intel oneAPI Math Kernel Library
Fara da Intel® oneAPI Math Kernel Library
Intel® oneAPI Math Kernel Library (oneMKL) yana taimaka muku cimma iyakar aiki tare da ɗakin karatu na lissafin lissafi na ingantattun ayyukan yau da kullun na CPU da GPU. Laburaren yana da mu'amalar C da Fortran don yawancin abubuwan yau da kullun akan CPU, da mu'amalar DPC++ don wasu abubuwan yau da kullun akan CPU da GPU. Kuna iya samun cikakken tallafi don ayyukan lissafi da yawa a cikin mu'amala daban-daban gami da:
Don C da Fortran akan CPU
- Algebra na layi
- Saurin Canje-canje (FFT)
- Vector math
- Kai tsaye da masu warware matsalar rashin ƙarfi
- Masu samar da lambar bazuwar
Don DPC++ akan CPU da GPU (Duba Intel® oneAPI Math Kernel Laburaren—Data Parallel C++ Reference Developer don ƙarin cikakkun bayanai.)
- Algebra na layi
- BLAS
- Ayyukan Sparse BLAS da aka zaɓa
- Ayyukan LAPACK da aka zaɓa
- Saurin Canje-canje (FFT)
- 1D, 2D, da 3D
- Masu samar da lambar bazuwar
- Ayyukan da aka zaɓa
- Ayyukan Lissafi na Vector da aka zaɓa
Kafin Ka Fara
Ziyarci shafin Bayanan Bayanan Saki don Abubuwan da aka sani da kuma mafi yawan bayanai na zamani.
Ziyarci Intel® oneAPI Math Kernel System Bukatun Bukatun Laburare Shafi don buƙatun tsarin.
Ziyarci Farawa da Intel® oneAPI DPC++/C++ Compiler don buƙatun DPC++ Compiler.
Mataki 1: Shigar Intel® oneAPI Math Kernel Library
Zazzage Intel® oneAPI Math Kernel Library daga Intel® oneAPI Base Toolkit.
Don rarrabawar Python, koma zuwa Shigar da Rarraba Intel® don Python* da ɗakunan karatu na Performance Intel® tare da pip da PyPI.
Don rarraba Python, lura da iyakance mai zuwa:
Kunshin haɓaka ɗayaMKL (mkl-devel) don rarraba PIP akan Linux * da macOS * baya samar da alamar haɗin kai na ɗakunan karatu (don ƙarin bayani duba batun PIP GitHub #5919).
A cikin yanayin ɗakin karatu mai ƙarfi ko guda ɗaya mai haɗawa tare da fakitin devel na MKL ɗaya (don ƙarin bayani duba Mai ba da Shawarar Layin Link na ɗayaMKL) dole ne ku canza layin hanyar haɗin yanar gizo tare da cikakken sunaye da nau'ikan ɗakunan karatu naMKL ɗaya.
Koma zuwa Intel® oneAPI Math Kernel Library da kayan aikin pkg-config don bayani game da haɗawa da haɗin kai tare da kayan aikin pkg-config.
layin hanyar haɗin dayaMKL examptare da kayan aikin Base na API guda ɗaya ta hanyar haɗin gwiwar:
- Linux:
icc app.obj -L${MKLROOT}/lib/intel64 -lmkl_intel_lp64-lmkl_intel_thread -lmkl_core -liomp5 -lpthread -lm -ldl - macOS:
icc app.obj -L${MKLROOT}/lib -Wl,-rpath,${MKLROOT}/lib-lmkl_intel_lp64 -lmkl_intel_thread -lmkl_core -liomp5 -lpthread
-lm -ldl
Layin hanyar haɗin gwiwa ɗayaMKL examptare da kunshin haɓaka PIP ta hanyar ɗakunan karatu cikakkun sunaye da sigogin: Linux:
icc app.obj ${MKLROOT}/lib/intel64/libmkl_intel_lp64.so.1 ${MKLROOT}/lib/intel64/libmkl_intel_thread.so.1 ${MKLROOT}/lib/intel64/libmkl_core.so.1 -liomp5 -lpread -lm -ldl - macOS:
icc app.obj -Wl,-rpath,${MKLROOT}/lib${MKLROOT}/lib/intel64/libmkl_intel_lp64.1.dylib $ {MKLROOT}/lib/intel64/libmkl_intel_thread.1.dylib
${MKLROOT}/lib/intel64/libmkl_core.1.dylib -liomp5 -lpthread -lm-ldl
Mataki 2: Zaɓi Aiki ko Na yau da kullun
Zaɓi aiki ko na yau da kullun daga MKL ɗaya wanda ya fi dacewa da matsalar ku. Yi amfani da waɗannan albarkatun:
Hanyoyin haɗi: Abubuwan ciki
Jagoran Haɓaka OneMKL don Linux*
Jagoran Haɓaka OneMKL don Windows*
Jagoran Haɓaka OneMKL don MacOS*
Jagoran Mai Haɓakawa ya ƙunshi cikakkun bayanai kan batutuwa da yawa da suka haɗa da:
- Haɗa da haɗa aikace-aikace
- Gina DLLs na al'ada
- Zare
- Gudanar da ƙwaƙwalwar ajiya
Maganar Mai Haɓakawa ɗayaMKL - C
Harshe oneMKL Magana Mai Haɓaka - Harshen Fortran
Maganar Mai Haɓakawa ɗayaMKL - DPC++ Harshe
- Maganar Mai Haɓakawa (a cikin tsarin C, Fortran, da DPC++) ya ƙunshi cikakkun bayanai na ayyuka da mu'amala ga duk wuraren ɗakin karatu.
