Векторизация процессов параллельной обработки сейсмических сигналов с использованием технологии «parallel studio»
Аннотация
Широкая популярность сплайн-методов в задачах анализа и обработки сейсмических и гео физических сигналов объясняется тем, что они служат универсальным инструментом приближения и по сравнению с другими математическими методами при равных с ними информационных и аппаратных затратах обеспечивают большую точность. С другой стороны, применяемые в таких системах аппаратные средства также должны отвечать требованиям высокой скорости обработки. Для достижения высокой скорости обработки необходимо разработать параллельные алгоритмы с использованием технологии «Parallel Studio» и реализовать их на многоядерных архитектурах процессоров. Статья предлагает программную реализацию сплайн-методов цифровой обработки сигналов с использованием технологии « Parallel Studio» на основе многоядерной архитектуры процессоров. Это позволяет в целом повысить эффективность функционирования систем за счет увеличения скорости обработки данных при установленных показателях точности.
Сейсмик ва геофизик сигналларни тахлил килишда ва канта ишлашда сплайн усуллари кенг оммалашган. Сплайн усуллар ушбу сигналларга универсал ёндашув воситаси булиб хизмат килади хамда бошка математик усуллар билан таккослаганда ахборотларнинг тенглиги ва аппарат харажатларига эга булиш билан купрок аникликка эга. Бошка томондан, бундай тизимларда ишлатиладиган аппарат воситалар хам юкори ишлаш тезлиги талабларига жавоб бериши керак. Юкори тезликка эришиш учун «Parallel Studio» технологияларни ишлатиш асосида параллел алгоритмларни ишлаб чикилади ва уларни куп ядроли протсессорларнинг архитектурасига тадбик килинади. Ушбу макола куп ядроли протсессорлар архитектураси асосида «Parallel Studio» технологиясини ишлатган холатда сигналларни ракамли ишлашнинг сплайн-усулларининг дастурий тадбикини тавсия килади. Бу урнатилган аниклик курсаткичларида маълумотларни кайта ишлаш тезлигини ошириш хисобига тизимнинг ишлаш самарадорлигини кутаришга имкон беради.
The wide popularity of spline methods in the problems of analysis and processing of seismic and geophysical signals is explained by the fact that they serve as a universal tool of approximation and in comparison with other mathematical methods at equal information and hardware costs provide greater accuracy[7-9]. On the other hand, the hardware used in such systems must also meet the requirements of high processing speed. In order to achieve high processing speed, it is necessary to develop parallel algorithms using «Parallel Studio» technology and implement them on multi-core processor architectures. The article offers software implementation of spline methods of digital signal processing using technology «Parallel Studio» based on multi-core architecture of processors. This allows to improve the overall efficiency of the system by increasing processing speed when the established indicators of accuracy. All the functionality of the tools Intel Parallel Studio seems to be a natural extension of Visual Studio and the expansion of its capabilities, which previously lacked to develop good parallel programs. There are two approaches to writing parallel programs. The first is parallelization, partially or completely, of already ready serial applications to speed up the work of some algorithms. In this case, the developer simply analyzes the application and determines the parts of the program that consume the maximum amount of processor resources. The second approach assumes the original design, taking into account the requirements of parallel execution of the load. And if in essence the project can be divided into sections that must be executed simultaneously, then starting its implementation in the form of a program is often a difficult task for beginners. It is especially difficult to write a project so that you do not have to resort to the first approach. Here the Parallel Advisor comes to the rescue. This is a class of tools that bears the methodology of creating parallel programs from scratch using the correct approaches to their implementation, including using parallel libraries. Intel Parallel Composer is not only Intel’s C ++ compiler. It is integrated in Visual Studio with the Integrated Performance Primitives (IPP) and Parallel TBB (Threading Building Blocks), which greatly facilitates the development of parallel code. Normal engineering practice involves testing the program for errors and the developer himself, at least at the level of unit tests. Parallel Inspector helps to detect two classes of errors: errors of multithreading and errors of work with memory, and the analysis for each class is launched separately. The last class of errors is well known to programmers who until recently used various tools to find memory leaks, a stack integrity violation or access to non-existent addresses. The second class of errors is related to the smoothed nature of the programs. The mechanism for detecting memory errors is based on the analysis of absolutely all read / write instructions and their addresses at the binary code level using binary tools.The Performance Profiler is designed to find out how efficiently the application uses the multiprocessor platform, and where are the bottlenecks in the program that prevent it from scaling and increase performance with the growth of compute cores in the system.
