US5200915A
Expired / Lapsed
US
⏱️ 0.0 years remaining
Device for rapidly solving a symmetric linear system on a supercomputer
Assignee: NEC Corp
•
Filing: 1991-06-12
•
Granted: 1993-04-06
•
Expires: 2011-06-12
Abstract
Abstract In order to solve a symmetric linear system given by AU=b, a supercomputer or vector processor calculates a solution vector u using a right-hand side vector b, a diagonal matrix, and an upper triangular matrix. Supplied with an array (JA) representing a column number of the upper triangular matrix, a pointer array constructing section (11) constructs a pointer array (JL) which points to a lower triangular matrix. Supplied with an array (AA) and the array (JA) which collectively represent a combination of the diagonal matrix and the upper triangular matrix, a matrix decomposing section (12) decomposes the array (AA) into an approximate matrix (M) which approximates a symmetric coefficient matrix A. A first product calculating section (16) calculates a first product vector (y) using the array (AA), the array (JA), a first vector (x), and the pointer array (JL). A second product calculating section (17) calculates a second product vector (y') using the approximate matrix (M) and a second vector (x'). An iterative calculating section (18) carries out iterative calculation on the array (AA), the array (JA), an array (B) representing the right-hand side vector, and the first and the second product vectors (y) and (y') to iteratively provide the first and the second product calculating sections with the first and the second vectors, respectively, and to finally produce the solution vector (u).
Inventors
- 👤 Ken Hayami
- 👤 Hiroshi Watanabe
Bibliographic Details
Application Number
—
Jurisdiction
US
Data Source
google_patents
Last Synchronized
2026-11-06 17:52
CPC Classifications
G; G06; G06F; G06F17/00; G06F17/10; G06F17/11; G06F17/12; G06F17/13
Claims Analysis
Total Claims: 8
Claim 1
Independent Claim
393 words
A supercomputer for solving a symmetric linear system given by Au=b, where A represents a symmetric coefficient matrix, b represents a right-hand side vector, and u represents a solution vector, said symmetric coefficient matrix being equal to a three-term sum of a diagonal matrix plus an upper triangular matrix plus a lower triangular matrix which is equal to a transpose matrix of said upper triangular matrix, said supercomputer being supplied with said right-hand side vector, said diagonal matrix, and said upper triangular matrix, said supercomputer calculating said solution vector using said right-hand side vector, said diagonal matrix, and said upper triangular matrix, said supercomputer comprising: a pointer array constructing section supplied with said upper triangular matrix to construct a pointer array which points to said lower triangular matrix to thereby speed up vector processing on said supercomputer; a matrix decomposing section supplied with said diagonal matrix and said upper triangular matrix to decompose said diagonal matrix and said upper triangular matrix into an approximate matrix which approximates said symmetric coefficient matrix; a first product calculating section, supplied with said diagonal matrix, said upper triangular matrix, and a first vector and connected to said pointer array constructing section, to calculate a first product vector using said diagonal matrix, said upper triangular matrix, said first vector, and said pointer array, said first product vector being equal to a product of said symmetric coefficient matrix and said first vector; a second product calculating section, supplied with a second vector and connected to said matrix decomposing section, to calculate a second product vector using said approximate matrix and said second vector, said second product vector being equal to a product of an inverse matrix of said approximate matrix and said second vector; and an iterative calculating section, supplied with said diagonal matrix, said upper triangular matrix, said right-hand side vector and connected to said first and said second product calculating sections, to carry out iterative calculation using said supercomputer on said diagonal matrix, said upper triangular matrix, said right-hand side vector, said first and said second product vectors, said iterative calculating section iteratively providing said first and said second product calculating sections with said first and said second vectors to make said first and said second product calculating sections iteratively calculate said first and said second product vectors, respectively, said iterative calculating section producing said solution vector.
Claim 2
Dependent Claim
393 words
A supercomputer for solving a symmetric linear system given by Au=b, where A represents a symmetric coefficient matrix, b represents a right-hand side vector, and u represents a solution vector, said symmetric coefficient matrix being equal to a three-term sum of a diagonal matrix plus an upper triangular matrix plus a lower triangular matrix which is equal to a transpose matrix of said upper triangular matrix, said supercomputer being supplied with said right-hand side vector, said diagonal matrix, and said upper triangular matrix, said supercomputer calculating said solution vector using said right-hand side vector, said diagonal matrix, and said upper triangular matrix, said supercomputer comprising: a pointer array constructing section supplied with said upper triangular matrix to construct a pointer array which points to said lower triangular matrix to thereby speed up vector processing on said supercomputer; a matrix decomposing section supplied with said diagonal matrix and said upper triangular matrix to decompose said diagonal matrix and said upper triangular matrix into an approximate matrix which approximates said symmetric coefficient matrix; a first product calculating section, supplied with said diagonal matrix, said upper triangular matrix, and a first vector and connected to said pointer array constructing section, to calculate a first product vector using said diagonal matrix, said upper triangular matrix, said first vector, and said pointer array, said first product vector being equal to a product of said symmetric coefficient matrix and said first vector; a second product calculating section, supplied with a second vector and connected to said matrix decomposing section, to calculate a second product vector using said approximate matrix and said second vector, said second product vector being equal to a product of an inverse matrix of said approximate matrix and said second vector; and an iterative calculating section, supplied with said diagonal matrix, said upper triangular matrix, said right-hand side vector and connected to said first and said second product calculating sections, to carry out iterative calculation using said supercomputer on said diagonal matrix, said upper triangular matrix, said right-hand side vector, said first and said second product vectors, said iterative calculating section iteratively providing said first and said second product calculating sections with said first and said second vectors to make said first and said second product calculating sections iteratively calculate said first and said second product vectors, respectively, said iterative calculating section producing said solution vector.
Claim 3
Dependent Claim
391 words
A supercomputer for solving a symmetric linear system given by Au=b, where A represents a symmetric coefficient matrix, b represents a right-hand side vector, and u represents a solution vector, said symmetric coefficient matrix being equal to a three-term sum of a diagonal matrix plus an upper triangular matrix plus a lower triangular matrix which is equal to a transpose matrix of said upper triangular matrix, said supercomputer being supplied with said right-hand side vector, said diagonal matrix, and said lower triangular matrix, said supercomputer calculating said solution vector using said right-hand side vector, said diagonal matrix, and said lower triangular matrix, said supercomputer comprising: a pointer array constructing section supplied with said lower triangular matrix to construct a pointer array which points to said upper triangular matrix to thereby speed up vector processing on said supercomputer; a matrix decomposing section, supplied with said diagonal matrix and said lower triangular matrix, to decompose said diagonal matrix and said lower triangular matrix into an approximate matrix which approximates said symmetric coefficient matrix; a first product calculating section, supplied with said diagonal matrix, said lower triangular matrix, and a first vector and connected to said pointer array constructing section, to calculate a first product vector using said diagonal matrix, said lower triangular matrix, said first vector, and said pointer array, said first product vector being equal to a product of said symmetric coefficient matrix and said first vector; a second product calculating section, supplied with a second vector and connected to said matrix decomposing section, to calculate a second product vector using said approximate matrix and said second vector, said second product vector being equal to a product of an inverse matrix of said approximate matrix and said second vector; and an iterative calculating section, supplied with said diagonal matrix, said lower triangular matrix, and said right-hand side vector and connected to said first and said second product calculating sections, to carry out iterative calculation on said diagonal matrix, said lower triangular matrix, said right-hand side vector, said first and said second product vectors, said iterative calculating section iteratively providing said first and said second product calculating sections with said first and said second vectors to make said first and said second product calculating sections iteratively calculate said first and said second product vectors, respectively, said iterative calculating section producing said solution vector.
Claim 4
Dependent Claim
391 words
A supercomputer for solving a symmetric linear system given by Au=b, where A represents a symmetric coefficient matrix, b represents a right-hand side vector, and u represents a solution vector, said symmetric coefficient matrix being equal to a three-term sum of a diagonal matrix plus an upper triangular matrix plus a lower triangular matrix which is equal to a transpose matrix of said upper triangular matrix, said supercomputer being supplied with said right-hand side vector, said diagonal matrix, and said lower triangular matrix, said supercomputer calculating said solution vector using said right-hand side vector, said diagonal matrix, and said lower triangular matrix, said supercomputer comprising: a pointer array constructing section supplied with said lower triangular matrix to construct a pointer array which points to said upper triangular matrix to thereby speed up vector processing on said supercomputer; a matrix decomposing section, supplied with said diagonal matrix and said lower triangular matrix, to decompose said diagonal matrix and said lower triangular matrix into an approximate matrix which approximates said symmetric coefficient matrix; a first product calculating section, supplied with said diagonal matrix, said lower triangular matrix, and a first vector and connected to said pointer array constructing section, to calculate a first product vector using said diagonal matrix, said lower triangular matrix, said first vector, and said pointer array, said first product vector being equal to a product of said symmetric coefficient matrix and said first vector; a second product calculating section, supplied with a second vector and connected to said matrix decomposing section, to calculate a second product vector using said approximate matrix and said second vector, said second product vector being equal to a product of an inverse matrix of said approximate matrix and said second vector; and an iterative calculating section, supplied with said diagonal matrix, said lower triangular matrix, and said right-hand side vector and connected to said first and said second product calculating sections, to carry out iterative calculation on said diagonal matrix, said lower triangular matrix, said right-hand side vector, said first and said second product vectors, said iterative calculating section iteratively providing said first and said second product calculating sections with said first and said second vectors to make said first and said second product calculating sections iteratively calculate said first and said second product vectors, respectively, said iterative calculating section producing said solution vector.
