DDFgraph allows extraction and visualization of dynamic data-flow graphs computed at runtime. It currently relies on a Lisp interpreter modified to monitor data use (both of global and local variables) during program ...
详细信息
DDFgraph allows extraction and visualization of dynamic data-flow graphs computed at runtime. It currently relies on a Lisp interpreter modified to monitor data use (both of global and local variables) during program execution and to detect the corresponding dataflow, namely the value of variables 'flowing' from one point of the program to another. A companion database stores this flow and produces, on demand, its representation as a graph for dot. Furthermore, the flow graph can be manipulated, in order to reduce the size of the graph, through extraction of some parts of particular interest and/or through compaction of given subparts leaving visible only global flow.
Duplicate and redundant workflows can be avoided by encouraging workflow reuse. In this paper, we present how workflow similarity matching approach can be used to further enhance existing workflow modeling tools. Most...
详细信息
The data-flow form of processor architecture has been investigated extensively in the context of general purpose computers and its applicability to signal processing has been suggested, but instances of the applicatio...
详细信息
The data-flow form of processor architecture has been investigated extensively in the context of general purpose computers and its applicability to signal processing has been suggested, but instances of the application of these principles are rare. In contrast to the conventional Von Neumann architecture in which a single program is executed in a sequence prescribed by the programmer and even to processors with multiple-instruction streams, a data-flow processor has no specified instruction ordering. Instead, instructions act on specific operands and the sequence of instruction execution is determined at run-time by the order in which the various operands become available. This technique of data-driven dynamic scheduling extends naturally to multiprocessors and is the basis for control of multiprocessor configurations of the SPS-1000 high-speed FFT processor.
A new algorithm for an optimized design of the FDTD (Finite Difference Time Domain) computations involved in simulations of electromagnetic wave propagation in irregular areas is presented in the paper. Computational ...
详细信息
A new algorithm for an optimized design of the FDTD (Finite Difference Time Domain) computations involved in simulations of electromagnetic wave propagation in irregular areas is presented in the paper. Computational programs are represented as macro dataflow graphs (MDFG) which are to be partitioned and assigned to processors for optimal parallel execution. The proposed hierarchical optimization is based on two steps. The first step consists of a computation cell redeployment algorithm in irregular meshes where optimization is performed in three main phases: generation of an initial MDFG based on wave propagation area partitioning, MDFG nodes merging with load balancing to obtain given number of macro nodes and communication optimization to minimize and balance inter-node data transmissions. The second step is a connectivity-based distributed node clustering which creates an optimized MDFG. The hierarchical algorithm combines both methods to speedup parallel execution of the FDTD computations and to reduce the execution time of the optimization algorithm.
SimVis is a novel technology for the interactive visual analysis of large and complex flowdata which results from computational fluid dynamics (CFD) simulation. The new technology which has been researched and develo...
详细信息
ISBN:
(纸本)9781424413065
SimVis is a novel technology for the interactive visual analysis of large and complex flowdata which results from computational fluid dynamics (CFD) simulation. The new technology which has been researched and developed over the last years at the VRVis Research Center in Vienna, introduces a new approach for interactive graphical exploration and analysis of time-dependent data (computed on large three-dimensional grids, and resulting in a multitude of different scalar/vector values for each cell of these grids). In this paper the major new technological concepts of the SimVis approach are presented and real-world application examples are given.
dataflow is a method for parallel computation in which the elementary operations of the computer are activated by the flow of data. Computation is carried out in some autonomous processors which are connected to each ...
详细信息
dataflow is a method for parallel computation in which the elementary operations of the computer are activated by the flow of data. Computation is carried out in some autonomous processors which are connected to each other by some communication lines. The author shows how this model of computation can lead to interesting ideas in the management of knowledge bases. Each of the processors in the system is seen as a database machine which makes use of the modern sophisticated storage devices. Thus parallelism is gained both at the storage level (as is customary in database literature) and in the overall computation process.< >
Reducing communication overhead is extremely important for parallelizing compiler to generate efficient codes for distributed memory machines. In this paper, a redundant parallel execution model (RPEM) is used as the ...
详细信息
Reducing communication overhead is extremely important for parallelizing compiler to generate efficient codes for distributed memory machines. In this paper, a redundant parallel execution model (RPEM) is used as the model for target programs. The extend dataflow graph is introduced, and optimization algorithms based on the data-flow analysis are discussed. The overhead of dataflow analysis can be reduced by performing analysis on the extend dataflow graph. The analysis helps to reduce the redundant communication overhead. These optimization algorithms are able to perform inter-loop and inter-procedure analysis. Experimental results prove that these optimizations algorithms are effective in reducing both the number of communications and the communication volume.
The authors present the ALFA architecture, a dataflow machine with 16384 functional units (FUs) grouped in 128 clusters. ALFA is based on the Backus FFP computational model and uses the static dataflow execution mod...
详细信息
The authors present the ALFA architecture, a dataflow machine with 16384 functional units (FUs) grouped in 128 clusters. ALFA is based on the Backus FFP computational model and uses the static dataflow execution model. This machine's behavior is deterministic and asynchronous. Consequently, after compile time, instructions and data are no longer related. In this machine, even though its behavior is deterministic, no control token is generated during the computation, but only data tokens. Furthermore, during the execution phase, no memory is required to contain the partial results exchanged among FUs. A cluster with 128 FUs has been simulated, and some results are presented.< >
High-level synthesis compilers often produce reoccurring patterns in intermediate CDFGs during translation. By identifying large reoccurring patterns, one may reduce area and communication overhead by efficiently reus...
详细信息
High-level synthesis compilers often produce reoccurring patterns in intermediate CDFGs during translation. By identifying large reoccurring patterns, one may reduce area and communication overhead by efficiently reusing hardware for multiple operations. This paper presents an algorithm for dynamically generating templates of reoccurring patterns for resource sharing in CDFGs. Results show 40-80% resource reduction using small, incremental template growth, and variations within a 5% margin among varying look-ahead depths.
暂无评论