J.-F. Collard. Space-time transformation of while-loops using speculative execution. In proc. of the Scalable High-Performance Computing Conf.(SHPCC'94), pages 429--436. IEEE, May 1994.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....parallelizer X and is called Meta X. The PAF [PAF90] parallelizer is a good candidate for LPARAD since it performs a very efficient and a very costly parallelism detection. Moreover, a set of specialized compilers is currently under development in the scope of the PAF project[Dum92, Red92, Col94] Our current implementation of LPARAD in the scope of the PAF project, Meta PAF, validates our approach. The source language of PAF and Meta PAF is a subset of Fortran 77: EQUIVALENCE, ENTRY, computed and assigned goto s are forbidden. The target machine is an Encore Multimax and EPF (Encore ....

....forms of parallelism (and generally more costly) might succeed. The following table presents the set of potential causes of the parallelization inefficiency currently indexed by LPARAD: Symptoms Inefficiency causes Remedies While loop While loop not parallelizable by Parallelize the while loop [Col94] or a classical parallelizer manually transform it into a do loop Non linear array references No detectable parallelism by a Parallelize the nest using [Dum92] classical parallelizer Cycle in the DG of the nest Possible presence of a recurrence Submit the nest to [Red92] Flow dependence ....

J.-F. Collard. Space-time transformation of while-loops using speculative execution. In Proc. of the 1994 Scalable High Performance Conputing Conf., Knoxville, Tenn., May 1994. To appear. Also available as LIP Report 93-38.

....synchronous parallelism because, in this case, there are no scannable space time mappings for a loop nest with at least one while loop that is not at the outermost level. Our approach is conservative in the sense that the target program does not execute unnecessary while loop iterations. Collard [3] permits the speculative execution of iterations that might afterwards turn out to be unnecessary. This may actually be beneficial in certain applications: Collard restricts himself to nests of for loops with exactly one outer while loop. Both his and our method achieve linear execution time, ....

J.-F. Collard. Space-time transformation of while-loops using speculative execution. In Proc. 1994 Scalable High Performance Computing Conf. IEEE Computer Society Press, May 1994. 26

....and barrier, which can impose substantial performance penalties on some systems. In a few cases, such as when a natural ordering is used in a relaxation algorithm, the inner loops carry dependencies and cannot be run in parallel. To exploit parallelism in these loops, a number of researchers [1,4,5,2,7,3] have proposed speculative execution: a wavefront technique is used to execute the program in parallel, despite the fact that all loops carry dependences. Since this ignores the termination condition of the while loop, iterations of the while loop are executed speculatively until each iteration ....

....is necessary to determine if the computation has converged. Our technique also allows us to provide efficient doacross pipelined parallelism when the body of a while loop contains cross processor dependencies. We believe the technique we propose is more practical than speculative execution [1,4,5,2,7,3]. In a language like HPF, the transformation we describe has to be performed by the compiler; there is no way for the user to express a reduction over local data and make a decision based on that. In the experiments we performed, for most of the computation, local data alone was sufficient to ....

J.-F. Collard. Space-time transformation of while-loops using speculative execution. In Proc. of the 1994 Scalable High Performance Computing Conf., pages 429--436, Knoxville, TN, May 1994. IEEE.

....As before, E and (20) imply (18) Thus Q 1 SR (w) fw 0 j 1 w 0 ; w 0 = wg and K 1 SR (w) w. Note that if the only while in the source program is the outermost loop, then k p holds, and Property 1 proves that the dataflow analysis is exact. This result justifies a conjecture in [4]. Another observation is that for static control programs the hypotheses of this section are trivially verified: this is the stage at which fuzzy ADA and exact ADA meet. The case of conditionals A similar result holds for conditionals. Let C be a conditional at depth k enclosing two statements S ....

J.-F. Collard. Space-time transformation of while-loops using speculative execution. In Proc. of the 1994 Scalable High Performance Computing Conf., pages 429--436, Knoxville, Tenn., May 1994. IEEE. 30

....This can be enforced by the introduction of compensating dependences. For more details, the reader is referred to [31] Let us focus on the second problem. When there is a single and outermost while loop, a necessary and sufficient condition for correctness is that fronts must be finite [15]. Our aim here is to give a termination criterion for more general speculative programs. Together with the preceding theorem, it will entail the total correctness of the object program. Firstly, we must have some idea about the structure of the target code. As it is the case for synchronous ....

....are finite. The complete sufficient condition for the total correctness of the target speculative program is given described in [31] Obviously, the more speculative the schedule, the larger the memory needed to store intermediate uncommitted result. For simple cases, this issue is addressed in [15]. The interplay of expansion and speculative scheduling, in the general case, is left for future work. In the rest of this paper, schedules are supposed to be non speculative. 5.4 Related Work The scheduling problem has been widely studied since the first Kennedy and Allen algorithm. It is not ....

J.-F. Collard. Space-time transformation of while-loops using speculative execution. In Proc. of the 1994 Scalable High Performance Computing Conf., pages 429--436, Knoxville, TN, May 1994. IEEE.

....hS0 ; j Gamma 2i else (23) and S 1 S 0 S 1 (i; w; j) if j = i 2 then hS0 ; ii else : 24) Note that if the only while in the source program is the outermost loop, then d p always holds, and Property 1 proves that the dataflow analysis is exact. This result justifies a conjecture in [3]. The case of conditionals. A similar result holds for conditionals. Let C be a conditional at depth c enclosing two statements S and R. S and R are thus governed by the same predicate , meaning that the environment includes ( y[1: c] while the definition of the candidate set Q p SR ....

J.-F. Collard. Space-time transformation of whileloops using speculative execution. In Proc. of the 1994 Scalable High Performance Computing Conf., pages 429--436, Knoxville, Tenn., May 1994. IEEE.

....predicate, this entails that y[1: M ] is a parameter of the maximum and the analysis is exact. An exemple of such an exact case is when the only while loop in the source program is the outermost statement. This result was proved by other, less general means in[4, 3] and justifies a conjecture in [2]. 4.4 Constructing properties on parameters The problem is now to deduce properties of the parameters from properties of their domains. This is a rather technical question, which is treated in detail in the appendix in the case of relationnal properties. Just to give a feeling of what is ....

J.-F. Collard. Space-time transformation of while-loops using speculative execution. In Proc. of the 1994 Scalable High Performance Computing Conf., pages 429--436, Knoxville, TN, May 1994. IEEE.

....this entails that y[1: M Sk ] is a parameter of the maximum and the analysis is exact. An example of such an exact case is when the only while loop in the source program is the outermost statement. This result was proved by other, less general means in [5, 6] and justifies a conjecture in [4]. 6 Constructing Properties on Parameters In the previous section, the purpose was to extract properties P on the parameter domains. The purpose of this section is to derive properties P on parameters of the maximum from properties P on parameter domains, without forgetting sources ....

J.-F. Collard. Space-time transformation of while-loops using speculative execution. In Proc. of the 1994 Scalable High Performance Computing Conf., pages 429--436, Knoxville, TN, May 1994. IEEE.

No context found.

J.-F. Collard. Space-time transformation of while-loops using speculative execution. In proc. of the Scalable High-Performance Computing Conf.(SHPCC'94), pages 429--436. IEEE, May 1994.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC