search

nordlow / compiler-benchmark

Benchmarks compilation speeds of different combinations of languages and compilers.
MIT License
138 stars 18 forks source link

Availability of data #24

Open Derek-Jones opened 2 years ago

Derek-Jones commented 2 years ago

I'm interested in the data, or rather the model of performance that might best fit the data. Here is one I did earlier: https://shape-of-code.com/2019/01/29/modeling-visual-studio-c-compile-times/

Is detailed data available to be shared?

How would you expect performance to depend on function count and depth?

For small values I would expect it to be linear.

run-count ought to be at least 10, to handle noise effects.

nordlow commented 2 years ago

Is detailed data available to be shared?

All the data is printed to standard out.

How would you expect performance to depend on function count and depth?

Ideally linearly on function count and depth. For some compiler implementations this is not the case, though.

run-count ought to be at least 10, to handle noise effects.

Are you saying we should force truncate it to 10?

Derek-Jones commented 2 years ago

By data availability, I meant results from previous runs made by yourself or others.

I guess quadratic is the obvious candidate for performance, when a simple implementation is used.

At least, i.e., at least 10.