trixi-framework / Trixi.jl / 34516258671

97%

push

github

Add 3D flux-differencing GPU turbo kernels (#3229)

* add full sweep kernel and selection

* fix tests

* cover CPU KA backend

* Apply suggestions from code review

Co-authored-by: Marco Artiano <57838732+MarcoArtiano@users.noreply.github.com>

* Update NEWS.md

Co-authored-by: Hendrik Ranocha <ranocha@users.noreply.github.com>

* rename flux_differencing_kernel

* rename get_node_vars_local

* add local data share for half sweep kernel

* Update NEWS.md

Co-authored-by: Benedict <135045760+benegee@users.noreply.github.com>

* use get node vars

* fix args

* remove Atomix

* add docs

* fix CI

* add turbo kernels

* add tests

* fix float32 turbo

* fix docstring

* Update src/semidiscretization/semidiscretization.jl

Co-authored-by: Hendrik Ranocha <ranocha@users.noreply.github.com>

* fix docs and add no lds kernel

* Apply batched suggestions from code review

Co-authored-by: Hendrik Ranocha <ranocha@users.noreply.github.com>

* Apply suggestion from @MarcoArtiano

* fix docs

* re-add kernels + global turbo

* add warning regarding LDS

* remove evalpoly

* improve warning message

* some more improvements in the error message

* improve warning and introduce error for workgroup size

* remove SVector redundancy, improve warning error

* Apply batched suggestions from code review

Co-authored-by: Hendrik Ranocha <ranocha@users.noreply.github.com>

* fix check function for weak kernels

* rename flux diff ka kernel with turbo

* fix check function

* remove turbo FullSweepGlobal

* restore FluxTurbo FluxSweepGlobal fallback test case

* fix fallback

* Apply batched suggestions from code review

Co-authored-by: Hendrik Ranocha <ranocha@users.noreply.github.com>

---------

Co-authored-by: Hendrik Ranocha <ranocha@users.noreply.github.com>
Co-authored-by: Benedict <135045760+benegee@users.noreply.github.com>

26 of 225 new or added lines in 4 files covered. (11.56%)

50781 of 52526 relevant lines covered (96.68%)

42159447.92 hits per line