[2003] | 1 | MODULE lib_fortran |
---|
| 2 | !!====================================================================== |
---|
| 3 | !! *** MODULE lib_fortran *** |
---|
| 4 | !! Fortran utilities: includes some low levels fortran functionality |
---|
| 5 | !!====================================================================== |
---|
| 6 | !! History : 3.2 ! 2010-05 Michael Dunphy, Rachid BENSHILA Original code |
---|
| 7 | !!---------------------------------------------------------------------- |
---|
| 8 | !! OPA 9.0 , LOCEAN-IPSL (2005) |
---|
| 9 | !! $Id: $ |
---|
| 10 | !! This software is governed by the CeCILL licence see modipsl/doc/NEMO_CeCILL.txt |
---|
| 11 | !!---------------------------------------------------------------------- |
---|
| 12 | USE par_oce |
---|
| 13 | USE par_kind |
---|
| 14 | USE lib_mpp ! distributed memory computing |
---|
| 15 | USE dom_oce |
---|
| 16 | USE in_out_manager |
---|
| 17 | |
---|
| 18 | IMPLICIT NONE |
---|
| 19 | PRIVATE |
---|
| 20 | |
---|
| 21 | PUBLIC glob_sum |
---|
| 22 | #if defined key_nosignedzeo |
---|
| 23 | PUBLIC SIGN |
---|
| 24 | #endif |
---|
| 25 | |
---|
| 26 | INTERFACE glob_sum |
---|
| 27 | #if defined key_mpp_rep1 |
---|
| 28 | MODULE PROCEDURE mpp_sum_indep |
---|
| 29 | #elif defined key_mpp_rep2 |
---|
| 30 | MODULE PROCEDURE mpp_sum_cmpx |
---|
| 31 | #else |
---|
| 32 | MODULE PROCEDURE glob_sum_2d, glob_sum_3d,glob_sum_2d_a, glob_sum_3d_a |
---|
| 33 | #endif |
---|
| 34 | END INTERFACE |
---|
| 35 | |
---|
| 36 | #if defined key_nosignedzeo |
---|
| 37 | INTERFACE SIGN |
---|
| 38 | MODULE PROCEDURE SIGN_SCALAR, SIGN_ARRAY_1D, SIGN_ARRAY_2D, SIGN_ARRAY_3D, & |
---|
| 39 | SIGN_ARRAY_1D_A, SIGN_ARRAY_2D_A, SIGN_ARRAY_3D_A, & |
---|
| 40 | SIGN_ARRAY_1D_B, SIGN_ARRAY_2D_B, SIGN_ARRAY_3D_B |
---|
| 41 | END INTERFACE |
---|
| 42 | #endif |
---|
| 43 | |
---|
| 44 | CONTAINS |
---|
| 45 | |
---|
| 46 | FUNCTION glob_sum_2d( ptab ) |
---|
| 47 | !!----------------------------------------------------------------------- |
---|
| 48 | !! *** FUNCTION glob_sum_2D *** |
---|
| 49 | !! |
---|
| 50 | !! ** Purpose : perform a sum on the global domain of a 2D array |
---|
| 51 | !!----------------------------------------------------------------------- |
---|
| 52 | REAL(wp), DIMENSION(:,:),INTENT(in) :: ptab |
---|
| 53 | REAL(wp) :: glob_sum_2d |
---|
| 54 | !!----------------------------------------------------------------------- |
---|
| 55 | |
---|
| 56 | glob_sum_2d = SUM( ptab(:,:)*tmask_i(:,:) ) |
---|
| 57 | IF( lk_mpp ) CALL mpp_sum( glob_sum_2d ) |
---|
| 58 | |
---|
| 59 | END FUNCTION glob_sum_2d |
---|
| 60 | |
---|
| 61 | FUNCTION glob_sum_3d( ptab ) |
---|
| 62 | !!----------------------------------------------------------------------- |
---|
| 63 | !! *** FUNCTION glob_sum_3D *** |
---|
| 64 | !! |
---|
| 65 | !! ** Purpose : perform a sum on the global domain of a 3D array |
---|
| 66 | !!----------------------------------------------------------------------- |
---|
| 67 | REAL(wp), DIMENSION(:,:,:) :: ptab |
---|
| 68 | REAL(wp) :: glob_sum_3d |
---|
| 69 | ! |
---|
| 70 | INTEGER :: jk |
---|
| 71 | !!----------------------------------------------------------------------- |
---|
| 72 | |
---|
| 73 | GLOB_SUM_3D = 0.e0 |
---|
| 74 | DO jk = 1, jpk |
---|
| 75 | glob_sum_3d = glob_sum_3d + SUM( ptab(:,:,jk)*tmask_i(:,:) ) |
---|
| 76 | END DO |
---|
| 77 | IF( lk_mpp ) CALL mpp_sum( glob_sum_3d ) |
---|
| 78 | |
---|
| 79 | END FUNCTION glob_sum_3d |
