Mamã le Stack la dzi

Nusitae Stack Allocation Gagale Vevie le Egbegbe Kɔmpiutadziɖoɖowo Me

Ɣesiaɣi si wò dɔwɔɖoɖoa wɔ dɔ tso biabia aɖe ŋu, wɔ tɔtrɔ aɖe, alo yɔ dɔwɔwɔ aɖe la, wole nyametsotso aɖe wɔm le ɖoɖoezizi me le megbe: afikae wòle be nyatakaka sia nanɔ le ŋkuɖodzinu me? Ƒe bla nanewoe nye sia la, stack allocation nye ŋkuɖodzinu siwo le kabakaba wu, siwo woate ŋu agblɔ ɖi wu siwo li na ɖoɖowɔlawo dometɔ ɖeka — ke hã wogase egɔme bubui le afisiafi. Le ɣeyiɣi si me wokpɔa dɔwɔɣiwo dzi, gbeɖuɖɔ nuƒoƒoƒulawo, kple alilikpo me xɔtuɖaŋuwo gɔme la, alesi woamae kple ɣeyiɣi si woamae ɖe stack la dzi gɔmesese ate ŋu afia vovototo si le dɔwɔnu si kpɔa ame 10,000 siwo zãnɛ le ɣeyiɣi ɖeka me gbɔ kple esi buckles le 500 te dome xexlẽdzesiwo.

Stack vs. Heap: Asitsatsa Vevitɔ

Ŋkuɖodzi le ɖoɖowɔɖi ƒe nɔnɔme akpa gãtɔ me ma ɖe nuto vevi eve me: stack kple heap. Stack la wɔa dɔ abe nyatakaka mamlɛtɔ si gena ɖe eme, gbãtɔ si dona (LIFO) ene. Ne woyɔ dɔwɔwɔ aɖe la, wotua "frame" yeye ɖe stack si me local variables, return addresses, kple function parameters le la dzi. Ne dɔwɔwɔ ma trɔ gbɔ la, woɖea frame bliboa ɖa enumake. Didi aɖeke meli o, agbalẽdzikpɔkpɔ aɖeke meli o, mama aɖeke meli o — asitɔtrɔ le nufiamɔ̃ ɖeka ŋu ko.

To vovo na ema la, heap la nye ŋkuɖodzinu ƒe ta gã aɖe si me mama kple mama ɖeɖeɖa ateŋu adzɔ le ɖoɖo ɖesiaɖe nu. Trɔtrɔ ɖe nɔnɔmewo ŋu sia xɔa ga geɖe: ele be mamala nalé ŋku ɖe blɔka siwo le ablɔɖe me ŋu, akpɔ mamawo gbɔ, eye le gbegbɔgblɔ geɖe me la, wòaɖo ŋu ɖe gbeɖuɖɔƒomɔ̃ ŋu be wòagaxɔ ŋkuɖodzinu si wometsɔ wɔ dɔe o. Heap ƒe mama le C ɖoɖowɔɖi si bɔ me xɔa ɣeyiɣi si ade zi gbɔ zi 10 va ɖo 20 wu stack mama. Le gbegbɔgblɔ siwo me woƒoa gbeɖuɖɔ nu ƒu abe Java alo C# ene me la, ga si woatsɔ awɔ dɔe ateŋu asɔ gbɔ wu ne wobu nuƒoƒoƒu ƒe ɖiɖiɖemewo ŋu.

Asitsatsa sia gɔmesese menye sukudede dzro aɖe ko o. Ne èle kɔmpiuta dɔwɔɖoɖo si wɔa dɔ akpe geɖewo ŋudɔ le sɛkɛnd ɖeka me — eɖanye mɔ̃ si xɔa adzɔxexe, ɣeyiɣi ŋutɔŋutɔ me numekuku ƒe dashboard, alo CRM si kpɔa kadodo geɖe ƒe nusiwo wotsɔ tso duta vɛ gbɔ o — tiatia si sɔ na mama ƒe mɔnu nyuitɔ na mɔ dzodzoewo kpɔa ŋusẽ ɖe ŋuɖoɖo ƒe ɣeyiɣiwo kple xɔtuɖaŋu ƒe gazazãwo dzi tẽ.

