UMichael Scott - sele ena 34 iminyaka njengonjingalwazi weNzululwazi yeKhompyutha kwiDyunivesithi yaseRochester, kwaye ekhayeni lakhe kwiYunivesithi yaseWisconsin-Madison wayengumongameli iminyaka emihlanu. Uphanda kwaye ufundisa abafundi malunga nokuhambelana nokusasazwa kwenkqubo kunye noyilo lolwimi.

Ihlabathi liyamazi uMichael kwincwadi yesifundo "Iipragmatiki zoLwimi lokucwangcisa", kuthekani ngomsebenzi "Ii-algorithms zolungelelwaniso olunokukhawuleza kwi-multiprocessors ekwabelwanayo kwimemori" ufumene ibhaso leDijkstra njengelinye lawona adumileyo kwicandelo lekhompyuter esasazwayo. Usenokumazi njengombhali waloo algorithm UMichael-Scott.

Ekunye noDoug Lee, waphuhlisa i-algorithms engathinteliyo kunye nemigca ehambelanayo enika amandla amathala eencwadi eJava. Ukuphunyezwa "iziseko zedatha ezimbini" kwiJavaSE 6 iphuculwe ukusebenza ngamaxesha e-10 ThreadPoolExecutor.

Iziqulatho:

• Umsebenzi wokuqala, kwiYunivesithi yaseRochester. Iprojekthi iCharlotte, ulwimi lweLynx;
• IEEE Scalable Coherent Interface, ukutshixa kwe-MCS;
• Ukusinda kwihlabathi elihlala litshintsha;
• Ngaba abafundi baba zizidenge? Iindlela zehlabathi, ukuthengiswa kwamanye amazwe;
• Umsebenzi osebenzayo kunye nabafundi;
• Indlela yokuqhubeka nokulungiselela iikhosi ezintsha kunye neencwadi;
• Amakhonkco phakathi kweshishini kunye nezemfundo;
• Ukuphunyezwa okusebenzayo kweengcamango. MCS, MS, CLH, JSR 166, ukusebenza noDoug Lee kunye nokunye;
• Inkumbulo yentengiselwano;
• Uyilo olutsha. Uloyiso lwenkumbulo yentengiselwano lukufuphi;
• Imemori engaguqukiyo, i-Optane DIMM, izixhobo ezikhawulezayo;
• Umkhwa omkhulu olandelayo. Izakhiwo zedatha ezimbini. Hydra.

Udliwano-ndlebe luqhutywa ngu:

Vitaly Aksenov - okwangoku i-postdoc kwi-IST Austria kunye nelungu leSebe le-Computer Technologies kwiYunivesithi ye-ITMO. Uqhuba uphando kwinkalo yethiyori kunye nokuziqhelanisa nezakhiwo zedatha ezikhuphisanayo. Ngaphambi kokusebenza kwi-IST, wafumana i-PhD yakhe kwiYunivesithi yaseParis Diderot kunye neYunivesithi ye-ITMO phantsi kolawulo lukaNjingalwazi uPeter Kuznetsov.

Alexey Fedorov - Umvelisi kwi-JUG Ru Group, inkampani yaseRashiya eququzelela iinkomfa zabaphuhlisi. U-Alexey uthathe inxaxheba ekulungiseleleni iinkomfa ezingaphezu kwe-50, kwaye ukuqhubeka kwakhe kubandakanya yonke into esuka kwisikhundla sonjiniyela wophuhliso kwi-Oracle (JCK, i-Java Platform Group) kwindawo yomphuhlisi e-Odnoklassniki.

UVladimir Sitnikov -Njineli kwiNetcracker. Iminyaka elishumi yomsebenzi ekusebenzeni kunye nokulinganisa kwe-NetCracker OS, isofthiwe esetyenziswa ngabaqhubi be-telecom ukwenza i-automate network kunye neenkqubo zolawulo lwezixhobo zenethiwekhi. Ndinomdla kwimiba yokusebenza kweJava kunye neOracle Database. Umbhali ongaphezulu kweshumi elinesibini lokuphucula ukusebenza kumqhubi osemthethweni wePostgreSQL JDBC.

Umsebenzi wokuqala, kwiYunivesithi yaseRochester. Iprojekthi kaCharlotte, ulwimi lweLynx.

Алексей: Ukuqala, bendifuna ukukuxelela ukuba eRashiya sonke siyithanda ngokwenene iSayensi yeKhompyutha, iSayensi yeDatha kunye ne-algorithms. Ingamanyala. Sifunde yonke into Incwadi kaCormen, Leiserson kunye noRivest. Ngoko ke, inkomfa ezayo, isikolo kunye nodliwano-ndlebe ngokwalo kufuneka ludume kakhulu. Sifumene imibuzo emininzi yolu dliwano-ndlebe kubafundi, abadwelisi benkqubo, kunye namalungu oluntu, ngoko ke sinombulelo kakhulu ngeli thuba. Ngaba iNzululwazi yeKhompyutha ifumana uthando olufanayo e-US?

UMichael: Intsimi yethu yahluke kakhulu, ineendlela ezininzi, kwaye ichaphazela uluntu ngeendlela ezininzi ezahlukeneyo kangangokuba kunzima ukuba ndikunike impendulo ecacileyo. Kodwa inyaniso yeyokuba izise iinguqu ezinkulu kushishino, kwimizi-mveliso, kubugcisa nakuluntu ngokubanzi kule minyaka ingama-30 idlulileyo.

Vitaliy: Masiqale ngento ekude. Kwiiyunivesithi ezininzi kukho into efana nobuchule kwindawo ethile. KwiYunivesithi yaseCarnegie Mellon le yi-computing ehambelanayo, ye-MIT yi-cryptography, iirobhothi kunye ne-multithreading. Ngaba kukho ubungcali obunjalo kwiDyunivesithi yaseRochester?

UMichael: Ukunyaniseka, ndingatsho ukuba i-CMU kunye ne-MIT zikhethekileyo kuzo zonke iindawo. Isebe lethu belisoloko linika ingqwalasela enkulu kubukrelekrele bokwenziwa. Isiqingatha sabantu esisisebenzelayo sibandakanyeka kwi-AI okanye kwintsebenziswano yabantu nekhompyutha - esi sabelo siphezulu kunamanye amasebe, kwaye bekusoloko kunjalo. Kodwa xa ndandiseyunivesithi, ndandingenazo naziphi na izifundo kwi-AI, kwaye andizange ndisebenze kweli candelo. Ngoko ke isebe lam ligxile kwingxaki endingenanto yakwenza nayo. Intuthuzelo yeyokuba eyesibini eyona ngxaki ibalulekileyo kwisebe lethu yinkqubo ehambelanayo nenemisonto emininzi, oko kukuthi, ubungcali bam.

Vitaliy: Uqale ukusebenza kwiNzululwazi yeKhompyutha xa intsimi yeprogram ye-multi-threaded yayisanda kuvela. Uluhlu lopapasho lwakho lubonisa ukuba imisebenzi yakho yokuqala yayijongene noluhlu olubanzi ngokufanelekileyo lwemiba: ulawulo lwenkumbulo kwiinkqubo ezinemisonto emininzi, iinkqubo zeefayile ezisasaziweyo, iinkqubo zokusebenza. Kutheni kukho izinto ezininzi kangaka? Ngaba uzama ukufumana indawo yakho kuluntu lophando?

UMichael: Njengomfundi, ndathatha inxaxheba Iprojekthi yeCharlotte kwiYunivesithi yaseWisconsin, apho enye yeenkqubo zokuqala ezisasazwayo zaphuhliswa. Apho ndasebenza kunye noRafael Finkel (Raphael Finkel) kunye noMarvin Solomon (UMarvin Solomon). I-dissertation yam yayinikezelwe kuphuhliso lolwimi lwenkqubo yesoftware yeenkqubo ezisasazwayo - ngoku wonke umntu sele elibele ngayo, kwaye ndiyambulela uThixo. Ndidale ulwimi lwenkqubo lweLynx, olwalujongiwe ukwenza kube lula ukwenza iiseva zenkqubo yokusebenza esasazwe ngokukhululekileyo. Ekubeni ngelo xesha ndandibandakanyeke ngokuyintloko kwiinkqubo zokusebenza, ndacinga ukuba umsebenzi wam wawuya kunxulumana ngokuyintloko nazo. Kodwa iRochester yayiyunivesithi encinci kakhulu, kwaye ngenxa yoko, amaqela ahlukeneyo apho asebenzisana ngokusondeleyo kakhulu. Bekungekho shumi elinesibini lezinye iinkqubo zokusebenza abantu apho ukuba ndithethe nabo, ngoko ke bonke oonxibelelwano babekunye nabantu abasebenza kwiindawo ezahlukeneyo ngokupheleleyo. Ndikonwabele kakhulu, ukuba yi-all-rounder yinzuzo enkulu kum. Ukuba sithetha ngokuthe ngqo malunga nezakhiwo zedatha ezinemisonto emininzi kunye ne-algorithms yokuvumelanisa, ngoko ke ndaqala ukusebenza kuzo ngokupheleleyo ngengozi.

IEEE Scalable Coherent Interface, ukutshixa kwe-MCS.

Vitaliy: Ungandixelela kancinci ngalento?

