_p2", 1, &vd2,
+	       NULL);
     }
 }
 
@@ -264,13 +306,6 @@ get_target_table (int device, int &num_funcs, int &num_vars, void **&table)
 static void
 offload_image (const void *target_image)
 {
-  struct TargetImage {
-    int64_t size;
-    /* 10 characters is enough for max int value.  */
-    char name[sizeof ("lib0000000000.so")];
-    char data[];
-  } __attribute__ ((packed));
-
   void *image_start = ((void **) target_image)[0];
   void *image_end   = ((void **) target_image)[1];
 
@@ -278,9 +313,8 @@ offload_image (const void *target_image)
 	 target_image, image_start, image_end);
 
   int64_t image_size = (uintptr_t) image_end - (uintptr_t) image_start;
-  TargetImage *image
-    = (TargetImage *) malloc (sizeof (int64_t) + sizeof ("lib0000000000.so")
-			      + image_size);
+  TargetImageDesc *image = (TargetImageDesc *) malloc (offsetof (TargetImageDesc, data)
+						       + image_size);
   if (!image)
     {
       fprintf (stderr, "%s: Can't allocate memory\n", __FILE__);
@@ -293,6 +327,8 @@ offload_image (const void *target_image)
 
   TRACE ("() __offload_register_image %s { %p, %d }",
 	 image->name, image_start, image->size);
+  /* Do not check the return value, because old versions of liboffloadmic did
+     not have return values.  */
   __offload_register_image (image);
 
   /* Receive tables for target_image from all devices.  */
@@ -327,11 +363,11 @@ offload_image (const void *target_image)
 	}
 
       dev_table.push_back (curr_dev_table);
+      delete [] table;
     }
 
   address_table->insert (std::make_pair (target_image, dev_table));
-
-  free (image);
+  image_descriptors->insert (std::make_pair (target_image, image));
 }
 
 /* Return the libgomp version number we're compatible with.  There is
@@ -382,26 +418,31 @@ GOMP_OFFLOAD_unload_image (int device, unsigned version,
 
   TRACE ("(device = %d, target_image = %p)", device, target_image);
 
-  /* TODO: Currently liboffloadmic doesn't support __offload_unregister_image
-     for libraries.  */
+  /* liboffloadmic unloads the image from all available devices.  */
+  if (image_descriptors->count (target_image) > 0)
+    {
+      TargetImageDesc *image_desc = (*image_descriptors)[target_image];
+      __offload_unregister_image (image_desc);
+      free (image_desc);
 
-  address_table->erase (target_image);
+      address_table->erase (target_image);
+      image_descriptors->erase (target_image);
+    }
 }
 
 extern "C" void *
 GOMP_OFFLOAD_alloc (int device, size_t size)
 {
-  TRACE ("(size = %d)", size);
+  TRACE ("(device = %d, size = %d)", device, size);
 
   void *tgt_ptr;
-  VarDesc vd1[2] = { vd_host2tgt, vd_tgt2host };
-  vd1[0].ptr = &size;
-  vd1[0].size = sizeof (size);
-  vd1[1].ptr = &tgt_ptr;
-  vd1[1].size = sizeof (void *);
-  VarDesc2 vd1g[2] = { { "size", 0 }, { "tgt_ptr", 0 } };
+  VarDesc vd[2] = { vd_host2tgt, vd_tgt2host };
+  vd[0].ptr = &size;
+  vd[0].size = sizeof (size);
+  vd[1].ptr = &tgt_ptr;
+  vd[1].size = sizeof (void *);
 
-  offload (__FILE__, __LINE__, device, "__offload_target_alloc", 2, vd1, vd1g);
+  offload (__FILE__, __LINE__, device, "__offload_target_alloc", 2, vd, NULL);
 
   return tgt_ptr;
 }
@@ -409,21 +450,21 @@ GOMP_OFFLOAD_alloc (int device, size_t size)
 extern "C" void
 GOMP_OFFLOAD_free (int device, void *tgt_ptr)
 {
-  TRACE ("(tgt_ptr = %p)", tgt_ptr);
+  TRACE ("(device = %d, tgt_ptr = %p)", device, tgt_ptr);
 
