ࡱ> -/,bjbjUU >?? ((kkkkk  (JLkkkkk E0(k (( 1:Resumo: Pre-stack Kirchhoff Time Migration (PKTM) uma parte central do processo de explorao de petrleo. Como o PKTM computacionalmente intensivo, trabalhos anteriores propuseram verses paralelas do PKTM para memria compartilhada, memria distribuda e aceleradores para reduzir o tempo de execuo. Os processadores modernos, no entanto, tm capacidade de processamento de vetores que geralmente so negligenciados na otimizao de desempenho. Neste trabalho, propomos uma anlise detalhada do uso da vetorizao em trs verses paralelas do PKTM: uma implementao de memria compartilhada com OpenMP (Open Multi-Processing), uma implementao de passagem de mensagem com MPI (Message Passing Interface) e uma implementao hbrida que usa o OpenMP e o MPI. Como a computao em nuvem provou ser uma alternativa promissora para executar aplicativos HPC, tambm avaliamos o impacto negativo que a sobrecarga introduzida pela camada de virtualizao pode ter nas execues implementadas do PKTM. Assim, analisamos o desempenho dessas verses sequenciais/paralelas e vetorizadas do PKTM quando executadas em tecnologias de virtualizao comumente usadas em nuvens: KVM (Mquina virtual baseada em kernel), Docker e Singularity. Nossos testes experimentais mostraram que a combinao da vetorizao com a paralelizao apresentou os melhores resultados, produzindo speedups de at 24,9 quando executado com 6 ncleos. Com relao aos overheads de virtualizao, observamos que o KVM apresentou uma pequena degradao no desempenho enquanto que os ambientes baseados em continer apresentavam um desempenho quase nativo. Palavras-chave: Kirchhoff, paralelismo, vetorizao, virtualizao. Abstract: Pre-stack Kirchhoff Time Migration (PKTM) is a central part of the oil exploration process. Since PKTM is computationally intensive, previous works have proposed parallel versions of PKTM for shared memory, distributed memory and accelerators to reduce the execution time. Modern processors, however, have vector processing abilities that are usually neglected in performance optimization. In this work, we propose a detailed analysis of the use of vectorization in three parallel versions of PKTM: a shared memory implementation with OpenMP (Open Multi-Processing), a message passing implementation with MPI (Message Passing Interface) and a hybrid implementation that uses both OpenMP and MPI. As cloud computing has proven to be a promising alternative to execute HPC applications, we also evaluated the negative impact that the overhead introduced by virtualization layer may have on the implemented PKTM executions. Thus, we analyzed the performance of these sequential/parallel and vectorized versions of PKTM when executed upon virtualization technologies commonly used on clouds: KVM (Kernel-based Virtual Machine), Docker and Singularity. Regarding the vectorization analysis, our experimental tests showed that the combination of the vectorization with the parallelism offered the best results, producing a speedup up to 24.9 when executing in a processor with six cores. Concerning the virtualization overheads, we observed that KVM showed a small degradation in performance while container-based environments presented near native performance. Keywords: Kirchhoff, Parallelism, Vectorizing, Virtualization.   \ d Z[\jŰŰŰכׅldhP(mH sH 1hP(5B*CJOJQJ\^JaJmH phsH +hP(B*CJOJQJ^JaJmH phsH )hP(5B*CJOJQJ\^JaJph)hP(6B*CJOJQJ]^JaJph#hP(B*CJOJQJ^JaJphhP(CJOJQJ^JaJhP(,hP(5>*B*CJOJQJ\^JaJph [\$a$ <P1h:pP(. A!"#$% Dp^ 666666666vvvvvvvvv66666686666666666666666666666666666666666666666666666666hH6666666666666666666666666666666666666666666666666666666666666666662 0@P`p2( 0@P`p 0@P`p 0@P`p 0@P`p 0@P`p 0@P`p8XV~_HmHnHsHtHH`H Normal*$CJ_HaJmHnHsHtHDA D 0Default Paragraph FontRiR 0 Table Normal4 l4a (k ( 0No List XX 0HTML Preformatted CharCJOJQJ^JaJJJ 0Heading $xCJOJQJ^JaJ8B8 0 Body Text d D/!D /0Body Text CharCJaJnHtH$/2$ 0List<"B< 0Caption  $xx6]*R* 0Index $e@b 0HTML Preformatted7 2( Px 4 #\'*.25@9CJOJQJ^JaJb/qb /0HTML Preformatted Char1CJOJQJ^JaJnHtHPK![Content_Types].xmlj0Eжr(΢Iw},-j4 wP-t#bΙ{UTU^hd}㨫)*1P' ^W0)T9<l#$yi};~@(Hu* Dנz/0ǰ $ X3aZ,D0j~3߶b~i>3\`?/[G\!-Rk.sԻ..a濭?PK!֧6 _rels/.relsj0 }Q%v/C/}(h"O = C?hv=Ʌ%[xp{۵_Pѣ<1H0ORBdJE4b$q_6LR7`0̞O,En7Lib/SeеPK!kytheme/theme/themeManager.xml M @}w7c(EbˮCAǠҟ7՛K Y, e.|,H,lxɴIsQ}#Ր ֵ+!,^$j=GW)E+& 8PK!Ptheme/theme/theme1.xmlYOo6w toc'vuر-MniP@I}úama[إ4:lЯGRX^6؊>$ !)O^rC$y@/yH*񄴽)޵߻UDb`}"qۋJחX^)I`nEp)liV[]1M<OP6r=zgbIguSebORD۫qu gZo~ٺlAplxpT0+[}`jzAV2Fi@qv֬5\|ʜ̭NleXdsjcs7f W+Ն7`g ȘJj|h(KD- dXiJ؇(x$( :;˹! I_TS 1?E??ZBΪmU/?~xY'y5g&΋/ɋ>GMGeD3Vq%'#q$8K)fw9:ĵ x}rxwr:\TZaG*y8IjbRc|XŻǿI u3KGnD1NIBs RuK>V.EL+M2#'fi ~V vl{u8zH *:(W☕ ~JTe\O*tHGHY}KNP*ݾ˦TѼ9/#A7qZ$*c?qUnwN%Oi4 =3ڗP 1Pm \\9Mؓ2aD];Yt\[x]}Wr|]g- eW )6-rCSj id DЇAΜIqbJ#x꺃 6k#ASh&ʌt(Q%p%m&]caSl=X\P1Mh9MVdDAaVB[݈fJíP|8 քAV^f Hn- "d>znNJ ة>b&2vKyϼD:,AGm\nziÙ.uχYC6OMf3or$5NHT[XF64T,ќM0E)`#5XY`פ;%1U٥m;R>QD DcpU'&LE/pm%]8firS4d 7y\`JnίI R3U~7+׸#m qBiDi*L69mY&iHE=(K&N!V.KeLDĕ{D vEꦚdeNƟe(MN9ߜR6&3(a/DUz<{ˊYȳV)9Z[4^n5!J?Q3eBoCM m<.vpIYfZY_p[=al-Y}Nc͙ŋ4vfavl'SA8|*u{-ߟ0%M07%<ҍPK! ѐ'theme/theme/_rels/themeManager.xml.relsM 0wooӺ&݈Э5 6?$Q ,.aic21h:qm@RN;d`o7gK(M&$R(.1r'JЊT8V"AȻHu}|$b{P8g/]QAsم(#L[PK-![Content_Types].xmlPK-!֧6 +_rels/.relsPK-!kytheme/theme/themeManager.xmlPK-!Ptheme/theme/theme1.xmlPK-! ѐ' theme/theme/_rels/themeManager.xml.relsPK]    P(x  @ @@UnknownG*Ax Times New Roman5Symbol3" Arial?= *Cx Courier NewG& xP!Liberation SansACambria Math",3i,3i !0$P x!xxResumo:HelioHelioOh+'0x  4 @ LX`hpResumo:HelioNormal_WordconvHelio2Microsoft Office Outlook@@mE@mE ՜.+,0 hp|   Resumo: Title  !"#%&'()*+.Root Entry F+E01TableWordDocument>SummaryInformation(DocumentSummaryInformation8$CompObjy  F'Microsoft Office Word 97-2003 Document MSWordDocWord.Document.89q  F#Documento do Microsoft Office Word MSWordDocWord.Document.89q