“@ Wansiator » Ary en Y Progiam writien In a programming language must be translated before it can be execu i ion ls 4yti ect. This Wanstation is typically accomplished by a software system called “Wanslator . ~ joa High-level language and “the target language ¥ @ compiler : Ff ‘the Source axjact code Than ‘tie translator is colled compiler pinta prea; Certain offer ranstater ‘ansfrm™, © programming tao frguoge 7 egled interpreter - an intermediate code , which can be chroctty executed - compiler Interpreter = Faster — shee =S one lin at a Time > whible SIC prog large and tomper a small’ ond Simple probem 7 ~ => probeate : -) 22, OTanvatet :: Any proatan ver con be weve Nogtar wivition In a programming language must be Wanslated befove it Heck. This ae een is Wanslation is typically accomplished by 2 sollware system called “Wanslator . Target program Ocompijer : Jj the sowce is a High-level language and “The target language. 1 ovject code “Thon Tie translator is coed & complet gramming neo Janguoge ‘into interpreter ; certain other ran stator hiansfrm 0 0 coed wnterpreler - an ‘intermediate code , which can be chucetty executed - Sic program Intermediate Code “opject Code couple Interpreter — slower = Faster . > whok Sic p19 —s one line at a ‘Time 3 srl and Simple -s large and complex Probiom prokem# imcde a define Machine conte | code translater - | a , Se source coxe: Source code i the input fo % compiler or otntt uctions thay the processor Con @ Object Code: Contains a sequence of instr 4o read of modify « “understand bur that difficult foro, human |. exe — object code (05/ application software) Pb can be executed by 4 code and perform The platform ‘nde pendent « 1@ Byte code : Byte code is the “Kind of code: thal | virtual mocking . The virtual machine reads the byt “| cpetations ; it specifies . Byte, Code is typically slower but tis directly executable by the |@ Machine code: Machine code is code tha Load into memory and s computer’s Physical prowssor without further cranstation + ‘ Wate: 08: 6-22 “fh Execution of a program i> level tang Machine dependent Hochine cede (Relocated) Machine code.ie fom : aa A { Symot tabe Wtermectiate Code gen manager Code optimization end » bependent / smineats, Part ‘Assembly code | Qualities of a Good compiler > What qualities would you want in a compiles ? + Gonorates correct code (first and feremst) a Semantic analy | pase “hie (Gemaeh Fioy handir ut larg indereetf Use oF compiler Technology program 10 object code symthesi | |-s Most common use: “Hranslate a high-level + progiam Translation : binary transtation , nardware IP Optimizations fr computer architoctuses © |e tmprove progam performane. , fare ine paralla ism - etc account “hardware > Autematic parallisation ov voctorisation | 5 performance nsumentation «eg " ~P9 oplios oF ce oY HE 5 inlerpretors + e.g + Python, Ruby 1 Perl ) matlab Sh. Ly software productivity tb {| = Detwugging aids + e.g. purdy L, security : Jove VM Uses compiler ainalysls to Prove walety’ of yava woke Lo ctext fivmatters just -in— tine compilation for jova “power global list butect computing Monagemnr , ] fey 5 Ability extract properties of a Source program (analysis) and ranyiom “A te construct a Target program Gynthes) « | Surnmasy +i lexical Analysis C scanning) Reacts Character in the Sour. program afd groups them inte word> Coaric unit 6 syrlax) Produces words and recognises what sort they are >The output is called token and s/o palrof tne form © oF orth = 103s» attibute > 4 rath stele msn ‘Tih Syntax (or syntactic) Analysis CParsing) aunty > imposes © Fierrarchicol structure on the toten steam > This hiertarchical structuie ts usually ‘expressed by recursive: tules - > context-free grammar > formalise “these recursff Ast for b*b-44#ax*c “ /\ aN ceo sida i #1 = An abstract Syntax Tree (45T) is a more useful data structure tor intemal representation. df is a compressect version d the parse area .C54R +ASTs ate one form of IR. (fh Semantic Analysis (context handling) > collects context (semantic )infosmation , cheeks for semantic error s- and annvtes nodes oF the tree with the results. Examples : —4ype checking : report efor if an operator is applied to an ‘ncompete operand fi Ly fout t 5 =Check flow - Of - contols = uniqueness ‘or name-related checks. Th Intermediok code generation :Th coce Optimisation: | sate gpay 6 tb improve ane intermediate, code, and THe tre. effectiveness oF cece gereration ond Perfor ane danger (ose * e from trivial (e:9- mance of = optirigations can yang constant folirg ) te highly sophisticated Ce-9- ‘m- fintyg) . wor exampb : replace the fist two stetemants ‘in the example Gg he || previous bide slide with » Imp? = 4%a fn compilers perform such a range of optimisations “rar one “Moder could argue for : on eee aajacee | een |_ “eget __ lltersamaal Ph rise Coat | — (ode - 56 Cout) \| Aranstier i Exampe ef cflont + HPHPC oor (input) c++ [ation] ¢ Coutput PHP — [ararto C++ |B crows comper : A crass compiler is compiler sha) capable of creating pre “exe code for a Platform otha than ‘the one on’ which “the compiler is vunning « | @ decompiler : Machine code to Src Code 6 Beolstiopring : Bootsharring is 4 Yechniaue fr Cis a compiter ov assemater + that it intends” to compiler) producing @ self compiling complies en mn “The any source. programming janguge languoge = Target=> Pavol translator wriflen in ¢ programm™ng | ereate another postay transabr written in” C+ + tang q pe —3/P of | soe peo P . woiitien by > € iC r Og de ee — we ves earns 7 < at way <— Skeam of toKen “Prim (Emer a numer”): 5 seanf (" %ad", &0)> nA toten Ciype, attrinue) int! — ‘keyword | U/— wr Tndided “in toxon 2 — Identifier 5 integer 10 —_ Number . integer Leet RE of Wentifier >, elulaty* x ds ANN s4stetatels Imeger sca!) * Real nunte : (qd) , Coy|p Lexical Analyzer “gp Big picture | Former tang instead of natural lang testa! onilyeer —s recogeiea. parts "of speech (dic Nona?) | —> Keyword , Identifier» | ait + Language ° # Tolan. Lexims, Pattern # Toketizaton : lexim to pian # Wny all this 2 # wical Anolyas (code) ot “y Roguiar Fxpressiom 4 ob is a ee denoting the empty set qf Sf RFs ty automate scanner Constructo 1 Example, (RE FA Nextiot analyzer @ is a regular expesion “thal represent the regular language ve G ap eo \ O (aex) % =o ee: Oo a + +0 4O © Alphabet firite) woywrd —Lfinits) Identifies — Infinite Sentence. — infinite operator — ih Fite. Mate + 22.6.22, PRL = Af embly lang *Rl= QE} ervty oti “ rRL= 4a= Concalination + > Unwn + I (> Cbwsure @ aM ta anoten RE CO Ley = 80% los Abc? ha roy + Rory = {ap ae, bb Bcf le = ey —a Syate elimination methos - Mulliple finite final stares + 4 2. Starting state Should no 3. Final state should not have outgoing . Lk ae TOES +-O) a tbo)? vt have incoming Aransi tion ees — a__O Fr = albaO--O - a+ Gori®Ssitya Left linear Right linear 7 2 osep | alatyr ~ A ¢ ' r * B . j ] . . t A> oB : > B— aBlesle + /\ oN wm) e b* ‘ i Right to left Conversim uy poo lore P— Qal Pole ea ty el OR] Ob @ aR Ibe R— Pal Re R— aPloR _ Fecogrine which, ‘Grammer . Right linea! b?+ (abt ab* a) pad lop @— oR] bale Rap lpRie <>ff Ganol structure of a comaler ares weal | sap snk Aneatysis Cale “ton _| ctenerafion ima Targer code Analysis optimisation Target Code aaesten -— 3 Back-end Fion) -end “peconceptual Structure ; two major ehares \ Inlermtiate sorger Code sauce tcode —> [Fant-end Ferran | Back-ene | mast Front-end : performs “the analysis of the source Language « ~ Rcegriie? legal and illegal programs And reports errors = rundeskinds the mput program and collects tk semantics in an IR = -prdues IR ani Shapes the Code for the loatk-ona = Much can be automated Back-end + dor» the taiget language synthesis - = chooses instructions +o implement each IR operation, — Transtak. wR ine target code ~ Needy to cortiim with system interfacy - Aupmation oy been Ley successfulinxn compilers with Mtn Componenis Torget 1 Fortran | Postal Target 4 SAM language specific Knowledge. must be encoded ,in, Tne ion! end + All aiget specific fnowlece must be 6 encoded, m We back-end dt Qualities of a good compiler what qualile, wouky ‘you. want in a compiler 2 = generate correct code Cfirst and .Joramost ) . = gemrate ‘fast “code ! ~ conforms to the specifications of “the input larquage ~ copes with esserttialy arbitary inp size» variables ete compilation tims Cincarly) proportional to stze of: source = good diayonss diagnostics ceosistent opfimisations = works well wilh the debuggerot Prinapls ot compilation me compiler must: > Proserve the tmparing of the program béIng compiled > "improve the source ‘wde in some war. Other issues (depending on the setting): + Speed (of compiled tode) + Space (Size ‘of ‘Compiled code) 1 Peedeacr tinfor matin provided ‘to “the i) © Debugging (ranfrmatiom’ dbsewa’ “ha “tetadionship soures Code. vs, t494) + compilation time efficiency (fast ov slue compiler) : # tistorical Note * “The fi « y a Ao Higher- Level Programming = Machine lar . ine. Language (1st generation) ee larguages (2% 9e)- eorly *thigh- level tangueige (288° genoratwor)- 19505 1950 / 4th genorafion igher level languages GAL - postscript) a 5th gonsration tang [logic boved » eg, Prolog) + otter classifications : nn Sibaatie = \mperative (neu) > declorative (whos) = oWject- oriented Languages = Scripfina languagesjam charuky —> divided Jormol tang inlo sub cliviiors- RG FA . L CFG Pehebun L esa (OTM v Recurrsively enumerate $4 (RES) -TH Context ~ “free Grammer (CFG) Lomtext ~ free Grammer (CFG) => Recunsive = Sy detor mine syntactic /riot production Jue G= lv. ts. 