Skip to content

Split packed binary field arithmetic benchmark into several

Dmitry Gordon requested to merge dgordon/refactor_benchmarks into main

Compilation of packed_binary_field_arithmetic took about 4.5 minutes on machine. The reason was that macro was generating huge functions and all of them were in the same file. And with LTO enabled most of the time is spent in a single thread.

In this MR:

  1. The benchmark was split into several ones, one per operation. That gives the biggest effect on compilation time.
  2. Macro was changed to generate one function per benchmark case.

With these changes compilation time is reduced to ~1.5m, which is not perfect but I have doubts if we can any better with LTO enabled

Merge request reports