Automatic Array Partitioning Based on the Smith Normal Form

We investigate the lattice-based array partitioning based on the theory of the Smith Normal Form and we present two elegant techniques for partitioning arrays in parallel DoAll loops for message-passing parallel machines: (1) DoAll loops with constant dependencies for communication-free partitioning: a general solution of all possible communication-free partitioning is derived where the dependencies among array references are described in constant distance vectors. (2) DoAll loops with non-constant dependencies for block-communication partitioning: the dependencies among array references are described in non-constant distance vectors. We derive the partitioning equations which allocate all remote data to a unique processor such that only one block-communication can obtain all the remote data for the computation. By using the Smith Normal Form decomposition, we are also able to verify our partitioning results.     

Compile-time techniques for data distribution in distributed memorymachines

A solution to the problem of partitioning data for distributed memory machines is discussed. The solution uses a matrix notation to describe array accesses in fully parallel loops, which allows the derivation of sufficient conditions for communication-free partitioning (decomposition) of arrays. A series of examples that illustrate the effectiveness of the technique for linear references, the use of loop transformations in deriving the necessary data decompositions, and a formulation that aids in deriving heuristics for minimizing a communication when communication-free partitions are not feasible are presented     

A thread partitioning approach for speculative multithreading

Speculative multithreading (SpMT) is a thread-level automatic parallelization technique, which partitions sequential programs into multithreads to be executed in parallel. This paper presents different thread partitioning strategies for nonloops and loops. For nonloops, we propose a cost estimation based on combined run-time effects of various speculation factors to predict the resulting performance of candidate threads to guide the thread partitioning. For loops, we parallelize all the profitable loops that can potentially offer additional performance benefits by multilevel spawning in loop bodies, loop iterations, and inner loops. Then we select a proper thread boundary located in the front of loop branch instruction to reduce invalid spawning threads that waste core resources. Experimental results show that the proposed approach can obtain a significant increase in speedup and Olden benchmarks reach a performance improvement of 6.62 % on average.     

Communication-free data allocation techniques for parallelizingcompilers on multicomputers

In distributed memory multicomputers, local memory accesses are much faster than those involving interprocessor communication. For the sake of reducing or even eliminating the interprocessor communication, the array elements in programs must be carefully distributed to local memory of processors for parallel execution. We devote our efforts to the techniques of allocating array elements of nested loops onto multicomputers in a communication-free fashion for parallelizing compilers. We first analyze the pattern of references among all arrays referenced by a nested loop, and then partition the iteration space into blocks without interblock communication. The arrays can be partitioned under the communication-free criteria with nonduplicate or duplicate data. Finally, a heuristic method for mapping the partitioned array elements and iterations onto the fixed-size multicomputers under the consideration of load balancing is proposed. Based on these methods, the nested loops can execute without any communication overhead on the distributed memory multicomputers. Moreover, the performance of the strategies with nonduplicate and duplicate data for matrix multiplication is studied     

面向SIMD机器的全局自动数据分割

提出了一种面向ＳＩＭＤ机器的全局数据自动分割算法,该算法能处理多个非紧嵌折循环嵌套,并且数组下标存取为循环变量的线性式,首先通过数据与迭代映射抽象了计算中的通信方式,然事提出识别规则模式通信模式的形式比条件,接着建立包含对准信息和相应通信开销的数据迭代图,并在数据迭代图的基础上提出了一个启发式算法来计算较优的数据分布和迭代分布,以优化处理单元之间的通信开销,通过发析多个循环嵌套所涉及的多个数组映和     

Redundant synchronization elimination for DOACROSS loops

Cross-iterations data dependences in DOACROSS loops require explicit data synchronizations to enforce them. However, the composite effect of some data synchronizations may cover the other dependences and make the enforcement of those covered dependences redundant. In this paper, we propose an efficient and general algorithm to identify redundant synchronizations in multiply nested DOACROSS loops which may have multiple statements and loop-exit control branches. Eliminating redundant synchronizations in DOACROSS loops allows more efficient execution of such loops. We also address the issues of enforcing data synchronizations in iterations near the boundary of the iteration space. Because some dependences may not exist in those boundary iterations, it adds complexity in determining the redundant synchronizations for those boundary iterations. The necessary and sufficient condition under which the synchronization is uniformly redundant is also studied. These results allow a parallelizing compiler to generate efficient data synchronization instructions for DOACROSS loops     