---|
| 80 | |
---|
| 81 | FUNCTION glob_sum_2d_a( ptab1, ptab2 ) |
---|
| 82 | !!----------------------------------------------------------------------- |
---|
| 83 | !! *** FUNCTION glob_sum_2D _a *** |
---|
| 84 | !! |
---|
| 85 | !! ** Purpose : perform a sum on the global domain of two 2D array |
---|
| 86 | !!----------------------------------------------------------------------- |
---|
| 87 | REAL(wp), DIMENSION(:,:) :: ptab1, ptab2 |
---|
| 88 | REAL(wp), DIMENSION(2) :: glob_sum_2d_a |
---|
| 89 | !!----------------------------------------------------------------------- |
---|
| 90 | |
---|
| 91 | glob_sum_2d_a(1) = SUM( ptab1(:,:)*tmask_i(:,:) ) |
---|
| 92 | glob_sum_2d_a(2) = SUM( ptab2(:,:)*tmask_i(:,:) ) |
---|
| 93 | IF( lk_mpp ) CALL mpp_sum( glob_sum_2d_a,2 ) |
---|
| 94 | |
---|
| 95 | END FUNCTION glob_sum_2d_a |
---|
| 96 | |
---|
| 97 | FUNCTION glob_sum_3d_a( ptab1, ptab2 ) |
---|
| 98 | !!----------------------------------------------------------------------- |
---|
| 99 | !! *** FUNCTION glob_sum_3D_a *** |
---|
| 100 | !! |
---|
| 101 | !! ** Purpose : perform a sum on the global domain of two 3D array |
---|
| 102 | !!----------------------------------------------------------------------- |
---|
| 103 | REAL(wp), DIMENSION(:,:,:) :: ptab1, ptab2 |
---|
| 104 | REAL(wp), DIMENSION(2) :: glob_sum_3d_a |
---|
| 105 | ! |
---|
| 106 | INTEGER :: jk |
---|
| 107 | !!----------------------------------------------------------------------- |
---|
| 108 | |
---|
| 109 | glob_sum_3d_a(:) = 0.e0 |
---|
| 110 | DO jk = 1, jpk |
---|
| 111 | glob_sum_3d_a(1) = glob_sum_3d_a(1) + SUM( ptab1(:,:,jk)*tmask_i(:,:) ) |
---|
| 112 | glob_sum_3d_a(2) = glob_sum_3d_a(2) + SUM( ptab2(:,:,jk)*tmask_i(:,:) ) |
---|
| 113 | END DO |
---|
| 114 | IF( lk_mpp ) CALL mpp_sum( glob_sum_3d_a,2 ) |
---|
| 115 | |
---|
| 116 | END FUNCTION glob_sum_3d_a |
---|
| 117 | |
---|
| 118 | #if defined key_mpp_rep2 |
---|
| 119 | FUNCTION mpp_sum_cmpx( pval ) |
---|
| 120 | !!---------------------------------------------------------------------- |
---|
| 121 | !! *** FUNCTION mpp_sum_cmpx *** |
---|
| 122 | !! |
---|
| 123 | !! ** Purpose : perform a sum in calling DDPDD routine |
---|
| 124 | !! |
---|
| 125 | !!---------------------------------------------------------------------- |
---|
| 126 | REAL(wp) :: mpp_sum_cmpx |
---|
| 127 | ! |
---|
| 128 | REAL(wp), DIMENSION(jpi,jpj), INTENT(IN) :: & |
---|
| 129 | & pval |
---|
| 130 | COMPLEX(wp):: ctmp |
---|
| 131 | REAL(wp) ::ztmp |
---|
| 132 | INTEGER :: ji,jj |
---|
| 133 | !!----------------------------------------------------------------------- |
---|
| 134 | |
---|
| 135 | ztmp = 0.e0 |
---|
| 136 | ctmp = CMPLX(0.e0,0.e0,wp) |
---|
| 137 | DO jj = 1,jpj |
---|
| 138 | DO ji =1, jpi |
---|
| 139 | ztmp = pval(ji,jj) * tmask_i(ji,jj) |
---|
| 140 | CALL DDPDD(CMPLX(ztmp,0.e0,wp),ctmp) |
---|
| 141 | END DO |
---|
| 142 | END DO |
---|
| 143 | IF( lk_mpp ) CALL mpp_sum( ctmp ) ! sum over the global domain |
---|
| 144 | mpp_sum_cmpx= REAL(ctmp,wp) |
---|
| 145 | |
---|
| 146 | END FUNCTION mpp_sum_cmpx |
---|
| 147 | |
---|
| 148 | SUBROUTINE DDPDD( ydda, yddb ) |
---|
| 149 | !!---------------------------------------------------------------------- |
---|
| 150 | !! *** ROUTINE DDPDD *** |
---|
| 151 | !! |
---|
| 152 | !! ** Purpose : Add a scalar element to a sum |
---|
| 153 | !! |
---|
| 154 | !! |
---|
| 155 | !! ** Method : The code uses the compensated summation with doublet |
---|
| 156 | !! (sum,error) emulated useing complex numbers. ydda is the |
---|
| 157 | !! scalar to add to the summ yddb |
---|
| 158 | !! |
---|
| 159 | !! ** Action : This does only work for MPI. |
---|
| 160 | !! |
---|
| 161 | !! References : Using Acurate Arithmetics to Improve Numerical |
---|
| 162 | !! Reproducibility and Sability in Parallel Applications |
---|
| 163 | !! Yun HE and Chris H. Q. DING, Journal of Supercomputing |
---|
| 164 | !! 18, 259-277, 2001 |
---|
| 165 | !!---------------------------------------------------------------------- |
---|
| 166 | |
---|
| 167 | COMPLEX(wp), INTENT(in) :: ydda |
---|
| 168 | COMPLEX(wp), INTENT(inout) :: yddb |
---|
| 169 | |
---|
| 170 | REAL(wp) :: zerr, zt1, zt2 ! local work variables |
---|
| 171 | |
---|
| 172 | ! Compute ydda + yddb using Knuth's trick. |
---|
| 173 | zt1 = real(ydda) + real(yddb) |
---|
| 174 | zerr = zt1 - real(ydda) |
---|
| 175 | zt2 = ((real(yddb) - zerr) + (real(ydda) - (zt1 - zerr))) & |
---|
| 176 | + aimag(ydda) + aimag(yddb) |
---|
| 177 | |
---|
| 178 | ! The result is t1 + t2, after normalization. |
---|
| 179 | yddb = cmplx ( zt1 + zt2, zt2 - ((zt1 + zt2) - zt1),wp ) |
---|
| 180 | |
---|
| 181 | END SUBROUTINE DDPDD |
---|
| 182 | #endif |
---|
| 183 | |
---|
| 184 | #if defined key_mpp_rep1 |
---|
| 185 | FUNCTION mpp_sum_indep( pval ) |
---|
| 186 | !!---------------------------------------------------------------------- |
---|
| 187 | !! *** ROUTINE mpp_sum_indep *** |
---|
| 188 | !! |
---|
| 189 | !! ** Purpose : Sum all elements in the pval array in |
---|
| 190 | !! an accurate order-independent way. |
---|
| 191 | !! |
---|
| 192 | !! ** Method : The code iterates the compensated summation until the |
---|
| 193 | !! result is guaranteed to be within 4*eps of the true sum. |
---|
| 194 | !! It then rounds the result to the nearest floating-point |
---|
| 195 | !! number whose last three bits are zero, thereby |
---|
| 196 | !! guaranteeing an order-independent result. |
---|
| 197 | !! |
---|
| 198 | !! ** Action : This does only work for MPI. |
---|
| 199 | !! It does not work for SHMEM. !! |
---|
| 200 | !! References : M. Fisher (ECMWF): IFS code + personal communication |
---|
| 201 | !! The algorithm is based on Ogita et al. (2005) |
---|
| 202 | !! SIAM J. Sci. Computing, Vol.26, No.6, pp1955-1988. |
---|
| 203 | !! This is based in turn on an algorithm |
---|
| 204 | !! by Knuth (1969, seminumerical algorithms). |
---|
| 205 | !! |
---|
| 206 | !! History : |
---|
| 207 | !! ! 07-07 (K. Mogensen) Original code heavily based on IFS. |
---|
| 208 | !!---------------------------------------------------------------------- |
---|
| 209 | REAL(wp) mpp_sum_indep |
---|
| 210 | REAL(wp), DIMENSION(jpi,jpj), INTENT(IN) :: pval |
---|
| 211 | ! |
---|
| 212 | REAL(wp), DIMENSION(3) :: zbuffl |
---|
| 213 | REAL(wp), DIMENSION(:), ALLOCATABLE :: zpsums, zperrs, zpcors, zbuffg, zp |
---|
| 214 | REAL(wp) :: zcorr, zerr, zolderr, zbeta, zres |
---|
| 215 | INTEGER, DIMENSION(:), allocatable :: irecv, istart |
---|
| 216 | INTEGER :: ikn, jj |
---|
| 217 | |
---|
| 218 | ! initialise to avoid uninitialised variables trapping of some compilers to complain. |
---|
| 219 | zres = 0.0_wp ; zerr = 0.0_wp ; zbuffl(:) = 0.0_wp |
---|
| 220 | ! Get global number of elements |
---|
| 221 | ikn = SIZE(pval) |
---|
| 222 | # ifdef key_mpp |
---|
| 223 | CALL mpp_sum( ikn ) |
---|
| 224 | # endif |
---|
| 225 | ! Check that the the algorithm can work |