Intel® oneAPI Math Kernel Mai Ba da Shawarar Neman Ayyukan Laburare
- Yi amfani da Mai ba da Shawarar Neman Ayyukan LAPACK don bincika ayyukan LAPACK da ke da amfani ga wata matsala. Domin misaliample, idan kun ayyana aiki kamar:
- Nau'in yau da kullun: Lissafi
- Matsala ta lissafi: Ƙarƙashin ƙira
- Nau'in Matrix: Gabaɗaya
- Aiki: Yi QR factorization
Mataki 3: Haɗa lambar ku
Yi amfani da mai ba da Shawarar Layin Haɗin kai na ɗayaMKL don saita umarnin hanyar haɗin gwargwadon fasalin shirin ku.
Wasu iyakoki da ƙarin buƙatu:
Intel® oneAPI Math Kernel Library na DPC++ kawai yana goyan bayan amfani da ɗakin karatu na mkl_intel_ilp64 da jerin abubuwa ko zaren TBB.
Don musaya na DPC++ tare da haɗin kai tsaye akan Linux
icpx -fsycl -fsycl-device-code-split=per_kernel -DMKL_ILP64 ${MKLROOT}/lib/intel64/libmkl_sycl.a -Wl,-rukunin farawa ${MKLROOT}/lib/intel64/libmkl_intel_ilp64.a ${MKLROOT}/lib/intel64/
libmkl_ .a ${MKLROOT}/lib/intel64/libmkl_core.a -Wl,–ƙarshen-group -lsycl -lOpenCL -lpthread -ldl -lm
Don misaliample, gini/mai haɗa main.cpp tare da ilp64 musaya da zaren TBB:
icpx -fsycl -fsycl-device-code-split=per_kernel -DMKL_ILP64 -I${MKLROOT}/hade main.cpp $
{MKLROOT}/lib/intel64/libmkl_sycl.a -Wl,–rukunin farawa ${MKLROOT}/lib/intel64/
libmkl_intel_ilp64.a ${MKLROOT}/lib/intel64/libmkl_tbb_thread.a ${MKLROOT}/lib/intel64/
libmkl_core.a -Wl,–Ƙungiyar-ƙarshen -L${TBROOT}/lib/intel64/gcc4.8 -ltbb -lsycl -lOpenCL -lpthread -lm -ldl
Don mu'amalar DPC++ tare da haɗin kai mai ƙarfi akan Linux
icpx -fsycl -DMKL_ILP64 -L$ {MKLROOT}/lib/intel64 -lmkl_sycl -lmkl_intel_ilp64 -lmkl_ -lmkl_core -lsycl -lOpenCL -lpthread -ldl -lm
Don misaliample, gini/mai haɗa main.cpp tare da ilp64 musaya da zaren TBB:
icpx -fsycl -DMKL_ILP64 -I ${MKLROOT}/hada da main.cpp -L${MKLROOT}/lib/intel64 -lmkl_sycl -lmkl_intel_ilp64 -lmkl_tbb_thread -lmkl_core -lsycl -l-Olplthld -bbl
Don mu'amalar DPC++ tare da haɗin kai tsaye akan Windows
icpx -fsycl -fsycl-device-code-split=per_kernel -DMKL_ILP64 "%MKLROOT%"\lib\intel64\mkl_sycl.lib
mkl_intel_ilp64.lib mkl_ .lib mkl_core_lib sycl.lib BudeCL.lib
Don misaliample, gini/mai haɗa main.cpp tare da ilp64 musaya da zaren TBB:
icpx -fsycl -fsycl-device-code-split=per_kernel -DMKL_ILP64 -I"%MKLROOT%" main.cpp"%MKLROOT%"\lib\intel64 mkl_sycl.lib mkl_intel_ilp64.lib mkl_intel_ilpXNUMX.lib mkllib_core mkllib. .lib BudeCL.lib tbb.lib
