光并行計(jì)算及其在圖像處理和計(jì)算分子生物學(xué)中的應(yīng)用.pdf

1、揚(yáng)州大學(xué)碩士學(xué)位論文光并行計(jì)算及其在圖像處理和計(jì)算分子生物學(xué)中的應(yīng)用姓名：徐曉華申請(qǐng)學(xué)位級(jí)別：碩士專業(yè)：計(jì)算機(jī)應(yīng)用技術(shù)指導(dǎo)教師：陳崚20050301徐曉華光并行計(jì)算及其在圖像處理和計(jì)算分子生物學(xué)中的廊用AbstractOpticalparallelcomputationhasadvantagesofprecisedelay，hi曲communicationspeedandbandwidth，highreliabilityandabili

2、tyofprocessinglargeamountofdatasimultaneouslyInthisthesis，westudyopticalparallelcomputingandpresentalgorithmsforimageprocessingmadcomputationalmolecularbiologyontheopticalbuscomputationalmodelnamedLARPBS(LinearArraywitha

3、ReconfigurablePipelinedOpticalBusSystem)WefirstintroducetheopticalcomputationalmodelLARPBSmadtheprimitiveoperationsandalgorithmsofmatrixesmultiplication，sortingonthemodelIntheareaofimageprocessing，wepresentfastLARPBSalgo

4、ritlunsofHoughtransformandEuclideandistancetransformIntheareaofcomputationalmolecularbiologywepresentLARPBSalgorithmsforlongestcommonsubsequenceandsequencesalignmentForallimagewith胛。胛pixelsollrLARPBSHoughtransformalgorit

5、hmCarlbecompletedinD(1)timeusing卅∥processorsandgettheoptimalspeedandefficiencyForanimagewithn?！皃ixelsthefirstEDTalgorithmonLARPBScanbecompletedinO(109nloglogn／logloglogn、timeusing∥processors，whileoursecondEDTalgorithmcan

6、computetheEDTin0(109nloglogn)timeusingO(n。／loglogn)processorsComparedwiththebestparallelEDTalgorithmsonLARPBSreportedSOfar，ouralgorithmshavethelowesttime—areacostandaremoreefficientOurthirdalgorithmcarlprocessEDTtransfor

7、minO(nlogn／(《“)logd(n)))timewith“。馥國(guó)’c積)processors，wherec(療)，a(n)sarisfylSc(矸)S以，l硪玎)：卸respectivelyByselectingdifferentc(n)and碳竹)，thetimecomplexityandthenumberofprocessorsusedcanbeadjustedThismakesthealgorithmhi曲lyscalab

8、leandflexibleThealgorithmisalsoageneralframeworkofEDTalgorithmsonLARPBS，manyexistingandunknownparallelEDTalgorithmscallbededucedfromthisframework，Particularlyifweletc(“)=n，硪n)=月5thealgorithmcarlaccomplishinO(1)timewith“2