File: step_generic.cpp

package info (click to toggle)
meep-mpich2 1.7.0-3
  • links: PTS, VCS
  • area: main
  • in suites: bullseye, buster, sid
  • size: 25,824 kB
  • sloc: cpp: 27,370; python: 10,574; lisp: 1,213; makefile: 440; sh: 28
file content (555 lines) | stat: -rw-r--r-- 18,796 bytes parent folder | download | duplicates (5)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
#include "meep.hpp"
#include "meep_internals.hpp"
#include "config.h"

#define DPR double * restrict
#define RPR realnum * restrict

/* These macros get into the guts of the LOOP_OVER_VOL loops to
   efficiently construct the index k into a PML sigma array.
   Basically, k needs to increment by 2 for each increment of one of
   LOOP's for-loops, starting at the appropriate corner of the grid_volume,
   and these macros define the relevant strides etc. for each loop.
   KSTRIDE_DEF defines the relevant strides etc. and goes outside the
   LOOP, wheras KDEF defines the k index and goes inside the LOOP. */
#define KSTRIDE_DEF(dsig, k, corner)				\
     const int k##0 = corner.in_direction(dsig)			\
                      - gv.little_corner().in_direction(dsig);	\
     const int s##k##1 = gv.yucky_direction(0) == dsig ? 2 : 0; \
     const int s##k##2 = gv.yucky_direction(1) == dsig ? 2 : 0; \
     const int s##k##3 = gv.yucky_direction(2) == dsig ? 2 : 0
#define KDEF(k,dsig) const int k = ((k##0 + s##k##1*loop_i1) + s##k##2*loop_i2) + s##k##3*loop_i3
#define DEF_k KDEF(k,dsig)
#define DEF_ku KDEF(ku,dsigu)
#define DEF_kw KDEF(kw,dsigw)

using namespace std;

namespace meep {

#define SWAP(t,a,b) { t xxxx = a; a = b; b = xxxx; }

/* update step for df/dt = curl g,
   i.e. f += dt curl g = dt/dx (dg1 - dg2)
   where dgk = gk[i] - gk[i+sk].

   g = (g1,g2), where g1 or g2 may be NULL.  Note that dt/dx and/or s1
   and s2 may be negative to flip signs of derivatives.

   PML: sig[k] = sigma[k]*dt/2, siginv[k] = 1 / (kap[k] + sigma[k]*dt/2).
   Here, k is the index in the dsig direction.  if dsig ==
   NO_DIRECTION, then PML is not used.  (dsig is the sigma direction.)

   if non-NULL, then cnd is an array of conductivity values, changing
   the underlying PDE to:
       df/dt = curl g - cnd f
   which is updated as:
       f = [ dt * curl g + (1 - dt cnd/2) f ] / (1 + dt cnd/2)
   cndinv should be an array of 1 / (1 + dt cnd/2).  In the case
   of PML, cndinv should contain 1 / (1 + dt (cnd + sigma)/2).

   fcnd is an auxiliary field used ONLY when we simultaneously have
   PML (dsig != NO_DIR) and conductivity, in which case fcnd solves
       dfcnd/dt = curl g - cnd*fcnd
   and f satisfies
       df/dt = dfcnd/dt - sigma*f.

   fu is another auxiliary field used only in PML (dsigu != NO_DIR),
   in which case f solves:
       df/dt = dfu/dt - sigma_u * f
   and fu replaces f in the equations above (fu += dt curl g etcetera).
*/
void step_curl(RPR f, component c, const RPR g1, const RPR g2,
	       ptrdiff_t s1, ptrdiff_t s2, // strides for g1/g2 shift
	       const grid_volume &gv, double dtdx,
	       direction dsig, const DPR sig, const DPR kap, const DPR siginv,
	       RPR fu, direction dsigu, const DPR sigu, const DPR kapu, const DPR siginvu,
	       double dt,
	       const RPR cnd, const RPR cndinv, RPR fcnd)
{
  if (!g1) { // swap g1 and g2
    SWAP(const RPR, g1, g2);
    SWAP(ptrdiff_t, s1, s2);
    dtdx = -dtdx; // need to flip derivative sign
  }

