Matematik asoslar va raqamli modellashtirish usullari

Kompyuter tadqiqotlari
va modellashtirish 2010 T. 2 № 3 P. 231-272
MATEMATIK ASOSLAR VA RAQAMLI MODELLASHTIRISH USULLARI
UDK: 004.421, 519.712
Algoritmlarni parallelizatsiyalashga kirish
va dasturlar
Va Karpov
Moskva fizika va texnologiya instituti,
Rossiya, 141700, Dolgoprudniy, Per. Institut, 9
E-mail: [e-mail]
2010 yil 10 sentyabr kuni qabul qilingan.
Parallel hisoblash
tizimlari va ketma-ket dasturlash texnologiyalari uchun dasturiy texnologiyadagi farqlar tasvirlangan, texnologiyadagi yangi
bosqichlarning paydo bo'lishi: algoritmlarni dekompozitsiya qilish, ijrochilarga ishlarni tayinlash
, mantiqiy ijrochilarni jismoniy shaxslarga yo'naltirish va namoyish qilish. Keyin
algoritmlarning ishlashini baholash masalalari qisqacha ko'rib chiqiladi. Algoritmlarni
va parallel ravishda bajarilishi mumkin bo'lgan ish dasturlarini dekompozitsiya qilish masalalari muhokama qilinadi.
Kalit so'zlar: algoritmlar va dasturlarni parallellashtirish, dekompozitsiya,
asimptotik tahlil qilish, hisoblash, qatlamlik parallel shakllar, Bernstein shartlari, haqiqiy qaramlik,
qaramlik chiqish ma'lumotlari, qarama-qarshilik, tsikllarni parallellashtirish
Introduction to the parallelization of algorithms and programs
V. E. Karpov
Moscow Institute of Physics and Technology, 9 Institutskii per, Dolgoprudny, 141700, Russia
Abstract. —
Difference of software development for parallel computing technology from sequential
programming is dicussed. Arguements for introduction of new phases into technology of software engineering
are given. These phases are: decomposition of algorithms, assignment of jobs to performers, conducting and
mapping of logical to physical performers. Issues of performance evaluation of algorithms are briefly discussed.
Decomposition of algorithms and programs into parts that can be executed in parallel is dicussed.
Keywords: parallelization of algorithms and programs, decomposition, asymptotic analysis, graph, multilevel
structure, Bernstein conditions, true dependence, dependence on the output data, antidependence, parallelizing
cycles
Citation: Computer Research and Modeling, 2010, vol. 2, no. 3, pp. 231–272 (Russian).
o'tish: saytda harakatlanish, qidiruv
232
Va Karpov