eth-cscs / costa Goto Github PK
View Code? Open in Web Editor NEWDistributed Communication-Optimal Shuffle and Transpose Algorithm
License: BSD 3-Clause "New" or "Revised" License
Distributed Communication-Optimal Shuffle and Transpose Algorithm
License: BSD 3-Clause "New" or "Revised" License
This cmake specification disables cross-compilation of COSTA, e.g. CP2K which depends on COSTA cannot be compiled for other targets like generic
.
After building the library with:
CC=gcc-9 CXX=g++-9 cmake -DCOSTA_SCALAPACK=MKL ..
The CMake outputs:
-- Setting build type to 'Release' as none was specified.
-- Selected ScaLAPACK backend for COSTA: MKL
-- The CXX compiler identification is GNU 9.2.0
-- Checking whether CXX compiler has -isysroot
-- Checking whether CXX compiler has -isysroot - yes
-- Checking whether CXX compiler supports OSX deployment target flag
-- Checking whether CXX compiler supports OSX deployment target flag - yes
-- Detecting CXX compiler ABI info
-- Detecting CXX compiler ABI info - done
-- Check for working CXX compiler: /usr/local/bin/g++-9 - skipped
-- Detecting CXX compile features
-- Detecting CXX compile features - done
-- Found MPI_CXX: /usr/local/Cellar/mpich/3.3.1/lib/libmpicxx.dylib (found version "3.1")
-- Found MPI: TRUE (found version "3.1") found components: CXX
-- Found OpenMP_CXX: -fopenmp (found version "4.5")
-- Found OpenMP: TRUE (found version "4.5") found components: CXX
-- cxxopts version 2.2.0
-- Looking for C++ include pthread.h
-- Looking for C++ include pthread.h - found
-- Performing Test CMAKE_HAVE_LIBC_PTHREAD
-- Performing Test CMAKE_HAVE_LIBC_PTHREAD - Success
-- Found Threads: TRUE
-- Found OpenMP_CXX: -fopenmp (found version "4.5")
-- Found MKL: /opt/intel/compilers_and_libraries_2019.0.117/mac/mkl/include
CANNOT FIND TARGET: mkl::mkl_intel_32bit_omp_dyn
CANNOT FIND TARGET: mkl::scalapack_mpich_intel_32bit_omp_dyn
-- Configuring done
CMake Error at src/costa/CMakeLists.txt:77 (add_library):
Target "costa_prefixed_scalapack" links to target
"mkl::scalapack_mpich_intel_32bit_omp_dyn" but the target was not found.
Perhaps a find_package() call is missing for an IMPORTED target, or an
ALIAS target is missing?
CMake Error at src/costa/CMakeLists.txt:60 (add_library):
Target "costa_scalapack" links to target
"mkl::scalapack_mpich_intel_32bit_omp_dyn" but the target was not found.
Perhaps a find_package() call is missing for an IMPORTED target, or an
ALIAS target is missing?
CMake Error at miniapps/CMakeLists.txt:17 (add_executable):
Target "pxgemr2d_miniapp" links to target
"mkl::scalapack_mpich_intel_32bit_omp_dyn" but the target was not found.
Perhaps a find_package() call is missing for an IMPORTED target, or an
ALIAS target is missing?
CMake Error at miniapps/CMakeLists.txt:17 (add_executable):
Target "pxtran_miniapp" links to target
"mkl::scalapack_mpich_intel_32bit_omp_dyn" but the target was not found.
Perhaps a find_package() call is missing for an IMPORTED target, or an
ALIAS target is missing?
-- Generating done
CMake Generate step failed. Build files cannot be regenerated correctly.
https://github.com/eth-cscs/COSTA#performance-results says
The benchmark code is available in the provided miniapp.
The link (https://github.com/eth-cscs/COSTA#data-redistribution-with-pxgemr2d) is wrong.
I guess the correct link is https://github.com/eth-cscs/COSTA/blob/master/miniapps/pxgemr2d_miniapp.cpp.
It looks like naming convention for costa-v2.1 release has changed. This breaks the spack recipe.
Hi @kabicm !