-  VarDesc vd1 = vd_host2tgt;
-  vd1.ptr = &tgt_ptr;
-  vd1.size = sizeof (void *);
-  VarDesc2 vd1g = { "tgt_ptr", 0 };
+  VarDesc vd = vd_host2tgt;
+  vd.ptr = &tgt_ptr;
+  vd.size = sizeof (void *);
 
-  offload (__FILE__, __LINE__, device, "__offload_target_free", 1, &vd1, &vd1g);
+  offload (__FILE__, __LINE__, device, "__offload_target_free", 1, &vd, NULL);
 }
 
 extern "C" void *
 GOMP_OFFLOAD_host2dev (int device, void *tgt_ptr, const void *host_ptr,
 		       size_t size)
 {
-  TRACE ("(tgt_ptr = %p, host_ptr = %p, size = %d)", tgt_ptr, host_ptr, size);
+  TRACE ("(device = %d, tgt_ptr = %p, host_ptr = %p, size = %d)",
+	 device, tgt_ptr, host_ptr, size);
   if (!size)
     return tgt_ptr;
 
@@ -432,18 +473,16 @@ GOMP_OFFLOAD_host2dev (int device, void *tgt_ptr, const void *host_ptr,
   vd1[0].size = sizeof (void *);
   vd1[1].ptr = &size;
   vd1[1].size = sizeof (size);
-  VarDesc2 vd1g[2] = { { "tgt_ptr", 0 }, { "size", 0 } };
 
-  offload (__FILE__, __LINE__, device, "__offload_target_host2tgt_p1", 2,
-	   vd1, vd1g);
+  offload (__FILE__, __LINE__, device, "__offload_target_host2tgt_p1", 2, vd1,
+	   NULL);
 
   VarDesc vd2 = vd_host2tgt;
   vd2.ptr = (void *) host_ptr;
   vd2.size = size;
-  VarDesc2 vd2g = { "var", 0 };
 
-  offload (__FILE__, __LINE__, device, "__offload_target_host2tgt_p2", 1,
-	   &vd2, &vd2g);
+  offload (__FILE__, __LINE__, device, "__offload_target_host2tgt_p2", 1, &vd2,
+	   NULL);
 
   return tgt_ptr;
 }
@@ -452,7 +491,8 @@ extern "C" void *
 GOMP_OFFLOAD_dev2host (int device, void *host_ptr, const void *tgt_ptr,
 		       size_t size)
 {
-  TRACE ("(host_ptr = %p, tgt_ptr = %p, size = %d)", host_ptr, tgt_ptr, size);
+  TRACE ("(device = %d, host_ptr = %p, tgt_ptr = %p, size = %d)",
+	 device, host_ptr, tgt_ptr, size);
   if (!size)
     return host_ptr;
 
@@ -461,18 +501,16 @@ GOMP_OFFLOAD_dev2host (int device, void *host_ptr, const void *tgt_ptr,
   vd1[0].size = sizeof (void *);
   vd1[1].ptr = &size;
   vd1[1].size = sizeof (size);
-  VarDesc2 vd1g[2] = { { "tgt_ptr", 0 }, { "size", 0 } };
 
-  offload (__FILE__, __LINE__, device, "__offload_target_tgt2host_p1", 2,
-	   vd1, vd1g);
+  offload (__FILE__, __LINE__, device, "__offload_target_tgt2host_p1", 2, vd1,
+	   NULL);
 
   VarDesc vd2 = vd_tgt2host;
   vd2.ptr = (void *) host_ptr;
   vd2.size = size;
-  VarDesc2 vd2g = { "var", 0 };
 
-  offload (__FILE__, __LINE__, device, "__offload_target_tgt2host_p2", 1,
-	   &vd2, &vd2g);
+  offload (__FILE__, __LINE__, device, "__offload_target_tgt2host_p2", 1, &vd2,
+	   NULL);
 
   return host_ptr;
 }
@@ -481,36 +519,45 @@ extern "C" void *
 GOMP_OFFLOAD_dev2dev (int device, void *dst_ptr, const void *src_ptr,
 		      size_t size)
 {
-  TRACE ("(dst_ptr = %p, src_ptr = %p, size = %d)", dst_ptr, src_ptr, size);
+  TRACE ("(device = %d, dst_ptr = %p, src_ptr = %p, size = %d)",
+	 device, dst_ptr, src_ptr, size);
   if (!size)
     return dst_ptr;
 
-  VarDesc vd1[3] = { vd_host2tgt, vd_host2tgt, vd_host2tgt };
-  vd1[0].ptr = &dst_ptr;
-  vd1[0].size = sizeof (void *);
-  vd1[1].ptr = &src_ptr;
-  vd1[1].size = sizeof (void *);
-  vd1[2].ptr = &size;
-  vd1[2].size = sizeof (size);
-  VarDesc2 vd1g[3] = { { "dst_ptr", 0 }, { "src_ptr", 0 }, { "size", 0 } };
+  VarDesc vd[3] = { vd_host2tgt, vd_host2tgt, vd_host2tgt };
+  vd[0].ptr = &dst_ptr;
+  vd[0].size = sizeof (void *);
+  vd[1].ptr = &src_ptr;
+  vd[1].size = sizeof (void *);
+  vd[2].ptr = &size;
+  vd[2].size = sizeof (size);
 
-  offload (__FILE__, __LINE__, device, "__offload_target_tgt2tgt", 3, vd1,
-	   vd1g);
+  offload (__FILE__, __LINE__, device, "__offload_target_tgt2tgt", 3, vd, NULL);
 
   return dst_ptr;
 }
 
 extern "C" void
-GOMP_OFFLOAD_run (int device, void *tgt_fn, void *tgt_vars)
+GOMP_OFFLOAD_async_run (int device, void *tgt_fn, void *tgt_vars,
+			void *async_data)
 {
-  TRACE ("(tgt_fn = %p, tgt_vars = %p)", tgt_fn, tgt_vars);
+  TRACE ("(device = %d, tgt_fn = %p, tgt_vars = %p, async_data = %p)", device,
+	 tgt_fn, tgt_vars, async_data);
 
-  VarDesc vd1[2] = { vd_host2tgt, vd_host2tgt };
-  vd1[0].ptr = &tgt_fn;
-  vd1[0].size = sizeof (void *);
-  vd1[1].ptr = &tgt_vars;
-  vd1[1].size = sizeof (void *);
-  VarDesc2 vd1g[2] = { { "tgt_fn", 0 }, { "tgt_vars", 0 } };
+  VarDesc vd[2] = { vd_host2tgt, vd_host2tgt };
+  vd[0].ptr = &tgt_fn;
+  vd[0].size = sizeof (void *);
+  vd[1].ptr = &tgt_vars;
+  vd[1].size = sizeof (void *);
+
+  offload (__FILE__, __LINE__, device, "__offload_target_run", 2, vd,
+	   (const void **) async_data);
+}
+
+extern "C" void
+GOMP_OFFLOAD_run (int device, void *tgt_fn, void *tgt_vars)
+{
+  TRACE ("(device = %d, tgt_fn = %p, tgt_vars = %p)", device, tgt_fn, tgt_vars);
 
-  offload (__FILE__, __LINE__, device, "__offload_target_run", 2, vd1, vd1g);
+  GOMP_OFFLOAD_async_run (device, tgt_fn, tgt_vars, NULL);
 }
diff --git a/liboffloadmic/plugin/offload_target_main.cpp b/liboffloadmic/plugin/offload_target_main.cpp
index 18b014641b6..58438107684 100644
--- a/liboffloadmic/plugin/offload_target_main.cpp
+++ b/liboffloadmic/plugin/offload_target_main.cpp
@@ -1,6 +1,6 @@
 /* Plugin for offload execution on Intel MIC devices.
 
-   Copyright (C) 2014 Free Software Foundation, Inc.
+   Copyright (C) 2014-2015 Free Software Foundation, Inc.
 