Don mu'amalar DPC++ tare da haɗin kai mai ƙarfi akan Windows
icpx -fsycl -DMKL_ILP64 "%MKLROOT%"\lib\intel64\mkl_sycl_dll.lib mkl_intel_ilp64_dll.lib mkl_ _dll.lib mkl_core_dll.lib tbb.lib sycl.lib BudeCL.lib
Don misaliample, gini/mai haɗa main.cpp tare da ilp64 musaya da zaren TBB:
icpx -fsycl -fsycl-device-code-split=per_kernel -DMKL_ILP64 -I"%MKLROOT%\hada da" main.cpp "%MKLROOT%"\lib\intel64\mkl_sycl_dll.lib mkl_intel_ilp64_dllbb tllb mkl_intel_ilpXNUMX_dllbb. bb ba .lib sycl.lib BudeCL.lib
Don Mutuntun C/Fortran tare da Tallafin Buɗewar Buɗewa
Yi amfani da musayawar ɗakin karatu na C/Fotran Intel® oneAPI Math Kernel tare da fasalin saukar da OpenMP zuwa GPU.
Dubi Jagorar Haɓakawa na Buɗewar Buɗewa don ƙarin cikakkun bayanai game da wannan fasalin.
Ƙara waɗannan canje-canje masu zuwa zuwa C/Fortran oneMKL tarawa/layukan haɗin yanar gizo don ba da damar fasalin saukar da OpenMP zuwa GPU:
- Ƙarin zaɓuɓɓukan haɗawa / haɗin kai: -fiopenmp -fopenmp-targets=spir64 -mllvm -vpo-paropt-use-raw-dev-ptr -fsycl
- Ƙarin ɗakin karatu ɗayaMKL: ɗakin karatu ɗayaMKL DPC++
Don misaliample, gini/mai haɗa main.cpp akan Linux tare da ilp64 musaya da zaren OpenMP:
icx -fiopenmp -fopenmp-targets=spir64 -mllvm -vpo-paropt-use-raw-dev-ptr -fsycl -DMKL_ILP64 -m64 -I$(MKLROOT)/ya hada da main.cpp L${MKLROOT}/lib/intel64 - lmkl_sycl -lmkl_intel_ilp64 -lmkl_intel_thread -lmkl_core -liomp5 -lsycl -lOpenCL -lstdc++ -lpthread -lm -ldl
Don duk sauran saiti masu goyan baya, duba Intel® oneAPI Math Kernel Library Advisor Link Line.
Nemo Ƙari
Resource: Bayani
Koyarwa: Amfani da Intel® oneAPI Math Kernel Library don Matrix Multiplication:
- Koyarwa – C Harshe
- Koyarwa - Harshen Fortran
Wannan koyawa tana nuna yadda zaku iya amfani da OneMKL don ninka matrices, auna aikin haɓaka matrix, da sarrafa zaren.
Intel® oneAPI Math Kernel Library (oneMKL) Sakin zaren sarrafa bayanan kula.
Bayanan bayanan saki sun ƙunshi keɓantattun bayanai ga sabuwar sakin dayaMKL gami da sabbin abubuwa da canje-canje. Bayanan saki sun haɗa da hanyoyin haɗin kai zuwa manyan albarkatun bayanan kan layi masu alaƙa da sakin. Hakanan zaka iya samun bayanai akan:
- Menene sabo a cikin sakin
- Abubuwan da ke cikin samfur
- Samun goyon bayan fasaha
- Ma'anar lasisi
Intel® oneAPI Math Kernel Library
Shafin samfurin Intel® oneAPI Math Kernel Library (oneMKL). Duba wannan shafi don tallafi da takaddun kan layi.
