00001 /* sla_geamv.f -- translated by f2c (version 20061008). 00002 You must link the resulting object file with libf2c: 00003 on Microsoft Windows system, link with libf2c.lib; 00004 on Linux or Unix systems, link with .../path/to/libf2c.a -lm 00005 or, if you install libf2c.a in a standard place, with -lf2c -lm 00006 -- in that order, at the end of the command line, as in 00007 cc *.o -lf2c -lm 00008 Source for libf2c is in /netlib/f2c/libf2c.zip, e.g., 00009 00010 http://www.netlib.org/f2c/libf2c.zip 00011 */ 00012 00013 #include "f2c.h" 00014 #include "blaswrap.h" 00015 00016 /* Subroutine */ int sla_geamv__(integer *trans, integer *m, integer *n, real 00017 *alpha, real *a, integer *lda, real *x, integer *incx, real *beta, 00018 real *y, integer *incy) 00019 { 00020 /* System generated locals */ 00021 integer a_dim1, a_offset, i__1, i__2; 00022 real r__1; 00023 00024 /* Builtin functions */ 00025 double r_sign(real *, real *); 00026 00027 /* Local variables */ 00028 extern integer ilatrans_(char *); 00029 integer i__, j; 00030 logical symb_zero__; 00031 integer iy, jx, kx, ky, info; 00032 real temp; 00033 integer lenx, leny; 00034 real safe1; 00035 extern doublereal slamch_(char *); 00036 extern /* Subroutine */ int xerbla_(char *, integer *); 00037 00038 00039 /* -- LAPACK routine (version 3.2) -- */ 00040 /* -- Contributed by James Demmel, Deaglan Halligan, Yozo Hida and -- */ 00041 /* -- Jason Riedy of Univ. of California Berkeley. -- */ 00042 /* -- November 2008 -- */ 00043 00044 /* -- LAPACK is a software package provided by Univ. of Tennessee, -- */ 00045 /* -- Univ. of California Berkeley and NAG Ltd. -- */ 00046 00047 /* .. */ 00048 /* .. Scalar Arguments .. */ 00049 /* .. */ 00050 /* .. Array Arguments .. */ 00051 /* .. */ 00052 00053 /* Purpose */ 00054 /* ======= */ 00055 00056 /* SLA_GEAMV performs one of the matrix-vector operations */ 00057 00058 /* y := alpha*abs(A)*abs(x) + beta*abs(y), */ 00059 /* or y := alpha*abs(A)'*abs(x) + beta*abs(y), */ 00060 00061 /* where alpha and beta are scalars, x and y are vectors and A is an */ 00062 /* m by n matrix. */ 00063 00064 /* This function is primarily used in calculating error bounds. */ 00065 /* To protect against underflow during evaluation, components in */ 00066 /* the resulting vector are perturbed away from zero by (N+1) */ 00067 /* times the underflow threshold. To prevent unnecessarily large */ 00068 /* errors for block-structure embedded in general matrices, */ 00069 /* "symbolically" zero components are not perturbed. A zero */ 00070 /* entry is considered "symbolic" if all multiplications involved */ 00071 /* in computing that entry have at least one zero multiplicand. */ 00072 00073 /* Parameters */ 00074 /* ========== */ 00075 00076 /* TRANS - INTEGER */ 00077 /* On entry, TRANS specifies the operation to be performed as */ 00078 /* follows: */ 00079 00080 /* BLAS_NO_TRANS y := alpha*abs(A)*abs(x) + beta*abs(y) */ 00081 /* BLAS_TRANS y := alpha*abs(A')*abs(x) + beta*abs(y) */ 00082 /* BLAS_CONJ_TRANS y := alpha*abs(A')*abs(x) + beta*abs(y) */ 00083 00084 /* Unchanged on exit. */ 00085 00086 /* M - INTEGER */ 00087 /* On entry, M specifies the number of rows of the matrix A. */ 00088 /* M must be at least zero. */ 00089 /* Unchanged on exit. */ 00090 00091 /* N - INTEGER */ 00092 /* On entry, N specifies the number of columns of the matrix A. */ 00093 /* N must be at least zero. */ 00094 /* Unchanged on exit. */ 00095 00096 /* ALPHA - REAL */ 00097 /* On entry, ALPHA specifies the scalar alpha. */ 00098 /* Unchanged on exit. */ 00099 00100 /* A - REAL array of DIMENSION ( LDA, n ) */ 00101 /* Before entry, the leading m by n part of the array A must */ 00102 /* contain the matrix of coefficients. */ 00103 /* Unchanged on exit. */ 00104 00105 /* LDA - INTEGER */ 00106 /* On entry, LDA specifies the first dimension of A as declared */ 00107 /* in the calling (sub) program. LDA must be at least */ 00108 /* max( 1, m ). */ 00109 /* Unchanged on exit. */ 00110 00111 /* X - REAL array of DIMENSION at least */ 00112 /* ( 1 + ( n - 1 )*abs( INCX ) ) when TRANS = 'N' or 'n' */ 00113 /* and at least */ 00114 /* ( 1 + ( m - 1 )*abs( INCX ) ) otherwise. */ 00115 /* Before entry, the incremented array X must contain the */ 00116 /* vector x. */ 00117 /* Unchanged on exit. */ 00118 00119 /* INCX - INTEGER */ 00120 /* On entry, INCX specifies the increment for the elements of */ 00121 /* X. INCX must not be zero. */ 00122 /* Unchanged on exit. */ 00123 00124 /* BETA - REAL */ 00125 /* On entry, BETA specifies the scalar beta. When BETA is */ 00126 /* supplied as zero then Y need not be set on input. */ 00127 /* Unchanged on exit. */ 00128 00129 /* Y - REAL */ 00130 /* Array of DIMENSION at least */ 00131 /* ( 1 + ( m - 1 )*abs( INCY ) ) when TRANS = 'N' or 'n' */ 00132 /* and at least */ 00133 /* ( 1 + ( n - 1 )*abs( INCY ) ) otherwise. */ 00134 /* Before entry with BETA non-zero, the incremented array Y */ 00135 /* must contain the vector y. On exit, Y is overwritten by the */ 00136 /* updated vector y. */ 00137 00138 /* INCY - INTEGER */ 00139 /* On entry, INCY specifies the increment for the elements of */ 00140 /* Y. INCY must not be zero. */ 00141 /* Unchanged on exit. */ 00142 00143 /* Level 2 Blas routine. */ 00144 00145 /* .. */ 00146 /* .. Parameters .. */ 00147 /* .. */ 00148 /* .. Local Scalars .. */ 00149 /* .. */ 00150 /* .. External Subroutines .. */ 00151 /* .. */ 00152 /* .. External Functions .. */ 00153 /* .. */ 00154 /* .. Intrinsic Functions .. */ 00155 /* .. */ 00156 /* .. Executable Statements .. */ 00157 00158 /* Test the input parameters. */ 00159 00160 /* Parameter adjustments */ 00161 a_dim1 = *lda; 00162 a_offset = 1 + a_dim1; 00163 a -= a_offset; 00164 --x; 00165 --y; 00166 00167 /* Function Body */ 00168 info = 0; 00169 if (! (*trans == ilatrans_("N") || *trans == ilatrans_("T") || *trans == ilatrans_("C"))) { 00170 info = 1; 00171 } else if (*m < 0) { 00172 info = 2; 00173 } else if (*n < 0) { 00174 info = 3; 00175 } else if (*lda < max(1,*m)) { 