Alesi Stack Allocation wɔa dɔ ŋutɔŋutɔ

Le hardware ƒe ɖoɖo nu la, processor architecture akpa gãtɔ tsɔa register (stack pointer) ɖoa anyi be woatsɔ akpɔ stack la ƒe tame si li fifia. Memory mama ɖe stack la dzi le bɔbɔe abe pointer sia dzi ɖeɖe kpɔtɔ to byte xexlẽme si hiã ene. Deallocation nye nusi to vovo: dzi pointer la ɖe edzi. Metadata ƒe ta aɖeke meli o, xexlẽdzesi siwo le ablɔɖe me aɖeke meli o, block siwo te ɖe wo nɔewo ŋu ƒe ƒuƒoƒo aɖeke meli o. Esia tae woɖɔa stack allocation zi geɖe be O(1) constant-time dɔwɔwɔ le esi kple overhead si womate ŋu atsɔ ɖeke le eme o.

Bu dɔwɔwɔ si bua akɔnta le xexlẽme bliboa ŋu na adzɔxegbalẽvi ƒe fli me nu. Ate ŋu aɖe gbeƒã nutoa me tɔtrɔ ʋɛ aɖewo: agbɔsɔsɔme ƒe xexlẽdzesi blibo, unit price float, tax rate float, kple result float. Wotua asixɔxɔ eneawo katã ɖe stack la dzi ne woge ɖe dɔwɔwɔa me eye wogaxɔa wo le wo ɖokui si ne edo go. Agbenɔƒe bliboa nye nusi woɖo ɖi eye wòbia be ɖoɖowɔla alo gbeɖuɖɔ ƒola nade nu eme zero.

ƒe nyawo

Gbese vevi: Stack mama menye kabakaba ko o — woateŋu agblɔe ɖi. Le ɖoɖo siwo me dɔwɔwɔ le vevie me la, zi geɖe la, nusiwo woate ŋu agblɔ ɖi le vevie wu duƒuƒu xoxo. Dɔwɔwɔ si wu enu ɣesiaɣi le microsecond 2 me la xɔ asi wu esi xɔa microsecond 1 le mamã dedie nu gake ɣeaɖewoɣi la, edzina ɖe edzi va ɖoa microsecond 50 le gbeɖuɖɔwo nuƒoƒoƒu ƒe ɖiɖiɖeme ta.

ƒe nyawo

Ɣekaɣie Woalɔ̃ ɖe Stack Allocation dzi

Menye nyatakaka ɖesiaɖee nye etɔ le stack la dzi o. Stack memory seɖoƒe li na (zi geɖe la, ele MB 1 kple MB 8 dome le ka ɖeka me, le dɔwɔɖoɖoa nu), eye nyatakaka siwo woɖo ɖe stack la dzi mate ŋu anɔ agbe didi wu dɔwɔwɔ si wɔe o. Ke hã, nɔnɔme siwo me kɔ li siwo me stack allocation nye tiatia si de ŋgɔ wu.

ƒe nyawo
• Nutoa me tɔtrɔ siwo nɔa agbe kpuie: Xexlẽdzesiwo, nuƒoƒoƒulawo, ɣeyiɣi kpui aɖe ƒe buffer siwo le kilobyte ʋee aɖewo te, kple loop indices nye dzɔdzɔme nusiwo sɔ na stack la. Wowɔa wo, wozãa wo, eye wotsɔa wo ƒua gbe le dɔwɔwɔ ƒe akpa ɖeka me.
• Nyatakaka ƒe ɖoɖo siwo ƒe lolome woɖo ɖi: Woateŋu atsɔ ƒuƒoƒo siwo ƒe lolome nya nuƒoƒoƒu-ɣeyiɣi, xɔtuɖoɖo suewo, kple asixɔxɔ ƒomeviwo ade ƒuƒoƒoa dzi afɔku manɔmee be woayɔ fũ. 256-byte buffer na ŋkeke ƒe ka ƒe ɖoɖowɔwɔ nye ame si sɔ nyuie.
• Ememe ʋuƒo siwo le vevie na dɔwɔwɔ: Ne woyɔ dɔwɔwɔ zi miliɔn geɖe le sɛkɛnd ɖeka me — abe asixɔxɔ ƒe akɔntabubu mɔ̃ si gbugbɔna le adzɔnuwo ƒe agbalẽwo dzi ene — ne woɖe heap mamawo ɖa le loop ƒe ŋutilã me ateŋu ana ŋgɔyiyi ƒe ŋgɔyiyi zi gbɔ zi 3 va ɖo zi gbɔ zi 10.
• Ɣeyiɣi ŋutɔŋutɔ alo ɣeyiɣi si woatsɔ awɔ dɔe: Fexexe ƒe dɔwɔwɔ, dashboard ƒe tɔtrɔ yeyewo, kple nyatakakawo ɖoɖo ɖa katã ɖea vi ne woƒo asa na gbeɖuɖɔwo nuƒoƒoƒu ƒe ɖiɖiɖeme siwo menye nusi woɖo ɖi o.
• Recursive algorithms with bounded depth: Ne àteŋu aka ɖe edzi be recursion depth nɔa dedie le seɖoƒe siwo le dedie me la, stack-allocated frames naa recursive functions kabakaba eye wònɔa bɔbɔe.

