并行處理與體系結(jié)構(gòu)hitsz-lec08-review1_第1頁
并行處理與體系結(jié)構(gòu)hitsz-lec08-review1_第2頁
并行處理與體系結(jié)構(gòu)hitsz-lec08-review1_第3頁
并行處理與體系結(jié)構(gòu)hitsz-lec08-review1_第4頁
并行處理與體系結(jié)構(gòu)hitsz-lec08-review1_第5頁
已閱讀5頁，還剩21頁未讀， 繼續(xù)免費(fèi)閱讀

下載本文檔

版權(quán)說明：本文檔由用戶提供并上傳，收益歸屬內(nèi)容提供方，若內(nèi)容存在侵權(quán)，請進(jìn)行舉報(bào)或認(rèn)領(lǐng)

文檔簡介

MemoryHierarchyReview1Outline7.1IntroductionPrincipleofLocalityMemoryHierarchy7.2BasicofCacheDirectedMappedCacheBitsinaCacheCacheWriteCacheMissesMultiwordCacheBlock7.3CachePerformanceImproveCachePerformanceMemoryisaBottleneckFullyAssociativeCacheSetAssociativeCacheMulti-LevelCache7.1PrincipleofLocalityProgramaccessarelativelysmallportionoftheiraddressspaceatanyinstantoftimeTwokindoflocality:1.TemporallocalityIfanitemisreferenced,itwilltendtobereferencedagainsoon2.SpatiallocalityIfanitemisreferenced,itemswhoseaddressareclosebywilltendtobereferencedsoon7.1MemoryHierarchy–Speedvs.SizeMemoryTypeDiskDRAMSRAM7.1MemoryHierarchy-OperationIfdataisfound(hit)transfertoprocessor,otherwise(miss)transferdatatoupperlevel.AccesstimeHittimeMisspenaltyUserswantlargeandfastmemories!

SRAMaccesstimesare2-25nsatcostof$100to$250perMbyte.

DRAMaccesstimesare60-120nsatcostof$5to$10perMbyte.

Diskaccesstimesare10to20millionnsatcostof$.10to$.20perMbyte.

Tryandgiveittothemanywaybuildamemoryhierarchy7.1MemoryHierarchy7.2CacheCacheMemoryhierarchybetweenCPUandmainmemoryThestoragemanagedtotakeadvantageoflocalityofaccessCachetwoissuesHowdoweknowifadataitemisinthecache?TagandvalidbitIfitis,howdowefindit?MappedapproachesHowdoescacheworkexampleAddXntocacheDirectMappingmapmanymemorywordsontoonelocationincacheAddressismodulothenumberofblocksinthecache(blockaddress)modulo(no.ofcacheblocksinthecache)Example:Cachehas8wordMapping=blockaddressmodule87.2DirectMappedCache7.2DirectMappedCacheTwoissues1.Whichmemorywordinthecache?Usetagtoidentify2.Whetherthememoryblockisvalid?Ex.Initially,thecacheisemptyUsevalidbittoidentifyThusthecachedatastructurearevalidtagdataword…CacheIndex7.2DirectMappedCacheIfatagismatchedandvalidbitison

ThenarequesthitTagiscomparedwithupperportionofaddressReadhitsReadvaliddataoncacheReadmissesstalltheCPU,fetchblockfrommemory,delivertocache,restart

Writehitscanreplacedataincacheandmemory(write-through)writethedataonlyintothecache(write-backthecachelater)

Writemissesreadtheentireblockintothecache,thenwritetheword7.2CacheReadWriteTerminology7.2CacheWriteIssuesTwocachewritescheme:1.WritebackWhenwriteoccurs,onlywritetothecache2.WritethroughWhenwriteoccurs,writetothecacheandmemoryWrite-backproblemcacheandmemoryinconsistence,andcomplextoimplementEx.Whenacacheentryisreplaced,itmustupdatethecorrespondingmemoryaddressWrite-throughproblemWritingtomainmemoryslowsdowntheperformanceEx.CPIwithoutcachemiss=1.2clockcycleswritetomemorycausesextra10cycles13%storeinstructionsingccSolution:writebuffer,storethedataintowritebufferwhilethedataiswaitingtobewrittentomemoryTheprocesscancontinueexecutionafterwritingdataintocacheandwritebuffer7.2DirectMappedCacheCacheExampleonDECstation3100UseMIPSR2000CPU64KBdata98KBcachesize