Claim 5
Dependent Claim
395 words
A vector processor for solving a symmetric linear system given by Au=b, where A represents a symmetric coefficient matrix, b represents a right-hand side vector, and u represents a solution vector, said symmetric coefficient matrix being equal to a three-term sum of a diagonal matrix plus an upper triangular matrix plus a lower triangular matrix which is equal to a transpose matrix of said upper triangular matrix, said vector processor being supplied with said right-hand side vector, said diagonal matrix, and said upper triangular matrix, said vector processor calculating said solution vector using said right-hand side vector, said diagonal matrix, and said upper triangular matrix, said vector processor comprising: a pointer array constructing section supplied with said upper triangular matrix to construct a pointer array which points to said lower triangular matrix to thereby speed up vector processing on said vector processor; a matrix decomposing section supplied with said diagonal matrix and said upper triangular matrix to decompose said diagonal matrix and said upper triangular matrix into an approximate matrix which approximates said symmetric coefficient matrix; a first product calculating section, supplied with said diagonal matrix, said upper triangular matrix, and a first vector and connected to said pointer array constructing section, to calculate a first product vector using said diagonal matrix, said upper triangular matrix, said first vector, and said pointer array, said first product vector being equal to a product of said symmetric coefficient matrix and said first vector; a second product calculating section, supplied with a second vector and connected to said matrix decomposing section, to calculate a second product vector using said approximate matrix and said second vector, said second product vector being equal to a product of an inverse matrix of said approximate matrix and said second vector; and an iterative calculating section, supplied with said diagonal matrix, said upper triangular matrix, said right-hand side vector and connected to said first and said second product calculating sections, to carry out iterative calculation on said diagonal matrix, said upper triangular matrix, said right-hand side vector, said first and said second product vectors, said iterative calculating section iteratively providing said first and said second product calculating sections with said first and said second vectors to make said first and said second product calculating sections iteratively calculate said first and said second product vectors, respectively, said iterative calculating section producing said solution vector.
Claim 6
Dependent Claim
395 words
A vector processor for solving a symmetric linear system given by Au=b, where A represents a symmetric coefficient matrix, b represents a right-hand side vector, and u represents a solution vector, said symmetric coefficient matrix being equal to a three-term sum of a diagonal matrix plus an upper triangular matrix plus a lower triangular matrix which is equal to a transpose matrix of said upper triangular matrix, said vector processor being supplied with said right-hand side vector, said diagonal matrix, and said upper triangular matrix, said vector processor calculating said solution vector using said right-hand side vector, said diagonal matrix, and said upper triangular matrix, said vector processor comprising: a pointer array constructing section supplied with said upper triangular matrix to construct a pointer array which points to said lower triangular matrix to thereby speed up vector processing on said vector processor; a matrix decomposing section supplied with said diagonal matrix and said upper triangular matrix to decompose said diagonal matrix and said upper triangular matrix into an approximate matrix which approximates said symmetric coefficient matrix; a first product calculating section, supplied with said diagonal matrix, said upper triangular matrix, and a first vector and connected to said pointer array constructing section, to calculate a first product vector using said diagonal matrix, said upper triangular matrix, said first vector, and said pointer array, said first product vector being equal to a product of said symmetric coefficient matrix and said first vector; a second product calculating section, supplied with a second vector and connected to said matrix decomposing section, to calculate a second product vector using said approximate matrix and said second vector, said second product vector being equal to a product of an inverse matrix of said approximate matrix and said second vector; and an iterative calculating section, supplied with said diagonal matrix, said upper triangular matrix, said right-hand side vector and connected to said first and said second product calculating sections, to carry out iterative calculation on said diagonal matrix, said upper triangular matrix, said right-hand side vector, said first and said second product vectors, said iterative calculating section iteratively providing said first and said second product calculating sections with said first and said second vectors to make said first and said second product calculating sections iteratively calculate said first and said second product vectors, respectively, said iterative calculating section producing said solution vector.