ƒe nyawo

Le nuwɔna me la, egbegbe nuƒoƒoƒulawo bi ɖe stack zazã nyuie wu me wòɖe dzesi ŋutɔ. Mɔnuwo abe sisi ƒe numekuku le Go kple Java ƒe JIT nuƒoƒoƒula ateŋu atsɔ heap allocations ayi stack la me le eɖokui si ne nuƒoƒoƒula ɖo kpe edzi be nyatakakaawo mesi le dɔwɔwɔ ƒe kekeme nu o. Ne èse asitɔtrɔ siawo gɔme la, àte ŋu aŋlɔ kɔpi si le dzadzɛ wu esime nègale viɖe kpɔm tso stack ƒe dɔwɔwɔ me.

Mɔɖɔ Siwo Bɔ Kple Alesi Nàƒo Asa Na Wo

Gbegblẽ si xɔ ŋkɔ wu si do ƒome kple stack enye stack overflow — nyatakaka geɖe mama wu esi stack ateŋu alé, zi geɖe to unbounded recursion alo local arrays siwo lolo akpa dzi. Le nuwɔwɔ ƒe nɔnɔme me la, zi geɖe la, stack overflow gblẽa ka la alo dɔwɔwɔ bliboa me eye mɔ nyui aɖeke meli si dzi woato agbugbɔe axɔ o. Esia tae frameworks kple operating systems de seɖoƒe na stack ƒe lolome.

Mɔ̃ bubu si menya kpɔna dzea sii bɔbɔe o enye be woatrɔ asi le nufiamɔ̃wo alo nufiamewo ŋu na nyatakaka siwo woɖo ɖe stack-allocated. Esi wònye be wogaxɔa stack memory le ɣeyiɣi si me dɔwɔwɔ aɖe trɔ gbɔ ta la, nufiame ɖesiaɖe si fiaa ŋkuɖodzinu ma va zua nufiame si le tsatsam. Le C kple C++ me la, esia hea nuwɔna si womeɖe o vɛ si ate ŋu adze abe ɖe wòwɔa dɔ le dodokpɔ me ene gake wòdoa kpo nu vevie le ewɔwɔ me. Rust ƒe borrow checker léa vodada ƒe hatsotso sia le nuƒoƒoƒuɣi, si nye susu ɖeka si ta gbea xɔ traction na systems programming.

Nya etɔ̃lia ku ɖe ka ƒe dedienɔnɔ ŋu. Thread ɖesiaɖe xɔa eya ŋutɔ ƒe stack, si fia be stack-allocated data nye thread-local le dzɔdzɔme nu. Esia nye viɖe ŋutɔŋutɔ le go geɖe me — gaxɔ aɖeke mehiã be woakpɔ teƒea ƒe tɔtrɔwo o. Gake ɣeaɖewoɣi la, dɔwɔlawo wɔa vodada be wodzea agbagba be yewoama nyatakaka siwo woɖo ɖe stack me le kaawo dome, si hea duɖimekeke ƒe nɔnɔmewo alo use-after-free bugs vɛ. Ne ehiã be woama nyatakakawo le kawo me alo anɔ anyi wu dɔwɔwɔ ƒe yɔyɔ la, heap lae nye tiatia si sɔ.

