-
Notifications
You must be signed in to change notification settings - Fork 98
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Upwind FDSBP performance improvements #1282
Conversation
* strengthen the isentropic vortex initial conditions in TreeMesh elixirs * update test values for new isentropic vortex initial conditions * update vortex initial condition in special elixir and docs * fix typos in the IC * update tree_mpi test values * remove comment lines because them seem to break literate Co-authored-by: Michael Schlottke-Lakemper <michael@sloede.com>
…ng by ca. 1/4 for Taylor-Green (serial)
… and reduce duplications
25de6bc
to
0377211
Compare
0377211
to
d39ed86
Compare
Yes, I will have a look tomorrow morning. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Overall nice work! Returning the pair of fp
and fm
fluxes was something that Gregor and I had in mind from the beginning but couldn't work out the best way to do it.
I left comments / questions mainly on the 1D versions of the new implementation because the 2D/3D variants are very similar.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM!
* improve some docstrings (#1274) * fix convergence_test for elixirs without trailing new line (#1280) * Fix `initial_condition_isentropic_vortex` (#1279) * strengthen the isentropic vortex initial conditions in TreeMesh elixirs * update test values for new isentropic vortex initial conditions * update vortex initial condition in special elixir and docs * fix typos in the IC * update tree_mpi test values * remove comment lines because them seem to break literate Co-authored-by: Michael Schlottke-Lakemper <michael@sloede.com> * improve performance of 3D volume integral with steger_warming_splitting by ca. 1/4 for Taylor-Green (serial) * lose 2 percent of threaded volume terms performance but simplify code and reduce duplications * remove flux_upwind stub * 2D and 1D as well; improvement ca. 20% in 3D, 10% in 2D, not much in 1D * positive_part, negative_part for 3D Steger-Warming; improves serial PID of TGV by ca. 10% * positive_part, negative_part for 2D, 1D * simplify CSE for LLVM (ca. 2 % for TGV) * comments on f_minus_plus etc. * improve comment * comments on positive/negative part Co-authored-by: Andrew Winters <andrew.ross.winters@liu.se> Co-authored-by: Michael Schlottke-Lakemper <michael@sloede.com>
* improve some docstrings (#1274) * fix convergence_test for elixirs without trailing new line (#1280) * Fix `initial_condition_isentropic_vortex` (#1279) * strengthen the isentropic vortex initial conditions in TreeMesh elixirs * update test values for new isentropic vortex initial conditions * update vortex initial condition in special elixir and docs * fix typos in the IC * update tree_mpi test values * remove comment lines because them seem to break literate Co-authored-by: Michael Schlottke-Lakemper <michael@sloede.com> * improve performance of 3D volume integral with steger_warming_splitting by ca. 1/4 for Taylor-Green (serial) * lose 2 percent of threaded volume terms performance but simplify code and reduce duplications * remove flux_upwind stub * 2D and 1D as well; improvement ca. 20% in 3D, 10% in 2D, not much in 1D * positive_part, negative_part for 3D Steger-Warming; improves serial PID of TGV by ca. 10% * positive_part, negative_part for 2D, 1D * simplify CSE for LLVM (ca. 2 % for TGV) * comments on f_minus_plus etc. * improve comment * comments on positive/negative part Co-authored-by: Andrew Winters <andrew.ross.winters@liu.se> Co-authored-by: Michael Schlottke-Lakemper <michael@sloede.com>
positive_part
andnegative_part
instead of computingabs
etc.All of the changes above (including updating SummationByPartsOperators.jl to v0.5.27) increase the performance of the Taylor-Green vortex
on my system at work as follows:
dev
@andrewwinters5000 Could you please check whether this looks good to you?