6) E—eEte| EE | Id v= 2 Ta Ate id pt See ° pe JES ETE EO EYE ESid fp Vaid! sting FI FE eee Sontentiol forrr Sidtere (Wp Ceomtnation of terminals hon terinalt) DSidtidte MW) DS idtid tid (Hi) —> sentence /statement (70 variats) __ tate:2 7.6.22,Ro a Ln fo {| S EAEME (D > id4 EXE Ly Sententiak vm Sida idee (ni) sid+idt id Cw ow ww @ FR ELE EDENE x : 8 yew id € SEtECE > S ELErid ' - E+ id tid > E+ idtid > idtid tid id + id * id | | | BS Ete sid | | | mare | Amiiguaus r more than one left mat/ tight mat /devivatign —> Anbigims qraminer ] G cannot Wirt, | : > parser \| AN Must be depyminit Fe E& | 1) AZ i gal) [ECVE 1p down | \ «| Nett righ lia / ta piring gente Ww AS) bo ana? 2 i PA BxG 2 pad bint aed tue seo 2438 HG 14| vitor 5-4 as|sala 5 Anigucs grammmeat | MS IN “4 l fe / ( | | ee “te 5 a SbS| Sade bSas le I We abab a Ss (ss : ) ALS o ” HRA LR YR TRE | al ble. — (guar exprin) FI i“ SS R AS AN\\, | | } is Ss i? ¢ eel © AY atbte Otbrest Context Free Grammer (crn) | Arbigtiious © Asociatwity (samme precedent) G) Priority /procedenca Fo Ete Jia 24344 E E + & | eS E + 5 | \ 2 4 Right assodative Em Etidlia Ztats # same precadance + TeSUIT some a ~ E AN ° i " | | 4 Left associative. # ES idtelia AN iv JN W+oe swf t vay | ay Right rewtsive grammer7 ‘Say “pe precactonce | [2x545 T= tam Forse le*el2lag E>ERT(T qo TeF le Fo 21316 E Rt 3459 17 = TERE IESE] Ai3i5 : | E> Eatit TAOTHEIe Fo 21315 PRA | Sout 1 | jie priority > naar to toot higher prinity — lower lever rary ~ — Tighe Pr . 4 = Fight ossocias~ [# ES EFe/ FTE [ete | Ce) [-8 lid BO Fat (rT ; T— THE Ie b— GtFIE 6 —Oea Hid a> (G)\-Glid| # R>eeR | eR IR*alblc : \ Binary Clas pre # ESE ad [Eo e | Ne E> eE ow T(T to T and FIFE F— Wot F | Tru] Fate ee unambiguion frm of regular mp < ‘ Em ETIT To TFIF . Fo F* lalolc eu \ Aa AX Fe A> AMIE > fax 4,\ = Aceves. aN J Aye / ra’ co a A ‘= pet Aat CN 3 of A gy A> KAlB Ny i aod t eB Coe B AOR l= axepHf Eleminating left recursion [a= Aa 18) A — BAy AW wale LE aa) jw ES Fattt => Ta Ttelr Fo id L= pew replace with C_. pair ef Production A x recurs to Teme, Ries 6 rr LN ane A & | é€ ied > Context flex grammer =. uname gram oy Wert Fecursion free 9° Ero +TEevle TORT’ - “le pro eel eft mat top doun approachworivation | Ambigous Crammer + 7 rnore. than one left most geivation STE ) ' "4 4nere ove. (non one parse Wee _ Ambigou® | Foe +e fewe lid idtid+id id tid *id |> Ambigous to Unampidous 5 manually ; Cog no altervettive wor |: _ Same same | er, eT te Y | AN yi; + a | E NN I\s EOE id I | id id id id | Two woys of Ambignus Analysis AN. Eom & i 4. Order of precedence same 2. priority Ust tid eidtt Associativity + 2 = Gaeyyy ee left ~ Fecursi (ab > Right Recursive Ae = ‘sive ght ~ Yecursive 2. Right = Yecursive & 7’) o> left Recursive E —e+T |T | E * E es ~ . 3 . en AY AS" For Ambigow Analysis 2S a + T Fe Ls oii, JN IN a Fo + ot +4 & | |e tmambigous : level of Precectence Lowest precedence. —s rear fo rovt FOE4TIT a To TeFle Higher Precedence, — far away from tout = Blr— id AN BO EST IT no fesul F \ EF t+ T ™ T+) F | “ZN told T Tw F : | \ & +5 Deivation F F Context Free Commer | Loft Recurave Cran Rignt Recursive Gamera= (Wy. a. ae P) = ion > Comext fee grammer — Left Recursio if Eta E& E derivation of Ea —> Et => E© G Right recursive if sta OF bt ; left S— aAt A— Able S—s aAc A—s Ab|b S—> a@Ac A— Able | Fane [n> oy | farts | nya} | favre | 707 Hust haw. 4 Production for each non - terriinal aC-AbC /Obbt,.