    Contributed by Ilya Verbin <ilya.verbin@intel.com>.
 
@@ -139,14 +139,13 @@ __offload_target_table_p1 (OFFLOAD ofldt)
   int num_vars = (var_table_end - var_table_begin) / 2;
   TRACE ("(num_funcs = %d, num_vars = %d)", num_funcs, num_vars);
 
-  VarDesc vd1[2] = { vd_tgt2host, vd_tgt2host };
-  vd1[0].ptr = &num_funcs;
-  vd1[0].size = sizeof (num_funcs);
-  vd1[1].ptr = &num_vars;
-  vd1[1].size = sizeof (num_vars);
-  VarDesc2 vd2[2] = { { "num_funcs", 0 }, { "num_vars", 0 } };
+  VarDesc vd[2] = { vd_tgt2host, vd_tgt2host };
+  vd[0].ptr = &num_funcs;
+  vd[0].size = sizeof (num_funcs);
+  vd[1].ptr = &num_vars;
+  vd[1].size = sizeof (num_vars);
 
-  __offload_target_enter (ofldt, 2, vd1, vd2);
+  __offload_target_enter (ofldt, 2, vd, NULL);
   __offload_target_leave (ofldt);
 }
 
@@ -166,13 +165,11 @@ __offload_target_table_p2 (OFFLOAD ofldt)
   void **table = (void **) malloc (table_size);
   TRACE ("(table_size = %d)", table_size);
 
-  VarDesc vd1;
-  vd1 = vd_tgt2host;
-  vd1.ptr = table;
-  vd1.size = table_size;
-  VarDesc2 vd2 = { "table", 0 };
+  VarDesc vd = vd_tgt2host;
+  vd.ptr = table;
+  vd.size = table_size;
 
-  __offload_target_enter (ofldt, 1, &vd1, &vd2);
+  __offload_target_enter (ofldt, 1, &vd, NULL);
 
   void **p;
   int i = 0;
@@ -193,14 +190,13 @@ __offload_target_alloc (OFFLOAD ofldt)
   size_t size = 0;
   void *ptr = NULL;
 
-  VarDesc vd1[2] = { vd_host2tgt, vd_tgt2host };
-  vd1[0].ptr = &size;
-  vd1[0].size = sizeof (size);
-  vd1[1].ptr = &ptr;
-  vd1[1].size = sizeof (void *);
-  VarDesc2 vd2[2] = { { "size", 0 }, { "ptr", 0 } };
+  VarDesc vd[2] = { vd_host2tgt, vd_tgt2host };
+  vd[0].ptr = &size;
+  vd[0].size = sizeof (size);
+  vd[1].ptr = &ptr;
+  vd[1].size = sizeof (void *);
 
-  __offload_target_enter (ofldt, 2, vd1, vd2);
+  __offload_target_enter (ofldt, 2, vd, NULL);
   ptr = malloc (size);
   TRACE ("(size = %d): ptr = %p", size, ptr);
   __offload_target_leave (ofldt);
@@ -212,12 +208,11 @@ __offload_target_free (OFFLOAD ofldt)
 {
   void *ptr = 0;
 
-  VarDesc vd1 = vd_host2tgt;
-  vd1.ptr = &ptr;
-  vd1.size = sizeof (void *);
-  VarDesc2 vd2 = { "ptr", 0 };
+  VarDesc vd = vd_host2tgt;
+  vd.ptr = &ptr;
+  vd.size = sizeof (void *);
 
-  __offload_target_enter (ofldt, 1, &vd1, &vd2);
+  __offload_target_enter (ofldt, 1, &vd, NULL);
   TRACE ("(ptr = %p)", ptr);
   free (ptr);
   __offload_target_leave (ofldt);
@@ -231,14 +226,13 @@ __offload_target_host2tgt_p1 (OFFLOAD ofldt)
   void *var_ptr = NULL;
   size_t var_size = 0;
 