---|
| 226 | |
---|
| 227 | IF ( ( REAL( 2 * ikn ) * EPSILON( zres ) ) >= 1.0 ) THEN |
---|
| 228 | CALL ctl_stop('mpp_sum_indep:', & |
---|
| 229 | & 'size of array is too large to guarantee error bounds') |
---|
| 230 | ENDIF |
---|
| 231 | |
---|
| 232 | ALLOCATE( & |
---|
| 233 | & zp(MAX(ikn,1)), & |
---|
| 234 | & zbuffg(jpnij*SIZE(zbuffl)), & |
---|
| 235 | & zpsums(jpnij), & |
---|
| 236 | & zperrs(jpnij), & |
---|
| 237 | & zpcors(jpnij) & |
---|
| 238 | & ) |
---|
| 239 | |
---|
| 240 | zolderr = HUGE(zerr) |
---|
| 241 | |
---|
| 242 | ! Copy the input array. This avoids some tricky indexing, at the |
---|
| 243 | ! expense of some inefficency. |
---|
| 244 | |
---|
| 245 | IF ( ikn > 0 ) THEN |
---|
| 246 | zp(:) = RESHAPE(pval, (/ jpi * jpj /) ) |
---|
| 247 | ELSE |
---|
| 248 | zp(1) = 0.0_wp |
---|
| 249 | ENDIF |
---|
| 250 | |
---|
| 251 | k_loop: DO |
---|
| 252 | |
---|
| 253 | ! Transform local arrays |
---|
| 254 | |
---|
| 255 | IF ( ikn > 0 ) THEN |
---|
| 256 | CALL comp_sum ( zp, ikn, zcorr, zerr ) |
---|
| 257 | ENDIF |
---|
| 258 | |
---|
| 259 | ! Gather partial sums and error bounds to all processors |
---|
| 260 | |
---|
| 261 | zbuffl(1) = zp(MAX(ikn,1)) |
---|
| 262 | |
---|
| 263 | IF ( ikn > 0 ) THEN |
---|
| 264 | zbuffl(2) = zerr |
---|
| 265 | zbuffl(3) = zcorr |
---|
| 266 | ELSE |
---|
| 267 | zbuffl(2) = 0.0_wp |
---|
| 268 | zbuffl(3) = 0.0_wp |
---|
| 269 | ENDIF |
---|
| 270 | |
---|
| 271 | IF ( jpnij > 1 ) THEN |
---|
| 272 | ALLOCATE( & |
---|
| 273 | & irecv(jpnij), & |
---|
| 274 | & istart(jpnij) & |
---|
| 275 | & ) |
---|
| 276 | CALL mpp_allgatherv( zbuffl, SIZE(zbuffl), & |
---|
| 277 | & zbuffg, jpnij * SIZE(zbuffl), irecv, istart ) |
---|
| 278 | DEALLOCATE( & |
---|
| 279 | & irecv, & |
---|
| 280 | & istart & |
---|
| 281 | & ) |
---|
| 282 | |
---|
| 283 | DO jj = 1, jpnij |
---|
| 284 | zpsums(jj) = zbuffg(1+(jj-1)*SIZE(zbuffl)) |
---|
| 285 | zperrs(jj) = zbuffg(2+(jj-1)*SIZE(zbuffl)) |
---|
| 286 | zpcors(jj) = zbuffg(3+(jj-1)*SIZE(zbuffl)) |
---|
| 287 | END DO |
---|
| 288 | |
---|
| 289 | ELSE |
---|
| 290 | zpsums(1) = zbuffl(1) |
---|
| 291 | zperrs(1) = zbuffl(2) |
---|
| 292 | zpcors(1) = zbuffl(3) |
---|
| 293 | ENDIF |
---|
| 294 | |
---|
| 295 | ! Transform partial sums |
---|
| 296 | CALL comp_sum( zpsums, jpnij, zcorr, zerr ) |
---|
| 297 | zerr = zerr + SUM(zperrs) |
---|
| 298 | zcorr = zcorr + SUM(zpcors) |
---|
| 299 | |
---|
| 300 | ! Calculate final result |
---|
| 301 | zres = zpsums(jpnij) + zcorr |
---|
| 302 | |
---|
| 303 | ! Calculate error bound. This is corollary 4.7 from Ogita et al. |
---|
| 304 | ! (2005) |
---|
| 305 | zbeta = zerr *( REAL( 2*ikn, wp ) * EPSILON(zres) ) & |
---|
| 306 | & /(1.0_wp - REAL( 2*ikn, wp ) * EPSILON(zres) ) |
---|
| 307 | |
---|
| 308 | zerr = EPSILON(zres) * ABS(zres) & |
---|
| 309 | & +(zbeta + ( 2.0_wp * EPSILON(zres) * EPSILON(zres) * ABS(zres) & |
---|
| 310 | & +3.0_wp * TINY(zres) ) ) |
---|
| 311 | |
---|
| 312 | ! Update the last element of the local array |
---|
| 313 | zp(MAX(ikn,1)) = zpsums(nproc+1) |
---|
| 314 | |
---|
| 315 | ! Exit if the global error is small enough |
---|
| 316 | IF ( zerr < 4.0_wp * SPACING(zres) ) EXIT k_loop |
---|
| 317 | |
---|
| 318 | ! Take appropriate action if ZRES cannot be sufficiently refined. |
---|
| 319 | IF (zerr >= zolderr) THEN |
---|
| 320 | CALL ctl_stop('Failed to refine sum', & |
---|
| 321 | & 'Warning: Possiblity of non-reproducible results') |
