Matches in SemOpenAlex for { <https://semopenalex.org/work/W3216608520> ?p ?o ?g. }
- W3216608520 abstract "Compared to the classical Lanczos algorithm, the s-step Lanczos variant has the potential to improve performance by asymptotically decreasing the synchronization cost per iteration. However, this comes at a price; despite being mathematically equivalent, the s-step variant may behave quite differently in finite precision, potentially exhibiting greater loss of accuracy and slower convergence relative to the classical algorithm. It has previously been shown that the errors in the s-step version follow the same structure as the errors in the classical algorithm, but are amplified by a factor depending on the square of the condition number of the O ( s ) -dimensional Krylov bases computed in each outer loop. As the condition number of these s-step bases grows (in some cases very quickly) with s, this limits the s values that can be chosen and thus can limit the attainable performance. In this work, we show that if a select few computations in s-step Lanczos are performed in double the working precision, the error terms then depend only linearly on the conditioning of the s-step bases. This has the potential for drastically improving the numerical behavior of the algorithm with little impact on per-iteration performance. Our numerical experiments demonstrate the improved numerical behavior possible with the mixed precision approach, and also show that this improved behavior extends to mixed precision s-step CG. We present preliminary performance results on NVIDIA V100 GPUs that show that the overhead of extra precision is minimal if one uses precisions implemented in hardware." @default.
- W3216608520 created "2021-12-06" @default.
- W3216608520 creator A5009460108 @default.
- W3216608520 creator A5052397462 @default.
- W3216608520 creator A5072265294 @default.
- W3216608520 date "2021-11-23" @default.
- W3216608520 modified "2023-10-18" @default.
- W3216608520 title "Mixed precision <i>s</i> ‐step Lanczos and conjugate gradient algorithms" @default.
- W3216608520 cites W1996375718 @default.
- W3216608520 cites W2002054778 @default.
- W3216608520 cites W2030995892 @default.
- W3216608520 cites W2035080386 @default.
- W3216608520 cites W2066692739 @default.
- W3216608520 cites W2075351838 @default.
- W3216608520 cites W2078794610 @default.
- W3216608520 cites W2096714979 @default.
- W3216608520 cites W2099611016 @default.
- W3216608520 cites W2105745683 @default.
- W3216608520 cites W2117366131 @default.
- W3216608520 cites W2117686912 @default.
- W3216608520 cites W2125426869 @default.
- W3216608520 cites W2127022449 @default.
- W3216608520 cites W2145194992 @default.
- W3216608520 cites W2316564661 @default.
- W3216608520 cites W2577617996 @default.
- W3216608520 cites W2728074266 @default.
- W3216608520 cites W3007399011 @default.
- W3216608520 cites W3216608520 @default.
- W3216608520 cites W591632009 @default.
- W3216608520 cites W754069224 @default.
- W3216608520 doi "https://doi.org/10.1002/nla.2425" @default.
- W3216608520 hasPublicationYear "2021" @default.
- W3216608520 type Work @default.
- W3216608520 sameAs 3216608520 @default.
- W3216608520 citedByCount "4" @default.
- W3216608520 countsByYear W32166085202021 @default.
- W3216608520 countsByYear W32166085202022 @default.
- W3216608520 countsByYear W32166085202023 @default.
- W3216608520 crossrefType "journal-article" @default.
- W3216608520 hasAuthorship W3216608520A5009460108 @default.
- W3216608520 hasAuthorship W3216608520A5052397462 @default.
- W3216608520 hasAuthorship W3216608520A5072265294 @default.
- W3216608520 hasBestOaLocation W32166085204 @default.
- W3216608520 hasConcept C11413529 @default.
- W3216608520 hasConcept C119256216 @default.
- W3216608520 hasConcept C121332964 @default.
- W3216608520 hasConcept C134306372 @default.
- W3216608520 hasConcept C151201525 @default.
- W3216608520 hasConcept C158693339 @default.
- W3216608520 hasConcept C162324750 @default.
- W3216608520 hasConcept C203739276 @default.
- W3216608520 hasConcept C20501136 @default.
- W3216608520 hasConcept C2777303404 @default.
- W3216608520 hasConcept C28826006 @default.
- W3216608520 hasConcept C33923547 @default.
- W3216608520 hasConcept C35912277 @default.
- W3216608520 hasConcept C41008148 @default.
- W3216608520 hasConcept C45374587 @default.
- W3216608520 hasConcept C50522688 @default.
- W3216608520 hasConcept C62520636 @default.
- W3216608520 hasConcept C81184566 @default.
- W3216608520 hasConceptScore W3216608520C11413529 @default.
- W3216608520 hasConceptScore W3216608520C119256216 @default.
- W3216608520 hasConceptScore W3216608520C121332964 @default.
- W3216608520 hasConceptScore W3216608520C134306372 @default.
- W3216608520 hasConceptScore W3216608520C151201525 @default.
- W3216608520 hasConceptScore W3216608520C158693339 @default.
- W3216608520 hasConceptScore W3216608520C162324750 @default.
- W3216608520 hasConceptScore W3216608520C203739276 @default.
- W3216608520 hasConceptScore W3216608520C20501136 @default.
- W3216608520 hasConceptScore W3216608520C2777303404 @default.
- W3216608520 hasConceptScore W3216608520C28826006 @default.
- W3216608520 hasConceptScore W3216608520C33923547 @default.
- W3216608520 hasConceptScore W3216608520C35912277 @default.
- W3216608520 hasConceptScore W3216608520C41008148 @default.
- W3216608520 hasConceptScore W3216608520C45374587 @default.
- W3216608520 hasConceptScore W3216608520C50522688 @default.
- W3216608520 hasConceptScore W3216608520C62520636 @default.
- W3216608520 hasConceptScore W3216608520C81184566 @default.
- W3216608520 hasFunder F4320306084 @default.
- W3216608520 hasFunder F4320309755 @default.
- W3216608520 hasIssue "3" @default.
- W3216608520 hasLocation W32166085201 @default.
- W3216608520 hasLocation W32166085202 @default.
- W3216608520 hasLocation W32166085203 @default.
- W3216608520 hasLocation W32166085204 @default.
- W3216608520 hasOpenAccess W3216608520 @default.
- W3216608520 hasPrimaryLocation W32166085201 @default.
- W3216608520 hasRelatedWork W1979614781 @default.
- W3216608520 hasRelatedWork W2067819697 @default.
- W3216608520 hasRelatedWork W2099169296 @default.
- W3216608520 hasRelatedWork W2501847067 @default.
- W3216608520 hasRelatedWork W3136232294 @default.
- W3216608520 hasRelatedWork W3166616159 @default.
- W3216608520 hasRelatedWork W3216608520 @default.
- W3216608520 hasRelatedWork W4287268258 @default.
- W3216608520 hasRelatedWork W4386346348 @default.
- W3216608520 hasRelatedWork W591632009 @default.
- W3216608520 hasVolume "29" @default.
- W3216608520 isParatext "false" @default.