UMichael: Libali elihlekisayo eli endingadinwanga kulibalisela wonke umntu. Kwenzeke kwinkomfa ASPLOS e Boston - oku kwaba ekupheleni 80s okanye 90s ekuqaleni. UJohn Mellor-Crummey (UJohn Mellor-Crummey), ophumelele kwifakhalthi yethu. Ndandimazi, kodwa sasingazange silwenze uphando ngokubambisana ngaphambili. UMary Vernon (UMary Vernon) evela eWisconsin inike intetho malunga nenkqubo yemultiprocessor ababeyiphuhlisa eWisconsin: Wisconsin Multicube. Le Multicube yayinesixhobo songqamaniso kwinqanaba le-hardware ebizwa ngokuba yi-Q kwi-Sync Bit, kwaye kamva yathiywa ngokutsha ngokuba yi-Q kwi-Lock Bit kuba yayivakala ngathi yi-Colby cheese, eyayiyi-pun. Ukuba unomdla kwiindlela ezininzi zokufunda, uyazi ukuba iColby ekugqibeleni yaba yinjini yongqamaniso yomgangatho we-IEEE Scalable Coherent Interface. Le yayiyindlela yokutshixa eyadala izikhombisi ukusuka kwenye i-cache ukuya kwenye kwinqanaba le-hardware ukuze umnini wetshixo ngamnye azi ukuba leliphi ithuba. Xa mna noJohn besiva ngale nto, sajongana saza sathi: kutheni ukwenza oku kwinqanaba le-hardware? Ngaba ayinakwenzeka into enye kusetyenziswa ukuthelekisa kunye nokutshintshana? Sithathe enye yencwadana elele eklasini sabhala kuyo Ukuvalela kwe-MCS, ngoxa uMariya wayeqhubeka nengxelo yakhe. Emva koko, sayiphumeza, sazama, umbono waphumelela, kwaye sapapasha inqaku. Ngelo xesha, kum, esi sihloko sasibonakala sisiphazamiso esimnandi, emva koko ndiceba ukubuyela kwiinkqubo zokusebenza. Kodwa ke enye ingxaki ngendlela efanayo yavela, kwaye ekugqibeleni ungqamaniso, ufundo oluninzi, kunye nolwakhiwo lwedatha yaba yinto yam ekhethekileyo. Njengoko ubona, konke oku kwenzeka ngamabona-ndenzile.

Vitaliy: Kudala ndiqhelene ne-MCS blocking, kodwa kude kube ngoku bendingayazi ukuba ngumsebenzi wakho, kwaye andiyiqondi ukuba sisishunqulelo sefani yakho.

Indlela yokusinda kwihlabathi elihlala litshintsha?

Алексей: Ndinombuzo kwisihloko esihambelanayo. Kwiminyaka engama-30 okanye engama-40 eyadlulayo kwakukho inkululeko engakumbi kwizinto ezikhethekileyo. Ukuba ufuna ukuqala umsebenzi kwiinkqubo ezininzi zokufunda okanye ezisasazwayo, wamkelekile, ukuba ufuna ukungena kwiinkqubo zokusebenza, akukho ngxaki. Kwindawo nganye kwakukho imibuzo emininzi evulekileyo kunye neengcali ezimbalwa. Iingcali ezinqamlekileyo ziye zavela ngoku: akukho zingcali kuphela kwiinkqubo zokusebenza ngokubanzi, kukho iingcali kwiinkqubo zomntu ngamnye. Kuyafana nenkqubo yokufundwa kwemininzi kunye nokusasazwa. Kodwa ingxaki kukuba ubomi bethu abunasiphelo; Ukusinda njani kweli hlabathi litsha?

UMichael: Asikhethekile kulo mba; Ndinethamsanqa ukuba ndiqalise ukusebenza kwiSayensi yeKhompyutha xa intsimi yayineminyaka "yobutsha". Ezinye iziseko zazisele zibekiwe, kodwa yonke into yayingekavuthwa. Eli thuba aliveli rhoqo. Ubunjineli bombane sele bukhona ixesha elide kakhulu, ifiziksi nangaphezulu, imathematika phantse ukusukela ekuqaleni kwexesha. Kodwa oku akuthethi ukuba akusekho mntu wenza izinto ezibangela umdla kwimathematika. Kusekho iingxaki ezininzi ezivulekileyo, kodwa kwangaxeshanye, kufuneka kufundwe okuninzi. Uchanekile ukuqaphela ukuba ngoku kukho iingcali ezininzi ngakumbi kunangaphambili, kodwa oku kuthetha kuphela ukuba sizifumana sikwimeko efanayo nezinye iindawo ezininzi ezenziwa ngabantu.

Алексей: Ndinomdla kumba osebenzayo walo mba apha. Ndinemvelaphi yezibalo, kwaye ngexesha lezifundo zam bendidla ngokuya kwiinkomfa kwaye ndisebenza kwimixholo eyahlukeneyo yesayensi. Ndafumanisa ukuba akukho namnye kubaphulaphuli owayeziqonda iingxelo zam, yaye ngendlela efanayo, iingxelo zabanye abantu zaziqondakala kubo kuphela. Oku akunjalo kwizihloko ezikumgangatho ophezulu, kodwa kamsinya nje ukuba uqale ukukhangela into ethile, abaphulaphuli abanakukwazi ukuqhubeka nawe. Uhlangabezana njani nale nto?

UMichael: Ayisoloko iphumelela. Kutshanje ndilungiselele ingxelo apho ndingene nzulu kakhulu kwiinkcukacha zobuchwepheshe. Njengoko intetho yayiqhubeka, kwacaca ukuba inkoliso yabaphulaphuli yayingandiqondi, ngoko kwafuneka ndiqhelane nemeko ngoko nangoko. Izilayidi azikwazanga ukutshintshwa, ngoko ke akuzange kuhambe kakuhle kakhulu - ke, ngokuthetha ngokubanzi, ndiyazama ukungasebenzisi izilayidi. Ngokubanzi, ingcebiso yam kukuqwalasela abaphulaphuli bakho. Kufuneka wazi ukuba uthetha nabani, yintoni inqanaba labo lolwazi, kwaye kufuneka bave ntoni ukuze bawuxabise umsebenzi wakho.

Vitaliy: Ngaba ungasinika ufifi malunga nokuba lecture ibimalunga nantoni?

UMichael: Ukunyaniseka, ndingathanda ukunganabi kwesi sihloko ukuze ndishiye abantu ekuthethwa ngabo bengaziwa. Ingongoma yeyokuba sihlala singena nzulu kakhulu kubuntsokothi bengxaki esisebenza kuyo, ngoko kuba nzima kuthi ukuchaza ekuqaleni kwentetho ukuba kutheni le ngxaki inomdla kwaye ibalulekile kwaye inxulumana njani nemiba ethi abaphulaphuli sele besazi. Ngokokuqaphela kwam, abafundi banelona xesha linzima lokufunda obu buchule. Kwaye le yayiyeyona ndawo ibuthathaka kwingxelo yam yakutshanje. Ingxelo ecwangciswe ngokufanelekileyo kufuneka, kwasekuqaleni, ifumane uqhagamshelwano nabaphulaphuli, ibacacisele ukuba yintoni kanye kanye ingxaki kwaye inxulumana njani nezihloko esele zaziwa kuyo. Indlela enobuchule ngayo le ntshayelelo ixhomekeke kubaphulaphuli. Ukuba i-motley ngokupheleleyo, ngoko ingxelo ingaba ngamanqanaba amaninzi. Intshayelelo kufuneka ifikeleleke kuye wonke umntu, kwaye ekugqibeleni isiqwenga asinakukwazi ukuhambisana nawe, kodwa abantu abaqhelene nentsimi yakho baya kukwazi ukuyiqonda.

Ngaba abafundi baba zizidenge? Iindlela zehlabathi, ukuthengiswa kwamanye amazwe.

Алексей: Ubujonga abafundi amashumi eminyaka. Ngaba abafundi bafumana i-dumber okanye ubulumko ukusuka kwishumi ukuya kwishumi leminyaka okanye unyaka nonyaka? E-Russia, oonjingalwazi bahlala bekhalaza ukuba abafundi bafumana i-dumber minyaka le, kwaye ayicacanga into enokuyenza ngayo.

UMichael: Uyayiva nyani i-negativity eninzi kuthi bantu badala. Ngokuzithoba, sinomkhwa wokulindela ukuba abafundi bathathe yonke le minyaka ingama-30 yamava esele sinabo. Ukuba ndinokuqonda okunzulu kunoko ndandikwenza ngo-1985, kutheni bengenakho nje abafundi? Mhlawumbi ngenxa yokuba beneminyaka engama-20 ubudala, ucinga ntoni? Ndicinga ukuba olona tshintsho lubalulekileyo kumashumi eminyaka akutshanje lube kukwakheka kwabantu: ngoku sinabafundi abaninzi bamazwe aphesheya, ngaphandle kwabaseCanada. Bekukade kukho abantu abaninzi baseCanada kuba sikufutshane kakhulu kumda waseCanada kwaye abafundi abasuka apho banokugoduka ngeempelaveki. Kodwa ngoku zininzi iiyunivesithi ezilungileyo eCanada, kwaye abantu baseCanada bakhetha ukufunda apha, bambalwa kakhulu abaza e-USA.

Алексей: Ngaba ucinga ukuba le yindlela yasekhaya okanye yehlabathi?

UMichael: Andikhumbuli ukuba ngubani kanye kanye, kodwa omnye umntu wathi umhlaba uthe tyaba. Intsimi yethu iye yaba ngamazwe ngamazwe. IiNkomfa zeACM Ngaphambili, babebanjwe kuphela e-United States, emva koko bagqiba ekubeni babambe kanye emva kweminyaka emi-4 kwamanye amazwe, kwaye ngoku babanjwe emhlabeni jikelele. Olu tshintsho lwachaphazela ngakumbi IEEE, njengoko ibisoloko iyintlangano yamazwe ngamazwe ngaphezu kwe-ACM. Kwaye kukho izitulo zeprogram ezivela eChina, eIndiya, eRashiya, eJamani nakwamanye amazwe amaninzi, kuba kuninzi okwenzekayo kuyo yonke indawo ngoku.

Алексей: Kodwa, mhlawumbi, kukho imiba engentle yokwenziwa kwamazwe ngamazwe?

UMichael: Ndingathi yonke imiba engalunganga ayihambelani neteknoloji, kodwa kwezopolitiko. Kudala, eyona ngxaki yayiyeyokuba i-US yayiba abona bantu bakrelekrele nabanobuchule abavela kumazwe kwihlabathi liphela. Kwaye ngoku eyona ngxaki inkulu yimidlalo yezopolitiko phakathi kwamazwe ahlukeneyo ajikeleze ii-visa kunye nokufudukela kwelinye ilizwe.

Алексей: Oko kukuthi, imiqobo kunye nezinto ezinjalo. Icacile.

Владимир: Mna ngokwam, ndinomdla kwindlela oyithathayo xa ufundisa isifundo esitsha kubafundi. Kukho iindlela ezahlukeneyo zokukhetha: unokuzama kuqala kubo bonke ukubakhuthaza ukuba bazame into entsha, okanye unokunikela ingqalelo ngakumbi kwiinkcukacha zendlela iteknoloji ethile isebenza ngayo. Ukhetha ntoni?

