–ѕа°±б>ю€ %'ю€€€$€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€м•Ѕq`рњ)bjbjqPqP8::)€€€€€€§®®®®®®®Љdddd p Љ36ИИИИИИЮ ™≤іііііі$ih— јЎ®≤ИИ≤≤Ў®®ИИн^^^≤L®И®И≤^≤≤^^®®^И| p≠ЬJ7—dю@^Ц03^С >С ^С ®^8≤≤^≤≤≤≤≤ЎЎN≤≤≤3≤≤≤≤ЉЉЉd DЉЉЉ ЉЉЉ®®®®®®€€€€ RESUMO Version control system is largely used for controlling modification over artifacts across the time, presenting a valuable source of information. However, depending on the repository size and its lifetime, extracting such information is not an easy task, as a large amount of data needs to be processed. For solving this problem, some authors reduce the grain of analysis (such as just using files) or the amount of history that can be analyzed to be feasible its processing in a reasonable amount of time. However, such constrains make the result less accurate. Besides that, version control system is normally used for managing textual artifacts, where diff, patch, and merge operations make sense. When artifacts cannot be represented as textual, they are treated as binary artifacts in order to avoid jeopardizing the user productivity processing them. For these artifacts, the aforementioned operations are not available. Additionally, such files are saved as a whole for each revision made over these artifacts. As a consequence, binary artifacts do not provide sufficient data for extracting knowledge over them. In this case, repository projects composed mostly of these artifacts will not provide accurate information during analysis. In this thesis we aim to provide a GPU approach for allowing large knowledge repository exploration, considering fine grain data like methods, classes, and files. On the other hand, we also provide mechanisms for allowing version control over binary artifacts (specially image and video) through specialized diff, patch, and merge operations. Version control over binary files allows a better comprehension of modifications and storage space and network bandwidth reduction. Palavras Chave: Version Control; Image; Video; Expertise; Dependencies; GPU; CUDA B H у щ ‘÷з)ся“ њ њ њ“µЃ h€ЊhgV{h€ЊhgV{5Б\Бh%:_hgV{mH sH hgV{mH sH h%:_hgV{^JmH sH "h€ЊhgV{5БCJ \БaJ mH sH h0M`5БCJ \БaJ mH sH  @ B с у ‘’÷з)чтккккктттт$a$gd€Њgd%:_$a$gd€Њ )э61РhP:pЙ=№∞|. ∞»A!∞•"∞•#РЙ$РЙ%∞∞ƒ∞ƒ РƒЖЬ 666666666666666666666666666666666666666666 6666666666 666666666666 666666666666666666666666666666666666666666666666666666666666666666D@с€D IЭNormalCJ^J_HaJmHsHtH >Aт€°> 0Fonte parбg. padrгoTiу€≥T 0 Tabela normalц4÷ l4÷aц ,kф€Ѕ, 0 Sem lista )€€€€@Bсу‘’÷з+Ш0ААШ0ААШ0ААШ0ААШ0ААШ0ААШ0ААШ0ААШ0ААШ0ААШ0АА+KИ0IИ0а/)) ) €€_GoBack÷+ж+зопцшэ€+ў№}Бю$(+333++е%:_0M`gV{IЭ€Њ%qƒЙ=№(<у÷з+€@А4Cѕ)P@€€Unknown€€€€€€€€€€€€GР€:аAxј €Times New Roman5РАSymbol3&Р €:аCxј €ArialGРАMS ??MS Mincho7Р€а€@ЯCambria"AИр–©—2;ІI{