  /* The following are a bunch of special cases of the "MOST GENERAL CASE"
     loop below.  We make copies of the loop for each special case in
     order to keep the innermost loop efficient.  This is especially
     important because the non-PML cases are actually more common.
     (The "right" way to do this is by partial evaluation of the
      most general case, but that would require a code generator.) */

  if (dsig == NO_DIRECTION) { // no PML in f update
    if (dsigu == NO_DIRECTION) { // no fu update
      if (cnd) {
    	double dt2 = dt * 0.5;
    	if (g2) {
    	  LOOP_OVER_VOL_OWNED0(gv, c, i)
    	    f[i] = ((1 - dt2 * cnd[i]) * f[i] -
    		    dtdx * (g1[i+s1] - g1[i] + g2[i] - g2[i+s2])) * cndinv[i];
    	}
    	else {
    	  LOOP_OVER_VOL_OWNED0(gv, c, i)
    	    f[i] = ((1 - dt2 * cnd[i]) * f[i]
    		    - dtdx * (g1[i+s1] - g1[i])) * cndinv[i];
    	}
      }
      else { // no conductivity
    	if (g2) {
    	  LOOP_OVER_VOL_OWNED0(gv, c, i)
    	    f[i] -= dtdx * (g1[i+s1] - g1[i] + g2[i] - g2[i+s2]);
    	}
    	else {
    	  LOOP_OVER_VOL_OWNED0(gv, c, i)
    	    f[i] -= dtdx * (g1[i+s1] - g1[i]);
    	}
      }
    }
    else { // fu update, no PML in f update
      KSTRIDE_DEF(dsigu, ku, gv.little_owned_corner0(c));
      if (cnd) {
    	double dt2 = dt * 0.5;
    	if (g2) {
    	  LOOP_OVER_VOL_OWNED0(gv, c, i) {
    	    DEF_ku; double fprev = fu[i];
    	    fu[i] = ((1 - dt2 * cnd[i]) * fprev -
    		    dtdx * (g1[i+s1] - g1[i] + g2[i] - g2[i+s2])) * cndinv[i];
    	    f[i] = siginvu[ku] * ((kapu[ku] - sigu[ku]) * f[i] + fu[i] - fprev);
    	  }
    	}
    	else {
    	  LOOP_OVER_VOL_OWNED0(gv, c, i) {
    	    DEF_ku; double fprev = fu[i];
    	    fu[i] = ((1 - dt2 * cnd[i]) * fprev
    		    - dtdx * (g1[i+s1] - g1[i])) * cndinv[i];
    	    f[i] = siginvu[ku] * ((kapu[ku] - sigu[ku]) * f[i] + fu[i] - fprev);
    	  }
    	}
      }
      else { // no conductivity
    	if (g2) {
    	  LOOP_OVER_VOL_OWNED0(gv, c, i) {
    	    DEF_ku; double fprev = fu[i];
    	    fu[i] -= dtdx * (g1[i+s1] - g1[i] + g2[i] - g2[i+s2]);
    	    f[i] = siginvu[ku] * ((kapu[ku] - sigu[ku]) * f[i] + fu[i] - fprev);
    	  }
    	}
    	else {
    	  LOOP_OVER_VOL_OWNED0(gv, c, i) {
    	    DEF_ku; double fprev = fu[i];
    	    fu[i] -= dtdx * (g1[i+s1] - g1[i]);
    	    f[i] = siginvu[ku] * ((kapu[ku] - sigu[ku]) * f[i] + fu[i] - fprev);
    	  }
    	}
      }
    }
  }
  else { /* PML in f update */
    KSTRIDE_DEF(dsig, k, gv.little_owned_corner0(c));
    if (dsigu == NO_DIRECTION) { // no fu update
      if (cnd) {
	double dt2 = dt * 0.5;
