Simulation time: 1e3 seconds.
\endverbatim
+\subsection options_smpi_temps smpi/keep-temps: not cleaning up after simulation
+
+\b Default: 0 (false)
+
+Under some conditions, SMPI generates a lot of temporary files. They
+usually get cleaned, but you may use this option to not erase these
+files. This is for example useful when debugging or profiling
+executions using the dlopen privatization schema, as missing binary
+files tend to fool the debuggers.
+
\subsection options_model_smpi_lat_factor smpi/lat-factor: Latency factors
The motivation and syntax for this option is identical to the motivation/syntax
--cfg=smpi/papi-events:"default:PAPI_L3_LDM:PAPI_L2_LDM"
\endverbatim
-\subsection options_smpi_global smpi/privatize-global-variables: Automatic privatization of global variables
+\subsection options_smpi_privatization smpi/privatization: Automatic privatization of global variables
-MPI executables are meant to be executed in separated processes, but SMPI is
+MPI executables are usually meant to be executed in separated processes, but SMPI is
executed in only one process. Global variables from executables will be placed
-in the same memory zone and shared between processes, causing hard to find bugs.
-To avoid this, several options are possible :
- - Manual edition of the code, for example to add __thread keyword before data
- declaration, which allows the resulting code to work with SMPI, but only
- if the thread factory (see \ref options_virt_factory) is used, as global
- variables are then placed in the TLS (thread local storage) segment.
- - Source-to-source transformation, to add a level of indirection
- to the global variables. SMPI does this for F77 codes compiled with smpiff,
- and used to provide coccinelle scripts for C codes, which are not functional anymore.
- - Compilation pass, to have the compiler automatically put the data in
- an adapted zone.
- - Runtime automatic switching of the data segments. SMPI stores a copy of
- each global data segment for each process, and at each context switch replaces
- the actual data with its copy from the right process. This mechanism uses mmap,
- and is for now limited to systems supporting this functionnality (all Linux
- and some BSD should be compatible).
- Another limitation is that SMPI only accounts for global variables defined in
- the executable. If the processes use external global variables from dynamic
- libraries, they won't be switched correctly. To avoid this, using static
- linking is advised (but not with the simgrid library, to avoid replicating
- its own global variables).
-
- To use this runtime automatic switching, the variable \b smpi/privatize-global-variables
- should be set to yes
+in the same memory zone and shared between processes, causing intricate bugs.
+Several options are possible to avoid this, as described in the main
+<a href="https://hal.inria.fr/hal-01415484">SMPI publication</a>.
+SimGrid provides two ways of automatically privatizing the globals,
+and this option allows to choose between them.
+
+ - <b>no</b> (default): Do not automatically privatize variables.
+ - <b>mmap</b> or <b>yes</b>: Runtime automatic switching of the data segments.\n
+ SMPI stores a copy of each global data segment for each process,
+ and at each context switch replaces the actual data with its copy
+ from the right process. No copy actually occures as this mechanism
+ uses mmap for efficiency. As such, it is for now limited to
+ systems supporting this functionnality (all Linux and most BSD).\n
+ Another limitation is that SMPI only accounts for global variables
+ defined in the executable. If the processes use external global
+ variables from dynamic libraries, they won't be switched
+ correctly. The easiest way to solve this is to statically link
+ against the library with these globals (but you should never
+ statically link against the simgrid library itself).
+ - <b>dlopen</b>: Link multiple times against the binary.\n
+ SMPI loads several copy of the same binary in memory, resulting in
+ the natural duplication global variables. Since the dynamic linker
+ refuses to link the same file several times, the binary is copied
+ in a temporary file before being dl-loaded (it is erased right
+ after loading).\n
+ Note that this feature is somewhat experimental at time of writing
+ (v3.16) but seems to work.\n
+ This approach greatly speeds up the context switching, down to
+ about 40 CPU cycles with our raw contextes, instead of requesting
+ several syscalls with the \c mmap approach. Another advantage is
+ that it permits to run the SMPI contexts in parallel, which is
+ obviously not possible with the \c mmap approach.\n
+ Further work may be possible to alleviate the memory and disk
+ overconsumption. It seems that we could
+ <a href="https://lwn.net/Articles/415889/">punch holes</a>
+ in the files before dl-loading them to remove the code and
+ constants, and mmap these area onto a unique copy. This require
+ to understand the ELF layout of the file, but would
+ reduce the disk- and memory- usage to the bare minimum. In
+ addition, this would reduce the pressure on the CPU caches (in
+ particular on instruction one).
\warning
This configuration option cannot be set in your platform file. You can only
pass it as an argument to smpirun.
-
\subsection options_model_smpi_detached Simulating MPI detached send
This threshold specifies the size in bytes under which the send will return
The behavior and motivation for this configuration option is identical with \a smpi/test, see
Section \ref options_model_smpi_test for details.
