roboservant/benchmarks

Benchmarks

I've noticed compilation can be a little slow on realistic apps so I set up a test.

components

1. codegen

   given an integer input n, create an api with n endpoints. they should take a reasonably complicated input and return a reasonably complicated output

2. runner

   build dependencies for i from 1 to 100 generate api with i endpoints compile it, but timed print (i,time)

3. plot to see linearity/quadraticity?

possible gotchas

• we usually compile these in different modules rather than together: if we don't get slow compilation in one file, try out multiple files.

notes

• flat profile if you don't have the call to fuzz
• no difference between deriving via Atom vs Compound, so not a Generics thing.