4' 1 i 'b~c~ei: DISCLAIMER Thim re@ wus prepared ununum)unl ofworkrn~nmrdhy unugency ofthc(Jnitd Stntcs (kwcrnmerrt. Neither the (Jnitcd Statc#Govcrnmcnt nor any ugcncythcrcof, rmr unyof their cmplnyccm mukc~mry wurrunty. cxprcM or implied, or nxwmernmy Icgul Iiuhilily or ren~rrsihility fnr the ntxxmcy, wrlplctcrrcns,or uscfulncrworwry inrormulinn, nppnrntuw PIWIIUCI,m procwssdIrichMA or represents Ihui its uw wnold not inrringc prlvntcly owned rights. I:crcrcncc herein 10 wry specific

more »

... l producl, prtnmst t,r wrvicc hy trndc rrumc, trudcmnrk, mmmrwturcr, or olhcrwim dnm nol rrcecmnrily wrntitulc nr imply il~cndmcmcnt, rcctmmcmhriinn. or hwnring hy the (Irrikrl .StHIcn (invcrnmcnl or HIIy Ilgcncy Ihcrcwf. The vicwñ nd npininnrnor tiuthor~cRprcmal hcrcln do not mxcxmrily Ate (W rcflcul th(~c or Ihc []nilAStutcn (;otcrnmclll~)ronyugcncy thcrcnf. LosAllamilos DISIIIINIIH)NW llllSIW,lMEHl lSUHllMmD I compare discrete-ordinates and Monte Carlo techniques for solving integrodifferential equations and compare their relative adaptability to vector processors, 1 discuss the utility of multiprocessors for Monte Carlo calculations and describe a simple architecture (the monodirectional edgecoupled array or MECA) that seems ideally suited to Monte Carlo and overcomes many of the packaging problems associated with more general multiprocessors.