---|
| 322 | ENDIF |
---|
| 323 | |
---|
| 324 | zolderr = zerr |
---|
| 325 | |
---|
| 326 | ENDDO k_loop |
---|
| 327 | |
---|
| 328 | ! At this stage, we have guaranteed that ZRES less than 4*EPS |
---|
| 329 | ! away from the exact sum. There are only four floating point |
---|
| 330 | ! numbers in this range. So, if we find the nearest number that |
---|
| 331 | ! has its last three bits zero, then we have a reproducible result. |
---|
| 332 | |
---|
| 333 | mpp_sum_indep = fround(zres) |
---|
| 334 | |
---|
| 335 | DEALLOCATE( & |
---|
| 336 | & zpcors, & |
---|
| 337 | & zperrs, & |
---|
| 338 | & zpsums, & |
---|
| 339 | & zbuffg, & |
---|
| 340 | & zp & |
---|
| 341 | & ) |
---|
| 342 | |
---|
| 343 | END FUNCTION mpp_sum_indep |
---|
| 344 | |
---|
| 345 | SUBROUTINE comp_sum( pval, kn, pcorr, perr ) |
---|
| 346 | !!---------------------------------------------------------------------- |
---|
| 347 | !! *** ROUTINE comp_sum *** |
---|
| 348 | !! |
---|
| 349 | !! ** Purpose : To perform compensated (i.e. accurate) summation. |
---|
| 350 | !! |
---|
| 351 | !! ** Method : These routines transform the elements of the array P, |
---|
| 352 | !! such that: |
---|
| 353 | !! 1) pval(kn) contains sum(pval) |
---|
| 354 | !! 2) pval(1)...pval(kn-1) contain the rounding errors |
---|
| 355 | !! that were made in calculating sum(pval). |
---|
| 356 | !! 3) The exact sum of the elements of pval is unmodified. |
---|
| 357 | !! On return, pcorr contains the sum of the rounding errors, |
---|
| 358 | !! perr contains the sum of their absolute values. |
---|
| 359 | !! After calling this routine, an accurate sum of the |
---|
| 360 | !! elements of pval can be calculated as res=pval(n)+pcorr. |
---|
| 361 | !! |
---|
| 362 | !! ** Action : |
---|
| 363 | !! |
---|
| 364 | !! References : M. Fisher (ECMWF) IFS code + personal communications |
---|
| 365 | !! |
---|
| 366 | !! History : |
---|
| 367 | !! ! 07-07 (K. Mogensen) Original code heavily based on IFS |
---|
| 368 | !!-------------------------------------------------------------------- |
---|
| 369 | INTEGER, INTENT(IN) :: kn ! Number of elements in input array |
---|
| 370 | REAL(wp), DIMENSION(kn), INTENT(INOUT) :: pval ! Input array to be sum on input |
---|
| 371 | ! pval(kn) = sum (pval) on output |
---|
| 372 | ! pval(1)...pval(kn-1) = rounding errors on output |
---|
| 373 | REAL(wp) :: pcorr ! Sum of rounding errors |
---|
| 374 | REAL(wp) :: perr ! Sum of absolute rounding errors |
---|
| 375 | !! * Local declarations |
---|
| 376 | REAL(wp) :: zx, zz, zpsum |
---|
| 377 | INTEGER :: jj |
---|
| 378 | |
---|
| 379 | pcorr = 0.0_wp |
---|
| 380 | perr = 0.0_wp |
---|
| 381 | |
---|
| 382 | zpsum = pval(1) |
---|
| 383 | |
---|
| 384 | DO jj = 2, kn |
---|
| 385 | |
---|
| 386 | ! It is vital that these 4 lines are not optimized in any way that |
---|
| 387 | ! changes the results. |
---|
| 388 | zx = pval(jj) + zpsum |
---|
| 389 | zz = zx - pval(jj) |
---|
| 390 | pval(jj-1) = ( pval(jj) - ( zx - zz ) ) + ( zpsum - zz ) |
---|
| 391 | zpsum = zx |
---|
| 392 | |
---|
| 393 | ! Accumulate the correction and the error |
---|
| 394 | pcorr = pcorr + pval(jj-1) |
---|
| 395 | perr = perr + ABS( pval(jj-1) ) |
---|
| 396 | |
---|
| 397 | END DO |
---|
| 398 | |
---|
| 399 | pval(kn) = zpsum |
---|
| 400 | |
---|
| 401 | END SUBROUTINE comp_sum |
---|
| 402 | |
---|
| 403 | FUNCTION fround(pres) |
---|
| 404 | !!---------------------------------------------------------------------- |
---|
| 405 | !! *** ROUTINE fround *** |
---|
| 406 | !! |
---|
| 407 | !! ** Purpose : Rounding of floating-point number |
---|
| 408 | !! |
---|
| 409 | !! ** Method : Returns the value of PRES rounded to the nearest |
---|
| 410 | !! floating-point number that has its last three bits zero |
---|
| 411 | !! This works on big-endian and little-endian machines. |
---|
| 412 | !! |
---|
| 413 | !! ** Action : |
---|
| 414 | !! |
---|
| 415 | !! References : M. Fisher (ECMWF) IFS code + personal communication |
---|
| 416 | !! |
---|
| 417 | !! History : |
---|
| 418 | !! ! 07-07 (K. Mogensen) Original code heavily based on IFS. |
---|
| 419 | !!---------------------------------------------------------------------- |
---|
| 420 | REAL(wp) fround |
---|
| 421 | REAL(wp), INTENT(IN) :: pres ! Value to be rounded |
---|
| 422 | ! |
---|
| 423 | REAL(wp) :: zz(2), zup, zdown |
---|
| 424 | INTEGER :: ii(2), iequiv(8), ints_per_real, i_low_word |
---|
| 425 | INTEGER :: jj |
---|
| 426 | |
---|
| 427 | ii(:) = 1 |
---|
| 428 | zz(:) = 1.0_wp |
---|
| 429 | |
---|
| 430 | ! Warning: If wp = 64 bits (or 32 bits for key_sp) this will not work. |
---|
| 431 | |
---|
| 432 | #if defined key_sp |
---|
| 433 | ints_per_real = 32 / BIT_SIZE(ii) |
---|
| 434 | #else |
---|
| 435 | ints_per_real = 64 / BIT_SIZE(ii) |
---|
| 436 | #endif |
---|
| 437 | |
---|
| 438 | ! Test whether big-endian or little-endian |
---|
| 439 | |
---|
| 440 | zup = -1.0_wp |
---|
| 441 | iequiv(1:ints_per_real) = TRANSFER(zup,iequiv(1:ints_per_real)) |
---|
| 442 | |
---|
| 443 | IF ( iequiv(1) == 0 ) THEN |
---|
| 444 | i_low_word = 1 ! Little-endian |
---|
| 445 | ELSE |
---|
| 446 | i_low_word = ints_per_real ! Big-endian |
---|
| 447 | ENDIF |
---|
| 448 | |
---|
| 449 | ! Find the nearest number with all 3 lowest-order bits zeroed |
---|
| 450 | |
---|
| 451 | iequiv(1:ints_per_real) = transfer(pres,iequiv(1:ints_per_real)) |
---|
| 452 | zup = pres |
---|
| 453 | zdown = pres |
---|
| 454 | |
---|
| 455 | IF (IBITS(iequiv(i_low_word),0,3)/=0) THEN |
---|
| 456 | |
---|
| 457 | DO jj = 1, 4 |
---|
| 458 | |
---|
| 459 | zup = NEAREST( zup, 1.0_wp ) |
---|
| 460 | iequiv(1:ints_per_real) = TRANSFER( zup, iequiv(1:ints_per_real) ) |
---|
| 461 | IF ( IBITS( iequiv(i_low_word), 0, 3 ) == 0 ) EXIT |
---|
| 462 | |
---|
| 463 | zdown = NEAREST( zdown, -1.0 ) |
---|
| 464 | iequiv(1:ints_per_real) = TRANSFER( zdown, iequiv(1:ints_per_real)) |
---|
| 465 | IF ( IBITS( iequiv(i_low_word),0,3) == 0 ) EXIT |
---|
| 466 | |
---|
| 467 | END DO |
---|
| 468 | |
---|
| 469 | IF ( IBITS( iequiv( i_low_word ), 0, 3) /= 0 ) THEN |
---|
| 470 | CALL ctl_stop('Fround:','This is not possible') |
---|
| 471 | ENDIF |
---|
| 472 | |
---|
| 473 | ENDIF |
---|
| 474 | |
---|
| 475 | fround = TRANSFER( iequiv(1:ints_per_real), pres ) |
---|
| 476 | |
---|
| 477 | END FUNCTION fround |
---|
| 478 | #endif |
---|
| 479 | |
---|
| 480 | |
---|
| 481 | #if defined key_nosignedzero |
---|
| 482 | FUNCTION SIGN_SCALAR(pa,pb) |
---|
| 483 | !!----------------------------------------------------------------------- |
---|
| 484 | !! *** FUNCTION SIGN_SCALAR *** |
---|
| 485 | !! |
---|
| 486 | !! ** Purpose : overwrite f95 behaviour of intrinsinc sign function |
---|
| 487 | !!----------------------------------------------------------------------- |
---|
| 488 | REAL(wp) :: pa,pb ! input |
---|
| 489 | REAL(wp) :: SIGN_SCALAR ! result |
---|
| 490 | IF ( pb >= 0.e0) THEN |