	if (g2) {
	  LOOP_OVER_VOL_OWNED0(gv, c, i) {
	    DEF_k;
	    realnum fcnd_prev = fcnd[i];
	    fcnd[i] = ((1 - dt2 * cnd[i]) * fcnd[i] -
		       dtdx * (g1[i+s1]-g1[i] + g2[i]-g2[i+s2])) * cndinv[i];
	    f[i] = ((kap[k] - sig[k]) * f[i] + (fcnd[i] - fcnd_prev)) * siginv[k];
	  }
	}
	else {
	  LOOP_OVER_VOL_OWNED0(gv, c, i) {
	    DEF_k;
	    realnum fcnd_prev = fcnd[i];
	    fcnd[i] = ((1 - dt2 * cnd[i]) * fcnd[i] -
		       dtdx * (g1[i+s1] - g1[i])) * cndinv[i];
	    f[i] = ((kap[k] - sig[k]) * f[i] + (fcnd[i] - fcnd_prev)) * siginv[k];
	  }
	}
      }
      else { // no conductivity (other than PML conductivity)
	if (g2) {
	  LOOP_OVER_VOL_OWNED0(gv, c, i) {
	    DEF_k;
	    f[i] = ((kap[k] - sig[k]) * f[i] -
		    dtdx * (g1[i+s1] - g1[i] + g2[i] - g2[i+s2])) * siginv[k];
	  }
	}
	else {
	  LOOP_OVER_VOL_OWNED0(gv, c, i) {
	    DEF_k;
	    f[i] = ((kap[k] - sig[k]) * f[i] - dtdx * (g1[i+s1]-g1[i])) * siginv[k];
	  }
	}
      }
    }
    else { // fu update + PML in f update
      KSTRIDE_DEF(dsigu, ku, gv.little_owned_corner0(c));
      if (cnd) {
	double dt2 = dt * 0.5;
	if (g2) {
	  //////////////////// MOST GENERAL CASE //////////////////////
	  LOOP_OVER_VOL_OWNED0(gv, c, i) {
	    DEF_k; DEF_ku; double fprev = fu[i];
	    realnum fcnd_prev = fcnd[i];
	    fcnd[i] = ((1 - dt2 * cnd[i]) * fcnd[i] -
		       dtdx * (g1[i+s1]-g1[i] + g2[i]-g2[i+s2])) * cndinv[i];
	    fu[i] = ((kap[k] - sig[k]) * fu[i] + (fcnd[i] - fcnd_prev)) * siginv[k];
	    f[i] = siginvu[ku] * ((kapu[ku] - sigu[ku]) * f[i] + fu[i] - fprev);
	  }
	  /////////////////////////////////////////////////////////////
	}
	else {
	  LOOP_OVER_VOL_OWNED0(gv, c, i) {
	    DEF_k; DEF_ku; double fprev = fu[i];
	    realnum fcnd_prev = fcnd[i];
	    fcnd[i] = ((1 - dt2 * cnd[i]) * fcnd[i] -
		       dtdx * (g1[i+s1] - g1[i])) * cndinv[i];
	    fu[i] = ((kap[k] - sig[k]) * fu[i] + (fcnd[i] - fcnd_prev)) * siginv[k];
	    f[i] = siginvu[ku] * ((kapu[ku] - sigu[ku]) * f[i] + fu[i] - fprev);
	  }
	}
      }
      else { // no conductivity (other than PML conductivity)
	if (g2) {
	  LOOP_OVER_VOL_OWNED0(gv, c, i) {
	    DEF_k; DEF_ku; double fprev = fu[i];
	    fu[i] = ((kap[k] - sig[k]) * fu[i] -
		    dtdx * (g1[i+s1] - g1[i] + g2[i] - g2[i+s2])) * siginv[k];
	    f[i] = siginvu[ku] * ((kapu[ku] - sigu[ku]) * f[i] + fu[i] - fprev);
	  }
	}
	else {
	  LOOP_OVER_VOL_OWNED0(gv, c, i) {
	    DEF_k; DEF_ku; double fprev = fu[i];
	    fu[i] = ((kap[k] - sig[k]) * fu[i] - dtdx * (g1[i+s1]-g1[i])) * siginv[k];
	    f[i] = siginvu[ku] * ((kapu[ku] - sigu[ku]) * f[i] + fu[i] - fprev);
	  }
	}
      }
    }
  }
}

