|
Hardware and Software for Supercomputers
Advantages and disadvantages of using the pointer vector method in a vector dataflow processor
N. I. Dikarev, B. M. Shabanov, A. S. Shmelev Joint Supercomputer Center
Abstract:
The article is devoted to the analysis of the Quicksort program execution in the vector dataflow processor (VDP), which uses the pointer vectors method to store arrays. Earlier revealed deficiency of pointer vectors method was compensated by the introduction of split and fuse commands into the processor instruction set. In this article we analyze improved graph of the Quicksort program and the results of its simulation on the original and modernized instruction sets of VDP. We also compare simulation results with performance of Intel Skylake processor core.
Key words and phrases:
vector processor, dataflow architecture, sorting algorithms, parallelism,
vector performance, instruction set.
Received: 12.12.2021 Accepted: 30.12.2021
Citation:
N. I. Dikarev, B. M. Shabanov, A. S. Shmelev, “Advantages and disadvantages of using the pointer vector method in a vector dataflow processor”, Program Systems: Theory and Applications, 12:4 (2021), 65–83
Linking options:
https://www.mathnet.ru/eng/ps389 https://www.mathnet.ru/eng/ps/v12/i4/p65
|
|