starpu-max/doc/chapters/configuration.texi

@c -*-texinfo-*-

@c This file is part of the StarPU Handbook.
@c Copyright (C) 2009--2011  Universit@'e de Bordeaux 1
@c Copyright (C) 2010, 2011, 2012  Centre National de la Recherche Scientifique
@c Copyright (C) 2011, 2012 Institut National de Recherche en Informatique et Automatique
@c See the file starpu.texi for copying conditions.

@menu
* Compilation configuration::   
* Execution configuration through environment variables::  
@end menu

@node Compilation configuration
@section Compilation configuration

The following arguments can be given to the @code{configure} script.

@menu
* Common configuration::        
* Configuring workers::         
* Extension configuration::     
* Advanced configuration::      
@end menu

@node Common configuration
@subsection Common configuration

@table @code

@item --enable-debug
Enable debugging messages.

@item --enable-fast
Disable assertion checks, which saves computation time.

@item --enable-verbose
Increase the verbosity of the debugging messages.  This can be disabled
at runtime by setting the environment variable @code{STARPU_SILENT} to
any value.

@smallexample
% STARPU_SILENT=1 ./vector_scal
@end smallexample

@item --enable-coverage
Enable flags for the @code{gcov} coverage tool.

@end table

Additionally, the @command{configure} script recognize many variables, which
can be listed by typing @code{./configure --help}. For example,
@code{./configure NVCCFLAGS="-arch sm_13"} adds a flag for the compilation of
CUDA kernels.

@node Configuring workers
@subsection Configuring workers

@table @code

@item --enable-maxcpus=@var{count}
Use at most @var{count} CPU cores.  This information is then
available as the @code{STARPU_MAXCPUS} macro.

@item --disable-cpu
Disable the use of CPUs of the machine. Only GPUs etc. will be used.

@item --enable-maxcudadev=@var{count}
Use at most @var{count} CUDA devices.  This information is then
available as the @code{STARPU_MAXCUDADEVS} macro.

@item --disable-cuda
Disable the use of CUDA, even if a valid CUDA installation was detected.

@item --with-cuda-dir=@var{prefix}
Search for CUDA under @var{prefix}, which should notably contain
@file{include/cuda.h}.

@item --with-cuda-include-dir=@var{dir}
Search for CUDA headers under @var{dir}, which should
notably contain @code{cuda.h}. This defaults to @code{/include} appended to the
value given to @code{--with-cuda-dir}.

@item --with-cuda-lib-dir=@var{dir}
Search for CUDA libraries under @var{dir}, which should notably contain
the CUDA shared libraries---e.g., @file{libcuda.so}.  This defaults to
@code{/lib} appended to the value given to @code{--with-cuda-dir}.

@item --disable-cuda-memcpy-peer
Explicitly disable peer transfers when using CUDA 4.0.

@item --enable-maxopencldev=@var{count}
Use at most @var{count} OpenCL devices.  This information is then
available as the @code{STARPU_MAXOPENCLDEVS} macro.

@item --disable-opencl
Disable the use of OpenCL, even if the SDK is detected.

@item --with-opencl-dir=@var{prefix}
Search for an OpenCL implementation under @var{prefix}, which should
notably contain @file{include/CL/cl.h} (or @file{include/OpenCL/cl.h} on
Mac OS).

@item --with-opencl-include-dir=@var{dir}
Search for OpenCL headers under @var{dir}, which should notably contain
@file{CL/cl.h} (or @file{OpenCL/cl.h} on Mac OS).  This defaults to
@code{/include} appended to the value given to @code{--with-opencl-dir}.

@item --with-opencl-lib-dir=@var{dir}
Search for an OpenCL library under @var{dir}, which should notably
contain the OpenCL shared libraries---e.g. @file{libOpenCL.so}. This defaults to
@code{/lib} appended to the value given to @code{--with-opencl-dir}.

@item --enable-gordon
Enable the use of the Gordon runtime for Cell SPUs.
@c TODO: rather default to enabled when detected

@item --with-gordon-dir=@var{prefix}
Search for the Gordon SDK under @var{prefix}.