---|
| 491 | SIGN_SCALAR = ABS(pa) |
---|
| 492 | ELSE |
---|
| 493 | SIGN_SCALAR =-ABS(pa) |
---|
| 494 | ENDIF |
---|
| 495 | |
---|
| 496 | END FUNCTION SIGN_SCALAR |
---|
| 497 | |
---|
| 498 | FUNCTION SIGN_ARRAY_1D(pa,pb) |
---|
| 499 | !!----------------------------------------------------------------------- |
---|
| 500 | !! *** FUNCTION SIGN_ARRAY_1D *** |
---|
| 501 | !! |
---|
| 502 | !! ** Purpose : overwrite f95 behaviour of intrinsinc sign function |
---|
| 503 | !!----------------------------------------------------------------------- |
---|
| 504 | REAL(wp) :: pa,pb(:) ! input |
---|
| 505 | REAL(wp) :: SIGN_ARRAY_1D(SIZE(pb,1)) ! result |
---|
| 506 | WHERE ( pb >= 0.e0 ) |
---|
| 507 | SIGN_ARRAY_1D = ABS(pa) |
---|
| 508 | ELSEWHERE |
---|
| 509 | SIGN_ARRAY_1D =-ABS(pa) |
---|
| 510 | END WHERE |
---|
| 511 | |
---|
| 512 | END FUNCTION SIGN_ARRAY_1D |
---|
| 513 | |
---|
| 514 | FUNCTION SIGN_ARRAY_2D(pa,pb) |
---|
| 515 | !!----------------------------------------------------------------------- |
---|
| 516 | !! *** FUNCTION SIGN_ARRAY_2D *** |
---|
| 517 | !! |
---|
| 518 | !! ** Purpose : overwrite f95 behaviour of intrinsinc sign function |
---|
| 519 | !!----------------------------------------------------------------------- |
---|
| 520 | REAL(wp) :: pa,pb(:,:) ! input |
---|
| 521 | REAL(wp) :: SIGN_ARRAY_2D(SIZE(pb,1),SIZE(pb,2)) ! result |
---|
| 522 | |
---|
| 523 | WHERE ( pb >= 0.e0 ) |
---|
| 524 | SIGN_ARRAY_2D = ABS(pa) |
---|
| 525 | ELSEWHERE |
---|
| 526 | SIGN_ARRAY_2D =-ABS(pa) |
---|
| 527 | END WHERE |
---|
| 528 | |
---|
| 529 | END FUNCTION SIGN_ARRAY_2D |
---|
| 530 | |
---|
| 531 | FUNCTION SIGN_ARRAY_3D(pa,pb) |
---|
| 532 | !!----------------------------------------------------------------------- |
---|
| 533 | !! *** FUNCTION SIGN_ARRAY_3D *** |
---|
| 534 | !! |
---|
| 535 | !! ** Purpose : overwrite f95 behaviour of intrinsinc sign function |
---|
| 536 | !!----------------------------------------------------------------------- |
---|
| 537 | REAL(wp) :: pa,pb(:,:,:) ! input |
---|
| 538 | REAL(wp) :: SIGN_ARRAY_3D(SIZE(pb,1),SIZE(pb,2),SIZE(pb,3)) ! result |
---|
| 539 | WHERE ( pb >= 0.e0 ) |
---|
| 540 | SIGN_ARRAY_3D = ABS(pa) |
---|
| 541 | ELSEWHERE |
---|
| 542 | SIGN_ARRAY_3D =-ABS(pa) |
---|
| 543 | END WHERE |
---|
| 544 | |
---|
| 545 | END FUNCTION SIGN_ARRAY_3D |
---|
| 546 | |
---|
| 547 | FUNCTION SIGN_ARRAY_1D_A(pa,pb) |
---|
| 548 | !!----------------------------------------------------------------------- |
---|
| 549 | !! *** FUNCTION SIGN_ARRAY_1D_A *** |
---|
| 550 | !! |
---|
| 551 | !! ** Purpose : overwrite f95 behaviour of intrinsinc sign function |
---|
| 552 | !!----------------------------------------------------------------------- |
---|
| 553 | REAL(wp) :: pa(:),pb(:) ! input |
---|
| 554 | REAL(wp) :: SIGN_ARRAY_1D_A(SIZE(b,1)) ! result |
---|
| 555 | |
---|
| 556 | WHERE ( pb >= 0.e0 ) |
---|
| 557 | SIGN_ARRAY_1D_A = ABS(pa) |
---|
| 558 | ELSEWHERE |
---|
| 559 | SIGN_ARRAY_1D_A =-ABS(pa) |
---|
| 560 | END WHERE |
---|
| 561 | |
---|
| 562 | END FUNCTION SIGN_ARRAY_1D_A |
---|
| 563 | |
---|
| 564 | FUNCTION SIGN_ARRAY_2D_A(pa,pb) |
---|
| 565 | !!----------------------------------------------------------------------- |
---|
| 566 | !! *** FUNCTION SIGN_ARRAY_2D_A *** |
---|
| 567 | !! |
---|
| 568 | !! ** Purpose : overwrite f95 behaviour of intrinsinc sign function |
---|
| 569 | !!----------------------------------------------------------------------- |