Partitioning and mapping nested loops on multiprocessor systems

A method for executing nested loops with constant loop-carried dependencies in parallel on message-passing multiprocessor systems to reduce communication overhead is presented. In the partitioning phase, the nested loop is divided into blocks that reduce the interblock communication, without regard to the machine topology. The execution ordering of the iterations is defined by a given time function based on L. Lamport's (1974) hyperplane method. The iterations are then partitioned into blocks so that the execution ordering is not disturbed, and the amount of interblock communication is minimized. In the mapping phase, the partitioned blocks are mapped onto a fixed-size multiprocessor system in such a manner that the blocks that have to exchange data frequently are allocated to the same processor or neighboring processors. A heuristic mapping algorithm for hypercube machines is proposed     

Partitioning and mapping of nested loops for linear array multicomputers

In distributed-memory multicomputers, minimizing interprocessor communication is the key to the efficient execution of parallel programs. In order to reduce the amount of communication overhead, parallel programs on multicomputers must be carefully scheduled by parallelizing compilers. This paper proposes some compilation techniques for partitioning and mapping nested loops with constant data dependences onto linear array multicomputers. First, a systematic partition strategy is proposed to project ann-dimensional computational structure, representing ann-nested loop, onto a line to form a one-dimensional projected structure with low communication overhead. Then, a mapping algorithm is proposed for mapping the partitioned loops onto linear arrays in a way that balances the workload and minimizes the communication cost among processors. Finally, parallel execution codes can be automatically generated for such linear array multicomputers.     

Partitioning and labeling of loops by unimodular transformations

A general method for the identification of the independent subsets in loops with constant dependence vectors is presented. It is shown that the dependence relation remains invariant under a unimodular transformation. Then a unimodular transformation is used to bring the dependence matrix into a form where the independent subsets are obtained by a direct and inexpensive partitioning algorithm. This leads to a procedure for the automatic conversion of a serial loop into a nest of parallel DO-ALL loops. Another unimodular transformation results in an algorithm to label the dependent iterations of an n-fold nested loop in O(n²) time. This provides a multithreaded dynamic scheduling scheme requiring only one fork and one join primitive     

A Computation+Communication Load Balanced Loop Partitioning Method for Distributed Memory Systems

Due to a significant communication overhead of sending and receiving data, the loop partitioning approaches on distributed memory systems must guarantee not just the computation load balance but computation+communication load balance. The previous approaches in loop partitioning have achieved a communication-free, computation load balanced iteration space partitioning solution for a limited subset of DOALL loops. But a large category of DOALL loops inevitably result in communication and the trade-offs between computation and communication must be carefully analyzed for these loops in order to balance out the combined computation time and communication overheads. In this work, we describe a partitioning approach based on the above motivation for the general cases of DOALL loops. Our goal is to achieve a computation+communication load balanced partitioning through static data and iteration space distribution. Our approach first performs partitioning of iteration and data spaces of a loop nest by analyzing communication and parallelism; it then performs architecture-dependent analysis to adjust the granularity of partitions, load balance each partition with respect to total computation+communication, and then performs mapping of partitions onto the available number of processors. This multiphase partitioning method works as follows. First, the code partitioning phase analyzes the references in the body of the DOALL loop nest and determines a set of directions for reducing a larger degree of communication by trading a lesser degree of parallelism. The partitioning is carried out in the iteration space of the loop by cyclically following a set of direction vectors such that the data references are maximally localized and reused, eliminating a larger communication volume than parallelism. We then perform data space partitioning based on a new larger partition owns rule to minimize the communication overhead for a compute intensive partition by localizing its references relatively more than a smaller noncompute intensive partition. A partition interaction graph is then constructed which is used by the architecture-dependent analysis phase to merge the partitions to achieve granularity adjustment, computation+communication load balance, and mapping on the actual number of available processors. Relevant theory and algorithms are developed along with a performance evaluation on the Cray T3D.     

