Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.5/libbench2/verify-rdft2.c @ 42:2cd0e3b3e1fd

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Current fftw source
author Chris Cannam
date Tue, 18 Oct 2016 13:40:26 +0100
parents
children
rev   line source
Chris@42 1 /*
Chris@42 2 * Copyright (c) 2003, 2007-14 Matteo Frigo
Chris@42 3 * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology
Chris@42 4 *
Chris@42 5 * This program is free software; you can redistribute it and/or modify
Chris@42 6 * it under the terms of the GNU General Public License as published by
Chris@42 7 * the Free Software Foundation; either version 2 of the License, or
Chris@42 8 * (at your option) any later version.
Chris@42 9 *
Chris@42 10 * This program is distributed in the hope that it will be useful,
Chris@42 11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
Chris@42 12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
Chris@42 13 * GNU General Public License for more details.
Chris@42 14 *
Chris@42 15 * You should have received a copy of the GNU General Public License
Chris@42 16 * along with this program; if not, write to the Free Software
Chris@42 17 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
Chris@42 18 *
Chris@42 19 */
Chris@42 20
Chris@42 21
Chris@42 22 #include "verify.h"
Chris@42 23
Chris@42 24 /* copy real A into real B, using output stride of A and input stride of B */
Chris@42 25 typedef struct {
Chris@42 26 dotens2_closure k;
Chris@42 27 R *ra;
Chris@42 28 R *rb;
Chris@42 29 } cpyr_closure;
Chris@42 30
Chris@42 31 static void cpyr0(dotens2_closure *k_,
Chris@42 32 int indxa, int ondxa, int indxb, int ondxb)
Chris@42 33 {
Chris@42 34 cpyr_closure *k = (cpyr_closure *)k_;
Chris@42 35 k->rb[indxb] = k->ra[ondxa];
Chris@42 36 UNUSED(indxa); UNUSED(ondxb);
Chris@42 37 }
Chris@42 38
Chris@42 39 static void cpyr(R *ra, const bench_tensor *sza,
Chris@42 40 R *rb, const bench_tensor *szb)
Chris@42 41 {
Chris@42 42 cpyr_closure k;
Chris@42 43 k.k.apply = cpyr0;
Chris@42 44 k.ra = ra; k.rb = rb;
Chris@42 45 bench_dotens2(sza, szb, &k.k);
Chris@42 46 }
Chris@42 47
Chris@42 48 /* copy unpacked halfcomplex A[n] into packed-complex B[n], using output stride
Chris@42 49 of A and input stride of B. Only copies non-redundant half; other
Chris@42 50 half must be copied via mkhermitian. */
Chris@42 51 typedef struct {
Chris@42 52 dotens2_closure k;
Chris@42 53 int n;
Chris@42 54 int as;
Chris@42 55 int scalea;
Chris@42 56 R *ra, *ia;
Chris@42 57 R *rb, *ib;
Chris@42 58 } cpyhc2_closure;
Chris@42 59
Chris@42 60 static void cpyhc20(dotens2_closure *k_,
Chris@42 61 int indxa, int ondxa, int indxb, int ondxb)
Chris@42 62 {
Chris@42 63 cpyhc2_closure *k = (cpyhc2_closure *)k_;
Chris@42 64 int i, n = k->n;
Chris@42 65 int scalea = k->scalea;
Chris@42 66 int as = k->as * scalea;
Chris@42 67 R *ra = k->ra + ondxa * scalea, *ia = k->ia + ondxa * scalea;
Chris@42 68 R *rb = k->rb + indxb, *ib = k->ib + indxb;
Chris@42 69 UNUSED(indxa); UNUSED(ondxb);
Chris@42 70
Chris@42 71 for (i = 0; i < n/2 + 1; ++i) {
Chris@42 72 rb[2*i] = ra[as*i];
Chris@42 73 ib[2*i] = ia[as*i];
Chris@42 74 }
Chris@42 75 }
Chris@42 76
Chris@42 77 static void cpyhc2(R *ra, R *ia,
Chris@42 78 const bench_tensor *sza, const bench_tensor *vecsza,
Chris@42 79 int scalea,
Chris@42 80 R *rb, R *ib, const bench_tensor *szb)
Chris@42 81 {
Chris@42 82 cpyhc2_closure k;
Chris@42 83 BENCH_ASSERT(sza->rnk <= 1);
Chris@42 84 k.k.apply = cpyhc20;
Chris@42 85 k.n = tensor_sz(sza);
Chris@42 86 k.scalea = scalea;
Chris@42 87 if (!BENCH_FINITE_RNK(sza->rnk) || sza->rnk == 0)
Chris@42 88 k.as = 0;
Chris@42 89 else
Chris@42 90 k.as = sza->dims[0].os;
Chris@42 91 k.ra = ra; k.ia = ia; k.rb = rb; k.ib = ib;
Chris@42 92 bench_dotens2(vecsza, szb, &k.k);
Chris@42 93 }
Chris@42 94
Chris@42 95 /* icpyhc2 is the inverse of cpyhc2 */
Chris@42 96
Chris@42 97 static void icpyhc20(dotens2_closure *k_,
Chris@42 98 int indxa, int ondxa, int indxb, int ondxb)
Chris@42 99 {
Chris@42 100 cpyhc2_closure *k = (cpyhc2_closure *)k_;
Chris@42 101 int i, n = k->n;
Chris@42 102 int scalea = k->scalea;
Chris@42 103 int as = k->as * scalea;
Chris@42 104 R *ra = k->ra + indxa * scalea, *ia = k->ia + indxa * scalea;
Chris@42 105 R *rb = k->rb + ondxb, *ib = k->ib + ondxb;
Chris@42 106 UNUSED(ondxa); UNUSED(indxb);
Chris@42 107
Chris@42 108 for (i = 0; i < n/2 + 1; ++i) {
Chris@42 109 ra[as*i] = rb[2*i];
Chris@42 110 ia[as*i] = ib[2*i];
Chris@42 111 }
Chris@42 112 }
Chris@42 113
Chris@42 114 static void icpyhc2(R *ra, R *ia,
Chris@42 115 const bench_tensor *sza, const bench_tensor *vecsza,
Chris@42 116 int scalea,
Chris@42 117 R *rb, R *ib, const bench_tensor *szb)
Chris@42 118 {
Chris@42 119 cpyhc2_closure k;
Chris@42 120 BENCH_ASSERT(sza->rnk <= 1);
Chris@42 121 k.k.apply = icpyhc20;
Chris@42 122 k.n = tensor_sz(sza);
Chris@42 123 k.scalea = scalea;
Chris@42 124 if (!BENCH_FINITE_RNK(sza->rnk) || sza->rnk == 0)
Chris@42 125 k.as = 0;
Chris@42 126 else
Chris@42 127 k.as = sza->dims[0].is;
Chris@42 128 k.ra = ra; k.ia = ia; k.rb = rb; k.ib = ib;
Chris@42 129 bench_dotens2(vecsza, szb, &k.k);
Chris@42 130 }
Chris@42 131
Chris@42 132 typedef struct {
Chris@42 133 dofft_closure k;
Chris@42 134 bench_problem *p;
Chris@42 135 } dofft_rdft2_closure;
Chris@42 136
Chris@42 137 static void rdft2_apply(dofft_closure *k_,
Chris@42 138 bench_complex *in, bench_complex *out)
Chris@42 139 {
Chris@42 140 dofft_rdft2_closure *k = (dofft_rdft2_closure *)k_;
Chris@42 141 bench_problem *p = k->p;
Chris@42 142 bench_tensor *totalsz, *pckdsz, *totalsz_swap, *pckdsz_swap;
Chris@42 143 bench_tensor *probsz2, *totalsz2, *pckdsz2;
Chris@42 144 bench_tensor *probsz2_swap, *totalsz2_swap, *pckdsz2_swap;
Chris@42 145 bench_real *ri, *ii, *ro, *io;
Chris@42 146 int n2, totalscale;
Chris@42 147
Chris@42 148 totalsz = tensor_append(p->vecsz, p->sz);
Chris@42 149 pckdsz = verify_pack(totalsz, 2);
Chris@42 150 n2 = tensor_sz(totalsz);
Chris@42 151 if (BENCH_FINITE_RNK(p->sz->rnk) && p->sz->rnk > 0)
Chris@42 152 n2 = (n2 / p->sz->dims[p->sz->rnk - 1].n) *
Chris@42 153 (p->sz->dims[p->sz->rnk - 1].n / 2 + 1);
Chris@42 154 ri = (bench_real *) p->in;
Chris@42 155 ro = (bench_real *) p->out;
Chris@42 156
Chris@42 157 if (BENCH_FINITE_RNK(p->sz->rnk) && p->sz->rnk > 0 && n2 > 0) {
Chris@42 158 probsz2 = tensor_copy_sub(p->sz, p->sz->rnk - 1, 1);
Chris@42 159 totalsz2 = tensor_copy_sub(totalsz, 0, totalsz->rnk - 1);
Chris@42 160 pckdsz2 = tensor_copy_sub(pckdsz, 0, pckdsz->rnk - 1);
Chris@42 161 }
Chris@42 162 else {
Chris@42 163 probsz2 = mktensor(0);
Chris@42 164 totalsz2 = tensor_copy(totalsz);
Chris@42 165 pckdsz2 = tensor_copy(pckdsz);
Chris@42 166 }
Chris@42 167
Chris@42 168 totalsz_swap = tensor_copy_swapio(totalsz);
Chris@42 169 pckdsz_swap = tensor_copy_swapio(pckdsz);
Chris@42 170 totalsz2_swap = tensor_copy_swapio(totalsz2);
Chris@42 171 pckdsz2_swap = tensor_copy_swapio(pckdsz2);
Chris@42 172 probsz2_swap = tensor_copy_swapio(probsz2);
Chris@42 173
Chris@42 174 /* confusion: the stride is the distance between complex elements
Chris@42 175 when using interleaved format, but it is the distance between
Chris@42 176 real elements when using split format */
Chris@42 177 if (p->split) {
Chris@42 178 ii = p->ini ? (bench_real *) p->ini : ri + n2;
Chris@42 179 io = p->outi ? (bench_real *) p->outi : ro + n2;
Chris@42 180 totalscale = 1;
Chris@42 181 } else {
Chris@42 182 ii = p->ini ? (bench_real *) p->ini : ri + 1;
Chris@42 183 io = p->outi ? (bench_real *) p->outi : ro + 1;
Chris@42 184 totalscale = 2;
Chris@42 185 }
Chris@42 186
Chris@42 187 if (p->sign < 0) { /* R2HC */
Chris@42 188 int N, vN, i;
Chris@42 189 cpyr(&c_re(in[0]), pckdsz, ri, totalsz);
Chris@42 190 after_problem_rcopy_from(p, ri);
Chris@42 191 doit(1, p);
Chris@42 192 after_problem_hccopy_to(p, ro, io);
Chris@42 193 if (k->k.recopy_input)
Chris@42 194 cpyr(ri, totalsz_swap, &c_re(in[0]), pckdsz_swap);
Chris@42 195 cpyhc2(ro, io, probsz2, totalsz2, totalscale,
Chris@42 196 &c_re(out[0]), &c_im(out[0]), pckdsz2);
Chris@42 197 N = tensor_sz(p->sz);
Chris@42 198 vN = tensor_sz(p->vecsz);
Chris@42 199 for (i = 0; i < vN; ++i)
Chris@42 200 mkhermitian(out + i*N, p->sz->rnk, p->sz->dims, 1);
Chris@42 201 }
Chris@42 202 else { /* HC2R */
Chris@42 203 icpyhc2(ri, ii, probsz2, totalsz2, totalscale,
Chris@42 204 &c_re(in[0]), &c_im(in[0]), pckdsz2);
Chris@42 205 after_problem_hccopy_from(p, ri, ii);
Chris@42 206 doit(1, p);
Chris@42 207 after_problem_rcopy_to(p, ro);
Chris@42 208 if (k->k.recopy_input)
Chris@42 209 cpyhc2(ri, ii, probsz2_swap, totalsz2_swap, totalscale,
Chris@42 210 &c_re(in[0]), &c_im(in[0]), pckdsz2_swap);
Chris@42 211 mkreal(out, tensor_sz(pckdsz));
Chris@42 212 cpyr(ro, totalsz, &c_re(out[0]), pckdsz);
Chris@42 213 }
Chris@42 214
Chris@42 215 tensor_destroy(totalsz);
Chris@42 216 tensor_destroy(pckdsz);
Chris@42 217 tensor_destroy(totalsz_swap);
Chris@42 218 tensor_destroy(pckdsz_swap);
Chris@42 219 tensor_destroy(probsz2);
Chris@42 220 tensor_destroy(totalsz2);
Chris@42 221 tensor_destroy(pckdsz2);
Chris@42 222 tensor_destroy(probsz2_swap);
Chris@42 223 tensor_destroy(totalsz2_swap);
Chris@42 224 tensor_destroy(pckdsz2_swap);