7.2MultiwordCacheBlockTakeadvantageofspatiallocalityWithacachemiss,wewillfetchmultiplewordsthatareadjacentIncreasingtheblocksizetendstodecreasemissrate:

Usesplitcachesbecausethereismorespatiallocalityincode:

7.2Performance–Missratevs.BlocksizeImprovementoninstructionmissMakereadingmultiplewordseasierbyusingbanksofmemory

7.2Memorysystem-hardwareIssues7.3ImproveCachePerformanceThreewaysLargercacheSetassociativecacheReducecachemissrateNewplacementruleotherthandirectmappingMulti-levelcacheReducecachemisspenalty7.3FlexiblePlacementofBlocksTherearetwomoreflexibleschemes,thendirectedmappedSetassociativecacheFullyassociativecacheExample:block12addressisplacedin8blockcachePlace12%8=4Place12%4=0DirectmappedSetassociativeFullyassociativePlacedanyatblock7.3FullyAssociativecacheAnextremeschemeAmemorydatacanbeplacedinanyblockinthecacheDisadvantage:SearchallentriesinthecacheforamatchParallelcomparators7.3SetAssociativeCacheBetweendirectmappedandfull-associativeAmemorydatacanbeplacedinasetofblocksinthecache(address)modulo(numberofsetsincache)Ex:12modulo4=0Disadvantage:SearchallentriesinthesetforamatchParallelcomparators7.3EightBlockCacheConfigurationTotalsizeofcacheinblocksisequaltothenumberofsetsThus,forfixedcachesize,increaseassociativitydecreasesthenumberofset,butincreasenumberofelementinaset7.3MissRatewithAssociativityHigherdegreeofassociativityLowermissrateMorehardwarecosttosearch7.3Implementationof4-waySet-AssociativeCacheParallelcomparators7.3UseMulti-LevelCachetoReduceMissPenaltyAddasecondlevelcache:oftenprimarycacheisonthesamechipastheprocessoruseSRAMstoaddanothercacheaboveprimarymemory(DRAM)misspenaltygoesdownifdataisin2ndlevelcacheUsingmultilevelcaches:tryandoptimizethehittimeonthe1stlevelcachetryandoptimizethemissrateonthe2ndlevelcachePrimarycache(L1)Secondarycache(L2)L1cachemissL2cachemissCachehit7.3DecreasingMissPenaltywithMultilevelCachesAddasecondlevelcache:oftenprimarycacheisonthesamechipastheprocessoru

溫馨提示

  • 1. 本站所有資源如無特殊說明，都需要本地電腦安裝OFFICE2007和PDF閱讀器。圖紙軟件為CAD,CAXA,PROE,UG,SolidWorks等.壓縮文件請下載最新的WinRAR軟件解壓。
  • 2. 本站的文檔不包含任何第三方提供的附件圖紙等，如果需要附件，請聯(lián)系上傳者。文件的所有權(quán)益歸上傳用戶所有。
  • 3. 本站RAR壓縮包中若帶圖紙，網(wǎng)頁內(nèi)容里面會有圖紙預(yù)覽，若沒有圖紙預(yù)覽就沒有圖紙。
  • 4. 未經(jīng)權(quán)益所有人同意不得將文件中的內(nèi)容挪作商業(yè)或盈利用途。
  • 5. 人人文庫網(wǎng)僅提供信息存儲空間，僅對用戶上傳內(nèi)容的表現(xiàn)方式做保護(hù)處理，對用戶上傳分享的文檔內(nèi)容本身不做任何修改或編輯，并不能對任何下載內(nèi)容負(fù)責(zé)。
  • 6. 下載文件中如有侵權(quán)或不適當(dāng)內(nèi)容，請與我們聯(lián)系，我們立即糾正。
  • 7. 本站不保證下載資源的準(zhǔn)確性、安全性和完整性, 同時(shí)也不承擔(dān)用戶因使用這些下載資源對自己和他人造成任何形式的傷害或損失。

最新文檔

評論

0/150

提交評論