-- 3 JN a Aes /\ A rc | € S IN, infinite 4 7 X\ for Es lefy rowsya poe! Grane S—2 aAc., Aw vbAl & ~< Right Recursion < ,Poblem ih (ae Of Top-down pare Th Elimination of left Recursion : ds omer) C=" Aa BX D> ABE A— Ba IC Gi . B—> AB [p Indirect left Recursion —a aA’ IEgate: 8 22 (9 S—vaced cAd we ead" so eAq » Cod a) (i) A Abta Ww Axa r cad mal N 4 zt ¢ : AY sani CEG - | Right mot | or my Gshift reducing o7r1 oi | Scare (otto UG icra) Deteminisie. Non- dete umiach, LRRG Grammer Nerarnmes | zn p 5 sfasing—> pare ‘tree generation sHust be deterministic Monat <- A> XB, | XB. Jobs —> w= “X83” A A 4 “\) /*S « ) a » a A> aA’ {27 an rows of rans terry | Hon-det ‘0 determirotic = left - factoring 2 aw < No barwhacking "5 h- WM, 5 ders [ik +ses]q if expression then, | F>* ‘Statemeey i S—> iF+55/ [a wise] $9’ ebes + ss oe E—b Rate: 17-8: ‘n oo : vet fac terie cra G read —L — 7 y r bet Non-det, ie ue el 1a pdeton 40104 ene LF 1G un Cutt mont divans) | | Luk) ULC) —S prediction with 4 a | k ji tory with this don: tL(D Game errarnener a UU) Grammer | LLC) longuage | UCD Parsereee \ Preclictive tasers) ‘nar implemen IL ae) SB Class of parsers (also known | with a parsing table -Tabl. driven parser « LLU) Guarera —a—table-passer UG) Grammar + * Non - ambiguous % Non ~lett rectusive * Deterministic clatter eH recunimn ke ag Rectusive escent Parser i 4 Parner that wer a ser Of recursive. Procedums to recogrizt its input (may ov may nok back tracking) « © SEyt|T ETE’ a torte Ei +Te’ 1e€ yr | Ne r — (+ lid T— FT’ Wo er’ NE | ro ce) 18 using this —> Recursive Percent Parser Prowdure EC) 5 | begin TOs | F PRM (5 end | procedure EPRN () + i| if input Symo) = ‘4 then begin Abv ()3 || TOs i| EPRN (> endenn 4 poe eh Predictive PaIser : a A predictive parser is n Efficient way gf implementing recursive descent pausing by tnandling the stack, implementation records eaplicitty + . Pane tabe~32D a +———»> output & ineur i x Cratse tree) aaa 4 Top symbol of . Stack, ° Fig: Prédictivé parsor Model ‘3 The parser is controlled by a program that determines X+ a. “Where ave ‘three possible possibilities : ALaf R245 4,-he Parser halts and announces Sutcessdul completion of passing: 2. Mo K=a + 4- the passer pop off x from the staci, and advances the input (pointer to “the next Inpur Symbol. 3. 4$.X% is non- terminal ‘symbol . : : cet i - W MIkrals {x= Uvpl} the parser replace x a4 i 4rom ‘the top of stacls and push ms VU ps Tenis = S = Tal IY EF {| . ip MLX-a= error . thon the Parser call error fecuvery routineTy FI Vos were fe Fo (Ce) \id We" id+id id” Stacy WG idtidnia $ Fer id tid wid s SER idtidwidg = *. HET id idtidtigs + SEE +id vid $ t SETA @ id wid $ @ gis eee idy ids § ET id idv ids -z x + G EVER » idg 7 ae Non- terminals > ria YE>IE” TORT \poid { | Tor? [Es +1’ | | LU] 5 reprasenb 3 pase tree E= TE’ = FTE’ =, Stee > ide’ = id+TE’ =: id+ 17 E* > id+id VE’ = id4 fd FI/E® > idvidw idt/® > idtid wid &’ > td+idsid (etter Yo left west gerivoti”)