Anyị na-aga n'ihu na-ekwu maka ngwaọrụ maka ịlele arụmọrụ CPU na igwe Linux. Taa na ihe: temci, uarch-bench, likwid, perf-tools na lvm-mca.
Ihe nrịbama ndị ọzọ:
Ото - - Unsplash
Nke a bụ ngwá ọrụ na-atụle oge igbu nke mmemme abụọ. N'ezie, ọ na-enye gị ohere iji tụnyere oge igbu nke ngwa abụọ. Onye dere akụrụngwa ahụ bụ nwa akwụkwọ si Jamanị, Johannes Bechberger, onye mepụtara ya dịka akụkụ nke nzere bachelọ ya na 2016. Ngwa ngwa taa ikike n'okpuru GNU General Public License.
Johannes chọrọ ịmepụta ngwá ọrụ ga-enye ya ohere ịlele arụmọrụ nke usoro mgbakọ na mwepụ na gburugburu ebe a na-achịkwa. Ya mere, otu n'ime ihe ndị bụ isi nke temci bụ ike ịmepụta ebe nyocha. Ọmụmaatụ, : gbanwee ntọala njikwa ugboro CPU, gbanyụọ na cache L1 na L2, gbanyụọ turbo mode na Intel processors, wdg. Maka benchmarking temci na-eji ngwaọrụ. , и .
Nke a bụ ihe ngwa ngwa dị na nke mbụ:
# compare the run times of two programs, running them each 20 times
> temci short exec "sleep 0.1" "sleep 0.2" --runs 20
Benchmark 20 times [####################################] 100%
Report for single runs
sleep 0.1 ( 20 single benchmarks)
avg_mem_usage mean = 0.000, deviation = 0.0
avg_res_set mean = 0.000, deviation = 0.0
etime mean = 100.00000m, deviation = 0.00000%
max_res_set mean = 2.1800k, deviation = 3.86455%
stime mean = 0.000, deviation = 0.0
utime mean = 0.000, deviation = 0.0
sleep 0.2 ( 20 single benchmarks)
avg_mem_usage mean = 0.000, deviation = 0.0
avg_res_set mean = 0.000, deviation = 0.0
etime mean = 200.00000m, deviation = 0.00000%
max_res_set mean = 2.1968k, deviation = 3.82530%
stime mean = 0.000, deviation = 0.0
utime mean = 0.000, deviation = 0.0
Dabere na nsonaazụ benchmarking, sistemụ na-emepụta na eserese, tebụl na eserese, nke na-ekewa temci na ngwọta ndị yiri ya.
N'ime adịghị ike nke temci, "ntorobịa" ya pụtara ìhè. N'ihi nke a ọ nhazi ngwaike na ngwanrọ. Dịka ọmụmaatụ, ọ siri ike ịgba ọsọ na macOS, yana ụfọdụ atụmatụ adịghị na sistemụ dabere na ARM. N'ọdịnihu, ọnọdụ ahụ nwere ike ịgbanwe, ebe ọ bụ na onye edemede na-arụsi ọrụ ike na-emepụta ọrụ ahụ, na ọnụ ọgụgụ kpakpando na GitHub ji nwayọọ nwayọọ na-abawanye - ọ bụghị ogologo oge gara aga temci ọbụna na Akụkọ Hacker.
Ngwa maka inyocha arụmọrụ nke ọrụ CPU dị ala, nke injinia Travis Downs mepụtara (). N'oge na-adịbeghị anya ọ na-ede blọgụ na GitHub Pages, nke na-ekwu maka ngwaọrụ benchmarking na ihe ndị ọzọ metụtara ya. N'ozuzu, uarch-bench ka na-amalite inweta ewu ewu, mana ọ bụlarị nke a na-ahụkarị bi na Hacker News na thematic threads dị ka ihe aga-na maka benchmarking.
Uarch-bench na-enye gị ohere inyocha arụmọrụ ebe nchekwa, ọsọ nbudata data na ọrụ nhicha . Enwere ike ịchọta nsonaazụ benchmarking nke mmemme ahụ mepụtara na ala nke ibe.
Ọ dị mma ịmara na uarch-bench, dị ka temci, Intel Turbo Boost ọrụ (ọ na-akpaghị aka na-abawanye ọsọ elekere processor n'okpuru ibu) ka nsonaazụ ule na-agbanwe agbanwe.