/* field-update equation f += betadt * g (plus variants for conductivity
   and/or PML).  This is used in 2d calculations to add an exp(i beta z)
   time dependence, which gives an additional i \beta \hat{z} \times
   cross-product in the curl equations. */
void step_beta(RPR f, component c, const RPR g,
	       const grid_volume &gv, double betadt,
	       direction dsig, const DPR siginv,
	       RPR fu, direction dsigu, const DPR siginvu,
	       const RPR cndinv, RPR fcnd)
{
  if (!g) return;
  if (dsig != NO_DIRECTION) { // PML in f update
    KSTRIDE_DEF(dsig, k, gv.little_owned_corner0(c));
    if (dsigu != NO_DIRECTION) { // PML in f + fu
      KSTRIDE_DEF(dsigu, ku, gv.little_owned_corner0(c));
      if (cndinv) { // conductivity + PML
	//////////////////// MOST GENERAL CASE //////////////////////
	LOOP_OVER_VOL_OWNED0(gv, c, i) {
	  DEF_k; DEF_ku; double df;
	  double dfcnd = betadt * g[i] * cndinv[i];
	  fcnd[i] += dfcnd;
	  fu[i] += (df = dfcnd * siginv[k]);
	  f[i] += siginvu[ku] * df;
	}
	/////////////////////////////////////////////////////////////
      }
      else { // PML only
	LOOP_OVER_VOL_OWNED0(gv, c, i) {
	  DEF_k; DEF_ku; double df;
	  fu[i] += (df = betadt * g[i] * siginv[k]);
	  f[i] += siginvu[ku] * df;
	}
      }
    }
    else { // PML in f, no fu
      if (cndinv) { // conductivity + PML
	LOOP_OVER_VOL_OWNED0(gv, c, i) {
	  DEF_k;
	  double dfcnd = betadt * g[i] * cndinv[i];
	  fcnd[i] += dfcnd;
	  f[i] += dfcnd * siginv[k];
	}
      }
      else { // PML only
	LOOP_OVER_VOL_OWNED0(gv, c, i) {
	  DEF_k;
	  f[i] += betadt * g[i] * siginv[k];
	}
      }
    }
  }
  else { // no PML in f update
    if (dsigu != NO_DIRECTION) { // fu, no PML in f
      KSTRIDE_DEF(dsigu, ku, gv.little_owned_corner0(c));
      if (cndinv) { // conductivity, no PML
	LOOP_OVER_VOL_OWNED0(gv, c, i) {
	  DEF_ku; double df;
	  fu[i] += (df = betadt * g[i] * cndinv[i]);
	  f[i] += siginvu[ku] * df;
	}
      }
      else { // no conductivity or PML
	LOOP_OVER_VOL_OWNED0(gv, c, i) {
	  DEF_ku; double df;
	  fu[i] += (df = betadt * g[i]);
	  f[i] += siginvu[ku] * df;
	}
      }
    }
    else { // no PML, no fu
      if (cndinv) { // conductivity, no PML
	LOOP_OVER_VOL_OWNED0(gv, c, i)
	  f[i] += betadt * g[i] * cndinv[i];
      }
      else { // no conductivity or PML
	LOOP_OVER_VOL_OWNED0(gv, c, i)
	  f[i] += betadt * g[i];
      }
    }
  }
}

/* Given Dsqr = |D|^2 and Di = component of D, compute the factor f so
   that Ei = chi1inv * f * Di.   In principle, this would involve solving
   a cubic equation, but instead we use a Pade approximant that is
   accurate to several orders.  This is inaccurate if the nonlinear
   index change is large, of course, but in that case the chi2/chi3
   power-series expansion isn't accurate anyway, so the cubic isn't
   physical there either. */
inline double calc_nonlinear_u(const double Dsqr,
			       const double Di,
			       const double chi1inv,
			       const double chi2, const double chi3) {
  double c2 = Di*chi2*(chi1inv*chi1inv);
  double c3 = Dsqr*chi3*(chi1inv*chi1inv*chi1inv);
  return (1 + c2 + 2*c3)/(1 + 2*c2 + 3*c3);
}

/* Update E from D using epsilon and PML, *or* update H from B using
   mu and PML.

