2015
K computer simulation
> 1.24 PFlops
(shortest period ~ 1.2 s)on 82,134 nodes, 82,134 MPI ranks w/ 8 OpenMP threads, 657,072 cores
Mission
| The SPECFEM project is dedicated to Dimitri Komatitsch and open-source development. Our goal is to advance Earth sciences by providing computational tools to the community. |  |
Team
A major involvement in the development and maintenance of the SPECFEM project is done by:
 |
Jeroen Tromp, Princeton University Founder of SPECFEM codes (together with Dimitri Komatitsch) |
 |
Daniel Peter Lead developer and maintainer of SPECFEM codes |
 |
Vadim Monteiller, LMA Marseille Developer and maintainer of SPECFEM codes |
 |
Hom Nath Gharti, Queen’s University Developer and maintainer of SPECFEM codes |
 |
Carl Tape, University of Alaska Fairbanks User training and developer meeting organizer |
 |
Qinya Liu, University of Toronto Adjoint code capabilities, developer meeting organizer |
 |
Ebru Bozdag, Colorado School of Mines Global inversions and Mars in SPECFEM3D Globe |
 |
Jean-Paul Ampuero, Geoazur Laboratory Fault rupture dynamics in SPECFEM3D Cartesian code |
 |
Emanuele Casarotti, INGV Meshing expert |
 |
Rohit Kakodkar, Princeton University Software engineer for future code development |
 |
Bryant Chow, University of Alaska Fairbanks Lead developer of adjTomo tools |
Contributions to these open-source codes come from many authors. Please find the authors list in the corresponding code repositories.
Milestones
A short timeline of major SPECFEM achievements:
K computer simulation
(shortest period ~ 1.2 s)on 82,134 nodes, 82,134 MPI ranks w/ 8 OpenMP threads, 657,072 cores
2013
Blue Waters XE6 simulation
> 1 PFlops
(shortest period < 2 s)on 21,675 XE nodes, 693,600 MPI ranks, 693,600 cores
2010
BULL Joseph Fourier Prize winner
for the partial GPU port of
2008
ACM Gordon Bell Finalist
for
2008
Kraken XT5 simulation
> 160 TFlops
(shortest period ~1.72 s)on 149,784 cores
2003
ACM Gordon Bell Award
for Best Performance of
2003
Earth Simulator simulation
> 5 TFlops
(shortest period ~5 s)on 243 nodes, 1,944 MPI ranks, 1,944 cores