Skip to content
Snippets Groups Projects
Select Git revision
  • master default protected
1 result
You can move around the graph by using the arrow keys.
Created with Raphaël 2.2.015Feb141312111087654131Jan302928242322211816111098721Dec18171413121110754328Nov26232221151413931Oct30262524232219181615121110854228Sep2721201917141312629Aug282724232221201615223Jul20191064319Jun151488May426Apr25242319181716131110929Mar28272623222120191615141398762128Feb27232219refactoringadd option for tail ordering[skip ci] TODO Do proper cmake integration[skip ci] Add test[skip ci] Make it possible to use google-benchmark for autotuningAdd include for intrinsicspep8Make sure autotune scores are floatsuse loopy to add dependenciesMove patches for VCL into dune-codegen under a different name and introduce a switchensure ordering: vectorized code before tailadd TODOmake tail vectorization an optionremove wrong assertionChange scaling of autotune resultsfind suitable vcl sizecorrectly handle tail if tail_size % new_vcl_size != 0recursive vectorizationrenamingvectorize tailseparate c index array and vectorized index arraymake vectorization functions explicitly dependend on vector sizeensure tail is after vectorized block(unvectorized) tail works with vectorizationtail works unvectorizedadd tail explicitly[skip ci][WIP] Apply trasformation[skip ci][WIP] Tensor contraction reordering transformation[skip ci] Fix dependency bug in reduction removalLoopy transformation replacinrd reductions with inplace accumulation[!308] Bugfix/potential fixA bit more fixes for the autotune setup[fixup] Remove repr from the base classFix repr of ImmutableCuttingRecord to actually cut the underscore fields[skip ci] more fixup of autotune setuppep8Fixups on the autotuning naming scheme[skip ci][autotune] Use SHA256 in file anme generationFix typo[skip ci] PERFTOOL -> CODEGEN in autotune wrapper variable

Datenschutz und Impressum