Umsebenzi osebenzayo kunye nabafundi

Алексей: Kwaye njani ukufumana ibhalansi damn phakathi yokuqala neyesibini?

UMichael: Ingxaki kukuba iiklasi azisoloko zihamba ngendlela endifuna ngayo. Ndidla ngokunika abafundi imathiriyeli yokufunda kwangaphambili ukuze bacamngce kuyo, bayiqonde kangangoko banako, baze baqulunqe imibuzo malunga nezo ndawo bangenakuziqonda. Emva koko eklasini unokugxila kwizinto ezinzima kakhulu kwaye uzihlolisise kunye. Le yindlela endithanda ngayo ukufundisa iiklasi kakhulu. Kodwa ngenxa yomthwalo okhoyo ngoku kubafundi, andisoloko ndikwazi ukuqinisekisa ukuba balungiselela kwangaphambili. Ngenxa yoko, kuya kufuneka uchithe ixesha elininzi ekubaliseni ngokutsha izinto ngokubanzi kunokuba ungathanda. Ngaphandle koku, ndizama ukugcina iiklasi zethu zisebenzisana. Ngaphandle koko, kulula ukurekhoda ividiyo xa abafundi benokuyibukela ekhaya. Inqaku leeklasi eziphilayo kukunxibelelana kwabantu. Eklasini, ndikhetha ukusebenzisa itshokhwe nebhodi kunezilayidi, ngaphandle kwakwiimeko ezithile xa umzobo untsokothile kakhulu ukuba ungaboniswa ebhodini. Enkosi koku, akufuneki ndibambelele kwisicwangciso sesifundo esiqinileyo. Ekubeni kungekho myalelo ungqongqo endiwunikela ngawo umbandela, oku kundivumela ukuba ndiwenze ulungele abaphulaphuli ngokuxhomekeke kwimibuzo endiyifumanayo. Ngokubanzi, ndizama ukwenza iiklasi zisebenzisane kangangoko kunokwenzeka, ukwenzela ukuba izinto endizinikezelayo zixhomekeke kwimibuzo endiyibuzwayo.

Владимир: Ilunge kakhulu. Kumava am, kunzima ukufumana abaphulaphuli ukuba babuze imibuzo. Nokuba ubuza kwangaphambili ukuba ubuze nayiphi na imibuzo, nokuba usisidenge okanye ulumke kangakanani na, bayathula. Uhlangabezana njani nale nto?

UMichael: Uza kuhleka, kodwa ukuba uhlala uthule ngokwaneleyo, ngokukhawuleza okanye kamva wonke umntu uya kukhululeka kwaye umntu uya kubuza umbuzo. Okanye unokubuza umbuzo olula wobugcisa ngoewe okanye hayi impendulo ukuze ubone ukuba abantu bayayiqonda na le nto ithethwayo. Umzekelo, ngaba kukho ugqatso lwedatha kulo mzekelo ungentla? Ngubani ocinga njalo? Ngubani ongacingiyo? Ngubani ongayiqondiyo yonke into, kuba sisiqingatha sezandla kuphela esiye sanyuka?

Vitaliy: Kwaye ukuba uphendule ngokungachanekanga, uyagxothwa eklasini :)

UMichael: Ukuba awukaphenduli nto, kufuneka ubuze umbuzo. Kufuneka ndiqonde ukuba yintoni kanye kanye ekufuneka umfundi eyazi ukuze aphendule umbuzo endiwubuzayo. Ndiyabadinga bandincede ndibancede. Ndikulungele ukuziqhelanisa nabo ukuze bayiqonde ingxaki. Kodwa ukuba andazi ukuba kuqhubeka ntoni entlokweni yabo, andinakuyenza. Kwaye ukuba awuniki abafundi uxolo ixesha elide ngokwaneleyo, ngamanye amaxesha ekugqibeleni babuza imibuzo efanelekileyo, oko kukuthi, endivumela ukuba ndibone ukuba yintoni kanye eyenzekayo kwiintloko zabafundi. 

Алексей: Ngaba le mibuzo maxa wambi ikhokelela kwiingcamango ongazange uzicinge wena ngaphambili? Ngaba bezingalindelekanga? Ngaba bayakuvumela ukuba ujonge ingxaki ngendlela entsha?

UMichael: Kukho imibuzo evula indlela entsha yokunikela umbandela. Kudla ngokubakho imibuzo ekhokelela kwiingxaki ezibangela umdla endingazimiselanga ukuthetha ngazo. Abafundi bahlala bendixelela ukuba ndithanda ukuphuma ngaphandle kwesihloko xa oku kusenzeka. Kwaye, ngokutsho kwabo, le yeyona nto inomdla kakhulu kwisifundo. Kunqabile kakhulu, kumatyeli nje ambalwa, abafundi babuza imibuzo eyabangela umkhombandlela omtsha kuphando kwaye yakhula yaba linqaku. Oku kwenzeka rhoqo kwiincoko nabafundi kunokuba ngexesha leeklasi, kodwa ngamanye amaxesha kwakusenzeka ngexesha leeklasi. 

Алексей: Ke abafundi bakubuze imibuzo ngesiseko apho bekunokwenzeka ukupapasha inqaku?

UMichael: Ewe. 

Vitaliy: Uba nencoko kangaphi nabafundi? Bafuna nini ukufunda okungakumbi kunoko bekugutyungelwe ebudeni besifundo?

UMichael: Ndikunye nabafundi bam abaphumeleleyo - ngalo lonke ixesha. Ndinaba-5 okanye aba-6 kubo, kwaye sixoxa ngento ethile nabo ngalo lonke ixesha. Kwaye iincoko zolu hlobo nabafundi abahamba nje kwiiklasi zam aziqhelekanga kakhulu. Nangona ndinqwenela ukuba oku kwenzeke rhoqo. Ndiyarhana ukuba boyika nje ukuza kwifakhalthi ngexesha lomsebenzi. Kwisiqingatha sonyaka ngasinye, abanye abafundi bayakwazi ukoyisa lo mqobo wengqondo, kwaye kuhlala kunomdla kakhulu ukuthetha nabo emva kweklasi. Liyinyaniso elokuba, ukuba bonke abafundi bebenesibindi ngolo hlobo, ngendingenaxesha laneleyo. Ngoko mhlawumbi yonke into isebenza njengoko kufanele. 

Vitaliy: Ukwazi njani ukufumana ixesha lokunxibelelana nabafundi? Ngokokwazi kwam, e-USA ootitshala banomsebenzi omninzi-ukufaka izicelo zenkxaso-mali nokunye okunjalo. 

UMichael: Ngokunyaniseka, ukusebenza nabafundi yeyona nkalo yomsebenzi wam endiwuthanda kakhulu. Ngoko ndinenkuthazo eyaneleyo yoku. Ixesha elininzi endilichitha eofisini yam ndilichithela kwiintlanganiso zazo zonke iintlobo. Kusehlotyeni ngoku, ngoko ke ishedyuli yam ayixakekanga kangako, kodwa ebudeni bonyaka wesikolo, yonke imihla ukususela ngo-9 ukuya kweye-17 ndiphethe yonke into. Umsebenzi wophando, uphononongo, izibonelelo - kuzo zonke ezi kukho ngokuhlwa kunye neempelaveki kuphela. 

Indlela yokuqhubeka nokulungiselela iikhosi ezintsha kunye neencwadi.

Алексей: Ngaba ngoku uyaqhubeka nokufundisa naziphi na iikhosi ubuzifundisa ixesha elide? Into efana nentshayelelo kwiNzululwazi yeKhompyutha.

UMichael: Into yokuqala ethi qatha engqondweni apha sisifundo ngeelwimi zokuprograma. 

Алексей: Yahluke njani inguqulelo yanamhlanje yesi sifundo kuleyo yayiyi-10, 20, 30 kwiminyaka eyadlulayo? Mhlawumbi eyona nto inomdla apha ayizizo iinkcukacha zekhosi ethile, kodwa iindlela eziqhelekileyo.

UMichael: Ikhosi yam kwiilwimi zokucwangcisa yayingaqhelekanga ngexesha endandiyidala. Ndaqala ukuyifunda ekupheleni koo-1980, ndithatha indawo yogxa wam, uDoug Baldwin (UDoug Baldwin). Isihloko sekhosi sasinxulumene kuphela nobuchule bam, kodwa xa wayehamba, ndandingoyena mgqatswa ubalaseleyo wokufundisa ikhosi. Andizange ndithande naziphi na iincwadi zezifundo ezazikho ngelo xesha, ngoko ke ndagqibela ngokubhala incwadi yesi sifundo ngokwam. (Inqaku lomhleli: sithetha ngencwadi "Iipragmatiki zoLwimi lokucwangcisa") Ngoku isetyenziswa kwiiyunivesithi ezingaphezu kwama-200 ehlabathini lonke. Indlela yam esetyenziswayo ayiqhelekanga kuba ixuba ngabom iingxaki zoyilo nokuphunyezwa kolwimi, kwaye inika ingqwalasela emandla ekusebenzisaneni phakathi kwale miba kuzo zonke iinkalo ezinokwenzeka. Indlela esisiseko iye yahlala ingatshintshi, njengoko kukho iikhonsepthi ezininzi ezisisiseko: ukutsalwa, indawo yamagama, imodyuli, iindidi. Kodwa iseti yeelwimi apho ezi ngcamango zibonakaliswa zitshintshe ngokupheleleyo. Xa ikhosi yayiqala ukudalwa, kwakukho imizekelo emininzi ePascal, kodwa namhlanje uninzi lwabafundi bam abaluva nolu lwimi. Kodwa bayayazi iSwift, Go, Rust, ke kufuneka ndithethe ngeelwimi ezisetyenziswayo namhlanje. Kananjalo, abafundi ngoku bazazi kakuhle iilwimi zokubhala, kodwa xa ndandiqala ukufundisa le khosi, yayimalunga neelwimi eziqokelelweyo. Ngoku sifuna izinto ezininzi malunga nePython, iRuby kunye nePerl, kuba yiyiphi ikhowudi ebhaliweyo kule mihla, kwaye kukho izinto ezininzi ezinomdla ezenzekayo kwezi lwimi, kubandakanywa kwintsimi yoyilo lweelwimi. 

