Você está na página 1de 7

THE GRAPHICS PIPELINE

The Graphics Pipeline: Dalton Nell CS 2810 Dece !er "# 201$

THE GRAPHICS PIPELINE

A!stract The %raphics pipeline to&a' is (oc)se& on !ein% a!le to ren&er as possi!le as (ast as possi!le* This an' pol'%ons as

a+es %raphic car&s ,)ic+l' o!solete as the a o)nt o(

processin% po-er o( GP./s increase !' a!o)t a hal( a %i%a(lop e0er' 'ear* Ga e &e0elopers ha0e to !e a-are o( the &i((erences in %raphic car& capa!ilities so t'picall' their %a es ha0e pol'%on an& te1t)re !)&%ets that +eep the opti i2e& eno)%h to r)n on a ran%e o( har&-are* Ho-e0er

there is a ren&erin% techni,)e that ren&ers the c)rrent %raphics pipeline )seless* This ren&erin% etho& )ses point clo)& &ata rather than in&i0i&)al pol'%ons* Ren&erin% %eo etr' !' tra0ersin% the pointclo)& in e or' &oes not re,)ire an' processin% po-er# e0en i( the pointclo)& ,)alit'

or n) !er is increase&* Ren&erin% %eo etr' -ith pointclo)&s -o)l& t)rn the e((ecienc' o( %raphics ren&erin% (ro not re,)ire an' 34n25 to a si ple 3415 since increasin% points in the point clo)& -o)l&

ore processin% po-er*

THE GRAPHICS PIPELINE

The Graphics Pipeline E0er since $D %raphics %ot pop)lar 1660/s# %raphics har&-are ha0e !een &esi%ne& -ith one %oal in in&: to ren&er as an' pol'%ons as possi!le as (ast as possi!le* This lea&s to

%raphics processin% )nits not onl' ,)ic+l' !eco in% o!solete# !)t the' are also incre&i!l' e1pensi0e* Despite these ,)ic+ i pro0e ents the' are still li ite&* 7hene0er %a es are &esi%ne& the' are o(ten %i0en 8pol'%on !)&%ets9* Ga e -orl&s an& o!:ects are all create& o)t o( pol'%ons# )s)all' trian%les* ;)il&in% spheres or trees o)t o( pol'%ons o(ten lea0e the )n)s)all' sharp an& !loc+'* To sol0e this in to&a'/s o&ern pipeline is to si pl' )se loo+in% ore

pol'%ons# !)t &)e to the pol'%on !)&%et the %a e &e0elopers can/t sol0e all the sharp an& !loc+' o!:ects in this (asion* This not onl' li its the artist creatin% the %a e o!:ects since the' can/t %o o0er the pol'%on !)&%et# !)t the &e0elopers as the' nee& to le0el o( &etail a+e the %a e ore e((icient -ith

o&el s-itchin%# !inar' space partitionin% trees# an& other co plicate& al%orith s etho& that e1ists*

to &ra- less pol'%ons to the screen* No- pol'%ons aren/t the onl' ren&erin% There is a ren&erin%

etho& -ith spee& that isn/t a((ecte& !' the n) !er o( o!:ects on screen* etho& )ses 0ol) es also +no-n as 0o1els or ato s*

Instea& o( )sin% pol'%ons this ren&erin%

7ith )se o( co plicate& ra'tracin% techni,)es 'o) can ren&er an incre&i!le a o)nt o( o!:ects realti e -ith relati0e ease on nor al har&-are* This is !eca)se instea& o( loopin% thro)%h e0er' in&i0i&)al o!:ect an& &ra-in% it to the screen# instea& 'o)/re loopin% thro)%h )ncolore& pi1els an& %ra!!in% -hat color the' sho)l& !e (ro the 0ol) etric &ata* <ro )se o( octtrees#

,)a&trees# an& e0en %ri&s 'o) can %et !)ilt=in le0el o( &etail# (r)st)

c)llin%# an& !inar' space

partitionin%* 4>a-a )ra# Sa+a oto# ? >o'a a&a# 20105 Altho)%h 0ol) e !ase& ren&erin%