-  VarDesc vd1[2] = { vd_host2tgt, vd_host2tgt };
-  vd1[0].ptr = &var_ptr;
-  vd1[0].size = sizeof (void *);
-  vd1[1].ptr = &var_size;
-  vd1[1].size = sizeof (var_size);
-  VarDesc2 vd2[2] = { { "var_ptr", 0 }, { "var_size", 0 } };
+  VarDesc vd[2] = { vd_host2tgt, vd_host2tgt };
+  vd[0].ptr = &var_ptr;
+  vd[0].size = sizeof (void *);
+  vd[1].ptr = &var_size;
+  vd[1].size = sizeof (var_size);
 
-  __offload_target_enter (ofldt, 2, vd1, vd2);
+  __offload_target_enter (ofldt, 2, vd, NULL);
   TRACE ("(var_ptr = %p, var_size = %d)", var_ptr, var_size);
   last_var_ptr = var_ptr;
   last_var_size = var_size;
@@ -252,12 +246,11 @@ __offload_target_host2tgt_p2 (OFFLOAD ofldt)
   TRACE ("(last_var_ptr = %p, last_var_size = %d)",
 	 last_var_ptr, last_var_size);
 
-  VarDesc vd1 = vd_host2tgt;
-  vd1.ptr = last_var_ptr;
-  vd1.size = last_var_size;
-  VarDesc2 vd2 = { "var", 0 };
+  VarDesc vd = vd_host2tgt;
+  vd.ptr = last_var_ptr;
+  vd.size = last_var_size;
 
-  __offload_target_enter (ofldt, 1, &vd1, &vd2);
+  __offload_target_enter (ofldt, 1, &vd, NULL);
   __offload_target_leave (ofldt);
 }
 
@@ -269,14 +262,13 @@ __offload_target_tgt2host_p1 (OFFLOAD ofldt)
   void *var_ptr = NULL;
   size_t var_size = 0;
 
-  VarDesc vd1[2] = { vd_host2tgt, vd_host2tgt };
-  vd1[0].ptr = &var_ptr;
-  vd1[0].size = sizeof (void *);
-  vd1[1].ptr = &var_size;
-  vd1[1].size = sizeof (var_size);
-  VarDesc2 vd2[2] = { { "var_ptr", 0 }, { "var_size", 0 } };
+  VarDesc vd[2] = { vd_host2tgt, vd_host2tgt };
+  vd[0].ptr = &var_ptr;
+  vd[0].size = sizeof (void *);
+  vd[1].ptr = &var_size;
+  vd[1].size = sizeof (var_size);
 
-  __offload_target_enter (ofldt, 2, vd1, vd2);
+  __offload_target_enter (ofldt, 2, vd, NULL);
   TRACE ("(var_ptr = %p, var_size = %d)", var_ptr, var_size);
   last_var_ptr = var_ptr;
   last_var_size = var_size;
@@ -290,12 +282,11 @@ __offload_target_tgt2host_p2 (OFFLOAD ofldt)
   TRACE ("(last_var_ptr = %p, last_var_size = %d)",
 	 last_var_ptr, last_var_size);
 
-  VarDesc vd1 = vd_tgt2host;
-  vd1.ptr = last_var_ptr;
-  vd1.size = last_var_size;
-  VarDesc2 vd2 = { "var", 0 };
+  VarDesc vd = vd_tgt2host;
+  vd.ptr = last_var_ptr;
+  vd.size = last_var_size;
 
-  __offload_target_enter (ofldt, 1, &vd1, &vd2);
+  __offload_target_enter (ofldt, 1, &vd, NULL);
   __offload_target_leave (ofldt);
 }
 
@@ -307,16 +298,15 @@ __offload_target_tgt2tgt (OFFLOAD ofldt)
   void *dst_ptr = NULL;
   size_t size = 0;
 