Vitaliy: Emva koko umbuzo wam olandelayo uya kunxulumana nowangaphambili. Uhlala njani kule ndawo? Ndiyakrokrela ukuba ukuhlaziya ikhosi enje kufuna umsebenzi omninzi - kufuneka uqonde iilwimi ezintsha, uqonde iingcamango eziphambili. Uyenza njani le nto?

UMichael: Andikwazi ukuqhayisa ukuba ndihlala ndiphumelela i-100%. Kodwa ixesha elininzi ndenza nje into eyenziwa ngabanye abantu - funda i-Intanethi. Ukuba ndifuna ukuqonda iRust, ndiyiGoogle, yiya kwiphepha leMozilla kwaye ndifunde incwadi eposwe apho. Le yinxalenye yezinto ezenzekayo kuphuhliso lwezorhwebo. Ukuba sithetha ngesayensi, ngoko kufuneka ulandele iingxelo kwiinkomfa eziphambili. 

Ikhonkco phakathi kweshishini kunye nezemfundo

Vitaliy: Masithethe ngonxibelelwano phakathi kweshishini kunye nophando lwezenzululwazi. Kuluhlu lwakho lwemisebenzi, ndifumene amanqaku amaninzi malunga nokuhambelana kwe-cache. Ndiyaqonda ukuba i-cache consistency algorithms yayingazinzanga ngexesha lokupapashwa kwayo? Okanye ayisasazeki ngokwaneleyo. Zaziqheleke kangakanani izimvo zakho ekusebenzeni?

UMichael: Andiqinisekanga ncam ukuba zeziphi iimpapasho othetha ngazo. Ndenze umsebenzi omncinci kunye nabafundi bam uBill Bolosky (UWilliam Bolosky) kunye noLeonidas Kontotanassis (Leonidas Kontothanassis) ekuqaleni kweminyaka yoo-1990 ekulawuleni imemori koomatshini baseNeumann. Ngelo xesha, ishishini alizange libe nokuqonda malunga nendlela yokwenza ngokufanelekileyo inkqubo ye-multiprocessor: ngaba kufanelekile ukudala inkxaso yokufikelela kwimemori ekude kwinqanaba le-hardware, ngaba kufanelekile ukwenza imemori isasazwe, ngaba kunokwenzeka ukulayisha i-cache ukusuka? imemori ekude, okanye ngaba kuyimfuneko ukususa amaphepha kwigumbi lokusebenza? UBill noLeonidas bobabini basebenze kule ndawo kwaye bajonga iindlela ngaphandle kokulayishwa kwe-cache ekude. Oku bekunganxulumananga ngokuthe ngqo ne-cache coherence, kodwa ibisengumsebenzi kulawulo lwememori ye-NUMA, kwaye emva koko iindlela zangoku zokubekwa kwamaphepha kwiinkqubo zokusebenza zanamhlanje zakhula ukusuka koku. Lilonke, uBill noLeonidas benza umsebenzi obalulekileyo, nangona babengengowona unempembelelo kulo mmandla - babebaninzi abanye abantu ababesebenza kwinto enye ngelo xesha. Kamva, ndasebenza kwisihloko esihambelana nokuhambelana kwe-cache kumxholo wememori ye-hardware ye-hardware. Iqela endisebenze nalo kule ngxaki laphela lafumana amalungelo awodwa omenzi wechiza. Kukho iingcamango ezintle ezinomdla emva kwazo, kodwa andiqondi ukuba ziya kugqitywa ngokusetyenziswa. Enye indlela okanye enye, kunzima kum ukugweba inzuzo yabo. 

Алексей: Ngokuphathelele kulo mba, umbuzo othe ngqo wobuqu: kubaluleke kangakanani kuwe ukuba izimvo zakho zisetyenziswe? Okanye awucingi ngayo?

UMichael: Ndiyakuthanda ukubuza lo mbuzo kudliwanondlebe nabanye abantu, abafaki zicelo okanye abaviwa abafuna ukujoyina ifakhalthi. Andiqondi ukuba kukho impendulo echanekileyo kulo mbuzo. Abantu abenza izinto ezintle banokuba neentshukumisa ezahlukeneyo kakhulu. Nditsaleleka kwiingxaki kuba mna ngokobuqu ndizifumanisa zibangela umdla, kungekhona ngenxa yeengenelo zazo eziluncedo. Kodwa kwelinye icala, xa into enomdla isafumana isicelo, ndiyithanda kakhulu. Ngoko akukho lula apha. Kodwa ekuqaleni komsebenzi wam, andisayi kuqhutywa ngumbono wokuphela kokusetyenziswa kwehlabathi, kodwa ngokuvisisana kwengcamango kunye nomnqweno wokuyihlolisisa kwaye ndibone oko kuza kuyo. Ukuba ekugqibeleni inika iziphumo ezisebenzayo, inkulu. 

Алексей: Ngenxa yemfundo namava akho, ukwazi bhetele ukugweba ukuxabiseka kweengcamango zabanye abantu. Unokuzithelekisa kwaye ubone ukuba yeyiphi esebenza ngcono ngayo. Ndiqinisekile ukuba unoluvo malunga nezinto ezisetyenziswayo ngoku ngabavelisi abakhulu abafana ne-Intel. Ngokwembono yakho, ichaneke kangakanani ikhosi ezithathwa zezi nkampani?

UMichael: Ukuziqhelanisa kusoloko kujikeleza into enokuphumelela kwezorhwebo, oko kukuthi, ukwenza inzuzo, kwaye kungcono ubuze omnye umntu ngaloo nto. Umsebenzi wam uphumela kwiimpapasho, kwaye kwintsimi yeenkqubo zokusebenza zivavanywa ngokusekelwe kwizibonakaliso zokusebenza: isantya, ukusetyenziswa kwamandla, ubungakanani bekhowudi. Kodwa kum kwakusoloko kubonakala ngathi ezi ziphumo zinamandla zongezwa kumanqaku kuphela ukuze apapashwe, kwaye iinjongo zokwenene zabantu zomsebenzi zibuhle. Abaphandi bavavanya izisombululo ngokwembono yobugcisa, bayayikhathalela indlela ebukeka ngayo imibono, kwaye bazama ukwenza into engcono kuneendlela ezikhoyo. Abaphandi baqhutywa ziintshukumisa zobuqu, ezizimeleyo, zobuhle. Kodwa awukwazi ukubhala ngale nto kwinqaku ngokwayo; ezi zinto azikho iingxoxo zekomiti yeprogram. Ngethamsanqa, izisombululo ezintle zihlala zikhawuleza kwaye zingabizi. Ishumi elinesibini labalingane bam kunye nam saxoxa ngesi sihloko malunga ne-15 kwiminyaka eyadlulayo kwaye sagqiba ukubhala inqaku malunga nalo. Ndicinga ukuba usenokuyifumana ngoku, ibizwa ngokuba "Indlela yokuvavanya uphando lweenkqubo" okanye into enjalo, inababhali abangaphezu kweshumi elinesibini. Eli kuphela inqaku apho mna umbhali kunye Sasha Fedorova, ngoko ukuba ukhangela igama lakhe kuluhlu lwam lweempapasho, uya kuyifumana into oyifunayo. Ithetha malunga nokuvavanya iinkqubo zophando kunye nokuba kubaluleke kangakanani ubuhle. 

Алексей: Ngoko kukho umahluko phakathi komgangatho wento egqalwa njengelungileyo kwinzululwazi nakwishishini. Inzululwazi ivavanya ukusebenza, ukusetyenziswa kwamandla, i-TDP, ukukhululeka kokuphunyezwa, kunye nokunye okuninzi. Ngaba unalo ithuba lokuqhuba olu hlobo lophando eyunivesithi? Ngaba unayo ilabhoratri enoomatshini abohlukeneyo kunye nolwakhiwo olwahlukeneyo onokuqhuba kulo imifuniselo?

UMichael: Ewe, isebe lethu linoomatshini abaninzi abanomdla. Ubukhulu becala zincinci, sineqela elincinci kunye neenkqubo ezininzi ze-multiprocessor ezine-accelerators ezahlukeneyo. Ukongeza, ikhampasi ineziko elikhulu lekhompyuter elinceda izazinzulu ezivela kwiinkalo ezahlukeneyo ezahlukeneyo. Imalunga newaka leendawo kunye namashumi amabini amawaka eecores, zonke kwiLinux. Ukuba kukho imfuneko, unokuhlala uthenga ezinye ii-AWS. Ke asinazithintelo ezibalulekileyo ngehardware. 

Алексей: Kwakunjani kwiminyaka engamashumi amathathu eyadlulayo? Ngaba kwakukho iingxaki ngoko?

UMichael: Kwahluke kancinci ngoko. Phakathi ukuya kutsho ekupheleni kweminyaka yee-1980, inzululwazi yayithathwa njengento emfutshane kwizixhobo zekhompyutha. Ukulungisa le meko, iNational Science Foundation (Isiseko seSayensi yeSizwe) udale inkqubo yophando olulungelelanisiweyo lokulinga (uPhando loVavanyo oluLungeleleyo, iCER). Injongo yale nkqubo yayikukubonelela ngeziseko ezingundoqo zekhompyutha kumasebe eSayensi yeKhompyutha, kwaye ifumene utshintsho olubalulekileyo. Ngemali asinike yona, thina kwiYunivesithi yaseRochester sathenga iButterfly yeBBN enamaqhina ayi-1984 ngo-128, lo yayingunyaka ngaphambi kokuba ndifike apho. Ngelo xesha yayiyeyona nkqubo inkulu yehlabathi enenkumbulo ekwabelwana ngayo. Yayineeprosesa ze-128, nganye kwibhodi yomama eyahlukileyo, kwaye yayihlala iirack ezine. Iprosesa nganye yayine-megabyte yememori, i-128 megabytes ye-RAM yayisisixa esingenakucingelwa ngelo xesha. Kulo matshini siphumeze ukutshixa i-MCS okokuqala. 

Алексей: Ngoko, ukuba ndiyakuqonda ngokuchanekileyo, ngoko okwangoku ingxaki nge-hardware isonjululwe? 

