Merge branch 'development' into embb517_mutex_based_atomics

8633a17d · Marcus Winter · c015aa4c · 560a935d · 8633a17d · 8633a17d
Commit 8633a17d authored May 17, 2016 by Marcus Winter
8 changed files
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
 Embedded Multicore Building Blocks (EMB²)
 =========================================
+Version 0.4.0
+-------------
+### Features:
+- Added consistency checking functionality to dataflow
+- Added C++ wrapper for base_c logging functions
+- Reworked dataflow_cpp interface for easier usage
+### Changes and improvements:
+- Improved network plugin with better error checking and task cancellation support
+- Revised dataflow_cpp so that network token count can now be set at runtime
+- Added automatic determination of token count to dataflow_cpp
+- Added checks for NULL pointers in C interface functions
+- Extended mtapi_info_t to conform to the standard
+- Fixed CodeSonar warnings across the code base
+- Changed spinlock implementation to yield every 1024 spins now
+- Changed asserts on interface visible parameters to execeptions
+- Enabled reuse of main thread, which is configurable via node attributes
+### Bug fixes:
+- Fixed problem causing low performance on the Jetson TK1 board
+- Fixed bug in mtapi_c causing a task wait to hang
+- Fixes issue with the AMD APP SDK in the OpenCL plugin
+- Fixed problem with automatic initialization in tasks_cpp
+- Fixed memory leaks in tests
+### Build system:
+- Removed dependency on an installed OpenCL SDK
+- Moved all MTAPI plugins into folder mtapi_plugins_c
+- Resolved MSVC warnings and build problem
+### Documentation:
+- Updated README to reflect new directory structure
+- Updated Doxygen documentation of dataflow_cpp for the improved interface
+- Updated tutorial and examples to show the usage of dataflow_cpp
 Version 0.3.2
 -------------
@@ -63,34 +101,32 @@ Version 0.3.0
 -------------
 ### Features:
- mtapi_c:
+* MTAPI:
-    - Implemented action plugin API
+  - Implemented action plugin API
-    - Implemented load balancing for distributed/heterogeneous systems
+  - Implemented load balancing for distributed/heterogeneous systems
-    - Implemented OpenCL action plugin
+  - Implemented OpenCL action plugin
-    - Implemented network action plugin
+  - Implemented network action plugin
- mtapi_cpp:
+  - Added support for distributed/heterogeneous systems
-    - Added support for distributed/heterogeneous systems
 ### Changes and improvements:
- mtapi_c:
+* Base:
-    - Added multi-instance task support and test
-    - Improved notification of worker threads
- mtapi_cpp:
-    - Moved interface for homogeneous systems to tasks_cpp
- base_cpp:
    - Moved tick types to internal namespace and added duration typedefs
- dataflow_cpp:
+* MTAPI:
-    - Removed spinlocks
+  - Added multi-instance task support and test
-    - Simplified registration of processes (only sources need to be added)
+  - Improved notification of worker threads
-    - Increased number of task queues in unit test
+  - Moved interface for homogeneous systems to tasks_cpp
-    - Added assertion in unit test
+* Algorithms:
-    - Improved exception handling
+  - Restricted partitioners to random access iterators
-    - Removed stray include
+  - Added unit tests for partitioners on large ranges
-    - Refactored to use tasks_cpp
+  - Refactored to use tasks_cpp
- algorithms_cpp:
+* Dataflow:
-    - Restricted partitioners to random access iterators
+  - Removed spinlocks
-    - Added unit tests for partitioners on large ranges
+  - Simplified registration of processes (only sources need to be added)
-    - Refactored to use tasks_cpp
+  - Increased number of task queues in unit test
+  - Added assertion in unit test
+  - Improved exception handling
+  - Removed stray include
+  - Refactored to use tasks_cpp
 ### Bug fixes:
 - Fixed unit test for dataflow_cpp

--- a/CMakeLists.txt
+++ b/CMakeLists.txt
 # Copyright (c) 2014-2016, Siemens AG. All rights reserved.
-# 
+#
 # Redistribution and use in source and binary forms, with or without
 # modification, are permitted provided that the following conditions are met:
-# 
+#
 # 1. Redistributions of source code must retain the above copyright notice,
 # this list of conditions and the following disclaimer.
-# 
+#
 # 2. Redistributions in binary form must reproduce the above copyright notice,
 # this list of conditions and the following disclaimer in the documentation
 # and/or other materials provided with the distribution.
-# 
+#
 # THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
 # AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 # IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
@@ -27,8 +27,8 @@ cmake_minimum_required (VERSION 2.8.9)
 # Version number
 set (EMBB_BASE_VERSION_MAJOR 0)
-set (EMBB_BASE_VERSION_MINOR 3)
+set (EMBB_BASE_VERSION_MINOR 4)
-set (EMBB_BASE_VERSION_PATCH 2)
+set (EMBB_BASE_VERSION_PATCH 0)
 # Fix compilation for CMake versions >= 3.1
 #

