Implementation
void zpbtrf(
final String UPLO,
final int N,
final int KD,
final Matrix<Complex> AB_,
final int LDAB,
final Box<int> INFO,
) {
final AB = AB_.having(ld: LDAB);
const ONE = 1.0;
const NBMAX = 32, LDWORK = NBMAX + 1;
int I, I2, I3, IB, J, JJ, NB;
final WORK = Matrix<Complex>(LDWORK, NBMAX);
final II = Box(0);
// Test the input parameters.
INFO.value = 0;
if ((!lsame(UPLO, 'U')) && (!lsame(UPLO, 'L'))) {
INFO.value = -1;
} else if (N < 0) {
INFO.value = -2;
} else if (KD < 0) {
INFO.value = -3;
} else if (LDAB < KD + 1) {
INFO.value = -5;
}
if (INFO.value != 0) {
xerbla('ZPBTRF', -INFO.value);
return;
}
// Quick return if possible
if (N == 0) return;
// Determine the block size for this environment
NB = ilaenv(1, 'ZPBTRF', UPLO, N, KD, -1, -1);
// The block size must not exceed the semi-bandwidth KD, and must not
// exceed the limit set by the size of the local array WORK.
NB = min(NB, NBMAX);
if (NB <= 1 || NB > KD) {
// Use unblocked code
zpbtf2(UPLO, N, KD, AB, LDAB, INFO);
} else {
// Use blocked code
if (lsame(UPLO, 'U')) {
// Compute the Cholesky factorization of a Hermitian band
// matrix, given the upper triangle of the matrix in band
// storage.
// Zero the upper triangle of the work array.
for (J = 1; J <= NB; J++) {
for (I = 1; I <= J - 1; I++) {
WORK[I][J] = Complex.zero;
}
}
// Process the band matrix one diagonal block at a time.
for (I = 1; I <= N; I += NB) {
IB = min(NB, N - I + 1);
// Factorize the diagonal block
zpotf2(UPLO, IB, AB(KD + 1, I), LDAB - 1, II);
if (II.value != 0) {
INFO.value = I + II.value - 1;
return;
}
if (I + IB <= N) {
// Update the relevant part of the trailing submatrix.
// If A11 denotes the diagonal block which has just been
// factorized, then we need to update the remaining
// blocks in the diagram:
// A11 A12 A13
// A22 A23
// A33
// The numbers of rows and columns in the partitioning
// are IB, I2, I3 respectively. The blocks A12, A22 and
// A23 are empty if IB = KD. The upper triangle of A13
// lies outside the band.
I2 = min(KD - IB, N - I - IB + 1);
I3 = min(IB, N - I - KD + 1);
if (I2 > 0) {
// Update A12
ztrsm(
'Left',
'Upper',
'Conjugate transpose',
'Non-unit',
IB,
I2,
Complex.one,
AB(KD + 1, I),
LDAB - 1,
AB(KD + 1 - IB, I + IB),
LDAB - 1);
// Update A22
zherk(
'Upper',
'Conjugate transpose',
I2,
IB,
-ONE,
AB(KD + 1 - IB, I + IB),
LDAB - 1,
ONE,
AB(KD + 1, I + IB),
LDAB - 1);
}
if (I3 > 0) {
// Copy the lower triangle of A13 into the work array.
for (JJ = 1; JJ <= I3; JJ++) {
for (II.value = JJ; II.value <= IB; II.value++) {
WORK[II.value][JJ] = AB[II.value - JJ + 1][JJ + I + KD - 1];
}
}
// Update A13 (in the work array).
ztrsm('Left', 'Upper', 'Conjugate transpose', 'Non-unit', IB, I3,
Complex.one, AB(KD + 1, I), LDAB - 1, WORK, LDWORK);
// Update A23
if (I2 > 0) {
zgemm(
'Conjugate transpose',
'No transpose',
I2,
I3,
IB,
-Complex.one,
AB(KD + 1 - IB, I + IB),
LDAB - 1,
WORK,
LDWORK,
Complex.one,
AB(1 + IB, I + KD),
LDAB - 1);
}
// Update A33
zherk('Upper', 'Conjugate transpose', I3, IB, -ONE, WORK, LDWORK,
ONE, AB(KD + 1, I + KD), LDAB - 1);
// Copy the lower triangle of A13 back into place.
for (JJ = 1; JJ <= I3; JJ++) {
for (II.value = JJ; II.value <= IB; II.value++) {
AB[II.value - JJ + 1][JJ + I + KD - 1] = WORK[II.value][JJ];
}
}
}
}
}
} else {
// Compute the Cholesky factorization of a Hermitian band
// matrix, given the lower triangle of the matrix in band
// storage.
// Zero the lower triangle of the work array.
for (J = 1; J <= NB; J++) {
for (I = J + 1; I <= NB; I++) {
WORK[I][J] = Complex.zero;
}
}
// Process the band matrix one diagonal block at a time.
for (I = 1; I <= N; I += NB) {
IB = min(NB, N - I + 1);
// Factorize the diagonal block
zpotf2(UPLO, IB, AB(1, I), LDAB - 1, II);
if (II.value != 0) {
INFO.value = I + II.value - 1;
return;
}
if (I + IB <= N) {
// Update the relevant part of the trailing submatrix.
// If A11 denotes the diagonal block which has just been
// factorized, then we need to update the remaining
// blocks in the diagram:
// A11
// A21 A22
// A31 A32 A33
// The numbers of rows and columns in the partitioning
// are IB, I2, I3 respectively. The blocks A21, A22 and
// A32 are empty if IB = KD. The lower triangle of A31
// lies outside the band.
I2 = min(KD - IB, N - I - IB + 1);
I3 = min(IB, N - I - KD + 1);
if (I2 > 0) {
// Update A21
ztrsm('Right', 'Lower', 'Conjugate transpose', 'Non-unit', I2, IB,
Complex.one, AB(1, I), LDAB - 1, AB(1 + IB, I), LDAB - 1);
// Update A22
zherk('Lower', 'No transpose', I2, IB, -ONE, AB(1 + IB, I),
LDAB - 1, ONE, AB(1, I + IB), LDAB - 1);
}
if (I3 > 0) {
// Copy the upper triangle of A31 into the work array.
for (JJ = 1; JJ <= IB; JJ++) {
for (II.value = 1; II.value <= min(JJ, I3); II.value++) {
WORK[II.value][JJ] = AB[KD + 1 - JJ + II.value][JJ + I - 1];
}
}
// Update A31 (in the work array).
ztrsm('Right', 'Lower', 'Conjugate transpose', 'Non-unit', I3, IB,
Complex.one, AB(1, I), LDAB - 1, WORK, LDWORK);
// Update A32
if (I2 > 0) {
zgemm(
'No transpose',
'Conjugate transpose',
I3,
I2,
IB,
-Complex.one,
WORK,
LDWORK,
AB(1 + IB, I),
LDAB - 1,
Complex.one,
AB(1 + KD - IB, I + IB),
LDAB - 1);
}
// Update A33
zherk('Lower', 'No transpose', I3, IB, -ONE, WORK, LDWORK, ONE,
AB(1, I + KD), LDAB - 1);
// Copy the upper triangle of A31 back into place.
for (JJ = 1; JJ <= IB; JJ++) {
for (II.value = 1; II.value <= min(JJ, I3); II.value++) {
AB[KD + 1 - JJ + II.value][JJ + I - 1] = WORK[II.value][JJ];
}
}
}
}
}
}
}
}