THE GRAPHICS PIPELINE sho-s incre&i!le potential# it -on/t !e (easi!le co pare& to pol'%onal ren&erin% (or 'ears to co e* Spee&# it/s all that atters in %a e ren&erin%* Pol'%onal ren&erin% alrea&' has this )n&er c)llin%# an& an' others# -ith )ch e((ort on

"

its !elt* 7ith ren&erin% techni,)es li+e i itations# ;SP# (r)st)

the help o( har&-are accelleration# it can represent h)%e %a e -orl&s -itho)t

h)%e 1080p resol)tions* Ho-e0er 0ol) e !ase& ren&erin% isn/t as ol& an& certainl' isn/t as (easi!le (or %a e &e0elop ent (or reasons that -ill soon !e e1plaine&# an& as s)ch it &oesn/t ha0e as an' ren&erin% techni,)es or cle0er al%orith s* @ol) e !ase& ren&erin% c)rrentl' is

reall' slo-* Despite 0ario)s ren&erin% techni,)es li+e e pt' 0ol) e s+ippin%# an& earl' ra' ter ination== ost &e0elopers &o-nscale the resol)tion or &on/t shoot as an' ra's in or&er to

%et eno)%h (ra es per secon& to %et a s ooth &ispla'* Despite this slo-ness# it &oesn/t %et an' slo-er e0en i( 'o) a&& ore o!:ects* This )ni,)e (eat)re o( 0ol) e !ase& ren&erin% co)l& !e

e1ploite& to ha0e %a e -orl&s -ith al ost li itless &etail e0en &o-n to literall' e0er' %rain o( &irt -itho)t a((ectin% the spee& o( the %a e* .n(ort)natel' (or no-# pol'%onal ren&erin% spee& -ise is still (ar ore (easi!le than 0ol) e !ase& ren&erin%* Great ren&erin% techni,)es (or

0ol) e !ase& ren&erin% are o)t there 4>rA%er# 7ester ann# 200$5# !)t )ntil har&-are accelleration e1ists (or it: pol'%onal ren&erin% (or %a e &e0elop ent -ins o)t on spee&* Detail# ost people -on/t !)' a se,)al o( a %a e i( it &oesn/t ha0e !etter 8%raphics9 or aps# s+eletal

&etail* Pol'%onal ren&erin% is prett' (l)she& in this re%ar&* It has access to nor al ani ations# spec)lar a+e realistic aps# sha&o-s# an& pro%ra

a!le sha&ers that all co !ine to%ether to

o&els an& e((ects ,)ite easil'* Despite pol'%onal ren&erin% ha0in% a li ite&

n) !er o( pol'%ons an& te1t)res to )se# co !inin% all the e((ects onto a hi%h=pol'%on# HD=

THE GRAPHICS PIPELINE te1t)re&#

ain character can lea& to a 0er' realistic an& !ea)ti()l scene* Co) can/t sa' this (or all

the scenes in a pol'%onal %a e tho)%h !eca)se o( the pol'%on !)&%et* Co)/ll al-a's (in& a lo-= pol'%on scene in a %a e so e-here* @ol) e !ase& ren&erin% -o)l& !e ,)ite the opposite* E0er'thin% co)l& !e as &etaile& as the &e0elopers -ant# !)t there/s so e &o-nsi&es to the &etail o( 0ol) e ren&erin%* Ani ation (or 0ol) e !ase& ren&erin% si pl' &oesn/t e1ist# as least its not (easi!le to e1ist c)rrentl'* <or s+eletal ani ation# the &e0eloper -o)l& ha0e to co p)tate the position o( each in&i0i&)al AT3D o( a character e0er' (ra e !ase& on !one -ei%hts* 7hich is a tre en&o)s tas+ (or an' har&-are* <ra e !' (ra e ani ation isn/t so o)t o( this -orl& (or 0ol) e !ase& ren&erin%# !)t it -o)l&n/t !e 0er' since it -o)l&n/t e or' e((icient nor -o)l& it !e 0er' prett'