+\subsection options_model_smpi_iprobe_cpu_usage smpi/iprobe-cpu-usage: Reduce speed for iprobe calls
+
+\b Default value: 1 (no change from default behavior)
+
+MPI_Iprobe calls can be heavily used in applications. To account correctly for the energy
+cores spend probing, it is necessary to reduce the load that these calls cause inside
+SimGrid.
+
+For instance, we measured a max power consumption of 220 W for a particular application but
+only 180 W while this application was probing. Hence, the correct factor that should
+be passed to this option would be 180/220 = 0.81.
+
\subsection options_model_smpi_init smpi/init: Inject constant times for calls to MPI_Init
\b Default value: 0
\subsection options_model_smpi_shared_malloc smpi/shared-malloc: Factorize malloc()s
-\b Default: yes
+\b Default: global
If your simulation consumes too much memory, you may want to modify
your code so that the working areas are shared by all MPI ranks. For
the motivation for that feature, please refer to the
<a href="https://simgrid.github.io/SMPI_CourseWare/topic_understanding_performance/matrixmultiplication/">relevant
section</a> of the SMPI CourseWare (see Activity #2.2 of the pointed
-assignment).
+assignment). In practice, change the call to malloc() and free() into
+SMPI_SHARED_MALLOC() and SMPI_SHARED_FREE().
+
+SMPI provides 2 algorithms for this feature. The first one, called \c
+local, allocates one bloc per call to SMPI_SHARED_MALLOC() in your
+code (each call location gets its own bloc) and this bloc is shared
+amongst all MPI ranks. This is implemented with the shm_* functions
+to create a new POSIX shared memory object (kept in RAM, in /dev/shm)
+for each shared bloc.
-SMPI can use shared memory by calling shm_* functions; this might speed up the simulation.
-This opens or creates a new POSIX shared memory object, kept in RAM, in /dev/shm.
+With the \c global algorithm, each call to SMPI_SHARED_MALLOC()
+returns a new adress, but it only points to a shadow bloc: its memory
+area is mapped on a 1MiB file on disk. If the returned bloc is of size
+N MiB, then the same file is mapped N times to cover the whole bloc.
+At the end, no matter how many SMPI_SHARED_MALLOC you do, this will
+only consume 1 MiB in memory.
-If you want to disable this behavior (for example for debugging
-purposes), set its value to "no" or "0".
+You can disable this behavior and come back to regular mallocs (for
+example for debugging purposes) using \c "no" as a value.
+
+If you want to keep private some parts of the buffer, for instance if these
+parts are used by the application logic and should not be corrupted, you
+can use SMPI_PARTIAL_SHARED_MALLOC(size, offsets, offsets_count).
+
+As an example,
+
+\code{.unparsed}
+ mem = SMPI_PARTIAL_SHARED_MALLOC(500, {27,42 , 100,200}, 2);
+\endcode
+
+will allocate 500 bytes to mem, such that mem[27..41] and mem[100..199]
+are shared and other area remain private.
+
+Then, it can be deallocated by calling SMPI_SHARED_FREE(mem).
+
+When smpi/shared-malloc:global is used, it is possible to optimize even
+further the memory consumption and the simulation time by using huge pages.
+To do so, it is required to mount a hugetlbfs on your system and allocate
+at least one huge page:
+
+\code{.unparsed}
+ mkdir /home/huge
+ sudo mount none /home/huge -t hugetlbfs -o rw,mode=0777
+ sudo sh -c 'echo 1 > /proc/sys/vm/nr_hugepages'
+\endcode
+Then, you can pass the option --cfg=smpi/shared-malloc-hugepage:/home/huge
+to smpirun.
\subsection options_model_smpi_wtime smpi/wtime: Inject constant times for calls to MPI_Wtime
- \c smpi/host-speed: \ref options_smpi_bench
- \c smpi/IB-penalty-factors: \ref options_model_network_coefs
- \c smpi/iprobe: \ref options_model_smpi_iprobe
+- \c smpi/iprobe-cpu-usage: \ref options_model_smpi_iprobe_cpu_usage
- \c smpi/init: \ref options_model_smpi_init
+- \c smpi/keep-temps: \ref options_smpi_temps
- \c smpi/lat-factor: \ref options_model_smpi_lat_factor
- \c smpi/ois: \ref options_model_smpi_ois
- \c smpi/or: \ref options_model_smpi_or
- \c smpi/os: \ref options_model_smpi_os
- \c smpi/papi-events: \ref options_smpi_papi_events
-- \c smpi/privatize-global-variables: \ref options_smpi_global
+- \c smpi/privatization: \ref options_smpi_privatization
- \c smpi/send-is-detached-thresh: \ref options_model_smpi_detached
- \c smpi/shared-malloc: \ref options_model_smpi_shared_malloc
- \c smpi/simulate-computation: \ref options_smpi_bench
