TA::retile uses efficient (retile_v2) method, no more GEMM-based retiling by default (thanks, @calewis )

evaleev · evaleev · commit 80fb70e652bf · 2024-08-26T17:28:43.000-04:00
diff --git a/src/TiledArray/conversions/retile.h b/src/TiledArray/conversions/retile.h
@@ -164,7 +164,7 @@ auto retile_v2(const DistArray<Tile, Policy>& source_array,
 template <typename Tile, typename Policy>
 auto retile(const DistArray<Tile, Policy>& array,
             const TiledRange& target_trange) {
-  return detail::retile_v0(array, target_trange);
+  return detail::retile_v2(array, target_trange);
 }
 
 }  // namespace TiledArray

Original file line number	Diff line number	Diff line change
`@@ -164,7 +164,7 @@ auto retile_v2(const DistArray<Tile, Policy>& source_array,`
`164`	`164`	`template <typename Tile, typename Policy>`
`165`	`165`	`auto retile(const DistArray<Tile, Policy>& array,`
`166`	`166`	`const TiledRange& target_trange) {`
`167`		`- return detail::retile_v0(array, target_trange);`
	`167`	`+ return detail::retile_v2(array, target_trange);`
`168`	`168`	`}`
`169`	`169`
`170`	`170`	`} // namespace TiledArray`