2 \defgroup SimGrid_API SimGrid modules */
4 /** \defgroup XBT_API XBT
6 \brief The core toolbox of SimGrid, containing usefull datatypes,
7 portability support and so on.
11 /** \defgroup MSG_API MSG
13 \brief Simple programming environment
15 MSG was the first distributed programming environment provided within
16 SimGrid. While almost realistic, it remains quite simple (simplistic?).
18 \section MSG_who Who should use this (and who shouldn't)
20 You should use this module if you want to study some heuristics for a
21 given problem you don't really want to implement.
22 If you want to use DAGs, have a look at the \ref SD_API programming
24 If you want to get a real (but experimental) implementation of your solution, have a look
25 at the \ref GRAS_API one. If you want to study an existing MPI program,
26 have a look at the \ref SMPI_API one. If none of those programming
27 environments fits your needs, you may consider implementing your own
28 directly on top of \ref SURF_API (but you probably want to contact us
33 /** \defgroup SIMIX_API SIMIX
35 \brief POSIX-like interface for building simulation
37 This is a developer-level interface that should be useful only if you
38 plan to design a new interface for SimGrid.
42 /** \defgroup GRAS_API GRAS
44 \brief Realistic programming environment (Grid Reality And Simulation)
46 GRAS provides a complete API to implement distributed application on top
47 of heterogeneous plateforms. In addition to the SimGrid implementation
48 of this interface (allowing you to work on your application within the
49 comfort of the simulator), an implementation suited to real platforms is
50 also provided (allowing you to really use your application once you're
51 done with developing it). It may still contain rought corners as
52 GRAS is not the most used part of SimGrid, however.
54 GRAS thus constitute a complete grid application developement framework,
55 encompassing both developer helping tools (the simulator and associated
56 tools) and an efficient while portable execution runtime.
58 \section GRAS_who Who should use this (and who shouldn't)
60 You should use this programming environment if you want to develop real
61 applications, ie if the final result of your work is a program which
62 may eventually be distributed. Rember however that GRAS is
63 considered as experimental at this point. Help would be welcomed
64 to improve this sorry situation...
66 If you just want to study some heuristics for a given problem you don't
67 want to implement really (ie, if your result would be a theorem), have a
68 look at the \ref MSG_API one, or the \ref SD_API one if you need to use DAGs.
69 If you want to study an existing MPI program, have a look at the
71 If none of those programming environments fits your needs, you may
72 consider implementing your own directly on top of \ref SURF_API (but you
73 probably want to contact us before).
76 /** \defgroup AMOK_API AMOK
78 \brief Distributed toolkit built over \ref GRAS_API (Advanced Metacomputing Overlay Kit)
80 AMOK provides several tools useful to most applications built on top of GRAS,
81 but yet not belonging to GRAS itself. It is planned that those modules will be
82 changed to real plugins one day, allowing users to load only the needed parts at
83 run time. For now, they live in another library against which you should link your
87 /** \defgroup SMPI_API SMPI
89 \brief Programming environment for the simulation of MPI applications
91 This programming environment permits to study existing MPI application
92 by emulating them on top of the SimGrid simulator. In other words, it
93 will constitute an emulation solution for parallel codes. You don't
94 even have to modify your code for that, although that may help, as
97 \section SMPI_who Who should use SMPI (and who shouldn't)
99 You should use this programming environment of the SimGrid suite if
100 you want to study existing MPI applications. If you want to create a
101 distributed application, you may be interested in the \ref GRAS_API
102 environment instead (but note that GRAS is not very actively
103 maintained nowadays). If you want to study some heuristics for a given
104 problem (and if your goal is to produce theorems and publications, not
105 code), have a look at the \ref MSG_API environment, or the \ref SD_API
106 one if you need to use DAGs. If none of those programming environments
107 fits your needs, you may consider implementing your own directly on
108 top of \ref SURF_API (but you probably want to contact us before).
110 \section SMPI_what What can run within SMPI?
112 You can run unmodified MPI applications (both C and Fortran) within
113 SMPI, provided you only use MPI calls that we implemented in MPI. Our
114 coverage of the interface is not bad, but will probably never be
115 complete. One sided communications and I/O primitives are not targeted
116 for now. The full list of not yet implemented functions is available
117 in file <tt>include/smpi/smpi.h</tt> of the archive, between two lines
118 containing the <tt>FIXME</tt> marker. If you really need a missing
119 feature, please get in touch with us: we can guide you though the
120 SimGrid code to help you implementing it, and we'd glad to integrate
121 it in the main project afterward if you contribute them back.
123 \section SMPI_adapting Adapting your MPI code to the use of SMPI
125 As detailed in the reference article (available at
126 http://hal.inria.fr/inria-00527150), you may want to adapt your code
127 to improve the simulation performance. But these tricks may seriously
128 hinder the result qualtity (or even prevent the app to run) if used
129 wrongly. We assume that if you want to simulate an HPC application,
130 you know what you are doing. Don't prove us wrong!
