Add numerical support of other real types (`Float32`)

huiyuxie commented 2 months ago

Address parts of #591 and redo #1604.

Tasks:

[x] acoustic_perturbation_2d
[x] compressible_euler_1d, compressible_euler_2d, compressible_euler_3d
[x] unit tests for above

github-actions[bot] commented 2 months ago

Review checklist

This checklist is meant to assist creators of PRs (to let them know what reviewers will typically look for) and reviewers (to guide them in a structured review process). Items do not need to be checked explicitly for a PR to be eligible for merging.

Purpose and scope

[ ] The PR has a single goal that is clear from the PR title and/or description.
[ ] All code changes represent a single set of modifications that logically belong together.
[ ] No more than 500 lines of code are changed or there is no obvious way to split the PR into multiple PRs.

Code quality

[ ] The code can be understood easily.
[ ] Newly introduced names for variables etc. are self-descriptive and consistent with existing naming conventions.
[ ] There are no redundancies that can be removed by simple modularization/refactoring.
[ ] There are no leftover debug statements or commented code sections.
[ ] The code adheres to our conventions and style guide, and to the Julia guidelines.

Documentation

[ ] New functions and types are documented with a docstring or top-level comment.
[ ] Relevant publications are referenced in docstrings (see example for formatting).
[ ] Inline comments are used to document longer or unusual code sections.
[ ] Comments describe intent ("why?") and not just functionality ("what?").
[ ] If the PR introduces a significant change or new feature, it is documented in NEWS.md.

Testing

[ ] The PR passes all tests.
[ ] New or modified lines of code are covered by tests.
[ ] New or modified tests run in less then 10 seconds.

Performance

[ ] There are no type instabilities or memory allocations in performance-critical parts.
[ ] If the PR intent is to improve performance, before/after time measurements are posted in the PR.

Verification

[ ] The correctness of the code was verified using appropriate tests.
[ ] If new equations/methods are added, a convergence test has been run and the results are posted in the PR.

Created with :heart: by the Trixi.jl community.

huiyuxie commented 2 months ago

General question here: Is it better to conduct unit testing while making changes, or to perform the tests after completing all the changes? Are there any good examples of testing from this repository? Thanks!

ranocha commented 2 months ago

General question here: Is it better to conduct unit testing while making changes, or to perform the tests after completing all the changes? Are there any good examples of testing from this repository? Thanks!

When you're working on numerical fluxes, you can add some unit tests, e.g., to parts like https://github.com/trixi-framework/Trixi.jl/blob/1745df43b6f4aa5801fb1bc29f4d328f37b42035/test/test_unit.jl#L633-L660 for the compressible Euler equations. Here, you could add additional tests using inputs with eltype(u) == Float32.

We should also add some full integration tests once something works completely - maybe by adding one or two new example elixirs showing how to run simulations with Float32.

codecov[bot] commented 2 months ago

Codecov Report

All modified and coverable lines are covered by tests :white_check_mark:

Project coverage is 96.12%. Comparing base (3b52a30) to head (60836bb).

Additional details and impacted files

```diff @@ Coverage Diff @@ ## main #1909 +/- ## ========================================== + Coverage 96.11% 96.12% +0.01% ========================================== Files 460 460 Lines 36926 36952 +26 ========================================== + Hits 35490 35520 +30 + Misses 1436 1432 -4 ``` | [Flag](https://app.codecov.io/gh/trixi-framework/Trixi.jl/pull/1909/flags?src=pr&el=flags&utm_medium=referral&utm_source=github&utm_content=comment&utm_campaign=pr+comments&utm_term=trixi-framework) | Coverage Δ | | |---|---|---| | [unittests](https://app.codecov.io/gh/trixi-framework/Trixi.jl/pull/1909/flags?src=pr&el=flag&utm_medium=referral&utm_source=github&utm_content=comment&utm_campaign=pr+comments&utm_term=trixi-framework) | `96.12% <100.00%> (+0.01%)` | :arrow_up: | Flags with carried forward coverage won't be shown. [Click here](https://docs.codecov.io/docs/carryforward-flags?utm_medium=referral&utm_source=github&utm_content=comment&utm_campaign=pr+comments&utm_term=trixi-framework#carryforward-flags-in-the-pull-request-comment) to find out more.

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.

huiyuxie commented 2 months ago

Hi @ranocha I have revised the code based on our discussion. Please review it once more, and if this version is satisfactory, I will apply similar revisions to the subsequent tasks.

For dual numbers, I have no idea whether it will be copied to the GPU and I think CUDA.jl does not directly support this type. If they are to be executed on GPUs, I will explore alternative methods (probably through struct) to manage them.

For documentation, are you referring to a general overview that explains the purpose of using a function like oftype, or to detailed documentation that points to each specific instance in the code?

ranocha commented 2 months ago

@sloede Could you please have a look as well?

ranocha commented 1 month ago

Maybe we can even provide a developer-only convenience function such as isexactf0 with isexactf0(0.5) == true and isexactf0(0.4) == false such that people can easily check whether they need to write convert(RealT, 0.4) or can just use 0.5f0?

I'm not sure it's worth writing something that can be achieved easily by

julia> 0.5 == 0.5f0
true

julia> 0.4 == 0.4f0
false

sloede commented 1 month ago

Maybe we can even provide a developer-only convenience function such as isexactf0 with isexactf0(0.5) == true and isexactf0(0.4) == false such that people can easily check whether they need to write convert(RealT, 0.4) or can just use 0.5f0?

I'm not sure it's worth writing something that can be achieved easily by
julia> 0.5 == 0.5f0
true

julia> 0.4 == 0.4f0
false

Fair enough. If it's properly documented somewhere, this should be sufficient.

ranocha commented 1 month ago

Currently, I apply convert to integers case by case. But I think it is not friendly for later contributors to implement new functions (i.e., it is not something straightforward to think about).

If we want to have a simple rule that's easy to remember, we should just recommend using convert, one, zero, oftype etc. I'm not sure whether others havesimilar use cases, but convert etc. is at least used quite often, e.g., in the SciML ecosystem:

https://github.com/SciML/OrdinaryDiffEq.jl/blob/438f34ba24bab5a2f6aa478850a267478cc53dea/src/perform_step/low_order_rk_perform_step.jl#L749C6-L749C28 https://github.com/SciML/OrdinaryDiffEq.jl/blob/438f34ba24bab5a2f6aa478850a267478cc53dea/src/tableaus/low_order_rk_tableaus.jl#L539-L582

ranocha commented 1 month ago

I started some benchmarks locally to see whether there is any impact on our standard benchmark examples. I would like to check the results before merging this PR.