diff options
| -rw-r--r-- | Tasks.org | 19 |
1 files changed, 12 insertions, 7 deletions
@@ -80,18 +80,23 @@ We have a first pass of these benchmarks, although they may need refined over ti We can generate and run benchmarks from library specs using ~candelabra_cli::cost::benchmark::run_benchmarks~. -** TODO Caching and loading outputs +** DONE Caching and loading outputs -We should cache the output cost models per machine, ideally with a key of library file modification time. -Relevant crates: https://docs.rs/persistentcache/latest/persistentcache/ https://github.com/jaemk/cached +We cache the benchmark results by type and invalidate based on library modification time. +Relevant code is in ~candelabra_cli::cache~ and ~candelabra_cli::cost~. +We also cache primrose results in ~candelabra_cli::candidates~ -** TODO Build cost model from benchmark +** DONE Integrate with CLI -Fit polynomials for each operation from the benchmarking data. +The CLI should get cost models for all of the candidate types, and for now just print them out. -** TODO Integrate with CLI +** DOING Build cost model from benchmark -The CLI should get cost models for all of the candidate types, and for now just print them out. +Fit polynomials for each operation from the benchmarking data. +Possibly helpful crates: + - https://docs.rs/fitme/latest/fitme/fn.fit.html + - https://lib.rs/crates/enterpolation + - https://docs.rs/varpro/latest/varpro/ * BLOCKED Semantic profiler |