Stack Allocation le Gbegbɔgblɔwo Kple Dɔwɔɖoɖowo Me

Programming language vovovowo kpɔa stack allocation gbɔ kple transparency ƒe dzidzenu vovovowo. Le C kple C++ me la, ɖoɖowɔla la kpɔa ŋusẽ si dze ƒã: teƒea ƒe tɔtrɔwo yia stack la dzi, eye malloc alo new tsɔa nyatakakawo dea heap la dzi. Le Go me la, nuƒoƒoƒula la wɔa sisi ƒe numekuku be yeatso nya me le eɖokui si, eye goroutines dzea egɔme kple 2 KB ƒe ƒuƒoƒo suesuesue siwo tsina le ŋusẽ me — egbɔkpɔnu dzeani aɖe si da sɔ le dedienɔnɔ kple dɔwɔwɔ me. PHP, gbegbɔgblɔ ƒe ŋusẽdoame ƒe ɖoɖowo abe Laravel ene, maa asixɔxɔ akpa gãtɔ to eƒe ememe Zend Engine ŋkuɖodzinudzikpɔla dzi, gake gɔmeɖose siwo le ete gɔmesese kpena ɖe dɔwɔlawo ŋu be woaŋlɔ kɔpi si wɔa dɔ nyuie wu le dɔwɔwɔ ƒe ɖoɖo gɔ̃ hã me.

| Ne modules 207 ma dɔwɔwɔ ƒe ɣeyiɣi ɖeka la, ŋkuɖodzinu ƒe mama ɖe biabia ɖesiaɖe dzi dzi ɖeɖe kpɔtɔ 15% gɔ̃ hã ateŋu aɖe egɔme be gɔmesese le server ƒe gazazãwo dzi ɖeɖe kpɔtɔ kple ŋgɔyiyi siwo woateŋu adzidze le ŋuɖoɖo ɣeyiɣiwo me na nuwuwu zãla siwo le woƒe dɔwɔƒewo dzi kpɔm le mɔ̃a dzi.

JavaScript kple TypeScript, siwo naa ŋusẽ egbegbe ŋgɔgbekpa akpa gãtɔ kple Node.js megbenyawo, ɖoa ŋu ɖe V8 mɔ̃a ƒe gbeɖuɖɔ nudzɔla ŋu bliboe hena ŋkuɖodzinu dzikpɔkpɔ. Dɔwɔlawo mateŋu ama nu tẽ ɖe stack la dzi o, gake V8 ƒe optimizing compiler (TurboFan) wɔa stack allocation le ememe na asixɔxɔ siwo wòateŋu aɖo kpe edzi be wonɔa agbe kpuie. Dɔwɔwɔ sue dzadzɛwo ŋɔŋlɔ kple teƒea ƒe tɔtrɔwo naa mɔnukpɔkpɔ nyuitɔ mɔ̃a be wòawɔ ŋgɔyiyi siawo ŋudɔ.

Aɖaŋu nyui siwo woatsɔ aɖe heap ƒe nyaƒoɖeamenu dzi akpɔtɔ

|

ƒe nyawo
1. Ti asixɔxɔ ƒomeviwo wu nufiame ƒomeviwo afisi gbea doa alɔ wo le. Le C# me la, struct zazã ɖe class teƒe na nu sue siwo wowɔ enuenu nana wonɔa ƒuƒoƒoa dzi. Le Go me la, struct suewo tsɔtsɔ yi to asixɔxɔ dzi tsɔ wu to pointer dzi kpɔa ŋusẽ ma ke gbɔ.
2. Ƒo asa na mama ɖe loop sesẽwo me. Do ŋgɔ na buffers eye nàgazã wo le iterations katã me. Ne èhiã ɣeyiɣi kpui aɖe ƒe slice alo array le loop si zɔna zi 100,000 me la, mae zi ɖeka do ŋgɔ na loop la eye nàgbugbɔe aɖo ɖe iteration ɖesiaɖe dzi.
3. Zã nusiwo ƒe ƒuƒoƒo na nusiwo wowɔ kple esiwo wogblẽ enuenu. Nyatakakadzraɖoƒe ƒe kadodo ƒe ƒuƒoƒo nye kpɔɖeŋu si wonya nyuie, gake kpɔɖeŋua wɔa dɔ sɔsɔe na HTTP biabia ƒe nuwo, serialization buffers, kple computation context structs.
4. Profile hafi awɔe wòanyo wu. Dɔwɔnuwo abe Go ƒe pprof, Java ƒe async-profiler, alo PHP ƒe Blackfire ateŋu afia afisi tututu mamawo dzɔna le. Dɔwɔwɔ nyuie wu nyatakakawo ƒe nɔnɔmetata wɔwɔ manɔmee dea afɔku me be woazã agbagbadzedze ɖe mɔ fafɛ siwo ƒã hafi wowɔa dɔ dzi.
5. Leverage arena allocators for batch operations. Ne èle nuŋlɔɖiwo ƒe hatsotso aɖe ŋudɔ wɔm — abe adzɔxegbalẽvi 500 wɔwɔ alo kadodo 10,000 tsɔtsɔ va duta ene — arena allocator léa ŋkuɖodzinu gã ɖeka eye wòmae ɖe akpa vovovowo me kple duƒuƒu si le abe stack ene, emegbe eɖea asi le block bliboa ŋu zi ɖeka ne batch la wu enu.

