Dr. Pradeep K Sinha, Senior Director (Corporate Strategy and R&D), C-DAC, India
The Intel® Parallel Computing Center (Intel® PCC) at Center for Development of Advanced Computing (C-DAC) is focused to port, scale and optimize popularly used Climate Forecast System version2 (CFSv2) model on hybrid architecture based on Intel Xeon Phi™ coprocessor used in large scale production grade environments. The objective is to enhance the computational speed and accuracy of the model so as to gain accurate weather predictions well in advance.
Climate Forecast System version2 (CFSv2) is a fully coupled model representing the interaction between the Earth's oceans, land and atmosphere. It was developed at the Environmental Modeling Center at National Centre for Environmental Prediction (NCEP). This engagement shall facilitate an in-depth understanding of the model, support scientific understanding of the climate system and provide precise information which would be valuable in guiding policy decisions.
“High-performance computing systems have allowed weather researchers to make more accurate predictions and gain a better understanding of why weather patterns change direction or intensity” said Dr. Pradeep K. Sinha. He emphasized by saying “C-DAC is focusing on modernizing the CFSv2 model to increase parallelism and scalability through optimizations that control cores, caches, threads, and vector capabilities of Intel Xeon Phi™ coprocessor”. This activity will be the first of its kind to port the coupled model on Intel Xeon Phi™ coprocessor in collaboration with Intel Experts which will be done using C-DAC’s PARAM Yuva II based on Intel Xeon Phi™ coprocessor at The National PARAM Supercomputing Facility (NPSF), C-DAC, Pune, India.
The current version of Climate Forecast System (CFSv2) includes the latest developments in the global atmospheric model (GFS, used for numerical weather forecast), coupling with a new version of the global ocean model (MOM4), and a sea ice model with relevant changes for Indian conditions. The CFSv2 model is a hybrid programming model (Message Passing Interface (MPI) and Open specification of Multi Processing (OpenMP) ) written using FORTRAN and C with IBM extensions (API/macros), Parallel Operating Environment (POE) that could be compiled and run on IBM Power based systems earlier.
The CFSv2 follows the MPMD architecture during its execution where all the executables are executed at the same time sharing their intermediate results with each other at times when required, making the model slow. The Atmospheric Model (GFS) in CFSv2 is highly compute-intensive and the Ocean Model (MOM4) is input/output intensive, which takes more time putting the coupler and GFS in wait state till it finishes and passes the results to the coupler, which in turn passes to GFS.
Current Status: While maintaining the semantics of the code, modifications are done to make it run on the Linux x86_64 architecture using Intel C, FORTRAN compiler, Intel® MPI Library, Intel flags and environment settings. The code which has been compiled with proper optimization flags executes efficiently keeping the output data unchanged, hence giving a stable version of the code on X86_64.
Intel's compilers may or may not optimize to the same degree for non-Intel microprocessors for optimizations that are not unique to Intel microprocessors. These optimizations include SSE2, SSE3, and SSSE3 instruction sets and other optimizations. Intel does not guarantee the availability, functionality, or effectiveness of any optimization on microprocessors not manufactured by Intel. Microprocessor-dependent optimizations in this product are intended for use with Intel microprocessors. Certain optimizations not specific to Intel microarchitecture are reserved for Intel microprocessors. Please refer to the applicable product User and Reference Guides for more information regarding the specific instruction sets covered by this notice.
Notice revision #20110804