UMichael: Ngokubanzi, ewe. Kukho izilumkiso ezimbalwa: okokuqala, ukuba wenza i-architecture yekhompyutheni kwinqanaba le-chip, kunzima ukwenza kwindawo yezemfundo kuba kukho izixhobo ezingcono kakhulu zokwenza kwishishini. Ukuba ufuna nantoni na encinci kune-nanometers ezili-10, kuya kufuneka uyiodole komnye umntu. Kule ndawo kulula kakhulu ukuba ngumphandi kwi-Intel. Ukuba usebenza kunxibelelwano lwe-optical kwiichips okanye kwimemori yesimo esiqinileyo, uya kufumana itekhnoloji kwishishini elingekabikho kwisayensi, ke kuya kufuneka udale umanyano. Umzekelo, uStephen Swanson (USteven Swanson) yenziwe intsebenziswano enjalo kwitekhnoloji entsha yememori. Le fomu ayisoloko isebenza, kodwa kwezinye iimeko inokuphumelela kakhulu. Ukongeza, kwisayensi uphuhliso lwezona nkqubo zinamandla zekhompyuter zinzima kakhulu. Ezona projekthi zinkulu zinkulu ngoku e-US, eJapan nase China zonke zigxile kushishino. 

Ukuphunyezwa okusebenzayo kweengcamango. MCS, MS, CLH, JSR 166, esebenza noDoug Lee kunye nokunye.

Vitaliy: Sele uthethe malunga nendlela oqale ngayo ukusebenza kwi-algorithms yongqamaniso. Unamanqaku amabini adumileyo malunga Ukuvalela kwe-MCS и Umgca kaMichael-Scott (MS), eyathi ngengqiqo iphunyezwe kwiJava. (Inqaku lomhleli: zonke iimpapasho zinokujongwa ikhonkco). Apho oku kuthintelwa kwaphunyezwa ngotshintsho oluthile kwaye kwavela Isitshixo se-CLH, kwaye umgca waphunyezwa njengoko bekucetyiwe. Kodwa kwadlula iminyaka emininzi phakathi kokupapashwa kwamanqaku enu nendlela asebenza ngayo. 

Алексей: Kubonakala malunga neminyaka eyi-10 kwimeko yomgca.

UMichael: Ngaphambi kokuba ezi mpawu zivele kwilayibrari eqhelekileyo yeJava?

Vitaliy: Ewe. Wenza ntoni ukuze kwenzeke oku? Okanye akukho nto bayenzileyo?

UMichael: Ndingakuxelela indlela i-MS Queue eyangena ngayo kwiJava 5. Kwiminyaka embalwa ngaphambi kokuba iphume, ndasebenza kunye neqela likaMark Moyers e-Sun Microsystems kwilebhu yabo kufuphi neBoston. Waququzelela iworkshop yabantu abaziyo abasebenza kwiingxaki ezinomdla kwimultithreading kuba wayefuna ukufumana izihloko anokuthi azithengisele inkampani yabo. Kulapho ndaqala khona ukudibana noDoug Lea. Mna noDoug nabanye abantu abamalunga nama-25 baseSun sasikunye sixoxa ngentetho kaDoug JSR 166, eyathi kamva yaba yi-java.util.concurrent. Endleleni, uDoug wathi angathanda ukusebenzisa umgca we-MS, kodwa ngenxa yoku ufuna ikhawuntala yenani lezinto ezikumgca wojongano. Oko kukuthi, oku bekufanele ukuba kwenziwe ngendlela eyahlukileyo, i-athomu, echanekileyo kwaye ngokukhawuleza. Ndicebise ukuba ndongeze amanani e-serial kwi-node, ndithathe inani le-node yokuqala kunye neyokugqibela kwaye ndisuse enye kwenye. UDoug wazonwaya intloko, wathi “kutheni,” waza wagqiba ekubeni enze loo nto kanye. Saxoxa ngokuphumeza le ndlela kwithala leencwadi, kodwa uDoug wawenza ngokwakhe umsebenzi omninzi. Ngenxa yoko, wakwazi ukuseka inkxaso ye-multithreading ebalaseleyo kwiJava. 

Алексей: Ke, ukuba ndiqonda kakuhle, indlela .ubungakanani() bekumele ukuba yinxalenye yojongano olusemgangathweni lwefolo, kwaye bekufanele ukuba ibenobunzima bealgorithmic ye O(1)?

UMichael: Ewe, kwaye ukongeza kule nto, i-counter eyahlukileyo iyafuneka.

Алексей: Kuba ukuba ufowunela .ubungakanani () indlela kwiJava, isiphumo silindeleke ukuba sifumaneke ngokukhawuleza kwaye singasekelwanga kubungakanani bengqokelela. Ndiyabona, enkosi.

UMichael: Kwiminyaka embalwa kamva ndandisebenza kwizakhiwo zedatha ezimbini kunye nomfundi wam uBill Scherer - enyanisweni, yile nto ndiza kuthetha ngayo. ingxelo ngeHydra. UDoug weza kuthi waza wathi unokuzisebenzisa kwiSakhelo soMgwebi weJava. Bekunye noMthetho oYilwayo, badale ukuphunyezwa okubini, okubizwa ngokuba yimigca enobulungisa nengalunganga. Ndibacebise ngale projekthi, nangona ndingazange ndithathe inxaxheba ekubhaleni ikhowudi yokwenene. Ngenxa yoko, isantya sabenzi bokufa siye sanda kakhulu. 

Владимир: Ngaba ukhe wadibana nokuphunyezwa okungachanekanga kwee-algorithms zakho okanye izicelo zokongeza amanqaku amatsha? Ngokubanzi, ukuziqhelanisa kufuneka kuhambelane nethiyori, kodwa amaxesha amaninzi ziyahluka. Masithi ubhale i-algorithm, kwaye ephepheni iyasebenza, kodwa abantu ababandakanyekayo ekuphunyezweni baqala ukukubuza iimpawu ezingaphezulu okanye uhlobo oluthile lokulungisa i-algorithm. Ngaba wakha waba neemeko ezinjalo?

UMichael: Umzekelo kuphela apho umntu weza kum waza wabuza "indlela yokuyiphumeza" ngumbuzo kaDoug, esele ndithetha ngayo. Kodwa kuye kwakho iimeko ezimbalwa apho kuye kwenziwa utshintsho olubangel’ umdla ukuze lulungele iimfuno eziluncedo. Umzekelo, iqela le-K42 e-IBM liguqule isitshixo se-MCS kwaye lenza ujongano olusemgangathweni ngoko kungabikho mfuneko yokudlula indawo yomgca emva naphambili ukuya kukufumana kunye nokukhulula iindlela. Ngombulelo kolu jongano oluqhelekileyo, umbono owawumhle kwithiyori waqala ukusebenza ngokusebenza. Kuyamangalisa ukuba abazange bapapashe inqaku malunga nalo, kwaye nangona befumene i-patent, kamva bayishiya. Lo mbono wawumangalisa, yaye ndizama ukuthetha ngawo nanini na kunokwenzeka. 

Kukho ezinye iimeko apho abantu baye benza uphuculo kwii-algorithms endizipapashiweyo. Ngokomzekelo, umgca we-MS unendlela yokufakela amanyathelo amabini, nto leyo ethetha ukuba kukho ii-CAS ezimbini kwindlela ebalulekileyo yomgca. Kwiimoto ezindala, iCAS yayibiza kakhulu. I-Intel kunye nabanye abavelisi baye baphucula kakuhle kutshanje, kodwa kanye ngexesha le miyalelo ye-30-cycle, ngoko ke ukuba ngaphezu kwesinye kwindlela ebalulekileyo kwakungafuneki. Ngenxa yoko, kwaphuhliswa umgca owahlukileyo owawufana nomgca we-MS, kodwa owawunomsebenzi omnye weathom kwindlela ebalulekileyo. Oku kwaphunyezwa ngenxa yokuba ngexesha elithile umsebenzi unokuthatha u-O(n) ixesha, kunokuba u-O(1). Kwakungenakwenzeka, kodwa kunokwenzeka. Oku kwenzeka ngenxa yokuba ngamaxesha athile i-algorithm yadlula umgca ukusuka ekuqaleni ukuya kwindawo yangoku kulo mgca. Ngokubanzi, i-algorithm iye yaphumelela kakhulu. Ngokokwazi kwam, ayisetyenziswanga kakhulu, ngokuyinxenye ngenxa yokuba imisebenzi yeathom ifuna izixhobo ezimbalwa kakhulu kunangaphambili. Kodwa umbono wawumkhulu. Ndikwathanda kakhulu umsebenzi kaDave Dice ovela kwi-Oracle. Yonke into ayenzayo iyasebenza kwaye usebenzisa intsimbi ngobuchule. Wayenesandla kuninzi lwe-NUMA-aware synchronization algorithms kunye nezakhiwo zedatha ezinemisonto emininzi. 

Владимир: Xa ubhala i-algorithms okanye ufundisa abafundi, umphumo womsebenzi wakho awubonakali ngokukhawuleza. Kuthatha ixesha ukuba uluntu luqhelane, masithi, inqaku elitsha. I-algorithm entsha ayifumani ngokukhawuleza isicelo. 

