Messaging refactor develop (#310)

* Copy the python file from depricated_message_infrastructure/*.py

* change multiprocessing.py to adapt interface

* fix cpplint error

* Update cpplib side MultiProcessing:
1. message_infrastructure_logging.h: logging for development
2. abstract_actor.h&posix_actor.cc: achieve stop function for signal
3. shm.h: Add SharedMemoryManager to cpplib
4. others: Update CMake, test and fix some type error'

fix pylint error

* Fix MultiProcessing Error and Cpplint Error

* Remove Actor interface for python

* Fix package path error

* Update API Interface
1. multiprocessing.py: Fix bug for sharedmemmanager name
2. remove lava package dependency for message_infrastructure
3. multiprocessing.cc: Add shmm and actors function
4. abstract_actor.h: Modify the posix actor and interface design
5. add and update test for multiprocessing

Signed-off-by: Miao, Hongda <[email protected]>

* Update import from message_infrastructure:
1. Move message_interface_enum.py to message_infrastructure
2. Update import syntax
3. Others: change multiprocessing init

Signed-off-by: Miao, Hongda <[email protected]>

* Update test_runtime.py import

Signed-off-by: Miao, Hongda <[email protected]>

* Update bash setenv.sh to add message_infrastructure

Signed-off-by: Miao, Hongda <[email protected]>

src/lava/magma/compiler/builders/channel_builder.py

@@ -15,7 +15,7 @@
     ChannelType,
 )
 from lava.magma.compiler.utils import PortInitializer
-from lava.magma.runtime.message_infrastructure \
+from message_infrastructure \
     .message_infrastructure_interface import (MessageInfrastructureInterface)
 
 if ty.TYPE_CHECKING:

src/lava/magma/compiler/channels/pypychannel.py

@@ -16,8 +16,8 @@
 )
 
 if ty.TYPE_CHECKING:
-    from lava.magma.runtime.message_infrastructure \
-        .message_infrastructure_interface import (
+    from message_infrastructure.message_infrastructure_interface \
+        import (
             MessageInfrastructureInterface)
 
 

src/lava/magma/core/process/process.py

@@ -11,7 +11,7 @@
 
 from lava.magma.core.process.interfaces import \
     AbstractProcessMember, IdGeneratorSingleton
-from lava.magma.core.process.message_interface_enum import ActorType
+from message_infrastructure.message_interface_enum import ActorType
 from lava.magma.core.process.ports.ports import \
     InPort, OutPort, RefPort, VarPort
 from lava.magma.core.process.variable import Var

src/lava/magma/runtime/message_infrastructure/CMakeLists.txt

@@ -4,7 +4,10 @@ project(message_passing)
 set (CMAKE_CXX_STANDARD 17)
 include_directories(message_infrastructure/csrc)
 
-file (GLOB MULTI_PROC_SRCS "message_infrastructure/csrc/multiprocessing.cc")
+set(MULTI_PROC_SRCS 
+    "message_infrastructure/csrc/multiprocessing.cc"
+    "message_infrastructure/csrc/posix_actor.cc"
+)
 
 file (GLOB PY_WRAPPER "message_infrastructure/csrc/message_infrastructure_py_wrapper.cc")
 

src/lava/magma/runtime/message_infrastructure/message_infrastructure/__init__.py

@@ -0,0 +1,4 @@
+from MessageInfrastructurePywrapper import CppMultiProcessing
+from MessageInfrastructurePywrapper import SharedMemManager
+from MessageInfrastructurePywrapper import Actor
+from MessageInfrastructurePywrapper import ProcessType

src/lava/magma/runtime/message_infrastructure/message_infrastructure/csrc/abstract_actor.h

@@ -5,31 +5,41 @@
 #ifndef ABSTRACT_ACTOR_H_
 #define ABSTRACT_ACTOR_H_
 
+#include <functional>
+
 namespace message_infrastructure {
 
+enum ActorStatus {
+  StatsError = -1,
+  StatsRuning = 0,
+  StatsStopped = 1
+};
+
 class AbstractActor {
  public:
   virtual int GetPid() = 0;
   virtual int Stop() = 0;
-  virtual int Pause() = 0;
-
   int pid_;
 };
 
 class PosixActor : public AbstractActor {
  public:
-  explicit PosixActor(int pid) {
+  explicit PosixActor(int pid, std::function<void()> target_fn) {
     this->pid_ = pid;
+    this->target_fn_ = target_fn;
   }
   int GetPid() {
     return this->pid_;
   }
-  int Stop() {
-    return 0;
-  }
-  int Pause() {
-    return 0;
+  int Wait();
+  int Stop();
+  int GetStatus() {
+    return this->status_;
   }
+  // int Trace();
+ private:
+  std::function<void()> target_fn_ = NULL;
+  int status_ = StatsStopped;
 };
 
 using ActorPtr = AbstractActor *;

src/lava/magma/runtime/message_infrastructure/message_infrastructure/csrc/message_infrastructure_logging.h

