René's URL Explorer Experiment

{"@context":"https://schema.org","@type":"DiscussionForumPosting","headline":"Performance improvements","articleBody":"Following some guidelines of the [Rust Performance Book](https://nnethercote.github.io/perf-book/title-page.html) here are some things we can try to improve performance:\r\n\r\n- Add `codegen-units = 1` to release build\r\n- Use a faster allocator. E.g. [mimalloc](https://github.com/microsoft/mimalloc) works on all operating systems\r\n\r\nNot so easy:\r\n- properly profile to identify hot parts\r\n- remove clones/allocations where not needed\r\n- use profile-guided optimization (e.g. via [cargo-pgo](https://github.com/Kobzol/cargo-pgo))\r\n  - unfortunately this is currently not working with LTO and the PGO version is 10-20% slower than LTO\r\n  - might be available in the future in `maturin` directly, see [here](https://github.com/PyO3/maturin/issues/1840)\r\n\r\n---\r\n\r\nQuick tests with `codegen-units = 1` added to `release-lto` (see [here](https://doc.rust-lang.org/rustc/codegen-options/index.html#codegen-units)) show performance improvements of benchmarks of up to 12% (mean is about 7%) while for `dual_number`, changes are a bit smaller (see below).\r\n\r\nProper benchmarks (across all benchmarks) with comparison to current release workflow are needed but this might be an easy-to-get improvement if it turns out to be faster for all cases.\r\n\r\n- Benchmark: dual_numbers\r\n- System: methane/CO2\r\n- `main`: main branch + lto\r\n- `main_codegen`: main branch + lto + codegen-units = 1\r\n- `develop_`: like main\r\n\r\n**Execution times in µs**\r\n| name            |    f64 |   dual |   dual2 |   hyperdual |   dual3 |\r\n|:----------------|-------:|-------:|--------:|------------:|--------:|\r\n| main            | 1.1382 | 1.2325 |  1.4539 |      1.6267 |  1.7563 |\r\n| main_codegen    | 1.0229 | 1.1741 |  1.3708 |      1.5777 |  1.6316 |\r\n| develop         | 1.0138 | 1.1989 |  1.4465 |      1.589  |  1.7549 |\r\n| develop_codegen | 0.9761 | 1.1681 |  1.4195 |      1.5446 |  1.6304 |\r\n\r\n**Slowdown t_f64/t_d for each branch/option**\r\n|                 |   f64 |    dual |   dual2 |   hyperdual |   dual3 |\r\n|:----------------|------:|--------:|--------:|------------:|--------:|\r\n| main            |     1 | 1.08285 | 1.27737 |     1.42919 | 1.54305 |\r\n| main_codegen    |     1 | 1.14782 | 1.34011 |     1.54238 | 1.59507 |\r\n| develop         |     1 | 1.18258 | 1.42681 |     1.56737 | 1.73101 |\r\n| develop_codegen |     1 | 1.1967  | 1.45426 |     1.58242 | 1.67032 |\r\n\r\n**Relative difference in % w.r.t. main + lto for each dual number (t_d_branch - t_d_main) / t_d_main * 100**\r\n| name            |      f64 |     dual |     dual2 |   hyperdual |     dual3 |\r\n|:----------------|---------:|---------:|----------:|------------:|----------:|\r\n| main_codegen    | -10.13   | -4.74 | -5.72  |    -3.01 | -7.10  |\r\n| develop         | -10.93 | -2.73 | -0.51 |    -2.32 | -0.08 |\r\n| develop_codegen | -14.24 | -5.23 | -2.37  |    -5.05 | -7.17  |\r\n\r\n","author":{"url":"https://github.com/g-bauer","@type":"Person","name":"g-bauer"},"datePublished":"2023-06-27T13:17:41.000Z","interactionStatistic":{"@type":"InteractionCounter","interactionType":"https://schema.org/CommentAction","userInteractionCount":1},"url":"https://github.com/163/feos/issues/163"}