Changing int data size based on compiler used about embench-iot HOT 1 OPEN

Hi @TilakGirijeswar

Thanks for your comment. Embench benchmarks are a set of real programs/kernels and reflect a range of coding styles. There is a deliberate decision not to standardize into a unified programming style - we want the breadth. The only changes made are to ensure all programs are self-validating and that they work correctly on a range of architectures from 8-bit to 64-bit (more on that below).

We thus have some programs that use the native types (int, short etc) and some that use explicit stdint types. The native types give an advantage on smaller machines (where int is 16-bits) while using standard types (uint32_t etc) gives an advantage to 32/64-bit machines. Overall it all balances out.

However...

You have found a bug. While we checked for validity on 16-bit architectures, we failed to consider the issue of multiplying by CPU_MHz. I have marked this as a bug, and assigned to @PaoloS02 to investigate in due course.

Thanks for your contribution.

from embench-iot.