Compare NewtonsMethod with NewtonsMethodAD #29
Assignees
Comments
bors bot
referenced
this issue
in CliMA/ClimateMachine.jl
Dec 17, 2020
1869: Make swapping sat adjust numerical methods easier r=charleskawczynski a=charleskawczynski ### Description This PR is a step towards fixing #1831 by using the numerical method type, rather than a function, as an input to `saturation_adjustment`. This should help us more rigorously test/compare the performance and robustness of numerical methods in saturation adjustment for `ρ, e_int, q_tot` and `ρ, p, q_tot` thermo variables, which are the only two cases that are potentially used beyond initialization. I've also battled the formatter a bit in the `@print` statements, and added a test section for performance with BenchmarkTools. This also adds a convenience method, `PhaseEquil_dev_only`, _for development only_ so that we can more easily toggle one option at a time while maintaining synchronization with all other arguments. Co-authored-by: Charles Kawczynski <[email protected]>
bors bot
referenced
this issue
in CliMA/ClimateMachine.jl
Dec 17, 2020
1869: Make swapping sat adjust numerical methods easier r=charleskawczynski a=charleskawczynski ### Description This PR is a step towards fixing #1831 by using the numerical method type, rather than a function, as an input to `saturation_adjustment`. This should help us more rigorously test/compare the performance and robustness of numerical methods in saturation adjustment for `ρ, e_int, q_tot` and `ρ, p, q_tot` thermo variables, which are the only two cases that are potentially used beyond initialization. I've also battled the formatter a bit in the `@print` statements, and added a test section for performance with BenchmarkTools. This also adds a convenience method, `PhaseEquil_dev_only`, _for development only_ so that we can more easily toggle one option at a time while maintaining synchronization with all other arguments. Co-authored-by: Charles Kawczynski <[email protected]>
bors bot
referenced
this issue
in CliMA/ClimateMachine.jl
Dec 18, 2020
1869: Make swapping sat adjust numerical methods easier, add GPU tests r=charleskawczynski a=charleskawczynski ### Description This PR is a step towards fixing #1831 by using the numerical method type, rather than a function, as an input to `saturation_adjustment`. This should help us more rigorously test/compare the performance and robustness of numerical methods in saturation adjustment for `ρ, e_int, q_tot` and `ρ, p, q_tot` thermo variables, which are the only two cases that are potentially used beyond initialization. I've also battled the formatter a bit in the `@print` statements, and added a test section for performance with BenchmarkTools. This also: - adds a convenience method, `PhaseEquil_dev_only`, _for development only_ so that we can more easily toggle one option at a time while maintaining synchronization with all other arguments. - Adds GPU tests Co-authored-by: Charles Kawczynski <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Description
Per Tapio's suggestion, we should compare
NewtonsMethodwithNewtonsMethodADw.r.t. performance/accuracy, and explore:The text was updated successfully, but these errors were encountered: