Matches in SemOpenAlex for { <https://semopenalex.org/work/W4285504049> ?p ?o ?g. }
Showing items 1 to 70 of
70
with 100 items per page.
- W4285504049 abstract "The <tex xmlns:mml=http://www.w3.org/1998/Math/MathML xmlns:xlink=http://www.w3.org/1999/xlink>$s$</tex> -step Conjugate Gradient (CG) algorithm has the potential to reduce the communication cost of standard CG by a factor of <tex xmlns:mml=http://www.w3.org/1998/Math/MathML xmlns:xlink=http://www.w3.org/1999/xlink>$s$</tex> . However, though mathematically equivalent, <tex xmlns:mml=http://www.w3.org/1998/Math/MathML xmlns:xlink=http://www.w3.org/1999/xlink>$s$</tex> -step CG may be numerically less stable compared to standard CG in finite precision, exhibiting slower convergence and decreased attainable accuracy. This limits the use of <tex xmlns:mml=http://www.w3.org/1998/Math/MathML xmlns:xlink=http://www.w3.org/1999/xlink>$s$</tex> -step CG in practice. To improve the numerical behavior of <tex xmlns:mml=http://www.w3.org/1998/Math/MathML xmlns:xlink=http://www.w3.org/1999/xlink>$s$</tex> -step CG and overcome this potential limitation, we incorporate two techniques. First, we improve convergence behavior through the use of higher precision at critical parts of the <tex xmlns:mml=http://www.w3.org/1998/Math/MathML xmlns:xlink=http://www.w3.org/1999/xlink>$s$</tex> -step iteration and second, we integrate a residual replacement strategy into the resulting mixed precision <tex xmlns:mml=http://www.w3.org/1998/Math/MathML xmlns:xlink=http://www.w3.org/1999/xlink>$s$</tex> -step CG to improve attainable accuracy. Our experimental results on the Summit Supercomputer demonstrate that when the higher precision is implemented in hardware, these techniques have virtually no overhead on the iteration time while improving both the convergence rate and the attainable accuracy of <tex xmlns:mml=http://www.w3.org/1998/Math/MathML xmlns:xlink=http://www.w3.org/1999/xlink>$s$</tex> -step CG. Even when the higher precision is implemented in software, these techniques may still reduce the time-to-solution (speedups of up to <tex xmlns:mml=http://www.w3.org/1998/Math/MathML xmlns:xlink=http://www.w3.org/1999/xlink>$1.8times$</tex> in our experiments), especially when <tex xmlns:mml=http://www.w3.org/1998/Math/MathML xmlns:xlink=http://www.w3.org/1999/xlink>$s$</tex> -step CG suffers from numerical instability with a small step size and the latency cost becomes a significant part of its iteration time." @default.
- W4285504049 created "2022-07-15" @default.
- W4285504049 creator A5052397462 @default.
- W4285504049 creator A5072265294 @default.
- W4285504049 creator A5090182413 @default.
- W4285504049 date "2022-05-01" @default.
- W4285504049 modified "2023-10-16" @default.
- W4285504049 title "Mixed Precision $s$-step Conjugate Gradient with Residual Replacement on GPUs" @default.
- W4285504049 cites W1970524119 @default.
- W4285504049 cites W2002054778 @default.
- W4285504049 cites W2024149451 @default.
- W4285504049 cites W2043904766 @default.
- W4285504049 cites W2066692739 @default.
- W4285504049 cites W2078794610 @default.
- W4285504049 cites W2090547338 @default.
- W4285504049 cites W2096714979 @default.
- W4285504049 cites W2104380208 @default.
- W4285504049 cites W2106833662 @default.
- W4285504049 cites W2112979995 @default.
- W4285504049 cites W2155967869 @default.
- W4285504049 cites W2165439482 @default.
- W4285504049 cites W2316564661 @default.
- W4285504049 cites W2476946651 @default.
- W4285504049 cites W2997929922 @default.
- W4285504049 cites W591632009 @default.
- W4285504049 cites W754069224 @default.
- W4285504049 doi "https://doi.org/10.1109/ipdps53621.2022.00091" @default.
- W4285504049 hasPublicationYear "2022" @default.
- W4285504049 type Work @default.
- W4285504049 citedByCount "0" @default.
- W4285504049 crossrefType "proceedings-article" @default.
- W4285504049 hasAuthorship W4285504049A5052397462 @default.
- W4285504049 hasAuthorship W4285504049A5072265294 @default.
- W4285504049 hasAuthorship W4285504049A5090182413 @default.
- W4285504049 hasConcept C11413529 @default.
- W4285504049 hasConcept C154945302 @default.
- W4285504049 hasConcept C155512373 @default.
- W4285504049 hasConcept C162324750 @default.
- W4285504049 hasConcept C2777303404 @default.
- W4285504049 hasConcept C41008148 @default.
- W4285504049 hasConcept C50522688 @default.
- W4285504049 hasConcept C81184566 @default.
- W4285504049 hasConceptScore W4285504049C11413529 @default.
- W4285504049 hasConceptScore W4285504049C154945302 @default.
- W4285504049 hasConceptScore W4285504049C155512373 @default.
- W4285504049 hasConceptScore W4285504049C162324750 @default.
- W4285504049 hasConceptScore W4285504049C2777303404 @default.
- W4285504049 hasConceptScore W4285504049C41008148 @default.
- W4285504049 hasConceptScore W4285504049C50522688 @default.
- W4285504049 hasConceptScore W4285504049C81184566 @default.
- W4285504049 hasFunder F4320306084 @default.
- W4285504049 hasFunder F4320309755 @default.
- W4285504049 hasFunder F4320332359 @default.
- W4285504049 hasFunder F4320332369 @default.
- W4285504049 hasLocation W42855040491 @default.
- W4285504049 hasOpenAccess W4285504049 @default.
- W4285504049 hasPrimaryLocation W42855040491 @default.
- W4285504049 hasRelatedWork W1975200206 @default.
- W4285504049 hasRelatedWork W2000654528 @default.
- W4285504049 hasRelatedWork W2073004239 @default.
- W4285504049 hasRelatedWork W2083522123 @default.
- W4285504049 hasRelatedWork W2128886773 @default.
- W4285504049 hasRelatedWork W2280258626 @default.
- W4285504049 hasRelatedWork W2388528661 @default.
- W4285504049 hasRelatedWork W2973451922 @default.
- W4285504049 hasRelatedWork W3094412894 @default.
- W4285504049 hasRelatedWork W3107474891 @default.
- W4285504049 isParatext "false" @default.
- W4285504049 isRetracted "false" @default.
- W4285504049 workType "article" @default.