@item --enable-maximplementations=@var{count}
Allow for at most @var{count} codelet implementations for the same
target device.  This information is then available as the
@code{STARPU_MAXIMPLEMENTATIONS} macro.

@item --disable-asynchronous-copy
Disable asynchronous copies between CPU and GPU devices.
The AMD implementation of OpenCL is known to
fail when copying data asynchronously. When using this implementation,
it is therefore necessary to disable asynchronous data transfers.

@item --disable-asynchronous-cuda-copy
Disable asynchronous copies between CPU and CUDA devices.

@item --disable-asynchronous-opencl-copy
Disable asynchronous copies between CPU and OpenCL devices.
The AMD implementation of OpenCL is known to
fail when copying data asynchronously. When using this implementation,
it is therefore necessary to disable asynchronous data transfers.
@end table

@node Extension configuration
@subsection Extension configuration

@table @code

@item --disable-socl
Disable the SOCL extension (@pxref{SOCL OpenCL Extensions}).  By
default, it is enabled when an OpenCL implementation is found.

@item --disable-starpu-top
Disable the StarPU-Top interface (@pxref{StarPU-Top}).  By default, it
is enabled when the required dependencies are found.

@item --disable-gcc-extensions
Disable the GCC plug-in (@pxref{C Extensions}).  By default, it is
enabled when the GCC compiler provides a plug-in support.

@item --with-mpicc=@var{path}
Use the @command{mpicc} compiler at @var{path}, for starpumpi
(@pxref{StarPU MPI support}).

@item --enable-comm-stats
Enable communication statistics for starpumpi (@pxref{StarPU MPI
support}).

@end table

@node Advanced configuration
@subsection Advanced configuration

@table @code

@item --enable-perf-debug
Enable performance debugging through gprof.

@item --enable-model-debug
Enable performance model debugging.

@item --enable-stats
@c see ../../src/datawizard/datastats.c
Enable gathering of memory transfer statistics.

@item --enable-maxbuffers
Define the maximum number of buffers that tasks will be able to take
as parameters, then available as the @code{STARPU_NMAXBUFS} macro.

@item --enable-allocation-cache
Enable the use of a data allocation cache to avoid the cost of it with
CUDA. Still experimental.

@item --enable-opengl-render
Enable the use of OpenGL for the rendering of some examples.
@c TODO: rather default to enabled when detected

@item --enable-blas-lib
Specify the blas library to be used by some of the examples. The
library has to be 'atlas' or 'goto'.

@item --disable-starpufft
Disable the build of libstarpufft, even if fftw or cuFFT is available.

@item --with-magma=@var{prefix}
Search for MAGMA under @var{prefix}.  @var{prefix} should notably
contain @file{include/magmablas.h}.

@item --with-fxt=@var{prefix}
Search for FxT under @var{prefix}.
@url{http://savannah.nongnu.org/projects/fkt, FxT} is used to generate
traces of scheduling events, which can then be rendered them using ViTE
(@pxref{Off-line, off-line performance feedback}).  @var{prefix} should
notably contain @code{include/fxt/fxt.h}.

@item --with-perf-model-dir=@var{dir}
Store performance models under @var{dir}, instead of the current user's
home.

@item --with-goto-dir=@var{prefix}
Search for GotoBLAS under @var{prefix}, which should notably contain @file{libgoto.so} or @file{libgoto2.so}.

@item --with-atlas-dir=@var{prefix}
Search for ATLAS under @var{prefix}, which should notably contain
@file{include/cblas.h}.

@item --with-mkl-cflags=@var{cflags}
Use @var{cflags} to compile code that uses the MKL library.

@item --with-mkl-ldflags=@var{ldflags}
Use @var{ldflags} when linking code that uses the MKL library.  Note
that the
@url{http://software.intel.com/en-us/articles/intel-mkl-link-line-advisor/,
MKL website} provides a script to determine the linking flags.

@item --disable-build-examples
Disable the build of examples.