ƒe nyawo

Menye nukpɔsusuwo koe mɔnu siawo nye o. Ne SaaS mɔ̃wo kpɔa xexeame ŋutɔŋutɔ ƒe dɔwɔwɔ ƒe agbawo gbɔ — asitsaha sue aɖe tɔ si wɔa gaxɔgbalẽviwo ɣleti sia ɣleti, HR dzikpɔla si le fetu dzi kpɔm na dɔwɔla 200, asitsatsa ƒe ƒuƒoƒo si le ŋku lém ɖe dɔdzikpɔha ƒe dɔwɔwɔ ŋu le mɔnu vovovowo dzi — ŋkuɖodzinu dzikpɔkpɔ nyuie ƒe ŋusẽkpɔɖeamedzi si woƒo ƒu nyea nuteƒekpɔkpɔ si wɔa dɔ kabakaba wu, si wɔa dɔ nyuie wu si zãlawo sena le wo ɖokui me ne womebu nusi le edzi yim le ete ŋu gbeɖe o gɔ̃ hã.

Kɔmpiutadziɖoɖo siwo Bu Dɔwɔwɔ Ŋu Tutu le Dzidzenu

Stack allocation nye dɔwɔwɔ ƒe puzzle si lolo wu sã ƒe akpa ɖeka, gake enye gɔmeɖoanyi. Alesi ŋkuɖodzinu wɔa dɔe le ɖoƒe si bɔbɔ wu gɔmesese naa susu ƒe kpɔɖeŋu siwo wohiã be woawɔ nyametsotso nyuiwo le ƒuƒoƒoa ƒe akpa ɖesiaɖe me gɔmesese na mɔ̃ɖaŋudɔwɔlawo — tso nyatakakawo ƒe wɔwɔme tiatia kple APIwo ƒe ɖoɖowɔwɔ dzi va ɖo xɔtuɖoɖowo ƒe ɖoɖowɔwɔ kple nunɔamesiwo ƒe seɖoƒewo ɖoɖo na nugoe me dɔwɔnawo dzi.

Le asitsaha siwo ɖoa ŋu ɖe mɔ̃wo abe Mewayz ene ŋu be woawɔ woƒe gbesiagbedɔwo la, mɔ̃ɖaŋununya ƒe nyametsotso siawo ƒe fetu dzena ŋutɔŋutɔ: axawo ƒe agba kabakaba, kadodo siwo le bɔbɔe wu, kple kakaɖedzi be ɖoɖoa magblẽ le agba si sɔ gbɔ wu te o. Ne ehiã be agbalẽŋɔŋlɔ ƒe mɔ̃ aɖe nalé ŋku ɖe woƒe anyinɔnɔ ŋu le ɣletigbalẽ gbogbo aɖewo me le ɣeyiɣi ŋutɔŋutɔ me, alo numekuku ƒe dashboard ƒo nyatakakawo nu ƒu ɖe asitsaha geɖewo me la, ŋkuɖodzinu si le ete la le vevie wu alesi zãla akpa gãtɔ akpɔe.

Kɔmpiutadziɖoɖo nyuitɔ kekeake se le eɖokui me be agbagbadzedze aɖeke mele eme be yeazã o elabena ewɔlawo fifia fifia tsitotsito siwo gakpɔtɔ nye esiwo womekpɔna o. Stack allocation — si le kabakaba, si woɖo ɖi, eye wònya kpɔ le eƒe bɔbɔe me — nye nyatakaka mawo dometɔ ɖeka si gɔme wòle be woase egɔme nyuie, eɖanye be èle wò ɖoɖowɔɖi gbãtɔ ŋlɔm alo èle mɔ̃ aɖe si subɔa asitsaha akpe geɖe le xexeame katã la tum o.

