OpenBLAS/relapack/src/cgbtrf.c

233 lines
7.3 KiB
C

#include "relapack.h"
#include "stdlib.h"
static void RELAPACK_cgbtrf_rec(const blasint *, const blasint *, const blasint *,
const blasint *, float *, const blasint *, blasint *, float *, const blasint *, float *,
const blasint *, blasint *);
/** CGBTRF computes an LU factorization of a complex m-by-n band matrix A using partial pivoting with row interchanges.
*
* This routine is functionally equivalent to LAPACK's cgbtrf.
* For details on its interface, see
* http://www.netlib.org/lapack/explore-html/d0/d3a/cgbtrf_8f.html
* */
void RELAPACK_cgbtrf(
const blasint *m, const blasint *n, const blasint *kl, const blasint *ku,
float *Ab, const blasint *ldAb, blasint *ipiv,
blasint *info
) {
// Check arguments
*info = 0;
if (*m < 0)
*info = -1;
else if (*n < 0)
*info = -2;
else if (*kl < 0)
*info = -3;
else if (*ku < 0)
*info = -4;
else if (*ldAb < 2 * *kl + *ku + 1)
*info = -6;
if (*info) {
const blasint minfo = -*info;
LAPACK(xerbla)("CGBTRF", &minfo, strlen("CGBTRF"));
return;
}
// Constant
const float ZERO[] = { 0., 0. };
// Result upper band width
const blasint kv = *ku + *kl;
// Unskew A
const blasint ldA[] = { *ldAb - 1 };
float *const A = Ab + 2 * kv;
// Zero upper diagonal fill-in elements
blasint i, j;
for (j = 0; j < *n; j++) {
float *const A_j = A + 2 * *ldA * j;
for (i = MAX(0, j - kv); i < j - *ku; i++)
A_j[2 * i] = A_j[2 * i + 1] = 0.;
}
// Allocate work space
const blasint n1 = CREC_SPLIT(*n);
const blasint mWorkl = (kv > n1) ? MAX(1, *m - *kl) : kv;
const blasint nWorkl = (kv > n1) ? n1 : kv;
const blasint mWorku = (*kl > n1) ? n1 : *kl;
const blasint nWorku = (*kl > n1) ? MAX(0, *n - *kl) : *kl;
float *Workl = malloc(mWorkl * nWorkl * 2 * sizeof(float));
float *Worku = malloc(mWorku * nWorku * 2 * sizeof(float));
LAPACK(claset)("L", &mWorkl, &nWorkl, ZERO, ZERO, Workl, &mWorkl);
LAPACK(claset)("U", &mWorku, &nWorku, ZERO, ZERO, Worku, &mWorku);
// Recursive kernel
RELAPACK_cgbtrf_rec(m, n, kl, ku, Ab, ldAb, ipiv, Workl, &mWorkl, Worku, &mWorku, info);
// Free work space
free(Workl);
free(Worku);
}
/** cgbtrf's recursive compute kernel */
static void RELAPACK_cgbtrf_rec(
const blasint *m, const blasint *n, const blasint *kl, const blasint *ku,
float *Ab, const blasint *ldAb, blasint *ipiv,
float *Workl, const blasint *ldWorkl, float *Worku, const blasint *ldWorku,
blasint *info
) {
if (*n <= MAX(CROSSOVER_CGBTRF, 1)) {
// Unblocked
LAPACK(cgbtf2)(m, n, kl, ku, Ab, ldAb, ipiv, info);
return;
}
// Constants
const float ONE[] = { 1., 0. };
const float MONE[] = { -1., 0. };
const blasint iONE[] = { 1 };
// Loop iterators
blasint i, j;
// Output upper band width
const blasint kv = *ku + *kl;
// Unskew A
const blasint ldA[] = { *ldAb - 1 };
float *const A = Ab + 2 * kv;
// Splitting
const blasint n1 = MIN(CREC_SPLIT(*n), *kl);
const blasint n2 = *n - n1;
const blasint m1 = MIN(n1, *m);
const blasint m2 = *m - m1;
const blasint mn1 = MIN(m1, n1);
const blasint mn2 = MIN(m2, n2);
// Ab_L *
// Ab_BR
float *const Ab_L = Ab;
float *const Ab_BR = Ab + 2 * *ldAb * n1;
// A_L A_R
float *const A_L = A;
float *const A_R = A + 2 * *ldA * n1;
// A_TL A_TR
// A_BL A_BR