--- a/base_cpp/include/embb/base/mutex.h
+++ b/base_cpp/include/embb/base/mutex.h
@@ -44,7 +44,6 @@ namespace base {
 * \{
 * \par Description
 *
- 12345678901234567890123456789012345678901234567890123456789012345678901234567890
 * The mutex concept is used for thread synchronization and provides a lock.
 * At any point in time, only one thread can exclusively hold the lock and
 * the lock is held until the thread explicitly releases it.

--- a/mtapi_c/include/embb/mtapi/c/mtapi.h
+++ b/mtapi_c/include/embb/mtapi/c/mtapi.h
@@ -552,8 +552,9 @@ enum mtapi_node_attributes_enum {
                                            the node */
  MTAPI_NODE_MAX_ACTIONS_PER_JOB,      /**< maximum number of actions in a job
                                            allowed by the node */
-  MTAPI_NODE_MAX_PRIORITIES            /**< maximum number of priorities
+  MTAPI_NODE_MAX_PRIORITIES,           /**< maximum number of priorities
                                            allowed by the node */
+  MTAPI_NODE_REUSE_MAIN_THREAD         /**< reuse main thread as worker */
 };
 /** size of the \a MTAPI_NODE_CORE_AFFINITY attribute */
 #define MTAPI_NODE_CORE_AFFINITY_SIZE sizeof(embb_core_set_t)
@@ -577,6 +578,8 @@ enum mtapi_node_attributes_enum {
 #define MTAPI_NODE_MAX_ACTIONS_PER_JOB_SIZE sizeof(mtapi_uint_t)
 /** size of the \a MTAPI_NODE_MAX_PRIORITIES attribute */
 #define MTAPI_NODE_MAX_PRIORITIES_SIZE sizeof(mtapi_uint_t)
+/** size of the \a MTAPI_NODE_REUSE_MAIN_THREAD attribute */
+#define MTAPI_NODE_REUSE_MAIN_THREAD_SIZE sizeof(mtapi_boolean_t)
 /* example attribute value */
 #define MTAPI_NODE_TYPE_SMP 1
@@ -688,6 +691,8 @@ struct mtapi_node_attributes_struct {
  mtapi_uint_t max_actions_per_job;    /**< stores
                                            MTAPI_NODE_MAX_ACTIONS_PER_JOB */
  mtapi_uint_t max_priorities;         /**< stores MTAPI_NODE_MAX_PRIORITIES */
+  mtapi_boolean_t reuse_main_thread;   /**< stores
+                                            MTAPI_NODE_REUSE_MAIN_THREAD */
 };
 /**

--- a/mtapi_c/src/embb_mtapi_thread_context_t.c
+++ b/mtapi_c/src/embb_mtapi_thread_context_t.c
@@ -54,6 +54,7 @@ mtapi_boolean_t embb_mtapi_thread_context_initialize_with_node_worker_and_core(
  that->core_num = core_num;
  that->priorities = node->attributes.max_priorities;
  that->is_initialized = MTAPI_FALSE;
+  that->is_main_thread = (worker_index == 0) ? node->attributes.reuse_main_thread : MTAPI_FALSE;
  embb_atomic_store_int(&that->run, 0);
  that->queue = (embb_mtapi_task_queue_t**)embb_mtapi_alloc_allocate(
@@ -121,17 +122,29 @@ mtapi_boolean_t embb_mtapi_thread_context_start(
  embb_core_set_add(&core_set, that->core_num);
  /* create thread */
-  err = embb_thread_create(&that->thread, &core_set, worker_func, that);
+  if (that->is_main_thread) {
-  if (EMBB_SUCCESS != err) {
+    /* reuse main thread */
-    embb_mtapi_log_error(
+    that->thread = embb_thread_current();
-      "embb_mtapi_ThreadContext_initializeWithNodeAndCoreNumber() could not "
+    err = embb_tss_create(&that->tss_id);
-      "create thread %d on core %d\n", that->worker_index, that->core_num);
+    if (EMBB_SUCCESS != err) {
-    return MTAPI_FALSE;
+      /* report error to scheduler */
-  }
+      embb_atomic_store_int(&that->run, -1);
+      return MTAPI_FALSE;
-  /* wait for worker to come up */
+    }
-  while (0 == embb_atomic_load_int(&that->run)) {
+    embb_tss_set(&(that->tss_id), that);
-    embb_thread_yield();
+    embb_atomic_store_int(&that->run, 1);
+  } else {
+    err = embb_thread_create(&that->thread, &core_set, worker_func, that);
+    if (EMBB_SUCCESS != err) {
+      embb_mtapi_log_error(
+        "embb_mtapi_ThreadContext_initializeWithNodeAndCoreNumber() could not "
+        "create thread %d on core %d\n", that->worker_index, that->core_num);
+      return MTAPI_FALSE;
+    }
+    /* wait for worker to come up */
+    while (0 == embb_atomic_load_int(&that->run)) {
+      embb_thread_yield();
+    }
  }
  if (0 < embb_atomic_load_int(&that->run)) {
@@ -146,7 +159,9 @@ void embb_mtapi_thread_context_stop(embb_mtapi_thread_context_t* that) {
  if (0 < embb_atomic_load_int(&that->run)) {
    embb_atomic_store_int(&that->run, 0);
    embb_condition_notify_one(&that->work_available);
-    embb_thread_join(&(that->thread), &result);
+    if (MTAPI_FALSE == that->is_main_thread) {
+      embb_thread_join(&(that->thread), &result);
+    }
  }
 }
@@ -158,6 +173,9 @@ void embb_mtapi_thread_context_finalize(embb_mtapi_thread_context_t* that) {
  embb_mtapi_log_trace("embb_mtapi_thread_context_finalize() called\n");
  if (that->is_initialized) {
+    if (that->is_main_thread) {
+      embb_tss_delete(&that->tss_id);
+    }
    embb_condition_destroy(&that->work_available);
    embb_mutex_destroy(&that->work_available_mutex);
  }