@end table

@node Execution configuration through environment variables
@section Execution configuration through environment variables

@menu
* Workers::                     Configuring workers
* Scheduling::                  Configuring the Scheduling engine
* Misc::                        Miscellaneous and debug
@end menu

@node Workers
@subsection Configuring workers

@table @code

@item @code{STARPU_NCPU}
Specify the number of CPU workers (thus not including workers dedicated to control acceleratores). Note that by default, StarPU will not allocate
more CPU workers than there are physical CPUs, and that some CPUs are used to control
the accelerators.

@item @code{STARPU_NCUDA}
Specify the number of CUDA devices that StarPU can use. If
@code{STARPU_NCUDA} is lower than the number of physical devices, it is
possible to select which CUDA devices should be used by the means of the
@code{STARPU_WORKERS_CUDAID} environment variable. By default, StarPU will
create as many CUDA workers as there are CUDA devices.

@item @code{STARPU_NOPENCL}
OpenCL equivalent of the @code{STARPU_NCUDA} environment variable.

@item @code{STARPU_NGORDON}
Specify the number of SPUs that StarPU can use.

@item @code{STARPU_WORKERS_NOBIND}
Setting it to non-zero will prevent StarPU from binding its threads to
CPUs. This is for instance useful when running the testsuite in parallel.

@item @code{STARPU_WORKERS_CPUID}
Passing an array of integers (starting from 0) in @code{STARPU_WORKERS_CPUID}
specifies on which logical CPU the different workers should be
bound. For instance, if @code{STARPU_WORKERS_CPUID = "0 1 4 5"}, the first
worker will be bound to logical CPU #0, the second CPU worker will be bound to
logical CPU #1 and so on.  Note that the logical ordering of the CPUs is either
determined by the OS, or provided by the @code{hwloc} library in case it is
available.

Note that the first workers correspond to the CUDA workers, then come the
OpenCL and the SPU, and finally the CPU workers. For example if
we have @code{STARPU_NCUDA=1}, @code{STARPU_NOPENCL=1}, @code{STARPU_NCPU=2}
and @code{STARPU_WORKERS_CPUID = "0 2 1 3"}, the CUDA device will be controlled
by logical CPU #0, the OpenCL device will be controlled by logical CPU #2, and
the logical CPUs #1 and #3 will be used by the CPU workers.

If the number of workers is larger than the array given in
@code{STARPU_WORKERS_CPUID}, the workers are bound to the logical CPUs in a
round-robin fashion: if @code{STARPU_WORKERS_CPUID = "0 1"}, the first and the
third (resp. second and fourth) workers will be put on CPU #0 (resp. CPU #1).

This variable is ignored if the @code{use_explicit_workers_bindid} flag of the
@code{starpu_conf} structure passed to @code{starpu_init} is set.

@item @code{STARPU_WORKERS_CUDAID}
Similarly to the @code{STARPU_WORKERS_CPUID} environment variable, it is
possible to select which CUDA devices should be used by StarPU. On a machine
equipped with 4 GPUs, setting @code{STARPU_WORKERS_CUDAID = "1 3"} and
@code{STARPU_NCUDA=2} specifies that 2 CUDA workers should be created, and that
they should use CUDA devices #1 and #3 (the logical ordering of the devices is
the one reported by CUDA).

This variable is ignored if the @code{use_explicit_workers_cuda_gpuid} flag of
the @code{starpu_conf} structure passed to @code{starpu_init} is set.

@item @code{STARPU_WORKERS_OPENCLID}
OpenCL equivalent of the @code{STARPU_WORKERS_CUDAID} environment variable.

This variable is ignored if the @code{use_explicit_workers_opencl_gpuid} flag of
the @code{starpu_conf} structure passed to @code{starpu_init} is set.

@item @code{STARPU_SINGLE_COMBINED_WORKER}
If set, StarPU will create several workers which won't be able to work
concurrently. It will create combined workers which size goes from 1 to the
total number of CPU workers in the system.