@@ -0,0 +1,28 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: BSD-3-Clause
+// See: https://spdx.org/licenses/
+
+#ifndef MESSAGE_INFRASTRUCTURE_LOGGING_H_
+#define MESSAGE_INFRASTRUCTURE_LOGGING_H_
+
+#include <stdio.h>
+#define LOG_MP (1)  // log for multiprocessing
+
+
+#define LAVA_LOG(_cond, _fmt, ...) { \
+  if ((_cond)) { \
+    printf("[CPP INFO]" _fmt, ## __VA_ARGS__); \
+  } \
+}
+
+#define LAVA_LOG_WARN(_cond, _fmt, ...) { \
+  if ((_cond)) { \
+    printf("[CPP WARNING]" _fmt, __FUNCTION__, ## __VA_ARGS__); \
+  } \
+}
+
+#define LAVA_LOG_ERR(_fmt, ...) { \
+  printf("[CPP ERROR]" _fmt, __FUNCTION__, ## __VA_ARGS__); \
+}
+
+#endif  // MESSAGE_INFRASTRUCTURE_LOGGING_H_

src/lava/magma/runtime/message_infrastructure/message_infrastructure/csrc/message_infrastructure_py_wrapper.cc

@@ -4,9 +4,11 @@
 
 #include <pybind11/functional.h>
 #include <pybind11/pybind11.h>
+#include <pybind11/stl.h>
 
 #include "multiprocessing.h"
-// #include "shm.h"
+#include "abstract_actor.h"
+#include "shm.h"
 // #include "shmem_channel.h"
 // #include "shmem_port.h"
 
@@ -15,11 +17,27 @@ namespace message_infrastructure {
 namespace py = pybind11;
 
 PYBIND11_MODULE(MessageInfrastructurePywrapper, m) {
-  py::class_<MultiProcessing> (m, "MultiProcessing")
+  py::class_<MultiProcessing> (m, "CppMultiProcessing")
     .def(py::init<>())
     .def("build_actor", &MultiProcessing::BuildActor)
     .def("check_actor", &MultiProcessing::CheckActor)
+    .def("get_actors", &MultiProcessing::GetActors)
+    .def("get_shmm", &MultiProcessing::GetSharedMemManager)
     .def("stop", &MultiProcessing::Stop);
+  py::enum_<ProcessType> (m, "ProcessType")
+    .value("ErrorProcess", ErrorProcess)
+    .value("ChildProcess", ChildProcess)
+    .value("ParentProcess", ParentProcess);
+  py::class_<SharedMemManager> (m, "SharedMemManager")
+    .def(py::init<>())
+    .def("alloc_mem", &SharedMemManager::AllocSharedMemory)
+    .def("stop", &SharedMemManager::Stop);
+  py::class_<PosixActor> (m, "Actor")
+    .def("wait", &PosixActor::Wait)
+    .def("stop", &PosixActor::Stop)
+    .def("get_status", &PosixActor::GetStatus);
+    // .def("trace", &PosixActor::Trace);
+
   /*
   py::class_<ShmemSendPort> (m, "ShmemSendPort")
     .def(py::init<std::string, SharedMemory*, Proto*, size_t, sem_t*, sem_t*>())

src/lava/magma/runtime/message_infrastructure/message_infrastructure/csrc/multiprocessing.cc

@@ -3,43 +3,61 @@
 // See: https://spdx.org/licenses/
 
 #include "multiprocessing.h"
+#include "message_infrastructure_logging.h"
 
 #include <sys/wait.h>
 #include <unistd.h>
-#include <iostream>
 
 namespace message_infrastructure {
 
-#define CPP_INFO "[CPP_INFO] "
+MultiProcessing::MultiProcessing() {
+  shmm_ = new SharedMemManager();
+}
 
-void MultiProcessing::BuildActor(std::function<void()> target_fn) {
+int MultiProcessing::BuildActor(std::function<void()> target_fn) {
   pid_t pid = fork();
 
-  if (pid < 0) {
-    std::cout << CPP_INFO << "cannot allocate pid\n";
-  }
-
   if (pid > 0) {
-    std::cout << CPP_INFO << "parent, create actor\n";
-    ActorPtr actor = new PosixActor(pid);
+    LAVA_LOG(LOG_MP, "Parent Process, create child process %d\n", pid);
+    ActorPtr actor = new PosixActor(pid, target_fn);
     actors_.push_back(actor);
+    return ParentProcess;
   }
 
   if (pid == 0) {
-    std::cout << CPP_INFO << "child, new process\n";
+    LAVA_LOG(LOG_MP, "child, new process\n");
     target_fn();
     exit(0);
   }
 
+  LAVA_LOG_ERR("Cannot allocate new pid for the process\n");
+  return ErrorProcess;
+
 }
 