--- a/mtapi_c/src/embb_mtapi_thread_context_t.h
+++ b/mtapi_c/src/embb_mtapi_thread_context_t.h
@@ -68,6 +68,7 @@ struct embb_mtapi_thread_context_struct {
  embb_atomic_int run;
  mtapi_status_t status;
  mtapi_boolean_t is_initialized;
+  mtapi_boolean_t is_main_thread;
 };
 #include <embb_mtapi_thread_context_t_fwd.h>

--- a/mtapi_c/src/mtapi_node_attributes_t.c
+++ b/mtapi_c/src/mtapi_node_attributes_t.c
@@ -52,6 +52,7 @@ void mtapi_nodeattr_init(
    attributes->max_jobs = MTAPI_NODE_MAX_JOBS_DEFAULT;
    attributes->max_actions_per_job = MTAPI_NODE_MAX_ACTIONS_PER_JOB_DEFAULT;
    attributes->max_priorities = MTAPI_NODE_MAX_PRIORITIES_DEFAULT;
+    attributes->reuse_main_thread = MTAPI_FALSE;
    embb_core_set_init(&attributes->core_affinity, 1);
    attributes->num_cores = embb_core_set_count(&attributes->core_affinity);
@@ -143,6 +144,11 @@ void mtapi_nodeattr_set(
          &attributes->max_priorities, attribute, attribute_size);
        break;
+      case MTAPI_NODE_REUSE_MAIN_THREAD:
+        local_status = embb_mtapi_attr_set_mtapi_boolean_t(
+          &attributes->reuse_main_thread, attribute, attribute_size);
+        break;
      default:
        /* attribute unknown */
        local_status = MTAPI_ERR_ATTR_NUM;

--- a/mtapi_plugins_c/mtapi_network_c/test/embb_mtapi_network_test_socket.cc
+++ b/mtapi_plugins_c/mtapi_network_c/test/embb_mtapi_network_test_socket.cc
@@ -53,8 +53,12 @@ void NetworkSocketTest::TestBasic() {
  err = embb_mtapi_network_socket_initialize(&server_sock);
  PT_EXPECT(err != 0);
-  err = embb_mtapi_network_socket_bind_and_listen(
+  uint16_t port = 4700;
-    &server_sock, "127.0.0.1", 4711, 5);
+  do {
+    port++;
+    err = embb_mtapi_network_socket_bind_and_listen(
+      &server_sock, "127.0.0.1", port, 5);
+  } while (err == 0 && port < 4800);
  PT_EXPECT(err != 0);
  err = embb_mtapi_network_socket_select(&server_sock, 1, 1);
@@ -62,7 +66,7 @@ void NetworkSocketTest::TestBasic() {
  err = embb_mtapi_network_socket_initialize(&client_sock);
  PT_EXPECT(err != 0);
-  err = embb_mtapi_network_socket_connect(&client_sock, "127.0.0.1", 4711);
+  err = embb_mtapi_network_socket_connect(&client_sock, "127.0.0.1", port);
  PT_EXPECT(err != 0);
  err = embb_mtapi_network_socket_select(&server_sock, 1, -1);