Contractor Rebrt : 181763 ICASE REPORT

In this paper, we extend the class of problenis t h a t can be effcrtively m n piled by parallelizing compilers. This is accornplislictl with the doconsider coi l st ruct which would allow these compilers t o paralldizc~ many prohlcnis i n \v l i i t 11 siil)staiitial loop-lcvel parallelisin is available h i t cattnot bc tlctcctctl hy staiitl;irtl corn pile-t iriie analysis. \Ire describe and espcriincii till I J analyze mcdia nisni.; I I srd to parallclize the work required for these types of loops. I n each of t hcse met Iiotls. a new loop structure is produced hy modifying the loop to be parallclizcd. \\'P also prcscnt t he rules by wliicli these loop transformations may be autoiiiatcd i n oi t1c.r. tha t they be iriclridccl in langiiage conipilers. Tlic i t i a i i i application a1c.n of o u r wscarcli involves problems i n scientific computations aittl engineering. The wot kIo;it l iiscd i n our experirnents iitcludcs a mixture of real ~)i~)l)lcins as wc4l as synt 1101 i rally gcncratetl inputs. I2rorn our estcwsivc tests oii t l i t 1 I:iicorc hl i i l t i i t t ; i s / : l~ (~ , IVV have reaclicd t he conclusion tha t for t he types of \voi liloatls \vv Iiavc. i t ivfs l igat ( V I , self-execution almost always perforins better tltitii ~~tc~-sc~l iccl i~l i~ig. b'iirtlier. 1 I I V i r i i provcment in performance t h a t accriics as a resiil t of global topological sori irig of indices as opposed to the less experisivc local sortiiig. is not very sigiiifirant i r i t I I P case of self-execution. 'This research was supported by the U.S. Office of Naval Research under Grant N0001486K-0310, the United States Air Force Office of Scientific Research under Contract NO. AFOSR Grant No. AFOSR 88-0117, and the National Aeronautics and Space Administration under NASA Contract No. NAS1-18107 while the first author was in residence a t the Institute for Computer Applications in Science and Engineering (ICASE), NASA Langley Research Center, Hampton, VA 23665.