NSF Workshop on on Software
Development for Multicore and Heterogeneous Processors
Cairo, Egypt
June 21-23 2009
Freely Available Applications for Benchmarking and Porting
The majority of these applications and benchmarks have not been ported
to the Cell, GPUs, or other acceleration hardware. Some have
optimizations for multicore, or can make use of optimizations provided
by the compiler. For applications, there are many that were once
based on OpenMP but are now MPI. In those cases, OpenMP may still be
available, as a compile-time option. (WRF and a few others offer
mixed-mode, with both MPI and OpenMP.)
Large scale applications
- WRF: Weather research &
forecasting
- NAMD: Molecular dynamics, built on CHARM++
- ChaNGa: Astrophysics, built on CHARM++
- CCSM:
Coupled climate model; CCSM4 is imminent and uses the Earth System
Modeling Framework to couple. Component models are POP (physical
oceanography), CICE (ocean ice), LSM (land surface). For atmosphere,
CCSM3 uses CAM; CCSM4 will use WRF. The component models are
available separately.
- GCM ModelE: global
climate model from NASA
- ROMS: ocean model
- HYCOM: ocean model
- NWCHEM:
molecular dynamics
Libraries & SDKs
- OpenMPI: communication library
- IBM Cell
SDK: free for research
- BLAS, BLACS, LAPACK: numerical libraries (C + Fortran). The
repository at netlib.org has vast
quantities of other applications and libraries, too.
- celltask.sourceforge.net
(Cell low-level parallelization)
- Benchmarks:
see www.nsf.gov/pubs/2006/nsf0605/nsf0605.jsp
for application & system-level benchmarks
- FFTW: Fourier transforms. Has
some Cell optimizations.
- CUDA,
cuBLAS, OpenCL: available to registered developers (free),
for nVidia GPUs
- Brook++,
ACML: freely available for non-commercial use, for ATI GPUs. Registration
required.
- MAGMA: forthcoming from UTK. No released code yet.
gbn 2009-06-22