I'm running some tests with wave-function reshuffling for FFT layout using COSTA and I observe a very low throughput of ~1.5 Gb/s for costa::transform(). By throughput I assume the total (global) matrix size divided by the time of transform(). I'm running the test on 2 Daint nodes, 1 rank / node, 12 threads / rank and the test basically swaps the matrix partition from distributing rows
+---+
| |
+---+
| |
+----
to distributing columns:
+-+-+
| | |
| | |
+-+-+
I tried several matrix sizes and I observe the same low throughput. For example, 200'000 x 600 matrix will expose the problem. Am I doing something wrong? I would expect a much higher performance.
Hi, I am trying to build COSTA on Stampede2, which comprises of KNL nodes.
I follow the instructions in INSTALL.md, yet hit a crash.
Here is the output:
login1.stampede2(1164)$ cmake -DCMAKE_INSTALL_PREFIX=$HOME/install_dir/costa/ ..
-- Setting build type to 'Release' as none was specified.
-- Selected ScaLAPACK backend for COSTA: OFF
-- The CXX compiler identification is Intel 18.0.2.20180210
-- Detecting CXX compiler ABI info
-- Detecting CXX compiler ABI info - done
-- Check for working CXX compiler: /opt/apps/intel18/impi/18.0.2/bin/mpicxx - skipped
-- Detecting CXX compile features
-- Detecting CXX compile features - done
-- Found MPI_CXX: /opt/apps/intel18/impi/18.0.2/bin/mpicxx (found version "3.1")
-- Found MPI: TRUE (found version "3.1") found components: CXX
-- Found OpenMP_CXX: -qopenmp (found version "5.0")
-- Found OpenMP: TRUE (found version "5.0") found components: CXX
-- cxxopts version 2.2.0
CMake Warning at CMakeLists.txt:101 (message):
No ScaLAPACK interface is specified. Building COSTA without SCALAPACK
wrappers.
-- Configuring done
-- Generating done
-- Build files have been written to: /work2/05608/tg849075/costa/build
login1.stampede2(1165)$ make
[ 4%] Building CXX object src/costa/CMakeFiles/costa.dir/grid2grid/block.cpp.o
[ 8%] Building CXX object src/costa/CMakeFiles/costa.dir/grid2grid/grid2D.cpp.o
[ 12%] Building CXX object src/costa/CMakeFiles/costa.dir/grid2grid/interval.cpp.o
[ 16%] Building CXX object src/costa/CMakeFiles/costa.dir/grid2grid/scalapack_layout.cpp.o
[ 20%] Building CXX object src/costa/CMakeFiles/costa.dir/grid2grid/communication_data.cpp.o
/work2/05608/tg849075/costa/src/costa/grid2grid/communication_data.cpp(167): error: invalid entity for this variable list in omp clause
#pragma omp parallel for shared(mpi_messages, workspace, offset_per_message, buffer)
^
/work2/05608/tg849075/costa/src/costa/grid2grid/communication_data.cpp(195): error: invalid entity for this variable list in omp clause
#pragma omp parallel for shared(idx, package_ticks, mpi_messages, offset_per_message, buffer, workspace)
^
compilation aborted for /work2/05608/tg849075/costa/src/costa/grid2grid/communication_data.cpp (code 2)
make[2]: *** [src/costa/CMakeFiles/costa.dir/grid2grid/communication_data.cpp.o] Error 2
make[1]: *** [src/costa/CMakeFiles/costa.dir/all] Error 2
make: *** [all] Error 2
I set:
export CC=mpicc
export CXX=mpicxx
Currently-loaded modules:
Currently Loaded Modules:
1) intel/18.0.2 2) libfabric/1.7.0 3) impi/18.0.2 4) git/2.24.1 5) autotools/1.1 6) python2/2.7.15 7) cmake/3.20.2 8) xalt/2.10.37 9) TACC
Please advise as to how to fix this issue.
A COSTA build on Piz Daint fails with the error message
CMake Error at cmake/FindCRAY_LIBSCI.cmake:22 (set_properties):
Unknown CMake command "set_properties".
because set_properties instead of set_property
is used.
A declarative, efficient, and flexible JavaScript library for building user interfaces.
๐ Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. ๐๐๐
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google โค๏ธ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.