-void MultiProcessing::Stop() {
+int MultiProcessing::Stop() {
+  int error_cnts = 0;
+
+  for (auto actor : actors_) {
+    error_cnts += actor->Stop();
+  }
+
+  LAVA_LOG(LOG_MP, "Stop Actors, error: %d\n", error_cnts);
+  return error_cnts;
 }
 
 void MultiProcessing::CheckActor() {
-  for(auto actor : actors_){
-    std::cout << CPP_INFO << actor->pid_ << std::endl;
+  for (auto actor : actors_) {
+    LAVA_LOG(LOG_MP, "Actor info %d\n", actor->GetPid());
   }
 }
 
+SharedMemManager* MultiProcessing::GetSharedMemManager() {
+  return this->shmm_;
+}
+
+std::vector<ActorPtr>& MultiProcessing::GetActors() {
+  return this->actors_;
+}
+
 } // namespace message_infrastructure

src/lava/magma/runtime/message_infrastructure/message_infrastructure/csrc/multiprocessing.h

@@ -13,15 +13,24 @@
 
 namespace message_infrastructure {
 
+enum ProcessType {
+  ErrorProcess = -1,
+  ChildProcess = 0,
+  ParentProcess = 1
+};
+
 class MultiProcessing {
  public:
-  // stop each actor in vector actors;
-  void Stop();
-  void BuildActor(std::function<void()>);
+  MultiProcessing();
+  int Stop();
+  int BuildActor(std::function<void()>);
   void CheckActor();
+  std::vector<ActorPtr>& GetActors();
+  SharedMemManager* GetSharedMemManager();
+
  private:
   std::vector<ActorPtr> actors_;
-  // SharedMemManager shmm_;
+  SharedMemManager *shmm_;
 };
 
 }  // namespace message_infrastructure

src/lava/magma/runtime/message_infrastructure/message_infrastructure/csrc/posix_actor.cc

@@ -0,0 +1,33 @@
+// Copyright (C) 2021 Intel Corporation
+// SPDX-License-Identifier: BSD-3-Clause
+// See: https://spdx.org/licenses/
+
+#include "abstract_actor.h"
+#include "message_infrastructure_logging.h"
+
+#include <sys/wait.h>
+#include <sys/types.h>
+
+namespace message_infrastructure {
+
+int PosixActor::Stop() {
+  int status;
+  // TODO: Add the options can be as args of the function
+  int options = 0;
+  int ret = waitpid(this->pid_, &status, options);
+
+  if (ret < 0) {
+    LAVA_LOG_ERR("process %d waitpid error\n", this->pid_);
+    return -1;
+  }
+
+  // Check the status
+  return 0;
+
+}
+
+int PosixActor::Wait() {
+  return 0;
+}
+
+}

src/lava/magma/runtime/message_infrastructure/message_infrastructure/csrc/shm.h

@@ -13,11 +13,48 @@
 #include <semaphore.h>
 #include <unistd.h>
 
+#include <vector>
+
 namespace message_infrastructure {
 
 class SharedMemory {
 };
 
+class SharedMemManager {
+ public:
+  int AllocSharedMemory(size_t mem_size) {
+    int shmid = shmget(key_, mem_size, 0644|IPC_CREAT);
+    if (shmid < 0)
+      return -1;
+
+    shms_.push_back(shmid);
+    key_++;
+    return shmid;
+  }
+
+  int DeleteSharedMemory(int shmid) {
+    // Release specific shared memory
+    int del_cnt = 0;
+    for (auto it = shms_.begin(); it != shms_.end(); it++) {
+      if ((*it) == shmid) {
+        shms_.erase(it);
+        del_cnt++;
+      }
+    }
+    return del_cnt;
+  }
+
+  int Stop() {
+    int stop_cnt = shms_.size();
+    shms_.clear();
+    return stop_cnt;
+  }
+
+ private:
+  key_t key_ = 0xdead;
+  std::vector<int> shms_;
+};
+
 }  // namespace message_infrastructure
 
 #endif  // SHM_H_

src/lava/magma/runtime/message_infrastructure/message_infrastructure/factory.py

@@ -0,0 +1,18 @@
+# Copyright (C) 2021-22 Intel Corporation
+# SPDX-License-Identifier: LGPL 2.1 or later
+# See: https://spdx.org/licenses/
+from message_infrastructure.message_interface_enum import ActorType
+from message_infrastructure.multiprocessing import MultiProcessing
+
+"""Factory class to create the messaging infrastructure"""
+
+
+class MessageInfrastructureFactory:
+    """Creates the message infrastructure instance based on type"""
+    @staticmethod
+    def create(factory_type: ActorType):
+        """type of actor framework being chosen"""
+        if factory_type == ActorType.MultiProcessing:
+            return MultiProcessing()
+        else:
+            raise Exception("Unsupported factory_type")