Benchmark CPU and memory subsystem together by computing polynomials of specified degree over large in-memory arrays of data.

The idea of this test is to get STREAM-like memory access pattern, but with variable arithmetic intensity. Arithmetic intensity is the ratio (CPU time)/(memory time). By varying it one can find how balanced are compute and memory subsystems of the computer.

OpenMP-aware C compiler, GNU make

Adjust Makefile to meet your system configuration. Then run make in the source code directory.

Two types of memory access patterns are available:

The input data is stored in one large array in memory. The calculation results are discarded. To run benchmark, type

   ./poly_ro.x [polynomial degree] [array size] [repeats]

• [int] Polynomial degree: the higher the polynomial degree,
• the more is arithmetic intensity of the benchmark.
• [int] Array size: number of elements of an array
• [int] Repeats: number of times the test will run.

The input data and the results are stored in two arrays in memory. The memory requirements are thus 2x bigger than in read-only test. To run the benchmark, type

   ./poly_rw.x [polynomial degree] [array size] [repeats]

Options are same to read-only test.