float *const A_TL = A;
float *const A_TR = A + 2 * *ldA * n1;
float *const A_BL = A + 2 * m1;
float *const A_BR = A + 2 * *ldA * n1 + 2 * m1;
// ipiv_T
// ipiv_B
blasint *const ipiv_T = ipiv;
blasint *const ipiv_B = ipiv + n1;
// Banded splitting
const blasint n21 = MIN(n2, kv - n1);
const blasint n22 = MIN(n2 - n21, n1);
const blasint m21 = MIN(m2, *kl - m1);
const blasint m22 = MIN(m2 - m21, m1);
// n1 n21 n22
// m * A_Rl ARr
float *const A_Rl = A_R;
float *const A_Rr = A_R + 2 * *ldA * n21;
// n1 n21 n22
// m1 * A_TRl A_TRr
// m21 A_BLt A_BRtl A_BRtr
// m22 A_BLb A_BRbl A_BRbr
float *const A_TRl = A_TR;
float *const A_TRr = A_TR + 2 * *ldA * n21;
float *const A_BLt = A_BL;
float *const A_BLb = A_BL + 2 * m21;
float *const A_BRtl = A_BR;
float *const A_BRtr = A_BR + 2 * *ldA * n21;
float *const A_BRbl = A_BR + 2 * m21;
float *const A_BRbr = A_BR + 2 * *ldA * n21 + 2 * m21;
// recursion(Ab_L, ipiv_T)
RELAPACK_cgbtrf_rec(m, &n1, kl, ku, Ab_L, ldAb, ipiv_T, Workl, ldWorkl, Worku, ldWorku, info);
// Workl = A_BLb
LAPACK(clacpy)("U", &m22, &n1, A_BLb, ldA, Workl, ldWorkl);
// partially redo swaps in A_L
for (i = 0; i < mn1; i++) {
const blasint ip = ipiv_T[i] - 1;
if (ip != i) {
if (ip < *kl)
BLAS(cswap)(&i, A_L + 2 * i, ldA, A_L + 2 * ip, ldA);
else
BLAS(cswap)(&i, A_L + 2 * i, ldA, Workl + 2 * (ip - *kl), ldWorkl);
}
}
// apply pivots to A_Rl
LAPACK(claswp)(&n21, A_Rl, ldA, iONE, &mn1, ipiv_T, iONE);
// apply pivots to A_Rr columnwise
for (j = 0; j < n22; j++) {
float *const A_Rrj = A_Rr + 2 * *ldA * j;
for (i = j; i < mn1; i++) {
const blasint ip = ipiv_T[i] - 1;
if (ip != i) {
const float tmpr = A_Rrj[2 * i];
const float tmpc = A_Rrj[2 * i + 1];
A_Rrj[2 * i] = A_Rrj[2 * ip];
A_Rrj[2 * i + 1] = A_Rr[2 * ip + 1];
A_Rrj[2 * ip] = tmpr;
A_Rrj[2 * ip + 1] = tmpc;
}
}
}
// A_TRl = A_TL \ A_TRl
BLAS(ctrsm)("L", "L", "N", "U", &m1, &n21, ONE, A_TL, ldA, A_TRl, ldA);
// Worku = A_TRr
LAPACK(clacpy)("L", &m1, &n22, A_TRr, ldA, Worku, ldWorku);
// Worku = A_TL \ Worku
BLAS(ctrsm)("L", "L", "N", "U", &m1, &n22, ONE, A_TL, ldA, Worku, ldWorku);
// A_TRr = Worku
LAPACK(clacpy)("L", &m1, &n22, Worku, ldWorku, A_TRr, ldA);
// A_BRtl = A_BRtl - A_BLt * A_TRl
BLAS(cgemm)("N", "N", &m21, &n21, &n1, MONE, A_BLt, ldA, A_TRl, ldA, ONE, A_BRtl, ldA);
// A_BRbl = A_BRbl - Workl * A_TRl
BLAS(cgemm)("N", "N", &m22, &n21, &n1, MONE, Workl, ldWorkl, A_TRl, ldA, ONE, A_BRbl, ldA);
// A_BRtr = A_BRtr - A_BLt * Worku
BLAS(cgemm)("N", "N", &m21, &n22, &n1, MONE, A_BLt, ldA, Worku, ldWorku, ONE, A_BRtr, ldA);
// A_BRbr = A_BRbr - Workl * Worku
BLAS(cgemm)("N", "N", &m22, &n22, &n1, MONE, Workl, ldWorkl, Worku, ldWorku, ONE, A_BRbr, ldA);
// partially undo swaps in A_L
for (i = mn1 - 1; i >= 0; i--) {
const blasint ip = ipiv_T[i] - 1;
if (ip != i) {
if (ip < *kl)
BLAS(cswap)(&i, A_L + 2 * i, ldA, A_L + 2 * ip, ldA);
else
BLAS(cswap)(&i, A_L + 2 * i, ldA, Workl + 2 * (ip - *kl), ldWorkl);
}
}
// recursion(Ab_BR, ipiv_B)
//RELAPACK_cgbtrf_rec(&m2, &n2, kl, ku, Ab_BR, ldAb, ipiv_B, Workl, ldWorkl, Worku, ldWorku, info);
LAPACK(cgbtf2)(&m2, &n2, kl, ku, Ab_BR, ldAb, ipiv_B, info);
if (*info)
*info += n1;
// shift pivots
for (i = 0; i < mn2; i++)
ipiv_B[i] += n1;
}