   To be generic, here we set f = u * g, where u may
   be a tensor, and we also have a nonlinear susceptibility chi.
   Here, g = (g,g1,g2) where g1 and g2 are the off-diagonal
   components, if any (g2 may be NULL).

   In PML (dsigw != NO_DIR), we have an additional auxiliary field fw,
   which is updated by the equations:
          fw = u * g
          df/dt = kappaw dfw/dt - sigmaw * fw
   That is, fw is updated like the non-PML f, and f is updated from
   fw by a little ODE.  Here, sigw[k] = sigmaw[k]*dt/2, kappaw[k] = kapw[k]

*/

void step_update_EDHB(RPR f, component fc, const grid_volume &gv,
		      const RPR g, const RPR g1, const RPR g2,
		      const RPR u, const RPR u1, const RPR u2,
		      ptrdiff_t s, ptrdiff_t s1, ptrdiff_t s2,
		      const RPR chi2, const RPR chi3,
		      RPR fw, direction dsigw, const DPR sigw, const DPR kapw)
{
  if (!f) return;

  if ((!g1 && g2) || (g1 && g2 && !u1 && u2)) { /* swap g1 and g2 */
    SWAP(const RPR, g1, g2);
    SWAP(const RPR, u1, u2);
    SWAP(ptrdiff_t, s1, s2);
  }

  // stable averaging of offdiagonal components
#define OFFDIAG(u,g,sx) (0.25 * ((g[i]+g[i-sx])*u[i] \
		   	       + (g[i+s]+g[(i+s)-sx])*u[i+s]))

  /* As with step_curl, these loops are all essentially copies
     of the "MOST GENERAL CASE" loop with various terms thrown out. */

