MSG comes with an extensive set of examples. It is sometimes difficult to find the one you need. This list aims at helping you finding the example from which you can learn what you want to. Basic examples and features =========================== * sendrecv/sendrecv.c: Ping-pong example. It's hard to think of a simpler example. The tesh files laying in the directory are instructive concerning the way to pass options to the simulators (as described in "Simgrid options and configurations"). * masterslave/masterslave_forwarder.c: Master/slaves example. This good old example is also very simple. Its basic version is fully commented in the web documentation, but several variants can be found in the same directory. * masterslave/masterslave_console.c: demonstrate how to use lua files instead of XML for the platform and deployment declaration using MSG_load_platform_script(). The most interesting part is probably not the C code, but rather the masterslave/masterslave_script.lua, which demonstrates how to express the platform and deployment in lua. * migration/migration.c Demonstrates how to use the MSG_process_migrate() function to let processes change the host they run on after their start. * suspend/suspend.c: Demonstrates how to suspend and resume processes using MSG_process_suspend() and MSG_process_resume(). * properties/msg_prop.c Attaching arbitrary informations to host, processes and such, and retrieving them with MSG_host_get_properties(), MSG_host_get_property_value(), MSG_process_get_properties() and MSG_process_get_property_value(). Also make sure to read the platform and deployment XML files to see how to declare these data. * parallel_task/parallel_task.c: Demonstrates the use of MSG_parallel_task_create(), to create special tasks that run on several hosts at the same time. The resulting simulations are very close to what can be achieved in SimDag, but still allows to use the other features of MSG (it'd be cool to be able to mix interfaces, but it's not possible ATM). * priority/priority.c: Demonstrates the use of MSG_task_set_priority() to change the computation priority of a given task. Asynchronous communications =========================== There is several examples of asynchronous communications coming in the archive. In addition to the fully documented example [50]Asynchronous communications, there is several other examples in the archive: * msg/icomms/peer.c: basic example of async functions (MSG_task_isend, MSG_task_irecv, MSG_comm_wait) * msg/icomms/peer2.c: demonstrates the MSG_comm_waitall() function * msg/icomms/peer3.c: demonstrates the MSG_comm_waitany() function Tracing and vizualization features ================================== * tracing/simple.c very simple program that creates, executes and destroy a task * tracing/ms.c TODO * tracing/categories.c example with the declaration of multiple categories * tracing/procmig.c example to trace process migration using the mask TRACE_PROCESS * tracing/trace_platform.c: Demonstrates how to trace the platform * tracing/user_variables.c: Demonstrates how to trace user-provided variables Models-related examples ======================= Packet level simulators ----------------------- These examples demonstrate how to use the bindings to classical Packet-Level Simulators (PLS), as explained in the relevant part of the web documentation. The most interesting is probably not the C files since they are unchanged from the other simulations, but the associated files, such as the platform files to see how to declare a platform to be used with the PLS bindings of SimGrid and the tesh files to see how to actually start a simulation in these settings. * ns3: Simple ping-pong using ns3 instead of the SimGrid models * gtnets Simple ping-pong using GTNeTs instead of the SimGrid models Other resource kinds -------------------- This section contains some sparse examples of how to use the other kind of resources, such as disk or GPU. These resources are quite experimental for now, but here we go anyway. * io/file.c Example with the disk resource * gpu/test_MSG_gpu_task_create.c Example of use of the very experimental (for now) GPU resource. Trace driven simulations ======================== The actions/actions.c example demonstrates how to run trace-driven simulations. It is very handy when you want to test an algorithm or protocol that does nothing unless it receives some events from outside. For example, a P2P protocol reacts to requests from the user, but does nothing if there is no such event. In such situations, SimGrid allows to write your protocol in your C file, and the events to react to in a separate text file. Declare a function handling each of the events that you want to accept in your trace files, register them using MSG_action_register in your main, and then use MSG_action_trace_run to launch the simulation. You can either have one trace file containing all your events, or a file per simulated process. Check the tesh files in the example directory for details on how to do it. This example uses this approach to replay MPI-like traces. It comes with a set of event handlers reproducing MPI events. This is somehow similar to SMPI, yet differently implemented. This code should probably be changed to use SMPI internals instead, but wasn't, so far. Examples of full applications ============================= * token_ring/ring_call.c: Classical token ring communication, where a token is exchanged along a ring to reach every participant. * pmm/msg_pmm.c: Parallel Matrix Multiplication is a little application. This is something that most MPI developper have written during their class, here implemented using MSG instead of MPI. * chord/chord.c: Classical Chord P2P protocol This example implements the well known Chord P2P protocol. Its main advantage is that it constitute a fully working non-trivial example. In addition, its implementation is rather efficient, as demonstrated in [57]http://hal.inria.fr/inria-00602216/