Group :: Monitoring
RPM: stream-mem
Main Changelog Spec Patches Sources Download Gear Bugs and FR Repocop
%def_with fortran
%add_debuginfo_skiplist %_bindir
Name: stream-mem
Version: 5.10
Release: alt3
Summary: STREAM: Sustainable Memory Bandwidth in High Performance Computers
License: distributable
Group: Monitoring
Url: http://www.cs.virginia.edu/stream/
Source: stream-%version.tar.gz
Packager: Michael Shigorin <mike at altlinux.org>
BuildRequires: libgomp-devel
%if_with fortran
BuildRequires: gcc-fortran
%endif
%description
The STREAM benchmark is a simple synthetic benchmark program that
measures sustainable memory bandwidth (in MB/s) and the corresponding
computation rate for simple vector kernels.
%prep
%setup -n stream-%version
%build
%ifarch %e2k
# as of lcc 1.25.23, this gives slightly better result that our default -O3
%define _optlevel 4
%endif
%make CFLAGS="-O%_optlevel -fopenmp -D_OPENMP" stream_c.exe
%if_with fortran
%make FFLAGS="-O%_optlevel" stream_f.exe
%endif
cat > README.ALT << EOF
Multiprocessor runs (find out the optimal factor by experimenting):
OMP_NUM_THREADS=4 stream_c
See also http://www.cs.virginia.edu/stream/ref.html#runrules
EOF
%install
install -pDm755 stream_c.exe %buildroot%_bindir/stream_c
%if_with fortran
install -pDm755 stream_f.exe %buildroot%_bindir/stream_f
%endif
%files
%_bindir/*
%doc HISTORY.txt LICENSE.txt READ.ME TO_DO
%doc README.ALT
# TODO: cover MPI version as well
%changelog
…
Full changelog you can see here
%add_debuginfo_skiplist %_bindir
Name: stream-mem
Version: 5.10
Release: alt3
Summary: STREAM: Sustainable Memory Bandwidth in High Performance Computers
License: distributable
Group: Monitoring
Url: http://www.cs.virginia.edu/stream/
Source: stream-%version.tar.gz
Packager: Michael Shigorin <mike at altlinux.org>
BuildRequires: libgomp-devel
%if_with fortran
BuildRequires: gcc-fortran
%endif
%description
The STREAM benchmark is a simple synthetic benchmark program that
measures sustainable memory bandwidth (in MB/s) and the corresponding
computation rate for simple vector kernels.
%prep
%setup -n stream-%version
%build
%ifarch %e2k
# as of lcc 1.25.23, this gives slightly better result that our default -O3
%define _optlevel 4
%endif
%make CFLAGS="-O%_optlevel -fopenmp -D_OPENMP" stream_c.exe
%if_with fortran
%make FFLAGS="-O%_optlevel" stream_f.exe
%endif
cat > README.ALT << EOF
Multiprocessor runs (find out the optimal factor by experimenting):
OMP_NUM_THREADS=4 stream_c
See also http://www.cs.virginia.edu/stream/ref.html#runrules
EOF
%install
install -pDm755 stream_c.exe %buildroot%_bindir/stream_c
%if_with fortran
install -pDm755 stream_f.exe %buildroot%_bindir/stream_f
%endif
%files
%_bindir/*
%doc HISTORY.txt LICENSE.txt READ.ME TO_DO
%doc README.ALT
# TODO: cover MPI version as well
%changelog
…
Full changelog you can see here