This paper presents an improved interconnect network for Tree-based FPGA architecture that unifies two unidirectional programmable networks. New tools are developed to place and route the largest benchmark circuits, where different optimization techniques are used to get an optimized architecture. The effect of variation in LUT and cluster size on the area, performance, and power of the Tree-based architecture is analyzed. Experimental results show that an architecture with LUT size 4 and arity size 4 is the most efficient in terms of area and static power dissipation, whereas the architectures with higher LUT and cluster size are efficient in terms of performance. We also show that unifying a Mesh with this Tree topology leads to an architecture which has good layout scalability and better interconnect efficiency compared to VPR-style Mesh.

No abstract

No abstract

International audienceIn this paper we present a new clustered mesh FPGA architecture where each cluster local interconnect is implemented as an MFPGA tree network. Unlike previous clustered mesh architectures, the mesh of tree allows us to consider large clusters sizes (thanks to MFPGA depopulated local interconnect). Experimentation shows that we obtain a reduction of 14% in switches number and 2 times in the placement and routing run time. Furthermore, compared to MFPGA, the mesh of tree achieves full mutability of all MCNC benchmarks since we can easily control both clusters LUTs occupation and mesh channel width

For the last five years parallel reservoir simulation has enjoyed great interest at Saudi Aramco. Realism in modeling the fields in Arabia requires the use of massive simulation models. These massive models capture detailed geological heterogeneity at relatively fine resolution. In addition, some of these fields have been producing for the last 50 years and are expected to continue to produce for decades to come. Today, there are more than 20 massive simulation models ranging in size from 500,000 to 10 million cells and are routinely used in managing the Saudi fields. Models of this magnitude require sufficiently "massive" computational power in order to carry out simulation studies in efficient and reasonable time. Faced with the challenge of meeting computational requirements, we were motivated to find a cost effective computational platform. In this work, we report on our experience in investigating PC-Clusters for mega-cell reservoir simulation studies. Our investigation demonstrated that commodity off-the-shelf Xeon based PC-Clusterscan deliver an exceptional performance at minimal investment in computing hardware and software. Up to 9.6 million cells model running for 50 years have been tested. Currently Saudi Aramco uses PC-Clusters for its simulation work at not more than 15% of the cost of traditional solutions using supercomputers. Introduction The area of parallel reservoir simulation has been extensively investigated with earliest attempts dating back to the late 1980s (Ref.1 has good coverage of work done). Fundamentally, the objective behind parallelization effort is to drive simulation run time to absolute minimal. The needs and benefits behind such effort are numerous: Mega-Cell Reservoir Simulation. Relatively high-resolution million-cell model provide the necessary realism in modeling large or highly heterogeneous reservoirs. A grid size of 50 meters is common in modeling small to medium scale reservoirs. However, use of same grid size in large fields would easily result in models with more than a million cells. At the extreme; a 50×50 meters areal grid size on the Ghawar field (the largest oil field in the world) would require more than 100 million cells. High-resolution grid block swill also capture details of geological heterogeneity providing us with better understanding and modeling of fluid flow. Modeling Complexity. Compositional and dual-porosity-dual-permeability models, by nature of the mathematical formulations require extensive numeric computations. Jacobian building and solver solution require more processing time as the number of unknowns are more. Long Simulation Time. Simulation models for old-fields would also benefit from parallel reservoir simulation. Fields in the Arabian gulf, especially Saudi fields have been producing for the last 50 years and are expected to continue production for decades to come. Such models would require an extended simulation time in "serial" mode but would be more manageable with parallel simulation. Stochastic Modeling. Statistical approach to geological model building results in numerous simulation-model realizations. Parallel simulation can process the models with minimal or no up-scaling and provide us with capabilities to better utilize stochastic tools and define and evaluate risks and uncertainties in field operation. Automatic and Semi-Automatic History Matching. The success of utilizing history matching tools is highly dependent on the speed of running a single simulation and the evaluation of numerous models. Parallel reservoir simulation is a natural complement that provides the necessary simulation performance. Improved Turn Around Time. The longer a simulation study takes the more resource it requires and the more costly it becomes. Simulation runs that take days or weeks to complete can be improved to few hours if the technology of parallel computing is properly utilized.