  if (dsigw != NO_DIRECTION) { //////// PML case (with fw) /////////////
    KSTRIDE_DEF(dsigw, kw, gv.little_owned_corner0(fc));
    if (u1 && u2) { // 3x3 off-diagonal u
      if (chi3) {
	//////////////////// MOST GENERAL CASE //////////////////////
	LOOP_OVER_VOL_OWNED(gv, fc, i) {
	  double g1s = g1[i]+g1[i+s]+g1[i-s1]+g1[i+(s-s1)];
	  double g2s = g2[i]+g2[i+s]+g2[i-s2]+g2[i+(s-s2)];
	  double gs = g[i]; double us = u[i];
	  DEF_kw; double fwprev = fw[i], kapwkw = kapw[kw], sigwkw = sigw[kw];
	  fw[i] = (gs * us + OFFDIAG(u1,g1,s1) + OFFDIAG(u2,g2,s2))
	    * calc_nonlinear_u(gs * gs + 0.0625 * (g1s*g1s + g2s*g2s),
			       gs, us, chi2[i], chi3[i]);
	  f[i] += (kapwkw + sigwkw) * fw[i] - (kapwkw - sigwkw) * fwprev;
	}
	/////////////////////////////////////////////////////////////
      }
      else {
	LOOP_OVER_VOL_OWNED(gv, fc, i) {
	  double gs = g[i]; double us = u[i];
	  DEF_kw; double fwprev = fw[i], kapwkw = kapw[kw], sigwkw = sigw[kw];
	  fw[i] = (gs * us + OFFDIAG(u1,g1,s1) + OFFDIAG(u2,g2,s2));
	  f[i] += (kapwkw + sigwkw) * fw[i] - (kapwkw - sigwkw) * fwprev;
	}
      }
    }
    else if (u1) { // 2x2 off-diagonal u
      if (chi3) {
	LOOP_OVER_VOL_OWNED(gv, fc, i) {
	  double g1s = g1[i]+g1[i+s]+g1[i-s1]+g1[i+(s-s1)];
	  double gs = g[i]; double us = u[i];
	  DEF_kw; double fwprev = fw[i], kapwkw = kapw[kw], sigwkw = sigw[kw];
	  fw[i] = (gs * us + OFFDIAG(u1,g1,s1))
	    * calc_nonlinear_u(gs * gs + 0.0625 * (g1s*g1s),
			       gs, us, chi2[i], chi3[i]);
	  f[i] += (kapwkw + sigwkw) * fw[i] - (kapwkw - sigwkw) * fwprev;
	}
      }
      else {
	LOOP_OVER_VOL_OWNED(gv, fc, i) {
	  double gs = g[i]; double us = u[i];
	  DEF_kw; double fwprev = fw[i], kapwkw = kapw[kw], sigwkw = sigw[kw];
	  fw[i] = (gs * us + OFFDIAG(u1,g1,s1));
	  f[i] += (kapwkw + sigwkw) * fw[i] - (kapwkw - sigwkw) * fwprev;
	}
      }
    }
    else if (u2) { // 2x2 off-diagonal u
      abort("bug - didn't swap off-diagonal terms!?");
    }
    else { // diagonal u
      if (chi3) {
	if (g1 && g2) {
	  LOOP_OVER_VOL_OWNED(gv, fc, i) {
	    double g1s = g1[i]+g1[i+s]+g1[i-s1]+g1[i+(s-s1)];
	    double g2s = g2[i]+g2[i+s]+g2[i-s2]+g2[i+(s-s2)];
	    double gs = g[i]; double us = u[i];
	    DEF_kw; double fwprev = fw[i], kapwkw = kapw[kw], sigwkw = sigw[kw];
	    fw[i] = (gs*us)*calc_nonlinear_u(gs*gs+0.0625*(g1s*g1s+g2s*g2s),
					     gs, us, chi2[i], chi3[i]);
	    f[i] += (kapwkw + sigwkw) * fw[i] - (kapwkw - sigwkw) * fwprev;
	  }
	}
	else if (g1) {
	  LOOP_OVER_VOL_OWNED(gv, fc, i) {
	    double g1s = g1[i]+g1[i+s]+g1[i-s1]+g1[i+(s-s1)];
	    double gs = g[i]; double us = u[i];
	    DEF_kw; double fwprev = fw[i], kapwkw = kapw[kw], sigwkw = sigw[kw];
	    fw[i] = (gs*us)*calc_nonlinear_u(gs*gs + 0.0625*(g1s*g1s),
					     gs, us, chi2[i], chi3[i]);
	    f[i] += (kapwkw + sigwkw) * fw[i] - (kapwkw - sigwkw) * fwprev;
	  }
	}
	else if (g2) {
	  abort("bug - didn't swap off-diagonal terms!?");
	}
	else {
	  LOOP_OVER_VOL_OWNED(gv, fc, i) {
	    double gs = g[i]; double us = u[i];