Chris@42 225 }
Chris@42 226
Chris@42 227 void verify_rdft2(bench_problem *p, int rounds, double tol, errors *e)
Chris@42 228 {
Chris@42 229 C *inA, *inB, *inC, *outA, *outB, *outC, *tmp;
Chris@42 230 int n, vecn, N;
Chris@42 231 dofft_rdft2_closure k;
Chris@42 232
Chris@42 233 BENCH_ASSERT(p->kind == PROBLEM_REAL);
Chris@42 234
Chris@42 235 if (!BENCH_FINITE_RNK(p->sz->rnk) || !BENCH_FINITE_RNK(p->vecsz->rnk))
Chris@42 236 return; /* give up */
Chris@42 237
Chris@42 238 k.k.apply = rdft2_apply;
Chris@42 239 k.k.recopy_input = 0;
Chris@42 240 k.p = p;
Chris@42 241
Chris@42 242 if (rounds == 0)
Chris@42 243 rounds = 20; /* default value */
Chris@42 244
Chris@42 245 n = tensor_sz(p->sz);
Chris@42 246 vecn = tensor_sz(p->vecsz);
Chris@42 247 N = n * vecn;
Chris@42 248
Chris@42 249 inA = (C *) bench_malloc(N * sizeof(C));
Chris@42 250 inB = (C *) bench_malloc(N * sizeof(C));
Chris@42 251 inC = (C *) bench_malloc(N * sizeof(C));
Chris@42 252 outA = (C *) bench_malloc(N * sizeof(C));
Chris@42 253 outB = (C *) bench_malloc(N * sizeof(C));
Chris@42 254 outC = (C *) bench_malloc(N * sizeof(C));
Chris@42 255 tmp = (C *) bench_malloc(N * sizeof(C));
Chris@42 256
Chris@42 257 e->i = impulse(&k.k, n, vecn, inA, inB, inC, outA, outB, outC,
Chris@42 258 tmp, rounds, tol);
Chris@42 259 e->l = linear(&k.k, 1, N, inA, inB, inC, outA, outB, outC,
Chris@42 260 tmp, rounds, tol);
Chris@42 261
Chris@42 262 e->s = 0.0;
Chris@42 263 if (p->sign < 0)
Chris@42 264 e->s = dmax(e->s, tf_shift(&k.k, 1, p->sz, n, vecn, p->sign,
Chris@42 265 inA, inB, outA, outB,
Chris@42 266 tmp, rounds, tol, TIME_SHIFT));
Chris@42 267 else
Chris@42 268 e->s = dmax(e->s, tf_shift(&k.k, 1, p->sz, n, vecn, p->sign,
Chris@42 269 inA, inB, outA, outB,
Chris@42 270 tmp, rounds, tol, FREQ_SHIFT));
Chris@42 271
Chris@42 272 if (!p->in_place && !p->destroy_input)
Chris@42 273 preserves_input(&k.k, p->sign < 0 ? mkreal : mkhermitian1,
Chris@42 274 N, inA, inB, outB, rounds);
Chris@42 275
Chris@42 276 bench_free(tmp);
Chris@42 277 bench_free(outC);
Chris@42 278 bench_free(outB);
Chris@42 279 bench_free(outA);
Chris@42 280 bench_free(inC);
Chris@42 281 bench_free(inB);
Chris@42 282 bench_free(inA);
Chris@42 283 }
Chris@42 284
Chris@42 285 void accuracy_rdft2(bench_problem *p, int rounds, int impulse_rounds,
Chris@42 286 double t[6])
Chris@42 287 {
Chris@42 288 dofft_rdft2_closure k;
Chris@42 289 int n;
Chris@42 290 C *a, *b;
Chris@42 291
Chris@42 292 BENCH_ASSERT(p->kind == PROBLEM_REAL);
Chris@42 293 BENCH_ASSERT(p->sz->rnk == 1);
Chris@42 294 BENCH_ASSERT(p->vecsz->rnk == 0);
Chris@42 295
Chris@42 296 k.k.apply = rdft2_apply;
Chris@42 297 k.k.recopy_input = 0;
Chris@42 298 k.p = p;
Chris@42 299 n = tensor_sz(p->sz);
Chris@42 300
Chris@42 301 a = (C *) bench_malloc(n * sizeof(C));
Chris@42 302 b = (C *) bench_malloc(n * sizeof(C));
Chris@42 303 accuracy_test(&k.k, p->sign < 0 ? mkreal : mkhermitian1, p->sign,
Chris@42 304 n, a, b, rounds, impulse_rounds, t);
Chris@42 305 bench_free(b);
Chris@42 306 bench_free(a);
Chris@42 307 }