Nyabiase Siwo Wobiana Enuenu

Nukae nye stack allocation eye nukatae wòle vevie?

Stack allocation nye ŋkuɖodzinu dzikpɔkpɔ ƒe mɔnu si me wodzraa nyatakakawo ɖo le mamlɛtɔ-geɖe, gbãtɔ-do go ƒe ɖoɖo si dzi ɖoɖowɔɖia ƒe dɔwɔwɔ ƒe sisi kpɔa edzi le eɖokui si. Ele vevie elabena stack-allocated memory le kabakaba ŋutɔ wu heap allocation — gbeɖuɖɔ nudzɔla aɖeke mele ta dzi o, fragmentation aɖeke meli o, eye deallocation nye enumake ne function trɔ gbɔ. Le dɔwɔwɔ-vevietɔ dɔwɔwɔwo gome la, stack mama gɔmesese ateŋu aɖe latency dzi akpɔtɔ ŋutɔ eye wòana throughput nanyo ɖe edzi.

Ɣekaɣie wòle be mazã stack allocation ɖe heap allocation dzi?

Zã stack allocation na tɔtrɔ sue siwo nɔa anyi kpuie siwo ƒe lolome wonya le nuƒoƒoƒuɣi — abe teƒea ƒe xexlẽdzesi blibowo, structs, kple fixed-size arrays ene. Heap mama sɔ nyuie wu na nyatakaka ƒe ɖoɖo gãwo, nuƒoƒoƒu siwo ƒe lolome le tɔtrɔm, alo nusiwo wòle be woanɔ agbe didi wu dɔwɔwɔ si wɔ wo. Se vevitɔ: ne nyatakakaawo ƒe agbenɔƒe sɔ kple dɔwɔwɔ ƒe kekeme eye eƒe lolome nye nusi woateŋu agblɔ ɖi la, ɣesiaɣi kloe la, stack la nyea tiatia si le kabakaba wu.

Ðe woateŋu axe mɔ ɖe stack overflow vodadawo nu le dɔwɔwɔ ƒe dɔwɔwɔwo mea?

Ẽ, woate ŋu axe mɔ ɖe stack overflow vodadawo nu kple mɔ̃ɖaŋununya ƒe nuwɔna siwo ŋu woɖɔ ɖo. Ƒo asa na gbugbɔgadzɔ goglo alo seɖoƒemanɔsitɔ, ɖo seɖoƒe na teƒea ƒe tɔtrɔ gãwo mama, eye nàzã iterative algorithms le afisi wòanya wɔ le. Gbegbɔgblɔ kple dɔwɔɖoɖo akpa gãtɔ ɖea mɔ na wò be nàɖo stack ƒe lolome ƒe seɖoƒewo. Dɔwɔnu siwo léa ŋku ɖe nu ŋu kple mɔ̃dzikpɔnuwo abe Mewayz, si nye 207-module business OS si dzea egɔme tso $19/mo, ateŋu akpeɖe ƒuƒoƒowo ŋu woalé ŋku ɖe dɔwɔwɔ ƒe lãmesẽ ŋu eye woalé dɔwɔwɔ ƒe megbedede kaba.

Ðe egbegbe gbegbɔgblɔwo gakpɔtɔ kpɔa viɖe tso stack allocation mea?

Kakaɖedzitɔe. Gbegbɔgblɔ siwo ƒe dɔwɔwɔ ƒe ɣeyiɣiwo dzi kpɔkpɔ le gɔ̃ hã — abe Go, Rust, C#, kple Java — zãa escape analysis tsɔ nyaa nenye be woateŋu atsɔ tɔtrɔwo aɖo stack-allocated ɖe heap-allocated teƒe. Rust zi stack-first allocation dzi to eƒe aƒetɔ ƒe kpɔɖeŋu dzi, eye Go ƒe nuƒoƒoƒula la tsɔa dziku wɔa nu nyuitɔ kekeake nɛ. Mɔ̃ɖaŋu siawo gɔmesese kpena ɖe dɔwɔlawo ŋu be woaŋlɔ kɔpi siwo nuƒoƒoƒulawo ateŋu awɔ nyuie wu, si wɔnɛ be ŋkuɖodzinu zazã dzi ɖena kpɔtɔna eye dɔwɔwɔ ƒe ɣeyiɣiwo hã nɔa kabakaba wu.