基于GPU的无损压缩算法的研究与实现-计算机软件与理论专业论文.pdf.docxVIP

a 曰 a 曰.人 学 硕 ：丨.?学 位 论 文 A b s tr a c t W ith th e rap id d ev elopm ent of in form ation tech n ology, data is grow ing rap idly. It is in creasin gly u rgent for effectiv e inform ation storag e and transm ission du e to lim ited sto rag e an d n etw ork b an dw id th . D ata co m p ressio n is on e o f th e so lu tio n s. H ow ev er, d ata co m p ression o ften requ ires ad d ition al exp en ses, esp ecially th e h u g e am ou n ts o f d ata. T h erefore, n ew h igh -p erform an ce tech n o lo g y is n eed ed to d esign m ore efficient and h igh er qu ality algorithm s to ach iev e th e m assive data com p ression . H ow to redu ce com p ression tim e is v ery im p ortan t. In recen t years，h igh -p erform an ce g en eral com p u tin g o f G P U s is d ev elop in g qu ickly. G P U is far b eyond th e C P U in flo ating-poin t com pu tation , p arallel com pu tation and m em ory b andw idth .H eterog en eou s processing platform s of th e C P U + G P U h av e b ecom e m ain stream p arallel so lu tio n . T h is stu dy is to d esign an d op tim ization o f p arallel lossless com p ression alg orith m to enh ance th e com pression p erform an ce o f th e algorithm u sing high p arallel com p uting p ow er o f C P U + G P U . In th is stu d y，th e arch itectu re o f th e G P U w as d iscu ssed w ith d etailed d escrip tio n o f h ardw are arch itectu re an d th e so ftw are o f th e C U D A p ro gram m in g m o d el an d th e optim ization strategy for C U D A pro gram . T h en , th e b ase of th e com pression algorithm w as d escrib ed . F in ally, tw o lossless com p ression algorithm of G P U w as realized : th e L Z 7 7 an d b zip 2 algo rith m s. T h e alg orith m cou ld im p ro v e th e com p u tatio n al th rou gh p u t b y lo ad in g com p u te-inten siv e p art o f th e com p ressio n algorithm on G P U from C P U . A nd th en th e op tim ized d esign w as p erform ed com b in in g th e ch ara