Compiling C for the EARTH multithreaded architecture

Hendren, Laurie; Tang, Xinan; Zhu, Yingchun; Gao, Guang R.; Xue, Xu; Ouellet, P.

doi:10.1109/pact.1996.552551

Cited by 19 publications

(16 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The benchmarks pug and circuit use cyclic data structures, but perform majority of their computation inside acyclic list-traversing loops, which are detected as ]oreach loops by the heapLCD test. Finally, the hand-written Earth-C [10] versions of the benchmarks treeadd and power, that only use the parallelism detected by our dependence tests, respectively obtain speed-up by factors of 16 and 12 on the EARTH-MANNA multithreaded machine [10] using 16 processors. We are presently working on analyzing and collecting runtime performance improvement statistics for a larger set of benchmarks.…”

Section: Identifying Foratl Loopsmentioning

confidence: 97%

Detecting parallelism in C programs with recursive data structures

Ghiya

Hendren

Zhu

1998

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Abstract. In this paper we present techniques to detect three common patterns of parallelism in C programs that use recursive data structures. These patterns include, function calls that access disjoint sub-pieces of tree-like data structures, pointer-chasing loops that traverse list-like data structures, and array-based loops which operate on an array of pointers pointing to disjoint data structures. We design dependence tests using a family of three existing pointer analyses, namely points-to, connection and shape analyses, with special emphasis on shape analysis. To identify loop parallelism, we introduce special tests for detecting loop-carried dependences in the context of recursive data structures. We have implemented the tests in the framework of our McCAT C compiler, and we present some prehminary experimental results. I n t r o d u c t i o n a n d M o t i v a t i o nThis paper focuses on detecting three common patterns for parallel computations that use recursive data structures: (1) function-call parallelism including parallel recursive calls on tree-like structures; (2) forall parallelism for loops traversing arrays of list/tree-like structures; and (3) foreach parallelism for loops traversing list/tree-like structures, which is similar to doacross parallelism.In order to safely detect these patterns of parallelism in C programs, one must deal with dependences due to scalars, dependences due to pointers to stackallocated objects (stack-directed pointers), and dependences due to pointers to heap-allocated objects (heap-directed pointers). Thus, our approach uses the results of the family of pointer analyses that have been implemented in the McCAT optimizing/parallelizing C compiler: points-to analysis [l], connection analysis [2] and shape analysis [3]. Points-to analysis is used to detect dependences due to scalars and stack-directed pointers, while connection and shape analysis are used to detect dependences due to heap-directed pointers.The main focus of this paper is not the pointer analyses themselves, but rather how we can use the results of the analyses to detect parallelism. The remainder of the paper is structured as follows. In Section 2 we introduce the three parallelism patterns in more detail. In Section 3 we describe the overall setting of our approach, and present the rules to detect function call parallelism. * This research supported in part by NSERC and FCAR.

show abstract

Section: Identifying Foratl Loopsmentioning

confidence: 97%

Detecting parallelism in C programs with recursive data structures

Ghiya

Hendren

Zhu

1998

Lecture Notes in Computer Science

View full text Add to dashboard Cite

show abstract

“…The topics described in this paper have been inspired by our compiler research work for non-preemptive threads from programs written in explicitly parallel procedural languages designed for multithreaded machines such as the EARTH-C [15]. The compilation procedure is as follows:…”

Section: Spaa 98 Puerto Vallarta Mexicomentioning

confidence: 99%

“…In previous work, we have developed a framework to partition threads automatically from user programs [29], and implemented a simplified version on a research compiler [15]. We have also studied threaded-code optimizations [28], and compiled a suite of benchmarks to run on the multithreaded EARTH emulator [16].…”

Section: Spaa 98 Puerto Vallarta Mexicomentioning

confidence: 99%

See 1 more Smart Citation

How “hard” is thread partitioning and how “bad” is a list scheduling based partitioning algorithm?

Tang

Gao

1998

Proceedings of the Tenth Annual ACM Symposium on Parallel Algorithms and Architectures - SPAA '98

View full text Add to dashboard Cite

Adequate compiler support is essential to take advantage of the emerging multithreaded architecture. In this paper, we address two important questions in thread partitioning, which is a key step in compiler design for multithreaded architectures. The questions in which we are interested are: how "hard" is it to partition threads and how "bad" will a heuristic partitioning algorithm be? We propose a cost model for both multithreaded machines and user programs, and we formulate the thread partition problem as an optimization problem. Then, we answer the above two questions by proving that: 1) for the class of programs and architecture models we are interested in, the problem of thread partition for minimum execution time is NP-hard; 2) the run length produced by any list scheduling based thread partitioning algorithm is at most twice as long as that of an optimal solution.

show abstract

“…EARTH Threaded-C is the basic language dealing explicitly with threads and remote data accesses. EARTH-C is supported by the McCAT compiler [12] and hides remote data accesses and tbread handling, i.e. it traaslatm programs written at an abstract level (treelike parallelism with communication being hierarchical between parent and children but not taking place between siblings) into multithreaded code.…”

mentioning

confidence: 99%

Experiences with non-numeric applications on multithreaded architectures

Sodan¹,

Gao

Maquelin

et al. 1997

Proceedings of the Sixth ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming

View full text Add to dashboard Cite

Distributed-memory machines have proved successful for many challenging numerical programs that can be split into largely independent computation-intensive subtasks requiring little data exchange (although the amount of exchanged data may be large). However, many irregular applications -e.g. in the AI field -have a fairly tight data coupling that often results from the use of shared data structures, making them in many cases not amenable to parallelization on distributed-memory machines. EARTH is an efficient multithreaded architecture that supports in particular large numbers of small data exchanges by means of low start-up times and the ability of tolerance of even small latenci=. In this paper, we show the benefits provided by EARTH for applications of this sort by presenting experimental results from several AI applications run on the MANNA machine, which is a distributed-memory machine with a very highperformance communicantion network. EARTH-MANNA is shown to extend the range of programs that can be parallelized and run effectively on distributed-memory machines.

show abstract

Compiling C for the EARTH multithreaded architecture

Cited by 19 publications

References 13 publications

Detecting parallelism in C programs with recursive data structures

Detecting parallelism in C programs with recursive data structures

How “hard” is thread partitioning and how “bad” is a list scheduling based partitioning algorithm?

Experiences with non-numeric applications on multithreaded architectures

Contact Info

Product

Resources

About