UMichael: Kukude ukuba kucace ngokukhawuleza ukuba inqaku liya kubaluleka okanye hayi. Ndicinga ukuba kuya kuba mnandi ukwenza uphononongo lwamaphepha aphumelele amabhaso kwiinkomfa. Oko kukuthi, jonga amanqaku abantu abakwiikomiti zenkqubo ngaxa lithile babewagqala njengawona alungileyo. Kufuneka uzame ukubala ngenani lamakhonkco kunye nefuthe kwishishini ukuba la manqaku aphenjelelwa kangakanani na kwi-10, 20, 25 iminyaka. Ndiyathandabuza ukuba kuya kubakho unxibelelwano olukhulu phakathi kwezi zimbini. Ayizi kuba ngu-zero, kodwa kusenokwenzeka ukuba iya kuba buthathaka kakhulu kunokuba besingathanda. Iingcamango ezininzi zihlala zingabangwa ixesha elide ngaphambi kokuba zisasazeke. Umzekelo, makhe sithathe inkumbulo yetransekshini. Ngaphezulu kweminyaka eli-10 yadlula ukususela kwixesha lapapashwa inqaku lokuqala ukuya kwixesha abantu abaqala ngalo ukwakha oomatshini ngalo. Kwaye ngaphambi kokubonakala kwale nkumbulo kwiimveliso zorhwebo - kunye nazo zonke 20. Ixesha elide kakhulu akukho mntu wanikela ingqalelo kwinqaku, kwaye ke inani lamakhonkco kuyo landa ngokukhawuleza. Kuya kuba nzima ukuqikelela oku kwangaphambili. Kwelinye icala, ngamanye amaxesha izimvo zifumana ukuphunyezwa kwangoko. Kwiminyaka embalwa edlulileyo, ndabhala iphepha kunye noJoe Izraelevitz kwi-DISC ecebise inkcazo entsha esemthethweni yokuqinisekiswa kwezakhiwo zedatha eziqhubekayo ezinokuthi zisetyenziswe emva kokuba ikhomputha iqhuba ukuphahlazeka. Ndilithandile inqaku kwasekuqaleni, kodwa liye lathandwa ngakumbi kunokuba bendilindele. Yayisetyenziswa ngamaqela amaninzi ahlukeneyo kwaye ekugqibeleni yaba yinkcazo esemgangathweni yezakhiwo zokuzingisa. Yiyiphi, ngokuqinisekileyo, intle.

Владимир: Ngaba kukho naziphi na iindlela ozisebenzisayo xa uhlola? Ngaba uyazama ukuvavanya amanqaku akho kunye nabafundi bakho? Ngokubhekiselele ekubeni umntu omfundisileyo uhamba ngendlela eyiyo.

UMichael: Njengabo bonke abantu, ndinikela ingqalelo ngakumbi kwinto endiyenzayo okwangoku. Kwakhona, njengaye wonke umntu, ngamanye amaxesha ndiye ndijonge i-Google Scholar ukuze ndibone ukuba amaphepha am adlulileyo ayacatshulwa, kodwa oko kubangelwa kukufuna ukwazi. Ubukhulu becala ndithathekile kwinto eyenziwa ngabafundi bam ngoku. Xa kuziwa ekuvavanyeni umsebenzi wangoku, inxalenye yawo yingqwalasela yobuhle, yintoni entle nengeyiyo. Kwaye kwinqanaba lemihla ngemihla, imibuzo evulekileyo idlala indima enkulu. Umzekelo, umfundi uza kum kunye negrafu yeziphumo ezithile, kwaye sizama ukuqonda ukuba ukuziphatha okungaqhelekanga kwegrafu kuvela phi. Ngokuqhelekileyo, kumsebenzi wethu sizama rhoqo ukuqonda izinto esingaziqondi. 

Inkumbulo yentengiselwano

Vitaliy: Mhlawumbi singathetha kancinci malunga nememori yentengiselwano?

UMichael: Ndicinga ukuba kufanelekile ukuthetha kancinci kuba ndibeke umzamo omkhulu kuyo. Lo ngumxholo endineempapasho ezininzi kuwo nawuphi na omnye. Kodwa kwangaxeshanye, ngokungaqhelekanga, bendihlala ndithandabuza kakhulu malunga nenkumbulo yentengiselwano. Kwiingcinga zam, Inqaku likaHerlihy kunye noMoss (M. Herlihy, J. E. B. Moss) yapapashwa ngaphambi kwexesha layo. Ekuqaleni koo-1990, bacebise ukuba inkumbulo yentengiselwano inokunceda abadwelisi benkqubo abanetalente basebenze kwizakhiwo zedatha ezinemisonto emininzi, ukuze ezi zakhiwo zisetyenziswe njengamathala eencwadi ngabadwelisi benkqubo abaqhelekileyo. Oko kukuthi, kuya kuba luncedo ku-Doug Lee ukwenza i-JSR 166 yakhe. Kodwa imemori yokuthengiselana yayingajoliswanga ukwenza inkqubo enemisonto emininzi ibe lula. Kodwa le yindlela kanye eyaqala ngayo ukuqondwa ekuqaleni kweminyaka yoo-2000, xa yathi yasasazeka. Kwapapashwa njengendlela yokusombulula ingxaki yenkqubo ehambelanayo. Le ndlela isoloko ibonakala ingenathemba kum. Inkumbulo yotshintshiselwano inokwenza kube lula ukubhala izakhiwo zedatha ezinxuseneyo. Oku, kubonakala kum, yinto awayizuzileyo. 

Malunga nobunzima bokubhala ikhowudi enemisonto emininzi

Алексей: Inika umdla kakhulu. Kubonakala ngathi kukho umqobo othile phakathi kwabaprogram abaqhelekileyo kunye nabo banokubhala ikhowudi enemisonto emininzi. Kunyaka ophelileyo, ndathetha izihlandlo ezininzi nabantu ababesebenzisa isakhelo se-algorithmic. Umzekelo, kunye noMartin Thomson, kunye nabadwelisi benkqubo abasebenza kumathala eencwadi anemisonto emininzi. (Inqaku lomhleli: UMartin Thompson ngumphuhlisi odumileyo, wabhala Ukuphazamisa и IAeron. Kwaye naye uye wenjenjalo ingxelo kwinkomfa yethu ye-Joker 2015, ukurekhoda ividiyo iyafumaneka kuYouTube. Uyafana ivuliwe kule nkomfa ukurekhoda okubalulekileyo nayo ikhona). Owona mceli mngeni mkhulu, bathi, kukwenza ii-algorithms zikhawuleze kwaye zisebenziseke lula. Oko kukuthi, bazama ukunqoba lo mqobo kwaye batsale abantu abaninzi kangangoko kunokwenzeka kule ndawo. Ucinga ntoni ngayo?

UMichael: Le yeyona ngxaki iphambili yokuphindaphinda: indlela yokufezekisa ukusebenza okuphezulu ngaphandle kokunyusa ubunzima benkqubo. 

Алексей: Kuba xa bezama ukuphepha ukuntsonkotha, i-algorithm iba ngaphantsi jikelele.

UMichael: Isitshixo apha siyilwe ngokufanelekileyo imifuziselo. Kubonakala kum ukuba le yeyona nto iphambili kwiinkqubo zekhompyuter njengentsimi. UButler Lampson uyathanda ukusebenzisa eli gama, kwaye usibiza ngokuba "ngabarhwebi bezinto ezithathwayo." Itekhnoloji ezilula azikho namhlanje. Iiprosesa esizisebenzisayo zineebhiliyoni ezili-10 ze-transistors-ukulula akukho mbuzo. Kwangaxeshanye, i-ISA ilula kakhulu kuneprosesa, kuba sisebenze ixesha elide ukuyibonelela ngokusebenza okuphezulu kunye nojongano olulula. Kodwa ayizizo zonke izinto ezigudileyo naye. Ingxaki efanayo ikwi-accelerators ebonakala ngoku kwimarike. Imibuzo iyavela-ukwenziwa njani ujongano olululo lweGPU, indlela yoguqulelo oluntsonkothileyo, ucinezelo, indlela yokudlulisa iikhowudi, umatshini wealgebra womgca, okanye iFPGA ebhetyebhetye ngakumbi. Uyenza njani i-interface eyenza isixhobo sisebenziseke lula kwaye sifihle ubunzima? Ayiyi kuyisusa, kodwa kunokuba uyifihle kumdwelisi wenkqubo olula. 

Алексей: Njengoko ndiyiqonda, sisenomqobo ekuqondeni ama-abstracts. Masithathe imodeli yememori kwinqanaba lethu lophuhliso lwesayensi kunye nethekhnoloji, le yenye yezinto eziphambili. Ndiyabulela kuyo, bonke abaprogram bahlula ngamaqela amabini: inxalenye enkulu ngabo bangayiqondiyo, kwaye inxalenye encinci ngabo baqondayo, okanye bacinga ukuba bayaqonda. 

UMichael: Umbuzo olungileyo - ngaba kukho nabani na kuthi oyiqondayo imodeli yememori?

Vitaliy: Ngokukodwa kwi-C++.

UMichael: Thetha noHans Boehm ngelinye ixesha. Ngomnye wabona bantu bakrelekrele ndibaziyo, ingcaphephe ekhokelayo kwiimodeli zenkumbulo. Uya kukuxelela kwangoko ukuba kuninzi angaziqondiyo. Kodwa ukuba sibuyela kumbandela wokukhutshwa, ngoko, ngokombono wam, ingcamango ebaluleke kakhulu kwintsimi yeemodeli zememori kwiminyaka engama-30 edlulileyo yaboniswa. kwiDissertation kaSarita Adve. (Inqaku lomhleli: uluhlu olupheleleyo lopapasho luyafumaneka ikhonkco).

Алексей: Umbuzo wam ngowokuba: ngaba lo mqobo uvela kuhlobo lwalo mbono? 

UMichael: Hayi. USarita wafika kwisigqibo sokuba ngendlela efanelekileyo, unokufihla ngempumelelo yonke into enzima, ufumane ukusebenza okuphezulu kwaye unike umdwelisi we-API elula. Kwaye ukuba ulandela le API, unokufezekisa ukungaguquguquki. Ndicinga ukuba le yimodeli efanelekileyo. Bhala ikhowudi ngaphandle kwemidyarho yedatha kwaye ufumane ukulandelelana okulandelelanayo. Kakade ke, ukuze kuncitshiswe amathuba okubaleka, kufuneka izixhobo ezikhethekileyo, kodwa ngomnye umcimbi lowo. 

Владимир: Ngaba kukho amaxesha kumsebenzi wakho apho ingxaki eyayibonakala isonjululwe ngequbuliso yajika yaba yintlekele, okanye kwafumaniseka ukuba le ngxaki yayingenakulungiswa? Umzekelo, ngokwethiyori ungalinganisa naliphi na inani okanye ubone ukuba naliphi na inani liphambili. Kodwa ekusebenzeni oku kunokuba nzima ukwenza nge-hardware yangoku kunzima ukulinganisa amanani. Ngaba kwenzeka into efanayo nakuwe?