Intel® oneAPI Math Kernel Littafin dafa abinci
Laburaren Math Kernel na Intel® oneAPI ya ƙunshi ayyuka da yawa na yau da kullun don taimaka muku warware matsalolin lambobi daban-daban, kamar ninka matrices, warware tsarin daidaitawa, da yin canjin Fourier.
Bayanan kula don Intel® oneAPI Math Kernel Library Vector Statistics
Wannan takaddar ta ƙunshi ƙarewaview, samfurin amfani da sakamakon gwaji na masu samar da lambar bazuwar da aka haɗa a cikin VS.
Intel® oneAPI Math Kernel Library Vector Statistics Random Number Generator Data Performance
Bayanan aiki da aka samu ta amfani da ƙididdigar ƙididdiga (VS) janareta na bazuwar lamba (RNG) gami da CPE (clocks per element) naúrar ma'auni, ainihin ma'aunin bazuwar lambobi (BRNG), samar da janareta na rarrabawa, da tsayin abubuwan da aka samar.
Intel® oneAPI Math Kernel Laburare Ayyukan Lissafi na Vector Mathematics da Daidaitaccen Bayanai
Vector Mathematics (VM) yana ƙididdige ayyuka na farko akan gardamar vector. VM ya haɗa da saitin ingantattun aiwatarwa na manyan ayyuka na lissafi masu tsadar lissafi (ikon, trigonometric, exponential, hyperbolic, da sauransu) waɗanda ke aiki akan vectors.
Bayanan kula don Intel® oneAPI Math Kernel Laburaren Ƙididdiga na Ƙididdiga
Ƙididdiga taƙaice wani yanki ne na yankin Vector Statistics na Intel® oneAPI Math Kernel Library. Ƙididdiga taƙaice tana ba ku ayyuka don binciken ƙididdiga na farko, kuma yana ba da mafita don sarrafa layi ɗaya na manyan bayanai masu girma dabam.
LAPACK Examples
Wannan takaddar tana ba da lambar examples na MKL LAPACK (Linear Algebra PACKage) na yau da kullun.
Sanarwa da Rarrabawa
Software da kayan aikin da aka yi amfani da su a cikin gwaje-gwajen aiki ƙila an inganta su don aiki kawai akan microprocessors na Intel. Gwajin aiki, kamar SYSmark da MobileMark, ana auna su ta amfani da takamaiman tsarin kwamfuta, abubuwan haɗin gwiwa, software, ayyuka da ayyuka. Duk wani canji zuwa ɗayan waɗannan abubuwan na iya sa sakamakon ya bambanta. Ya kamata ku tuntuɓi wasu bayanai da gwaje-gwajen aiki don taimaka muku cikakken kimanta siyayyar da kuke tunani, gami da aikin wancan samfurin lokacin da aka haɗa shi da wasu samfuran. Don ƙarin cikakkun bayanai ziyarci www.intel.com/benchmarks.
Fasahar Intel na iya buƙatar kayan aikin da aka kunna, software ko kunnawa sabis.
Babu samfur ko sashi wanda zai iya zama cikakkiyar amintacce.
Kudin ku da sakamakon ku na iya bambanta.
© Kamfanin Intel. Intel, tambarin Intel, da sauran alamun Intel alamun kasuwanci ne na Kamfanin Intel Corporation ko rassan sa. Ana iya da'awar wasu sunaye da alamun a matsayin mallakin wasu.
Samfura da Bayanin Aiki
Aiki ya bambanta ta amfani, daidaitawa da sauran dalilai. Ƙara koyo a www.Intel.com/PerformanceIndex.
Bita na sanarwa #20201201
Babu lasisi (bayyana ko fayyace, ta estoppel ko akasin haka) ga kowane haƙƙin mallakar fasaha da aka bayar ta wannan takaddar.
Samfuran da aka siffanta na iya ƙunsar lahani na ƙira ko kurakurai da aka sani da errata wanda zai iya sa samfurin ya saba da ƙayyadaddun bayanai da aka buga. Ana samun siffa ta halin yanzu akan buƙata.
Intel yana ƙin duk cikakkun bayanai da garanti mai ma'ana, gami da ba tare da iyakancewa ba, garantin ciniki, dacewa don wata manufa, da rashin cin zarafi, da kowane garanti da ya taso daga hanyar aiki, hanyar mu'amala, ko amfani a kasuwanci.
Takardu / Albarkatu
 |
intel oneAPI Math Kernel Library [pdf] Jagorar mai amfani oneAPI Math Kernel Library, Math Kernel Library, Kernel Library, Library |