Claim 7
Dependent Claim
396 words
A vector processor for solving a symmetric linear system given by Au=b, where A represents a symmetric coefficient matrix, b represents a right-hand side vector, and u represents a solution vector, said symmetric coefficient matrix being equal to a three-term sum of a diagonal matrix plus an upper triangular matrix plus a lower triangular matrix which is equal to a transpose matrix of said upper triangular matrix, said vector processor being supplied with said right-hand side vector, said diagonal matrix, and said lower triangular matrix, said vector processor calculating said solution vector using said right-hand side vector, said diagonal matrix, and said lower triangular matrix, said vector processor comprising: a pointer array constructing section supplied with said lower triangular matrix to construct a pointer array which points to said upper triangular matrix to thereby speed up vector processing on said vector processor; a matrix decomposing section, supplied with said diagonal matrix and said lower triangular matrix, to decompose said diagonal matrix and said lower triangular matrix into an approximate matrix which approximates said symmetric coefficient matrix; a first product calculating section, supplied with said diagonal matrix, said lower triangular matrix, and a first vector and connected to said pointer array constructing section, to calculate a first product vector using said diagonal matrix, said lower triangular matrix, said first vector, and said pointer array, said first product vector being equal to a product of said symmetric coefficient matrix and said first vector; a second product calculating section, supplied with a second vector and connected to said matrix decomposing section, to calculate a second product vector using said approximate matrix and said second vector, said second product vector being equal to a product of an inverse matrix of said approximate matrix and said second vector; and an iterative calculating section, supplied with said diagonal matrix, said lower triangular matrix, and said right-hand side vector and connected to said first and said second product calculating sections, to carry out iterative calculation on said diagonal matrix, said lower triangular matrix, said right-hand side vector, said first and said second product vectors, said iterative calculating section iteratively providing said first and said second product calculating sections with said first and said second vectors to make said first and said second product calculating sections iteratively calculate said first and said second product vectors, respectively, said iterative calculating section producing said solution vector.
Claim 8
Dependent Claim
396 words
A vector processor for solving a symmetric linear system given by Au=b, where A represents a symmetric coefficient matrix, b represents a right-hand side vector, and u represents a solution vector, said symmetric coefficient matrix being equal to a three-term sum of a diagonal matrix plus an upper triangular matrix plus a lower triangular matrix which is equal to a transpose matrix of said upper triangular matrix, said vector processor being supplied with said right-hand side vector, said diagonal matrix, and said lower triangular matrix, said vector processor calculating said solution vector using said right-hand side vector, said diagonal matrix, and said lower triangular matrix, said vector processor comprising: a pointer array constructing section supplied with said lower triangular matrix to construct a pointer array which points to said upper triangular matrix to thereby speed up vector processing on said vector processor; a matrix decomposing section, supplied with said diagonal matrix and said lower triangular matrix, to decompose said diagonal matrix and said lower triangular matrix into an approximate matrix which approximates said symmetric coefficient matrix; a first product calculating section, supplied with said diagonal matrix, said lower triangular matrix, and a first vector and connected to said pointer array constructing section, to calculate a first product vector using said diagonal matrix, said lower triangular matrix, said first vector, and said pointer array, said first product vector being equal to a product of said symmetric coefficient matrix and said first vector; a second product calculating section, supplied with a second vector and connected to said matrix decomposing section, to calculate a second product vector using said approximate matrix and said second vector, said second product vector being equal to a product of an inverse matrix of said approximate matrix and said second vector; and an iterative calculating section, supplied with said diagonal matrix, said lower triangular matrix, and said right-hand side vector and connected to said first and said second product calculating sections, to carry out iterative calculation on said diagonal matrix, said lower triangular matrix, said right-hand side vector, said first and said second product vectors, said iterative calculating section iteratively providing said first and said second product calculating sections with said first and said second vectors to make said first and said second product calculating sections iteratively calculate said first and said second product vectors, respectively, said iterative calculating section producing said solution vector.
References Cited (Backward Citations)
US3621209A
Cited
USPTO Annuity & Maintenance Windows
| Annuity Stage | Window Opens | Payment Deadline | Surcharge Period Until | Estimated Fee | Status |
|---|---|---|---|---|---|
| 3.5 Years (Year 4) | 1996-04-05 | 1996-10-05 | 1997-04-06 | $2000.00 | Window Closed / Lapsed |
| 7.5 Years (Year 8) | 2000-04-05 | 2000-10-05 | 2001-04-06 | $3760.00 | Window Closed / Lapsed |
| 11.5 Years (Year 12) | 2004-04-05 | 2004-10-05 | 2005-04-06 | $7700.00 | Window Closed / Lapsed |