Maka ugbu a, ọrụ ahụ dị na mmalite nke mmepe, ya mere uarch-bench enweghị akwụkwọ zuru ezu, na ọrụ ya nwere ike ịnwe ahụhụ - dịka ọmụmaatụ, na mmalite na Ryzen. Ọzọkwa, naanị akara akara maka ụlọ ọrụ x86 ka akwadoro. Onye edemede ahụ na-ekwe nkwa itinyekwu ọrụ n'ọdịnihu ma na-akpọ gị òkù ka ị sonyere mmepe ahụ.
Nke a bụ otu ngwaọrụ maka iji Intel, AMD na ARMv8 nyochaa arụmọrụ nke igwe Linux. Emepụtara ya n'okpuru nkwado nke Federal Federal Ministry of Education and Research na 2017 wee wepụta ya na isi mmalite.
N'ime ngwa ọrụ likwid, anyị nwere ike ime ka likwid-powermeter pụta ìhè, nke na-egosiputa ozi sitere na ndekọ RAPL banyere ike nke sistemụ na-eri, yana likvid-setFrequencies, nke na-enye gị ohere ijikwa oge nhazi. Ị nwere ike ịhụ ndepụta zuru ezu .
Ndị injinia na-etinye aka na nyocha HPC na-eji ngwá ọrụ a. Dịka ọmụmaatụ, na likwid otu ndị ọkachamara sitere na Regional Computing Center nke Mahadum Erlangen-Nuremberg (RRZE) na Germany. Ọ na-ekerekwa òkè na-arụsi ọrụ ike n'imepụta ngwá ọrụ a.
Ото - - Unsplash
Ngwá ọrụ a maka nyochaa arụmọrụ nke sava Linux Brendan Gregg. Ọ bụ otu n'ime ndị mmepe - usoro nchọta siri ike maka ngwa debugging ozugbo.
Ngwa perf dabere na perf_events na ftrace kernel subsystems. Ọrụ ha na-enye gị ohere iji nyochaa latency I / O (iosnoop), soro usoro oku arụmụka (unccount, funcslower, funcgraph and functrace) na ịnakọta ọnụ ọgụgụ na "hits" na cache faịlụ (cachestat). N'okwu nke ikpeazụ, iwu ahụ dị ka nke a:
# ./cachestat -t
Counting cache functions... Output every 1 seconds.
TIME HITS MISSES DIRTIES RATIO BUFFERS_MB CACHE_MB
08:28:57 415 0 0 100.0% 1 191
08:28:58 411 0 0 100.0% 1 191
08:28:59 362 97 0 78.9% 0 8
08:29:00 411 0 0 100.0% 0 9
Otu nnukwu obodo etolitela gburugburu ngwa ahụ (). Ma enwere ụlọ ọrụ na-arụsi ọrụ ike na-eji ngwa perf, dịka ọmụmaatụ . Mana a na-emepewanye ngwa ahụ ma gbanwee ya (ọ bụ ezie na ewepụtala mmelite n'oge na-adịbeghị anya). Ya mere, njehie nwere ike ime na ọrụ ya - onye edemede dere na mgbe ụfọdụ perf-tools na-akpata ụjọ kernel.
Ngwa na-ebu amụma ole koodu igwe akụrụngwa ga-achọ na CPU dị iche iche. Nwanyị Ntuziaka kwa okirikiri () na ibu dị na ngwaike nke otu ngwa na-ewepụta.
ewepụtara lvm-mca na 2018 dịka akụkụ nke ọrụ a , nke na-emepụta usoro zuru ụwa ọnụ maka nyocha, mgbanwe na njikarịcha nke mmemme. A maara na ndị dere llvm-mca sitere n'ike mmụọ nsọ maka nyocha arụmọrụ ngwanrọ ma chọọ imepụta ihe ọzọ. Dịkwa ka ndị ọrụ si kwuo, mmepụta ngwá ọrụ ahụ (nhazi ha na ọnụ ọgụgụ ha) yiri IACA n'ezie - ọmụmaatụ. . Agbanyeghị, llvm-mca na-anabata naanị , yabụ ị ga-abụrịrị na ị ga-eji ndị ntụgharị na-arụ ọrụ na ya.
Ihe anyị na-ede na blọọgụ na netwọk mmekọrịta anyị:
Anyị na 1cloud.ru na-enye ọrụ efu "" Ị nwere ike ijikwa ndekọ DNS na otu akaụntụ nkeonwe.
isi: www.habr.com