Supporting Timing Analysis by Automatic Bounding of Loop Iterations

Static timing analyzers, which are used to analyze real-time systems, need to know the minimum and maximum number of iterations associated with each loop in a real-time program so accurate timing predictions can be obtained. This paper describes three complementary methods to support timing analysis by bounding the number of loop iterations. First, an algorithm is presented that determines the minimum and maximum number of iterations of loops with multiple exits. Even when the number of iterations cannot be exactly determined, it is desirable to know the lower and upper iteration bounds. Second, when the number of iterations is dependent on unknown values of variables, the user is asked to provide bounds for these variables. These bounds are used to determine the minimum and maximum number of iterations. Specifying the values of variables is less error prone than specifying the number of loop iterations directly. Finally, a method is given to tightly predict the execution time of inner loops whose number of iterations is dependent on counter variables of outer level loops. This is accomplished by formulating the total number of iterations of a loop in terms of summations and solving the resulting equation. These three methods have been successfully integrated in an existing timing analyzer that predicts the performance for optimized code on a machine that exploits caching and pipelining. The result is tighter timing analysis predictions and less work for the user.     

PFACC: An OpenACC-like programming model for irregular nested parallelism

OpenACC is a directive-based programming model which allows programmers to write graphic processing unit (GPU) programs by simply annotating parallel loops. However, OpenACC has poor support for irregular nested parallel loops, which are natural choices to express nested parallelism. We propose PFACC, a programming model similar to OpenACC. PFACC directives can be used to annotate parallel loops and to guide data movement between different levels of memory hierarchy. Parallel loops can be arbitrarily nested or be placed inside functions that would be (possibly recursively) called in other parallel loops. The PFACC translator translates C programs with PFACC directives into CUDA programs by inserting runtime iteration-sharing and memory allocation routines. The PFACC runtime iteration-sharing routine is a two-level mechanism. Thread blocks dynamically organize loop iterations into batches and execute the batches in a depth-first order. Different thread blocks share iterations among one another with an iteration-stealing mechanism. PFACC generates CUDA programs with reasonable memory usage because of the depth-first execution order. The two-level iteration-sharing mechanism is implemented purely in software and fits well with the CUDA thread hierarchy. Experiments show that PFACC outperforms CUDA dynamic parallelism in terms of performance and code size on most benchmarks.     

Software pipelining of loops by the method of modulo scheduling

Software pipelining is an efficient method of loop optimization that allows for parallelism of operations related to different loop iterations. Currently, most commercial compilers use loop pipelining methods based on modulo scheduling algorithms. This paper reviews these algorithms and considers algorithmic solutions designed for overcoming the negative effects of software pipelining of loops (a significant growth in code size and increase in the register pressure) as well as methods making it possible to use some hardware features of a target architecture. The paper considers global-scheduling mechanisms allowing one to pipeline loops containing a few basic blocks and loops in which the number of iterations is not known before the execution.     

Automatic partitioning of parallel loops and data arrays fordistributed shared-memory multiprocessors

Presents a theoretical framework for automatically partitioning parallel loops to minimize cache coherency traffic on shared-memory multiprocessors. While several previous papers have looked at hyperplane partitioning of iteration spaces to reduce communication traffic, the problem of deriving the optimal tiling parameters for minimal communication in loops with general affine index expressions has remained open. Our paper solves this open problem by presenting a method for deriving an optimal hyperparallelepiped tiling of iteration spaces for minimal communication in multiprocessors with caches. We show that the same theoretical framework can also be used to determine optimal tiling parameters for both data and loop partitioning in distributed memory multicomputers. Our framework uses matrices to represent iteration and data space mappings and the notion of uniformly intersecting references to capture temporal locality in array references. We introduce the notion of data footprints to estimate the communication traffic between processors and use linear algebraic methods and lattice theory to compute precisely the size of data footprints. We have implemented this framework in a compiler for Alewife, a distributed shared-memory multiprocessor     