UMichael: Andikhumbuli kwangoko into enjalo. Kuye kwakho amaxesha apho kwakubonakala ngathi akukho nto iseleyo yokwenza kwindawo ethile, kodwa emva koko kwenzeka into entsha nenomdla apho. Umzekelo, bendicinga ukuba indawo yomgca ongenamda sele ifikelele ekuvuthweni. Emva kophuculo oluninzi kumgca we-MNS, akukho nto ingako yenzekayo. Kwaye ke uMorrison (uAdam Morrison) kunye noAfek (Yehuda Afek) baqulunqa Umgca we-LCRQ. Kwacaca ukuba umgca ongenamkhawulo wemisonto emininzi wawunokwenzeka, apho ixesha elininzi kwakukho kuphela umyalelo wokulanda kunye nokunyusa kwindlela ebalulekileyo. Kwaye oku kwenza ukuba kube nokwenzeka ukuphumeza umyalelo wobungakanani bomsebenzi ongcono. Ayikuko ukuba asazi ukuba ukulanda kunye nokunyusa kuyinto eluncedo kakhulu. U-Eric Freudenthal wabhala malunga noku kumsebenzi wakhe kwi-Ultracomputer kunye no-Allan Gottlieb ekupheleni kweminyaka yee-1980, kodwa yayimalunga nemigca emincinci. UMorrison kunye no-Afek bakwazile ukusebenzisa i-fetch-and-increment kumgca ongenamida.

Uyilo olutsha. Ngaba uloyiso lwenkumbulo yentengiselwano ivaliwe?

Владимир: Ngaba ujonge izisombululo ezintsha zokwakha ezinokuba luncedo kwii-algorithms? 

UMichael: Ewe, zininzi izinto endifuna ukuzibona ziphunyezwa. 

Владимир: Luhlobo olunjani, umzekelo?

UMichael: Okokuqala, izandiso ezimbalwa ezilula kwimemori yethu ye-hardware-level transactional in Intel kunye ne-IBM processors. Ngokukodwa, ndingathanda ukuba umthwalo ongeyontengiselwano kunye nevenkile esanda kwenzeka ukuba ifumaneke ngokukhawuleza ngaphakathi kweentengiselwano. Ngokukhawuleza bakhokelela kwi-loops kwimeko-ngaphambi kokulandelelana, ngoko ke kunokuba nzima. Kodwa ukuba ugcina umaleko wokutsalwa, zininzi izinto ezinomdla kakhulu onokuzenza ngaphandle kwentengiselwano ngelixa isenzeka. Andazi ukuba kuya kuba nzima kangakanani ukuphumeza oku, kodwa kuya kuba luncedo kakhulu. 

Enye into eluncedo kukulayisha i-cache kwimemori ekude. Ndicinga ukuba kungekudala okanye kamva oku kuya kwenziwa. Le teknoloji iya kuvumela ukudalwa kweenkqubo ezinememori edibeneyo. Kuyakwenzeka ukugcina, yithi, i-100 terabytes yememori engaguqukiyo kwi-rack, kwaye inkqubo yokusebenza ngokwayo iya kuthatha isigqibo ngokuguquguqukayo ukuba ngawaphi amacandelo aloo nkumbulo afanele ahambelane nendawo yedilesi yendawo yeprosesa. Oku kuya kuba luncedo kakhulu kwi-computing yefu, njengoko kuya kuvumela inani elikhulu lememori ukuba linikezelwe kwimisebenzi eyidingayo. Ndicinga ukuba kukho umntu oza kuyenza.

Vitaliy: Ukugqiba ukuthetha ngememori yentengiselwano, ndinombuzo omnye ngaphezulu kwesi sihloko. Ngaba inkumbulo yentengiselwano ekugqibeleni iya kuthatha indawo yezakhiwo zedatha ezinemisonto emininzi?

UMichael: Hayi. Iintengiselwano yindlela eqikelelwayo. Kwinqanaba lenkqubo ezi zitshixo zeatom, kodwa ngaphakathi ziingqikelelo. Uqikelelo olunjalo lusebenza ukuba uninzi loqikelelo luchanekile. Ke ngoko, inkumbulo yentengiselwano isebenza kakuhle xa imisonto ingasebenzi kakuhle, kwaye kufuneka uqiniseke ukuba akukho lunxibelelwano. Kodwa ukuba umyalezo uqala phakathi kwemisonto, ukuthengiselana akuncedi kangako. Makhe ndichaze, sithetha ngetyala xa iintengiselwano zisongelwe kulo lonke umsebenzi weathom. Zisenokusetyenziswa ngempumelelo njengamacandelo olwakhiwo lwedatha enemisonto emininzi. Umzekelo, ukuba ufuna amagama amathathu CAS, kwaye kufuneka uphindaphinde izinto ezintathu ezincinci embindini wealgorithm enemisonto emininzi esebenza ngemisonto engamashumi amabini ngexesha elinye. Ngokubanzi, ukuthengiselana kunokuba luncedo, kodwa abayi kuphelisa isidingo sokuyila ngokufanelekileyo izakhiwo zedatha ezininzi. 

Imemori engaguqukiyo, i-Optane DIMM, izixhobo ezikhawulezayo.

Vitaliy: Into yokugqibela endifuna ukuthetha ngayo sisihloko sophando lwakho lwangoku: imemori engaguqukiyo. Yintoni esinokuyilindela kule nkalo kwixesha elizayo elingekude? Mhlawumbi uyazi naziphi na iinkqubo ezisebenzayo esele zikhona? 

UMichael: Andiyongcali yehardware, ndiyazi kuphela into endiyifunde ezindabeni kunye nento endiyixelelwa ngabalingane bam. Wonke umntu sele evile ukuba i-Intel iyathengisa Optane DIMM, enamaxesha angama-3 e-latency efundwayo kunye namaxesha e-10 e-latency yokubhala kune-RAM eguquguqukayo. Kungekudala ziya kufumaneka kwiinguqulelo zevolumu enkulu kakhulu. Kuyahlekisa ukucinga ukuba unokuba nelaptop ene-terabytes ezininzi ze-byte-addressable RAM. Kungenzeka ukuba kwiminyaka eyi-10 siya kuthatha isigqibo sokusebenzisa obu buchwepheshe butsha, kuba sisebenzisa i-DRAM - yandisa nje ivolumu. Kodwa ngenxa yokuzimela kwamandla, amathuba amatsha ngokupheleleyo avuleleka kuthi. Sinokutshintsha ngokwesiseko isitakhi sogcino ukuze kungabikho kwahlulo phakathi kwenkumbulo esebenzayo ene-byte-addressable kunye ne-block-structured memory eqhubekayo. Ngaloo ndlela, asiyi kudinga ukulungelelanisa yonke into efuna ukuhanjiswa ukusuka kwiprogram enye ukuya kwenye kwiifayile eziqingqiweyo. Kule nto sinokufumana imigaqo emininzi ebalulekileyo echaphazela iinkqubo zokusebenza, indawo yokusebenza, kunye nokusabalalisa izitolo zedatha. Lo mmandla unika umdla kakhulu ukusebenza kuwo. Ngokwam, kunzima kum ukuqikelela ukuba oku kuya kukhokelela phi, kodwa iingxaki apha ziyonwabisa kakhulu. Kusenokubakho utshintsho oluguquguqukayo apha, kwaye lulandela ngokwendalo kakhulu kumsebenzi wofundo oluninzi, kuba ukusilela ukubuyisela yinkqubo "yokuphindaphinda" ecaleni komsebenzi wesiqhelo wenkqubo. 

Isihloko sesibini esiphambili endisebenza kuso ngoku kukulawula izixhobo ezikhawulezayo kunye nokufikelela ngokukhuselekileyo kwizixhobo ukusuka kwindawo yomsebenzisi enolawulo lomgaqo-nkqubo wenkqubo. Kwiminyaka yakutshanje, kukho umkhwa wokuhambisa ukufikelela kwisixhobo kwindawo yomsebenzisi. Oku kwenziwa kuba i-TCP-IP kernel stack ayikwazi ukusebenza phezu kojongano lwenethiwekhi efuna ipakethe entsha rhoqo nge-5 microseconds ayizukuqhubeka. Ngoko ke, abavelisi banikezela ukufikelela ngokuthe ngqo kwizixhobo. Kodwa oku kuthetha ukuba inkqubo yokusebenza ilahlekelwa ulawulo lwenkqubo kwaye ayikwazi ukubonelela ngokufikelela ngokufanelekileyo kwisixhobo kwizicelo ezikhuphisanayo. Iqela lethu lophando likholelwa ukuba le ntsilelo inokuphetshwa. Siza kuba nenqaku malunga noku kwi-USENIX ATC kule nyanga. Inxulumene nomsebenzi wokuzingisa, kuba i-byte-addressable memory ehlala ixesha elide, eneneni, sisixhobo esine-I/O ekhawulezayo kakhulu ekufuneka ifikelelwe kwindawo yomsebenzisi. Olu phando lwenza ukuba kubekho iindlela ezintsha zokujonga iimicrokernels, exokernels, kunye nezinye iinzame zemveli zokuhambisa ngokukhuselekileyo ukusebenza ukusuka kwi-OS kernel ukuya kwindawo yomsebenzisi. 

Владимир: Imemori ye-Byte-addressable inkulu, kodwa kukho umda womzimba - isantya sokukhanya. Oku kuthetha ukuba ngokuqinisekileyo kuya kubakho ukulibaziseka xa usebenzisana nesixhobo. 

UMichael: Kulungile.

Владимир: Ngaba kuya kubakho umthamo owaneleyo wokumelana nemithwalo emitsha?

UMichael: Ngumbuzo oncomekayo lo, kodwa kuya kuba nzima ukuba ndiwuphendule. Umbono wokucubungula kwimemori sele ukho ixesha elide, unomdla kakhulu, kodwa unzima kakhulu. Andizange ndisebenze kule ndawo, kodwa kuya kuba kuhle ukuba ezinye izinto zifunyenwe apho. Ndiyoyika ukuba andinanto yokongeza. 

Владимир: Kukho enye ingxaki. Izixa esitsha, ezikhulu kakhulu ze-RAM azinakukwazi ukungena kwi-CPU. Ke ngoko, ngenxa yokulinganiselwa komzimba, le RAM kufuneka ibe yodwa. 

