1 #include "relapack.h"
2
3 static void RELAPACK_slauum_rec(const char *, const blasint *, float *,
4 const blasint *, blasint *);
5
6
7 /** SLAUUM computes the product U * U**T or L**T * L, where the triangular factor U or L is stored in the upper or lower triangular part of the array A.
8 *
9 * This routine is functionally equivalent to LAPACK's slauum.
10 * For details on its interface, see
11 * http://www.netlib.org/lapack/explore-html/dd/d5a/slauum_8f.html
12 * */
RELAPACK_slauum(const char * uplo,const blasint * n,float * A,const blasint * ldA,blasint * info)13 void RELAPACK_slauum(
14 const char *uplo, const blasint *n,
15 float *A, const blasint *ldA,
16 blasint *info
17 ) {
18
19 // Check arguments
20 const blasint lower = LAPACK(lsame)(uplo, "L");
21 const blasint upper = LAPACK(lsame)(uplo, "U");
22 *info = 0;
23 if (!lower && !upper)
24 *info = -1;
25 else if (*n < 0)
26 *info = -2;
27 else if (*ldA < MAX(1, *n))
28 *info = -4;
29 if (*info) {
30 const blasint minfo = -*info;
31 LAPACK(xerbla)("SLAUUM", &minfo, strlen("SLAUUM"));
32 return;
33 }
34
35 // Clean char * arguments
36 const char cleanuplo = lower ? 'L' : 'U';
37
38 // Recursive kernel
39 RELAPACK_slauum_rec(&cleanuplo, n, A, ldA, info);
40 }
41
42
43 /** slauum's recursive compute kernel */
RELAPACK_slauum_rec(const char * uplo,const blasint * n,float * A,const blasint * ldA,blasint * info)44 static void RELAPACK_slauum_rec(
45 const char *uplo, const blasint *n,
46 float *A, const blasint *ldA,
47 blasint *info
48 ) {
49
50 if (*n <= MAX(CROSSOVER_SLAUUM, 1)) {
51 // Unblocked
52 LAPACK(slauu2)(uplo, n, A, ldA, info);
53 return;
54 }
55
56 // Constants
57 const float ONE[] = { 1. };
58
59 // Splitting
60 const blasint n1 = SREC_SPLIT(*n);
61 const blasint n2 = *n - n1;
62
63 // A_TL A_TR
64 // A_BL A_BR
65 float *const A_TL = A;
66 float *const A_TR = A + *ldA * n1;
67 float *const A_BL = A + n1;
68 float *const A_BR = A + *ldA * n1 + n1;
69
70 // recursion(A_TL)
71 RELAPACK_slauum_rec(uplo, &n1, A_TL, ldA, info);
72
73 if (*uplo == 'L') {
74 // A_TL = A_TL + A_BL' * A_BL
75 BLAS(ssyrk)("L", "T", &n1, &n2, ONE, A_BL, ldA, ONE, A_TL, ldA);
76 // A_BL = A_BR' * A_BL
77 BLAS(strmm)("L", "L", "T", "N", &n2, &n1, ONE, A_BR, ldA, A_BL, ldA);
78 } else {
79 // A_TL = A_TL + A_TR * A_TR'
80 BLAS(ssyrk)("U", "N", &n1, &n2, ONE, A_TR, ldA, ONE, A_TL, ldA);
81 // A_TR = A_TR * A_BR'
82 BLAS(strmm)("R", "U", "T", "N", &n1, &n2, ONE, A_BR, ldA, A_TR, ldA);
83 }
84
85 // recursion(A_BR)
86 RELAPACK_slauum_rec(uplo, &n2, A_BR, ldA, info);
87 }
88