SUPPLE: An efficient run-time support for non-uniform parallel loops

This paper presents SUPPLE (SUPort for Parallel Loop Execution), an innovative run-time support for the execution of parallel loops with regular stencil data references and non-uniform iteration costs. SUPPLE relies upon a static block data distribution to exploit locality, and combines static and dynamic policies for scheduling non-uniform iterations. It adopts, as far as possible, a static scheduling policy derived from the owner computes rule, and moves data and iterations among processors only if a load imbalance actually occurs. SUPPLE always tries to overlap communications with useful computations by reordering loop iterations and prefetching remote ones in the case of workload imbalance. The SUPPLE approach has been validated by many experimental results obtained by running a multi-dimensional flame simulation kernel on a 64-node Cray T3D. We have fed the benchmark code with several synthetic input data sets built on the basis of a load imbalance model. We have compared our results with those obtained with a CRAFT Fortran implementation of the benchmark.     

Skewed Data Partition and Alignment Techniques for Compiling Programs on Distributed Memory Multicomputers

Minimizing data communication over processors is the key to compile programs for distributed memory multicomputers. In this paper, we propose new data partition and alignment techniques for partitioning and aligning data arrays with a program in a way of minimizing communication over processors. We use skewed alignment instead of the dimension-ordered alignment techniques to align data arrays. By developing the skewed scheme, we can solve more complex programs with minimized data communication than that of the dimension-ordered scheme. Finally, we compare the proposed scheme with the dimension-ordered alignment one by experimental results. The experimental results show that our proposed scheme has more opportunities to align data arrays such that data communications over processors can be minimized.     

Reducing Off-Chip Memory Access via Stream-Conscious Tiling on Multimedia Applications

The iteration space of a loop nest is the set of all loop iterations bounded by the loop limits. Tiling the iteration space can effectively exploit the available parallelism, which is essential to multiprocessor compiling and pipelined architecture design. Another improvement brought by tiling is the better data locality that can dramatically reduce memory access and, consequently, the relevant memory access energy consumptions. However, previous studies on tiling were based on the data dependence, thus arrays without dependencies such as input arrays (data streams) were not considered. In this paper, we extend the tiling exploration to also accommodate those dependence-free arrays, and propose a stream-conscious tiling scheme for off-chip memory access optimization. We show that input arrays are as important, if not more, as the arrays with data dependencies when the focus is on memory access optimization instead of parallelism extraction. Our approach is verified on TI’s low power C55X DSP with popular multimedia applications, exhibiting off-chip memory access reduction by 67% on average over the traditional iteration space tiling.     

Compiler Support for Array Distribution on NUMA Shared Memory Multiprocessors

Management of program data to improve data locality and reduce false sharing is critical for scaling performance on NUMA shared memory multiprocessors. We use HPF-like data decomposition directives to partition and place arrays in data-parallel applications on Hector, a shared-memory NUMA multiprocessor. We describe a compiler system for automating the partitioning and placement of arrays. The compiler exploits Hectors shared memory architecture to efficiently implement distributed arrays. Experimental results from a prototype implementation demonstrate the effectiveness of these techniques. They also demonstrate the magnitude of the performance improvement attainable when our compiler-based data management schemes are used instead of operating system data management policies; performance improves by up to a factor of 5.     

Programming language constructs for highly parallel operations on lists

A data structure called strips is described for representing linked lists, which enables unit time access of random list elements. Running parallel prefix on strips effectively converts a list into an array. When combined with nondeterministic statement sequencing and data operations, loops for performing iterations over lists, and insertions and deletions on lists can be parallelized yielding very efficient algorithms. The strips-based representation also allows efficient serial operations on lists, which is important both when loops cannot be parallelized or when there is more parallelism than processors.This work was supported in part under ONR Grant N00014-86-K-0215 and under NSF Grant DCR-8503610.     

Systolic opportunities for multidimensional data streams

Portable image processing applications require an efficient, scalable platform with localized computing regions. This paper presents a new class of area I/O systolic architecture to exploit the physical data locality of planar data streams by processing data where it falls. A synthesis technique using dependence graphs, data partitioning, and computation mapping is developed to handle planar data streams and to systematically design arrays with area I/O. Simulation results show that the use of area I/O provides a 16 times speedup over systems with perimeter I/O. Performance comparisons for a set of signal processing algorithms show that systolic arrays that consider planar data streams in the design process are up to three times faster than traditional arrays