132 If you get short on memory (the whole app is executed on a single node
133 when simulated), you should have a look at the SMPI_SHARED_MALLOC and
134 SMPI_SHARED_FREE macros. It allows to share memory areas between
135 processes. For example, matrix multiplication code may want to store
136 the blocks on the same area. Of course, the resulting computations
137 will useless, but you can still study the application behavior this
138 way. Of course, if your code is data-dependent, this won't work.
140 If your application is too slow, try using SMPI_SAMPLE_LOCAL,
141 SMPI_SAMPLE_GLOBAL and friends to indicate which computation loops can
142 be sampled. Some of the loop iterations will be executed to measure
143 their duration, and this duration will be used for the subsequent
144 iterations. These samples are done per processor with
145 SMPI_SAMPLE_LOCAL, and shared between all processors with
146 SMPI_SAMPLE_GLOBAL. Of course, none of this will work if the execution
147 time of your loop iteration are not stable.
149 Yes, that's right, these macros are not documented yet, but we'll fix
150 it as soon as time permits. Sorry about that -- patch welcomed!
151 Meanwhile, grep for them on the examples for more information.
153 \section SMPI_compiling Compiling your code
155 This is very simply done with the <tt>smpicc</tt> script. If you
156 already compiled any MPI code before, you already know how to use it.
157 If not, you should try to get your MPI code running on top of MPI
158 before giving SMPI a spin. Actually, that's very simple even if it's
159 the first time you use MPI code: just use smpicc as a compiler (in
160 replacement of gcc or your usual compiler), and you're set.
162 \section SMPI_executing Executing your code on top of the simulator
164 This is done though the <tt>smpirun</tt> script as follows.
165 <tt>my_hostfile.txt</tt> is a classical MPI hostfile (that is, this
166 file lists the machines on which the processes must be dispatched, one
167 per line) <tt>my_platform.xml</tt> is a classical SimGrid platform
168 file. Of course, the hosts of the hostfile must exist in the provided
169 platform. <tt>./program</tt> is the MPI program that you want to
170 simulate (must be compiled by <tt>smpicc</tt>) while <tt>-arg</tt> is
171 a command-line parameter passed to this program.
174 smpirun -hostfile my_hostfile.txt -platform my_platform.xml ./program -arg
177 smpirun accepts other parameters, such as <tt>-np</tt> if you don't
178 want to use all the hosts defined in the hostfile, <tt>-map</tt> to
179 display on which host each rank gets mapped of <tt>-trace</tt> to
180 activate the tracing during the simulation. You can get the full list
190 /** \defgroup SD_API SimDag
192 \brief Programming environment for DAG applications
194 SimDag provides some functionnalities to simulate parallel task scheduling
195 with DAGs models (Direct Acyclic Graphs).
196 The old versions of SimGrid were based on DAGs. But the DAG part (named SG)
197 was removed in SimGrid 3 because the new kernel (\ref SURF_API) was implemented. \ref SURF_API
198 was much faster and more flexible than SG and did not use DAGs.
199 SimDag is a new implementation of DAGs handling and it is built on top of \ref SURF_API.
201 \section SD_who Who should use this (and who shouldn't)
203 You should use this programming environment of the SimGrid suite if you want
204 to study algorithms and heuristics with DAGs of parallel tasks.
205 If you don't need to use DAGs for your simulation, have a look at the
206 \ref MSG_API programming environment.
207 If you want to implement a real distributed application, have a look at the
208 \ref GRAS_API programming environment.
209 If you want to study an existing MPI program, have a look at the
211 If none of those programming environments fits your needs, you may
212 consider implementing your own directly on top of \ref SURF_API (but you
213 probably want to contact us before).
218 @defgroup SURF_API SURF
220 @brief Internal kernel of all the simulators used in SimGrid, and associated models.
222 SURF provides the core functionnalities to simulate a virtual
223 platform. It is very low-level and is not intended to be used by end
224 users, but rather to serve as a basis for higher-level simulators. Its
225 interface are not frozen (and will probably never be), and the
226 structure emphasis on performance over ease of use. This module
227 contains the platform models. If you need a model that is not encoded
228 yet, please come to the devel mailing list so that we can discuss on
229 the feasibility of your idea.
231 Please note that as it is not really intended for public use, this
232 module is only partially documented.
237 @defgroup TRACE_API TRACE
239 @brief Tracing mechanism and its functions.
241 SimGrid can trace the resource (of hosts and links) utilization using
242 any of its programming interfaces (MSG, SimDAG and SMPI). This means
243 that the tracing will register how much power is used for each host
244 and how much bandwidth is used for each link of the platform.
246 The idea of the tracing facilities is to give SimGrid users to
247 possibility to classify MSG and SimDAG tasks by category, tracing the
248 platform utilization (hosts and links) for each of the categories.
249 The API enables the declaration of categories and a function to
250 associate them to the tasks (MSG and SD). The tasks that are not
251 classified according to a category are not traced. If no categories
252 are specified, simulations can still be traced using a special
253 parameter in the command line (see \ref tracing_tracing for details).