-  VarDesc vd1[3] = { vd_host2tgt, vd_host2tgt, vd_host2tgt };
-  vd1[0].ptr = &dst_ptr;
-  vd1[0].size = sizeof (void *);
-  vd1[1].ptr = &src_ptr;
-  vd1[1].size = sizeof (void *);
-  vd1[2].ptr = &size;
-  vd1[2].size = sizeof (size);
-  VarDesc2 vd1g[3] = { { "dst_ptr", 0 }, { "src_ptr", 0 }, { "size", 0 } };
+  VarDesc vd[3] = { vd_host2tgt, vd_host2tgt, vd_host2tgt };
+  vd[0].ptr = &dst_ptr;
+  vd[0].size = sizeof (void *);
+  vd[1].ptr = &src_ptr;
+  vd[1].size = sizeof (void *);
+  vd[2].ptr = &size;
+  vd[2].size = sizeof (size);
 
-  __offload_target_enter (ofldt, 3, vd1, vd1g);
+  __offload_target_enter (ofldt, 3, vd, NULL);
   TRACE ("(dst_ptr = %p, src_ptr = %p, size = %d)", dst_ptr, src_ptr, size);
   memcpy (dst_ptr, src_ptr, size);
   __offload_target_leave (ofldt);
@@ -329,14 +319,13 @@ __offload_target_run (OFFLOAD ofldt)
   void *fn_ptr;
   void *vars_ptr;
 
-  VarDesc vd1[2] = { vd_host2tgt, vd_host2tgt };
-  vd1[0].ptr = &fn_ptr;
-  vd1[0].size = sizeof (void *);
-  vd1[1].ptr = &vars_ptr;
-  vd1[1].size = sizeof (void *);
-  VarDesc2 vd2[2] = { { "fn_ptr", 0 }, { "vars_ptr", 0 } };
+  VarDesc vd[2] = { vd_host2tgt, vd_host2tgt };
+  vd[0].ptr = &fn_ptr;
+  vd[0].size = sizeof (void *);
+  vd[1].ptr = &vars_ptr;
+  vd[1].size = sizeof (void *);
 
-  __offload_target_enter (ofldt, 2, vd1, vd2);
+  __offload_target_enter (ofldt, 2, vd, NULL);
   TRACE ("(fn_ptr = %p, vars_ptr = %p)", fn_ptr, vars_ptr);
   void (*fn)(void *) = (void (*)(void *)) fn_ptr;
   fn (vars_ptr);
diff --git a/liboffloadmic/runtime/offload_host.cpp b/liboffloadmic/runtime/offload_host.cpp
index c10dc4efb55..e52019dfb28 100644
--- a/liboffloadmic/runtime/offload_host.cpp
+++ b/liboffloadmic/runtime/offload_host.cpp
@@ -64,6 +64,8 @@ static void __offload_fini_library(void);
 #define GET_OFFLOAD_NUMBER(timer_data) \
     timer_data? timer_data->offload_number : 0
 
+static void (*task_completion_callback)(void *);
+
 extern "C" {
 #ifdef TARGET_WINNT
 // Windows does not support imports from libraries without actually
@@ -2507,7 +2509,7 @@ extern "C" {
         const void *info
     )
     {
-	/* TODO: Call callback function, pass info.  */
+	task_completion_callback ((void *) info);
     }
 }
 
@@ -5669,6 +5671,11 @@ extern "C" void __offload_unregister_image(const void *target_image)
     }
 }
 
+extern "C" void __offload_register_task_callback(void (*cb)(void *))
+{
+    task_completion_callback = cb;
+}
+
 // Runtime trace interface for user programs
 
 void __offload_console_trace(int level)
diff --git a/liboffloadmic/runtime/offload_host.h b/liboffloadmic/runtime/offload_host.h
index afd5c9936ca..2a43fd6a68a 100644
--- a/liboffloadmic/runtime/offload_host.h
+++ b/liboffloadmic/runtime/offload_host.h
@@ -376,6 +376,9 @@ extern "C" bool __offload_target_image_is_executable(const void *target_image);
 extern "C" bool __offload_register_image(const void* image);
 extern "C" void __offload_unregister_image(const void* image);
 
+// Registers asynchronous task completion callback
+extern "C" void __offload_register_task_callback(void (*cb)(void *));
+
 // Initializes offload runtime library.
 DLL_LOCAL extern int __offload_init_library(void);