00176 info = 6; 00177 } else if (*incx == 0) { 00178 info = 8; 00179 } else if (*incy == 0) { 00180 info = 11; 00181 } 00182 if (info != 0) { 00183 xerbla_("SLA_GEAMV ", &info); 00184 return 0; 00185 } 00186 00187 /* Quick return if possible. */ 00188 00189 if (*m == 0 || *n == 0 || *alpha == 0.f && *beta == 1.f) { 00190 return 0; 00191 } 00192 00193 /* Set LENX and LENY, the lengths of the vectors x and y, and set */ 00194 /* up the start points in X and Y. */ 00195 00196 if (*trans == ilatrans_("N")) { 00197 lenx = *n; 00198 leny = *m; 00199 } else { 00200 lenx = *m; 00201 leny = *n; 00202 } 00203 if (*incx > 0) { 00204 kx = 1; 00205 } else { 00206 kx = 1 - (lenx - 1) * *incx; 00207 } 00208 if (*incy > 0) { 00209 ky = 1; 00210 } else { 00211 ky = 1 - (leny - 1) * *incy; 00212 } 00213 00214 /* Set SAFE1 essentially to be the underflow threshold times the */ 00215 /* number of additions in each row. */ 00216 00217 safe1 = slamch_("Safe minimum"); 00218 safe1 = (*n + 1) * safe1; 00219 00220 /* Form y := alpha*abs(A)*abs(x) + beta*abs(y). */ 00221 00222 /* The O(M*N) SYMB_ZERO tests could be replaced by O(N) queries to */ 00223 /* the inexact flag. Still doesn't help change the iteration order */ 00224 /* to per-column. */ 00225 00226 iy = ky; 00227 if (*incx == 1) { 00228 i__1 = leny; 00229 for (i__ = 1; i__ <= i__1; ++i__) { 00230 if (*beta == 0.f) { 00231 symb_zero__ = TRUE_; 00232 y[iy] = 0.f; 00233 } else if (y[iy] == 0.f) { 00234 symb_zero__ = TRUE_; 00235 } else { 00236 symb_zero__ = FALSE_; 00237 y[iy] = *beta * (r__1 = y[iy], dabs(r__1)); 00238 } 00239 if (*alpha != 0.f) { 00240 i__2 = lenx; 00241 for (j = 1; j <= i__2; ++j) { 00242 if (*trans == ilatrans_("N")) { 00243 temp = (r__1 = a[i__ + j * a_dim1], dabs(r__1)); 00244 } else { 00245 temp = (r__1 = a[j + i__ * a_dim1], dabs(r__1)); 00246 } 00247 symb_zero__ = symb_zero__ && (x[j] == 0.f || temp == 0.f); 00248 y[iy] += *alpha * (r__1 = x[j], dabs(r__1)) * temp; 00249 } 00250 } 00251 if (! symb_zero__) { 00252 y[iy] += r_sign(&safe1, &y[iy]); 00253 } 00254 iy += *incy; 00255 } 00256 } else { 00257 i__1 = leny; 00258 for (i__ = 1; i__ <= i__1; ++i__) { 00259 if (*beta == 0.f) { 00260 symb_zero__ = TRUE_; 00261 y[iy] = 0.f; 00262 } else if (y[iy] == 0.f) { 00263 symb_zero__ = TRUE_; 00264 } else { 00265 symb_zero__ = FALSE_; 00266 y[iy] = *beta * (r__1 = y[iy], dabs(r__1)); 00267 } 00268 if (*alpha != 0.f) { 00269 jx = kx; 00270 i__2 = lenx; 00271 for (j = 1; j <= i__2; ++j) { 00272 if (*trans == ilatrans_("N")) { 00273 temp = (r__1 = a[i__ + j * a_dim1], dabs(r__1)); 00274 } else { 00275 temp = (r__1 = a[j + i__ * a_dim1], dabs(r__1)); 00276 } 00277 symb_zero__ = symb_zero__ && (x[jx] == 0.f || temp == 0.f) 00278 ; 00279 y[iy] += *alpha * (r__1 = x[jx], dabs(r__1)) * temp; 00280 jx += *incx; 00281 } 00282 } 00283 if (! symb_zero__) { 00284 y[iy] += r_sign(&safe1, &y[iy]); 00285 } 00286 iy += *incy; 00287 } 00288 } 00289 00290 return 0; 00291 00292 /* End of SLA_GEAMV */ 00293 00294 } /* sla_geamv__ */