DDBTRF(3)
π DDBTRF π
okay so bestie, gather round, because this one is a linear algebra moment and it is lowkey iconic. This is the LAPACK routine DDBTRF β it computes an LU factorization of a real m-by-n band matrix A, and it does it without partial pivoting and without row interchanges. Yes, no pivoting. We're raw-dogging the elimination here. β¨
β¨ SYNOPSIS β¨
SUBROUTINE DDBTRF( M, N, KL, KU, AB, LDAB, INFO )
INTEGER INFO, KL, KU, LDAB, M, N
DOUBLE PRECISION AB( LDAB, * )
π¦ DESCRIPTION π¦
Ddbtrf computes an LU factorization of a real m-by-n band matrix A, and it does not use partial pivoting or row interchanges. Not even a little. π
This is the blocked version of the algorithm, and it calls Level 3 BLAS, which is the fast one, the main character one, the one you want doing your heavy lifting. π
π ARGUMENTS π
M(input)INTEGERβ the number of rows of the matrix A. Gotta beM >= 0.N(input)INTEGERβ the number of columns of the matrix A. Gotta beN >= 0.KL(input)INTEGERβ the number of subdiagonals within the band of A.KL >= 0, no exceptions.KU(input)INTEGERβ the number of superdiagonals within the band of A.KU >= 0too, periodt.AB(input/output)REALarray, dimension(LDAB,N)β okay this is the whole vibe right here, so listen up.
On entry, it holds the matrix A in band storage, in rows KL+1 to 2*KL+KU+1. Rows 1 to KL of the array don't need to be set β they're just chilling, unbothered, empty. π The j-th column of A is stored in the j-th column of the array AB like this:
AB(kl+ku+1+i-j,j) = A(i,j) for max(1,j-ku)<=i<=min(m,j+kl)
On exit, it holds the details of the factorization:
- U is stored as an upper triangular band matrix with
KL+KUsuperdiagonals, in rows1toKL+KU+1. - the multipliers used during the factorization are stored in rows
KL+KU+2to2*KL+KU+1.
See below for further details, bestie, I promise it's all there. π«Ά
LDAB(input)INTEGERβ the leading dimension of the array AB. It has to satisfyLDAB >= 2*KL+KU+1. Do not play with this one.INFO(output)INTEGERβ the receipts, and it can go three ways:= 0: successful exit. slay. π< 0: ifINFO = -i, the i-th argument had an illegal value. someone handed us garbage.> 0: ifINFO = +i,U(i,i)is exactly zero. The factorization has been completed, but the factor U is exactly singular, and division by zero will occur if it is used to solve a system of equations. π Not a drill. Not a vibe. Straight up undefined nonsense incoming.
πΈ THE BAND STORAGE SCHEME, ILLUSTRATED πΈ
okay so here's the visual, when M = N = 6, KL = 2, KU = 1:
On entry: On exit:
* a12 a23 a34 a45 a56 * u12 u23 u34 u45 u56
a11 a22 a33 a44 a55 a66 u11 u22 u33 u44 u55 u66
a21 a32 a43 a54 a65 * m21 m32 m43 m54 m65 *
a31 a42 a53 a64 * * m31 m42 m53 m64 * *
Array elements marked * are not used by the routine. They're just there. Decorative. Background characters. π
β οΈ real talk: because there's no pivoting, if that diagonal goes zero you getINFO > 0and the whole thing is singular. read theINFOvalue. always read theINFOvalue.
and that's the whole page, bestie π this one is a LAPACK original β the version and the date live down in the footer, so peek at yours if you need the receipts. now go forth and factor some bands. ππ¦β¨
π¦ where this page comes from
- ubuntu jammy: scalapack-doc 1.5-11
- π girlified by deepseek-ai/DeepSeek-V4.1-Flash on : a language model's rewrite of the original, under its terms