Сейсмик ва геофизик сигналларни тахлил килишда ва канта ишлашда сплайн усуллари кенг оммалашган. Сплайн усуллар ушбу сигналларга универсал ёндашув воситаси булиб хизмат килади хамда бошка математик усуллар билан таккослаганда ахборотларнинг тенглиги ва аппарат харажатларига эга булиш билан купрок аникликка эга. Бошка томондан, бундай тизимларда ишлатиладиган аппарат воситалар хам юкори ишлаш тезлиги талабларига жавоб бериши керак. Юкори тезликка эришиш учун «Parallel Studio» технологияларни ишлатиш асосида параллел алгоритмларни ишлаб чикилади ва уларни куп ядроли протсессорларнинг архитектурасига тадбик килинади. Ушбу макола куп ядроли протсессорлар архитектураси асосида «Parallel Studio» технологиясини ишлатган холатда сигналларни ракамли ишлашнинг сплайн-усулларининг дастурий тадбикини тавсия килади. Бу урнатилган аниклик курсаткичларида маълумотларни кайта ишлаш тезлигини ошириш хисобига тизимнинг ишлаш самарадорлигини кутаришга имкон беради.
The wide popularity of spline methods in the problems of analysis and processing of seismic and geophysical signals is explained by the fact that they serve as a universal tool of approximation and in comparison with other mathematical methods at equal information and hardware costs provide greater accuracy[7-9]. On the other hand, the hardware used in such systems must also meet the requirements of high processing speed. In order to achieve high processing speed, it is necessary to develop parallel algorithms using «Parallel Studio» technology and implement them on multi-core processor architectures. The article offers software implementation of spline methods of digital signal processing using technology «Parallel Studio» based on multi-core architecture of processors. This allows to improve the overall efficiency of the system by increasing processing speed when the established indicators of accuracy. All the functionality of the tools Intel Parallel Studio seems to be a natural extension of Visual Studio and the expansion of its capabilities, which previously lacked to develop good parallel programs. There are two approaches to writing parallel programs. The first is parallelization, partially or completely, of already ready serial applications to speed up the work of some algorithms. In this case, the developer simply analyzes the application and determines the parts of the program that consume the maximum amount of processor resources. The second approach assumes the original design, taking into account the requirements of parallel execution of the load. And if in essence the project can be divided into sections that must be executed simultaneously, then starting its implementation in the form of a program is often a difficult task for beginners. It is especially difficult to write a project so that you do not have to resort to the first approach. Here the Parallel Advisor comes to the rescue. This is a class of tools that bears the methodology of creating parallel programs from scratch using the correct approaches to their implementation, including using parallel libraries. Intel Parallel Composer is not only Intel’s C ++ compiler. It is integrated in Visual Studio with the Integrated Performance Primitives (IPP) and Parallel TBB (Threading Building Blocks), which greatly facilitates the development of parallel code. Normal engineering practice involves testing the program for errors and the developer himself, at least at the level of unit tests. Parallel Inspector helps to detect two classes of errors: errors of multithreading and errors of work with memory, and the analysis for each class is launched separately. The last class of errors is well known to programmers who until recently used various tools to find memory leaks, a stack integrity violation or access to non-existent addresses. The second class of errors is related to the smoothed nature of the programs. The mechanism for detecting memory errors is based on the analysis of absolutely all read / write instructions and their addresses at the binary code level using binary tools.The Performance Profiler is designed to find out how efficiently the application uses the multiprocessor platform, and where are the bottlenecks in the program that prevent it from scaling and increase performance with the growth of compute cores in the system.