UMichael: Konke kuxhomekeke kwinani leziphene kwimveliso yeesekethe ezidibeneyo. Ukuba kwakunokwenzeka ukwenza ii-wafers ze-semiconductor ngokupheleleyo ngaphandle kweziphene, ngoko kuya kwenzeka ukwenza yonke i-microcircuit kuyo. Kodwa ngoku asiyazi indlela yokwenza ii-microcircuits zibe zinkulu kunezitampu zokuposa. 

Владимир: Kodwa sisathetha ngeesayizi ezinkulu, malunga neesentimitha. Oku ngokuqinisekileyo kunempembelelo kwi-latency. 

UMichael: Ewe. Akukho nto unokuyenza malunga nesantya sokukhanya. 

Владимир: Ngelishwa. 

Umkhwa omkhulu olandelayo. Izakhiwo zedatha ezimbini. Hydra.

Vitaliy: Ngokokuqonda kwam, ubamba iindlela ezintsha ngokukhawuleza. Ungomnye wabokuqala ukusebenza kwimemori yentengiselwano, kwaye ungomnye wabokuqala ukusebenza kwimemori engaguquguqukiyo. Ucinga ukuba iya kuba yintoni intsingiselo enkulu elandelayo? Okanye mhlawumbi yimfihlo?

UMichael: Uthetha inyani, andazi. Ndiyathemba ukuba ndiza kukwazi ukuqaphela xa into entsha ivela. Andikhange ndibe nethamsanqa ngokwaneleyo lokuyila nayiphi na intsimi entsha ngokwam, kodwa ndibenethamsanqa elincinci kwaye ndiye ndakwazi ukuqalisa ukusebenza kwangoko kumacandelo amatsha adalwe ngabanye. Ndiyathemba ukuba ndiza kukwazi ukwenza oku kwixesha elizayo.

Алексей: Umbuzo wokugqibela kolu dliwano-ndlebe uya kuba malunga nokusebenza kwakho eHydra kunye nezinto ozenzayo esikolweni. Ukuba ndiyaqonda ngokuchanekileyo, ingxelo esikolweni iya kuba malunga ne-blocking-free algorithms, kunye nenkomfa malunga nezakhiwo zedatha kabini. Ngaba ungathetha amagama ambalwa ngezi ngxelo?

UMichael: Ngokuyinxenye, sele sichukumisile kule mibandela nawe kolu dliwano-ndlebe. Imalunga nomsebenzi endiwenzileyo nomfundi wam uBill Scherer. Wabhala ithisisi kuyo, kwaye uDoug Lee naye waba negalelo kuyo, kwaye ekugqibeleni yaba yinxalenye yemigca edibanisa imisonto emininzi kwilayibrari yeJava. Masicinge ukuba isakhiwo sedatha sifundwa kwaye sibhaliwe ngaphandle kokuthintela, oko kukuthi, umsebenzi ngamnye unenani elilinganiselwe lemiyalelo kwindlela ebalulekileyo. Ukuba uzama ukususa idatha kwisitya esingenanto, okanye uzame ukususa idatha ethile engekho kule ngqungquthela, ukwaziswa ngokukhawuleza ukuba oku akunakukwenziwa. Kodwa le ndlela yokuziphatha ayinakwamkeleka ukuba umsonto ufuna ngokwenene le datha. Emva koko into yokuqala efika engqondweni kukudala i-loop eya kuhlala ibuza ukuba idatha efunekayo ibonakala. Kodwa ke kukho ukuphazamiseka kuye wonke umntu. Ukongeza, ngale ndlela, unokulinda imizuzu eyi-10, kwaye enye intambo iya kuza, kwaye ngengozi iya kufumana idatha eyimfuneko kuqala. Izakhiwo zedatha ezimbini azikabinazo izitshixo, kodwa zivumela imisonto ukuba ilinde ngokufanelekileyo. Igama elithi "kabini" lithetha ukuba isakhiwo siqulethe idatha okanye izicelo zedatha, masibabize ngokuba yi-anti-data. Ke ukuba uzama ukubuyisa into kwisikhongozeli esingenanto, isicelo siyakufakwa kwisingxobo endaweni yoko. Ngoku umsonto unokulinda isicelo ngaphandle kokuphazamisa omnye umntu. Ukongeza, ulwakhiwo lwedatha lunika izinto eziphambili ngokubaluleka kwizicelo ukuze xa zifunyenwe, zigqithiselwe kumntu ofanelekileyo. Isiphumo sisixhobo sokungatshixwa esisenengcaciso esesikweni kunye nokusebenza kakuhle ekusebenzeni. 

Алексей: Ulindele ntoni kolu lwakhiwo lwedatha? Ngaba iya kuphucula ukusebenza kuzo zonke iimeko eziqhelekileyo, okanye ifaneleke ngakumbi kwiimeko ezithile? 

UMichael: Ku luncedo ukuba, okokuqala, ufuna isitya ngaphandle kokutshixa, kwaye, okwesibini, kufuneka ulinde kwimeko apho ufuna ukubuyisela idatha kwisitya esingekho kuyo. Ngokolwazi lwam olungcono, isakhelo sethu sibonelela ngokuziphatha okugqibeleleyo xa ezi meko zimbini zihlangatyezwa. Ngoko ke, kule meko ndincoma ukuyisebenzisa. Inzuzo ephambili yezakhiwo zedatha ezingatshixwanga kukuba ziphepha iingxaki zokusebenza. Kwaye ukulinda kubaluleke kakhulu kwii-algorithms ezininzi ukuba idatha idluliselwe ukusuka komnye umsonto ukuya kwenye.

Vitaliy: Makhe ndicacise: nizothetha ngento enye esikolweni nasenkomfeni?

UMichael: Esikolweni Ndiya kuthetha malunga nezakhiwo zedatha ezinemisonto emininzi ngokubanzi, kunye nemigaqo esisiseko echazwe ekuqaleni kwesifundo. Ndicinga ukuba abaphulaphuli bayazi ukuba yintoni imisonto kwaye baqhelene nezitshixo. Ngokusekwe kolu lwazi lusisiseko, ndiza kuthetha malunga nezakhiwo zedatha ezingatshixwanga. Ndiza kunika isishwankathelo sezona ngxaki zibalulekileyo kule ndawo, ukuchukumisa izihloko ezifana nolawulo lwememori. Andiqondi ukuba kuya kubakho nantoni na enzima ngakumbi kunomgca we-MS.

Алексей: Ngaba uceba ukufundisa malunga neziseko zedatha ezimbini ekupheleni kweklasi yakho esikolweni?

UMichael: Ndiza kuzikhankanya, kodwa andizukuchitha ixesha elininzi kuzo. Ingxelo yeHydra iya kunikezelwa kubo. Iza kugubungela iprojekthi ethe ekugqibeleni yayenza yaba yiJava, kunye nokusebenza noJoe Israelevich ukwenza umahluko ombini womgca weLCRQ, nokudala uyilo olukufutshane-jikelele lwezakhiwo zedatha ezimbini.

Алексей: Ke isifundo esikolweni sinokucetyiswa kwabaqalayo, kunye nesifundo kwizakhiwo zedatha kabini kwiHydra - kubantu abasele benamava?

UMichael: Ndilungise ukuba andilunganga, kodwa abaphulaphuli baseHydra baya kwahluka kakhulu, kuquka iingcali ezininzi zeJava, kunye nabantu ngokubanzi abangabandakanyekanga ngokuthe ngqo kwinkqubo enemisonto emininzi. 

Vitaliy: Ewe, yinyani.

Алексей: Noko sinethemba.

UMichael: Kule meko, ndiya kujongana nengxaki efanayo esaqala ngayo olu dliwano-ndlebe: indlela yokwenza ingxelo zombini ngokwaneleyo ngokwaneleyo kwiinkcukacha zobuchwepheshe kwaye ifikeleleke kubo bonke abaphulaphuli.

Vitaliy: Ngaba uya kunika ingxelo ngendlela efanayo owenza ngayo izifundo? Oko kukuthi, thetha nabaphulaphuli kwaye ulungelelanise imeko?

UMichael: Ndiyoyika ukuba ayizukwenzeka ngolo hlobo, kuba ingxelo iza kuba nezilayidi. Izilayidi zibalulekile xa abaphulaphuli beqala ukuthetha iilwimi ezahlukeneyo. Abantu abaninzi baya kukufumanisa kunzima ukundiva ngesiNgesi, ngakumbi ukuba ndithetha ngokukhawuleza. Ndikhethe ezi zihloko kuba Peter Kuznetsov undicele ukuba ndithethe malunga nolwakhiwo lwedatha olungatshixwanga kwiSikolo seSPTDC; kwaye ke ndandifuna ingxelo yenkomfa yeqela labasebenzisi beJava, kwaye ndandifuna ukukhetha into eya kuba nomdla ngakumbi kubadwelisi benkqubo beJava. Eyona ndlela ilula yayikukuthetha ngezo zinto kwithala leencwadi leJava ukuba ndibe nesandla ngendlela enye okanye enye. 

Алексей: Sicinga ukuba abaphulaphuli baseHydra sele beyazi into malunga nenkqubo yokutshixa kwaye mhlawumbi banamava kule ndawo. Kodwa le yintelekelelo kuphela; Ngapha koko, enkosi ngexesha lakho. Ndiqinisekile ukuba udliwano-ndlebe luya kuba nomdla kakhulu kubafundi bethu. Ndibulela kakhulu!

Vitaliy: Enkosi. 

UMichael: Ndingakuvuyela ukudibana nawe eSt. 

Алексей: Nathi sinomzi omhle. Ngaba ukhe waba apha?

UMichael: Hayi, andizange ndiye eRashiya konke konke. Kodwa iSt. Petersburg ibisoloko ikuluhlu lweendawo endingekabi kuzo, kodwa apho ndifuna ukuya khona ngokwenene, ngoko ndavuya kakhulu ngesi simemo. 

Алексей: Kakade ke, siza kuba nenkqubo yohambo lwezithethi. Enkosi kakhulu ngodliwano-ndlebe, kwaye ube nosuku oluhle!

Unokuqhubeka nencoko yakho noMichael kwinkomfa ye-Hydra 2019, eya kubanjwa ngoJulayi 11-12, 2019 eSt. Uya kuza nengxelo "Ulwakhiwo lwedatha ezimbini". Amatikiti angathengwa kwiwebhusayithi esemthethweni.

umthombo: www.habr.com