atch the (ra e rate* @ol) e ren&erin% &oes -in o)t in one aspect tho)%h*

7hen ren&erin% act)al 0ol) etric o!:ects li+e s o+e or clo)&s# 0ol) e !ase& ren&erin% co pletel' o)tper(or s pol'%onal ren&erin% 4I+its# >niss# Le(ohn# 20085* Dainl' !eca)se it is &i((ic)lt to represent clo)&s or s o+e -ith pol'%ons* Despite this# pol'%onal ren&erin% still -ins o)t !eca)se it is (ar easier to %et (ascinatin% e((ects li+e real=ti e sha&o-s# re(lections# phon%# transparenc'# an& others -hile +eepin% e0er'thin% 0is)all' consistent an& r)nnin% (ast* Ease o( )se# no!o&' -ill &e0elop -ith it i( it isn/t eas' to i ple ent* Pol'%onal ren&erin% has !een a stan&ar& since the !e%innin% o( 0i&eo %a es* It has tons o( &e0+its# (ra e-or+s# an& tools in or&er to create pol'%onal -orl&s -ith ani ate& o!:ects* Dost &oc) ente& reso)rces are all a!o)t ho- pol'%onal ren&erin% co)l& !e &one !etter# (aster# an& ore !ea)ti()l* All %a e

in&)stries )se it an& release en%ine licenses to share the technolo%'* Har&-are i ple ente& in o&ern co p)ters are &esi%ne& to accellerate pol'%onal !ase& ren&erin%* Pol'%onal ren&erin% is !' (ar the easiest to not onl' &e0elop -ith# !)t its also the (ar easiest to %et help -ith* @ol) e

THE GRAPHICS PIPELINE !ase& ren&erin% has a (e- articles (loatin% aro)n&# !)t there/s no tr)e reso)rces to create a ()ll' (le&%e& %a e !ase& on 0ol) es -ith ani ation# sha&o-s# re(lections# etc* This ren&erin% a%ain (ar ore (easi!le*

a+es pol'%onal

Altho)%h 0ol) e !ase& ren&erin% sho-s incre&i!le potential# it -on/t !e (easi!le co pare& to pol'%onal ren&erin% (or 'ears to co e* @ol) e ren&erin% is alrea&' )se& to&a' (or DRI an& CAT scannin% achines# !)t its )nli+el' to see an' +in& o( 0ol) e ren&erin% (or 0i&eo

%a es in the (orseea!le ()t)re* Hope()ll' &e0elopers -ill &e0elop har&-are that can han&le it -ith ani ation# sha&in%# re(raction# re(lection# an& aterial sha&ers as it -ill &rasticalll'

i pro0e 0i&eo %a es artisticall'* .n(ort)natel' -ith to&a's har&-are# pol'%onal ren&erin% per(or s !etter c)rrentl' in spee&# &etail# an& ease o( )se* In the ()t)re 0ol) e !ase& ren&erin% co)l& !e stan&ar&*

THE GRAPHICS PIPELINE Re(erences >a-a )ra# T*# Sa+a oto# N*# ? >o'a a&a# >*# 420105* Le0el=o(=Detail Ren&erin% o( Lar%e= Scale Irre%)lar @ol) e Datasets .sin% Particles* Journal of Computer Science and Technology* 2B 4B5# pp*60B=61B >rA%er# G*# 7ester ann# R*# 4200$5* Acceleration Techni,)es (or GP.=!ase& @ol) e Ren&erin%* IEEE Visualization 2003* 1" 45# pp*$8 I+its, D , >niss# G*# Le(ohn# A*# 420085* Volume !endering Techni"ues* H3NLINEI A0aila!le at: http:JJhttp*&e0eloper*n0i&ia*co JGP.Ge sJ%p)%e sKch$6*ht l* HLast Accesse& 2 Dece !er 201$I*

Você também pode gostar