|
COMPUTER SYSTEMS AND NETWORKS
Performance analysis of vectorized algorithms
A. E. Maslov, A. A. Zorin Join-stock company Scientific and technical center "Raduga", Moscow, Russia
Abstract:
This paper is devoted to evaluating the efficiency of vectorization for algorithms, which are used in various tasks in order to improve performance. Rational use cases for the SIMD extension are determined. The possibilities of achieving the declared theoretical limit of performance increase are determined. Comparison of use of SSE and AVX extensions for various data types (double, float, complex float and double) is made.
Keywords:
vectorization, SIMD, SSE, AVX, dot product, convolution, correlation.
Citation:
A. E. Maslov, A. A. Zorin, “Performance analysis of vectorized algorithms”, Informatsionnye Tekhnologii i Vychslitel'nye Sistemy, 2022, no. 4, 50–61
Linking options:
https://www.mathnet.ru/eng/itvs785 https://www.mathnet.ru/eng/itvs/y2022/i4/p50
|
Statistics & downloads: |
Abstract page: | 22 | Full-text PDF : | 1 | First page: | 1 |
|