@item @code{SYNTHESIZE_ARITY_COMBINED_WORKER}

Let the user decide how many elements are allowed between combined workers
created from hwloc information. For instance, in the case of sockets with 6
cores without shared L2 caches, if @code{SYNTHESIZE_ARITY_COMBINED_WORKER} is
set to 6, no combined worker will be synthesized beyond one for the socket
and one per core. If it is set to 3, 3 intermediate combined workers will be
synthesized, to divide the socket cores into 3 chunks of 2 cores. If it set to
2, 2 intermediate combined workers will be synthesized, to divide the the socket
cores into 2 chunks of 3 cores, and then 3 additional combined workers will be
synthesized, to divide the former synthesized workers into a bunch of 2 cores,
and the remaining core (for which no combined worker is synthesized since there
is already a normal worker for it).

The default, 2, thus makes StarPU tend to building a binary trees of combined
workers.

@item @code{STARPU_DISABLE_ASYNCHRONOUS_COPY}
Disable asynchronous copies between CPU and GPU devices.
The AMD implementation of OpenCL is known to
fail when copying data asynchronously. When using this implementation,
it is therefore necessary to disable asynchronous data transfers.

@item @code{STARPU_DISABLE_ASYNCHRONOUS_CUDA_COPY}
Disable asynchronous copies between CPU and CUDA devices.

@item @code{STARPU_DISABLE_ASYNCHRONOUS_OPENCL_COPY}
Disable asynchronous copies between CPU and OpenCL devices.
The AMD implementation of OpenCL is known to
fail when copying data asynchronously. When using this implementation,
it is therefore necessary to disable asynchronous data transfers.

@item @code{STARPU_DISABLE_CUDA_GPU_GPU_DIRECT}
Disable direct CUDA transfers from GPU to GPU, and let CUDA copy through RAM
instead. This permits to test the performance effect of GPU-Direct.

@end table

@node Scheduling
@subsection Configuring the Scheduling engine

@table @code

@item @code{STARPU_SCHED}
Choose between the different scheduling policies proposed by StarPU: work
random, stealing, greedy, with performance models, etc.

Use @code{STARPU_SCHED=help} to get the list of available schedulers.

@item @code{STARPU_CALIBRATE}
If this variable is set to 1, the performance models are calibrated during
the execution. If it is set to 2, the previous values are dropped to restart
calibration from scratch. Setting this variable to 0 disable calibration, this
is the default behaviour.

Note: this currently only applies to @code{dm}, @code{dmda} and @code{heft} scheduling policies.

@item @code{STARPU_BUS_CALIBRATE}
If this variable is set to 1, the bus is recalibrated during intialization.

@item @code{STARPU_PREFETCH}
@anchor{STARPU_PREFETCH}
This variable indicates whether data prefetching should be enabled (0 means
that it is disabled). If prefetching is enabled, when a task is scheduled to be
executed e.g. on a GPU, StarPU will request an asynchronous transfer in
advance, so that data is already present on the GPU when the task starts. As a
result, computation and data transfers are overlapped.
Note that prefetching is enabled by default in StarPU.

@item @code{STARPU_SCHED_ALPHA}
To estimate the cost of a task StarPU takes into account the estimated
computation time (obtained thanks to performance models). The alpha factor is
the coefficient to be applied to it before adding it to the communication part.

@item @code{STARPU_SCHED_BETA}
To estimate the cost of a task StarPU takes into account the estimated
data transfer time (obtained thanks to performance models). The beta factor is
the coefficient to be applied to it before adding it to the computation part.

@end table

@node Misc
@subsection Miscellaneous and debug

@table @code

@item @code{STARPU_SILENT}
This variable allows to disable verbose mode at runtime when StarPU
has been configured with the option @code{--enable-verbose}.

@item @code{STARPU_LOGFILENAME}
This variable specifies in which file the debugging output should be saved to.

@item @code{STARPU_FXT_PREFIX}
This variable specifies in which directory to save the trace generated if FxT is enabled. It needs to have a trailing '/' character.

@item @code{STARPU_LIMIT_GPU_MEM}
This variable specifies the maximum number of megabytes that should be
available to the application on each GPUs. In case this value is smaller than
the size of the memory of a GPU, StarPU pre-allocates a buffer to waste memory
on the device. This variable is intended to be used for experimental purposes
as it emulates devices that have a limited amount of memory.

@item @code{STARPU_GENERATE_TRACE}
When set to 1, this variable indicates that StarPU should automatically
generate a Paje trace when starpu_shutdown is called.

@end table