	    DEF_kw; double fwprev = fw[i], kapwkw = kapw[kw], sigwkw = sigw[kw];
	    fw[i] = (gs*us)*calc_nonlinear_u(gs*gs, gs,us, chi2[i],chi3[i]);
	    f[i] += (kapwkw + sigwkw) * fw[i] - (kapwkw - sigwkw) * fwprev;
	  }
	}
      }
      else if (u) {
	LOOP_OVER_VOL_OWNED(gv, fc, i) {
	  double gs = g[i]; double us = u[i];
	  DEF_kw; double fwprev = fw[i], kapwkw = kapw[kw], sigwkw = sigw[kw];
	  fw[i] = (gs * us);
	  f[i] += (kapwkw + sigwkw) * fw[i] - (kapwkw - sigwkw) * fwprev;
	}
      }
      else {
	LOOP_OVER_VOL_OWNED(gv, fc, i) {
	  DEF_kw; double fwprev = fw[i], kapwkw = kapw[kw], sigwkw = sigw[kw];
	  fw[i] = g[i];
	  f[i] += (kapwkw + sigwkw) * fw[i] - (kapwkw - sigwkw) * fwprev;
	}
      }
    }
  }
  else { /////////////// no PML (no fw) ///////////////////
    if (u1 && u2) { // 3x3 off-diagonal u
      if (chi3) {
	LOOP_OVER_VOL_OWNED(gv, fc, i) {
	  double g1s = g1[i]+g1[i+s]+g1[i-s1]+g1[i+(s-s1)];
	  double g2s = g2[i]+g2[i+s]+g2[i-s2]+g2[i+(s-s2)];
	  double gs = g[i]; double us = u[i];
	  f[i] = (gs * us + OFFDIAG(u1,g1,s1) + OFFDIAG(u2,g2,s2))
	    * calc_nonlinear_u(gs * gs + 0.0625 * (g1s*g1s + g2s*g2s),
			       gs, us, chi2[i], chi3[i]);
	}
      }
      else {
	LOOP_OVER_VOL_OWNED(gv, fc, i) {
	  double gs = g[i]; double us = u[i];
	  f[i] = (gs * us + OFFDIAG(u1,g1,s1) + OFFDIAG(u2,g2,s2));
	}
      }
    }
    else if (u1) { // 2x2 off-diagonal u
      if (chi3) {
	LOOP_OVER_VOL_OWNED(gv, fc, i) {
	  double g1s = g1[i]+g1[i+s]+g1[i-s1]+g1[i+(s-s1)];
	  double gs = g[i]; double us = u[i];
	  f[i] = (gs * us + OFFDIAG(u1,g1,s1))
	    * calc_nonlinear_u(gs * gs + 0.0625 * (g1s*g1s),
			       gs, us, chi2[i], chi3[i]);
	}
      }
      else {
	LOOP_OVER_VOL_OWNED(gv, fc, i) {
	  double gs = g[i]; double us = u[i];
	  f[i] = (gs * us + OFFDIAG(u1,g1,s1));
	}
      }
    }
    else if (u2) { // 2x2 off-diagonal u
      abort("bug - didn't swap off-diagonal terms!?");
    }
    else { // diagonal u
      if (chi3) {
	if (g1 && g2) {
	  LOOP_OVER_VOL_OWNED(gv, fc, i) {
	    double g1s = g1[i]+g1[i+s]+g1[i-s1]+g1[i+(s-s1)];
	    double g2s = g2[i]+g2[i+s]+g2[i-s2]+g2[i+(s-s2)];
	    double gs = g[i]; double us = u[i];
	    f[i] = (gs*us)*calc_nonlinear_u(gs*gs+0.0625*(g1s*g1s+g2s*g2s),
					gs, us, chi2[i], chi3[i]);
	  }
	}
	else if (g1) {
	  LOOP_OVER_VOL_OWNED(gv, fc, i) {
	    double g1s = g1[i]+g1[i+s]+g1[i-s1]+g1[i+(s-s1)];
	    double gs = g[i]; double us = u[i];
	    f[i] = (gs*us)*calc_nonlinear_u(gs*gs + 0.0625*(g1s*g1s),
						gs, us, chi2[i], chi3[i]);
	  }
	}
	else if (g2) {
	  abort("bug - didn't swap off-diagonal terms!?");
	}
	else {
	  LOOP_OVER_VOL_OWNED(gv, fc, i) {
	    double gs = g[i]; double us = u[i];
	    f[i] = (gs*us)*calc_nonlinear_u(gs*gs, gs,us, chi2[i],chi3[i]);
	  }
	}
      }
      else if (u) {
	LOOP_OVER_VOL_OWNED(gv, fc, i) {
	  double gs = g[i]; double us = u[i];
	  f[i] = (gs * us);
	}
      }
      else
	LOOP_OVER_VOL_OWNED(gv, fc, i) f[i] = g[i];
    }
  }
}

} // namespace meep