資源描述:
《大規(guī)模集成電路時(shí)鐘網(wǎng)絡(luò)布圖布線算法研究》由會(huì)員上傳分享,免費(fèi)在線閱讀,更多相關(guān)內(nèi)容在學(xué)術(shù)論文-天天文庫(kù)。
1、河北工業(yè)大學(xué)碩士學(xué)位論文大規(guī)模集成電路時(shí)鐘網(wǎng)絡(luò)布圖布線算法研究摘要當(dāng)前集成電路產(chǎn)業(yè)向深亞微米工藝不斷推進(jìn),正力圖突破32nm大關(guān)?,F(xiàn)有EDA工具難以應(yīng)付復(fù)雜度呈指數(shù)增長(zhǎng)的諸多VLSI設(shè)計(jì)難題,也缺乏對(duì)深亞微米工藝下一系列新問(wèn)題的考慮。另一方面,在計(jì)算智能領(lǐng)域,各種優(yōu)化技術(shù)日新月異,為解決非NP和NP復(fù)雜度的大規(guī)模、超大規(guī)模問(wèn)題展示了廣闊的前景。隨著VLSI的工藝向超深亞微米的推進(jìn),物理設(shè)計(jì)中的布線問(wèn)題,由于問(wèn)題規(guī)模的急劇增大,都迫切需要更有效的優(yōu)化算法解決方案。在本文中,我們采用解決不同條件下實(shí)際問(wèn)題的兩種模型
2、,即非均勻網(wǎng)格和無(wú)網(wǎng)格的兩種模型,使問(wèn)題的復(fù)雜度大大下降。然后介紹了一種改進(jìn)的蟻群算法。接下來(lái)把該算法用于解決兩端以及多端線網(wǎng)繞障礙的布線問(wèn)題,同時(shí)進(jìn)行實(shí)驗(yàn)仿真以及在此基礎(chǔ)上分析的結(jié)果。在VLSI布線過(guò)程中,時(shí)鐘網(wǎng)絡(luò)的布線非常重要,在同步數(shù)字系統(tǒng)中,作為數(shù)據(jù)流傳輸?shù)臅r(shí)間參考,時(shí)鐘信號(hào)控制著同步單元的操作。由于它關(guān)系到系統(tǒng)正確性和工作頻率,在時(shí)鐘信號(hào)的特性和時(shí)鐘網(wǎng)絡(luò)設(shè)計(jì)方面都展開(kāi)了許多研究。與普通控制信號(hào)相比,時(shí)鐘信號(hào)有著一些特殊性質(zhì):無(wú)論是作為控制信號(hào)還是作為數(shù)據(jù)信號(hào),時(shí)鐘信號(hào)都有很大的扇出負(fù)載,驅(qū)動(dòng)著成千上萬(wàn)
3、的同步單元;時(shí)鐘線網(wǎng)布線長(zhǎng)度長(zhǎng),從時(shí)鐘源點(diǎn)出發(fā)到時(shí)鐘接收端點(diǎn)可能要跨越整個(gè)芯片;工作頻率在整個(gè)系統(tǒng)中最高。因此,時(shí)鐘信號(hào)要求傳遞準(zhǔn)確、波形轉(zhuǎn)換迅速,同時(shí)也要求時(shí)鐘網(wǎng)絡(luò)具有抗干擾的魯棒性、較大的靈活性和較小的功耗。因此,本文中,提出一種基于蟻群算法的時(shí)鐘網(wǎng)絡(luò)布線算法,極大地減小了時(shí)鐘網(wǎng)絡(luò)布線的時(shí)鐘延時(shí)和時(shí)鐘偏差。關(guān)鍵詞:VLSI物理設(shè)計(jì),計(jì)算智能,蟻群算法,時(shí)鐘網(wǎng)絡(luò)布線,時(shí)鐘偏差i大規(guī)模集成電路電源/地線布圖布線算法研究VLSICLOCKNETWORKFLOORPLANNINGANDROUTINGALGORITH
4、MRESEARCHAbstractNowintegratedcircuitindustryisevolvingrapidlyindeepsub-microntechnologyaimedatovercomingthebarrierofwirewidthat32nm.Thistrendhasputtreatchallengesfortherecentlyavailabletoolsofelectronicdesignautomation.OneofthechallengesisthatforVLSIcircuit
5、s,manyNP-hardProblemsisimpossibleorverydifficulttobesolvedusingtraditionaloptimumalgorithms;theotheristhatmanynewandspecificdeepsub-microntechnologyproblemshadnotbeenconsidered,whichwillinfluencechip’sperformance.Andatthesamedine,inthefieldofcomputationalint
6、elligence,anumberofoptimizationtechniqueshaveshowntheirgreatpowerandpotentialinsolvinglarge-scalecomplexProblems.Withtherapidprogressindeepsub-microntechnology,mostoftheroutingproblemsraisedphysicaldesignofVLSIchipsaredemandingmoreefficientroutingalgorithms.
7、Inthisdissertation,Wehaveadoptedtwokindsofgraphicmodel,asymmetricgridgraphandgrid-offmodel,whichallowustoreducethespatial-temporalcomplexityofproblemsignificantly.Thenweputforwardanimprovedantcolonyalgorithm.Thenthealgorithmusedtosolvebothendsofthebarriersar
8、oundthelayoutofthewiringproblem,atthesametimegivesthesimulationexperimentonthisbasis,aswellasanalysisoftheresults.TheclocknetworkphysicaldesignisveryimportantinVLSIphysicaldesign.Insynchronousdi