---|
| 570 | REAL(wp) :: pa(:,:),pb(:,:) ! input |
---|
| 571 | REAL(wp) :: SIGN_ARRAY_2D_A(SIZE(pb,1),SIZE(pb,2)) ! result |
---|
| 572 | |
---|
| 573 | WHERE ( pb >= 0.e0 ) |
---|
| 574 | SIGN_ARRAY_2D_A = ABS(pa) |
---|
| 575 | ELSEWHERE |
---|
| 576 | SIGN_ARRAY_2D_A =-ABS(pa) |
---|
| 577 | END WHERE |
---|
| 578 | |
---|
| 579 | END FUNCTION SIGN_ARRAY_2D_A |
---|
| 580 | |
---|
| 581 | FUNCTION SIGN_ARRAY_3D_A(pa,pb) |
---|
| 582 | !!----------------------------------------------------------------------- |
---|
| 583 | !! *** FUNCTION SIGN_ARRAY_3D_A *** |
---|
| 584 | !! |
---|
| 585 | !! ** Purpose : overwrite f95 behaviour of intrinsinc sign function |
---|
| 586 | !!----------------------------------------------------------------------- |
---|
| 587 | REAL(wp) :: pa(:,:,:),pb(:,:,:) ! input |
---|
| 588 | REAL(wp) :: SIGN_ARRAY_3D_A(SIZE(pb,1),SIZE(pb,2),SIZE(pb,3)) ! result |
---|
| 589 | |
---|
| 590 | WHERE ( pb >= 0.e0 ) |
---|
| 591 | SIGN_ARRAY_3D_A = ABS(pa) |
---|
| 592 | ELSEWHERE |
---|
| 593 | SIGN_ARRAY_3D_A =-ABS(pa) |
---|
| 594 | END WHERE |
---|
| 595 | |
---|
| 596 | END FUNCTION SIGN_ARRAY_3D_A |
---|
| 597 | |
---|
| 598 | FUNCTION SIGN_ARRAY_1D_B(pa,pb) |
---|
| 599 | !!----------------------------------------------------------------------- |
---|
| 600 | !! *** FUNCTION SIGN_ARRAY_1D_B *** |
---|
| 601 | !! |
---|
| 602 | !! ** Purpose : overwrite f95 behaviour of intrinsinc sign function |
---|
| 603 | !!----------------------------------------------------------------------- |
---|
| 604 | REAL(wp) :: pa(:),pb ! input |
---|
| 605 | REAL(wp) :: SIGN_ARRAY_1D_B(SIZE(pa,1)) ! result |
---|
| 606 | |
---|
| 607 | IF ( pb >= 0.e0 ) THEN |
---|
| 608 | SIGN_ARRAY_1D_B = ABS(pa) |
---|
| 609 | ELSE |
---|
| 610 | SIGN_ARRAY_1D_B =-ABS(pa) |
---|
| 611 | ENDIF |
---|
| 612 | |
---|
| 613 | END FUNCTION SIGN_ARRAY_1D_B |
---|
| 614 | |
---|
| 615 | FUNCTION SIGN_ARRAY_2D_B(pa,pb) |
---|
| 616 | !!----------------------------------------------------------------------- |
---|
| 617 | !! *** FUNCTION SIGN_ARRAY_2D_B *** |
---|
| 618 | !! |
---|
| 619 | !! ** Purpose : overwrite f95 behaviour of intrinsinc sign function |
---|
| 620 | !!----------------------------------------------------------------------- |
---|
| 621 | REAL(wp) :: pa(:,:),pb ! input |
---|
| 622 | REAL(wp) :: SIGN_ARRAY_2D_B(SIZE(pa,1),SIZE(pa,2)) ! result |
---|
| 623 | |
---|
| 624 | IF ( pb >= 0.e0 ) THEN |
---|
| 625 | SIGN_ARRAY_2D_B = ABS(pa) |
---|
| 626 | ELSE |
---|
| 627 | SIGN_ARRAY_2D_B =-ABS(pa) |
---|
| 628 | ENDIF |
---|
| 629 | |
---|
| 630 | END FUNCTION SIGN_ARRAY_2D_B |
---|
| 631 | |
---|
| 632 | FUNCTION SIGN_ARRAY_3D_B(pa,pb) |
---|
| 633 | !!----------------------------------------------------------------------- |
---|
| 634 | !! *** FUNCTION SIGN_ARRAY_3D_B *** |
---|
| 635 | !! |
---|
| 636 | !! ** Purpose : overwrite f95 behaviour of intrinsinc sign function |
---|
| 637 | !!----------------------------------------------------------------------- |
---|
| 638 | REAL(wp) :: pa(:,:,:),pb ! input |
---|
| 639 | REAL(wp) :: SIGN_ARRAY_3D_B(SIZE(pa,1),SIZE(pa,2),SIZE(pa,3)) ! result |
---|
| 640 | |
---|
| 641 | IF (pb >= 0.e0 ) THEN |
---|
| 642 | SIGN_ARRAY_3D_B = ABS(pa) |
---|
| 643 | ELSE |
---|
| 644 | SIGN_ARRAY_3D_B =-ABS(pa) |
---|
| 645 | ENDIF |
---|
| 646 | |
---|
| 647 | END FUNCTION SIGN_ARRAY_3D_B |
---|
| 648 | #endif |
---|
| 649 | |
---|
| 650 | END MODULE lib_fortran |
---|