MFC
Exascale flow solver
Loading...
Searching...
No Matches
m_amr_advance.fpp.f90
Go to the documentation of this file.
1# 1 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
2!>
3!!@file
4!!@brief Contains module m_amr_advance
5
6# 7 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
7# 8 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
8# 1 "/home/runner/work/MFC/MFC/src/common/include/macros.fpp" 1
9# 1 "/home/runner/work/MFC/MFC/src/common/include/parallel_macros.fpp" 1
10# 1 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp" 1
11# 2 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
12# 3 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
13# 4 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
14# 5 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
15# 6 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
16
17# 8 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
18# 9 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
19# 10 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
20
21# 12 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
22# 1 "/home/runner/work/MFC/MFC/src/common/include/defaults/thermochem.fpp" 1
23! Default for Fypp runs outside the toolchain (e.g. documentation). Searched last, so the
24! build/include/<target>/thermochem.fpp the toolchain writes from m_thermochem takes precedence.
25# 4 "/home/runner/work/MFC/MFC/src/common/include/defaults/thermochem.fpp"
26# 5 "/home/runner/work/MFC/MFC/src/common/include/defaults/thermochem.fpp"
27# 13 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp" 2
28
29# 16 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
30# 17 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
31
32# 24 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
33
34# 53 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
35
36# 65 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
37
38# 75 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
39
40# 105 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
41
42# 117 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
43
44# 127 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
45
46# 174 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
47! New line at end of file is required for FYPP
48# 2 "/home/runner/work/MFC/MFC/src/common/include/parallel_macros.fpp" 2
49# 1 "/home/runner/work/MFC/MFC/src/common/include/omp_macros.fpp" 1
50# 1 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp" 1
51# 2 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
52# 3 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
53# 4 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
54# 5 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
55# 6 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
56
57# 8 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
58# 9 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
59# 10 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
60
61# 12 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
62# 1 "/home/runner/work/MFC/MFC/src/common/include/defaults/thermochem.fpp" 1
63! Default for Fypp runs outside the toolchain (e.g. documentation). Searched last, so the
64! build/include/<target>/thermochem.fpp the toolchain writes from m_thermochem takes precedence.
65# 4 "/home/runner/work/MFC/MFC/src/common/include/defaults/thermochem.fpp"
66# 5 "/home/runner/work/MFC/MFC/src/common/include/defaults/thermochem.fpp"
67# 13 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp" 2
68
69# 16 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
70# 17 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
71
72# 24 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
73
74# 53 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
75
76# 65 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
77
78# 75 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
79
80# 105 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
81
82# 117 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
83
84# 127 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
85
86# 174 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
87! New line at end of file is required for FYPP
88# 2 "/home/runner/work/MFC/MFC/src/common/include/omp_macros.fpp" 2
89
90# 4 "/home/runner/work/MFC/MFC/src/common/include/omp_macros.fpp"
91# 5 "/home/runner/work/MFC/MFC/src/common/include/omp_macros.fpp"
92# 6 "/home/runner/work/MFC/MFC/src/common/include/omp_macros.fpp"
93# 7 "/home/runner/work/MFC/MFC/src/common/include/omp_macros.fpp"
94# 8 "/home/runner/work/MFC/MFC/src/common/include/omp_macros.fpp"
95
96# 20 "/home/runner/work/MFC/MFC/src/common/include/omp_macros.fpp"
97
98# 49 "/home/runner/work/MFC/MFC/src/common/include/omp_macros.fpp"
99
100# 54 "/home/runner/work/MFC/MFC/src/common/include/omp_macros.fpp"
101
102# 59 "/home/runner/work/MFC/MFC/src/common/include/omp_macros.fpp"
103
104# 64 "/home/runner/work/MFC/MFC/src/common/include/omp_macros.fpp"
105
106# 69 "/home/runner/work/MFC/MFC/src/common/include/omp_macros.fpp"
107
108# 74 "/home/runner/work/MFC/MFC/src/common/include/omp_macros.fpp"
109
110# 87 "/home/runner/work/MFC/MFC/src/common/include/omp_macros.fpp"
111
112# 101 "/home/runner/work/MFC/MFC/src/common/include/omp_macros.fpp"
113
114# 106 "/home/runner/work/MFC/MFC/src/common/include/omp_macros.fpp"
115
116# 111 "/home/runner/work/MFC/MFC/src/common/include/omp_macros.fpp"
117
118# 116 "/home/runner/work/MFC/MFC/src/common/include/omp_macros.fpp"
119
120# 121 "/home/runner/work/MFC/MFC/src/common/include/omp_macros.fpp"
121
122# 126 "/home/runner/work/MFC/MFC/src/common/include/omp_macros.fpp"
123
124# 131 "/home/runner/work/MFC/MFC/src/common/include/omp_macros.fpp"
125
126# 136 "/home/runner/work/MFC/MFC/src/common/include/omp_macros.fpp"
127
128# 141 "/home/runner/work/MFC/MFC/src/common/include/omp_macros.fpp"
129
130# 171 "/home/runner/work/MFC/MFC/src/common/include/omp_macros.fpp"
131
132# 231 "/home/runner/work/MFC/MFC/src/common/include/omp_macros.fpp"
133
134# 245 "/home/runner/work/MFC/MFC/src/common/include/omp_macros.fpp"
135
136# 270 "/home/runner/work/MFC/MFC/src/common/include/omp_macros.fpp"
137
138# 281 "/home/runner/work/MFC/MFC/src/common/include/omp_macros.fpp"
139
140# 283 "/home/runner/work/MFC/MFC/src/common/include/omp_macros.fpp"
141# 294 "/home/runner/work/MFC/MFC/src/common/include/omp_macros.fpp"
142
143# 343 "/home/runner/work/MFC/MFC/src/common/include/omp_macros.fpp"
144
145# 353 "/home/runner/work/MFC/MFC/src/common/include/omp_macros.fpp"
146
147# 363 "/home/runner/work/MFC/MFC/src/common/include/omp_macros.fpp"
148
149# 372 "/home/runner/work/MFC/MFC/src/common/include/omp_macros.fpp"
150
151# 389 "/home/runner/work/MFC/MFC/src/common/include/omp_macros.fpp"
152
153# 399 "/home/runner/work/MFC/MFC/src/common/include/omp_macros.fpp"
154
155# 406 "/home/runner/work/MFC/MFC/src/common/include/omp_macros.fpp"
156
157# 412 "/home/runner/work/MFC/MFC/src/common/include/omp_macros.fpp"
158
159# 418 "/home/runner/work/MFC/MFC/src/common/include/omp_macros.fpp"
160
161# 424 "/home/runner/work/MFC/MFC/src/common/include/omp_macros.fpp"
162
163# 430 "/home/runner/work/MFC/MFC/src/common/include/omp_macros.fpp"
164
165# 436 "/home/runner/work/MFC/MFC/src/common/include/omp_macros.fpp"
166! New line at end of file is required for FYPP
167# 3 "/home/runner/work/MFC/MFC/src/common/include/parallel_macros.fpp" 2
168# 1 "/home/runner/work/MFC/MFC/src/common/include/acc_macros.fpp" 1
169# 1 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp" 1
170# 2 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
171# 3 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
172# 4 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
173# 5 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
174# 6 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
175
176# 8 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
177# 9 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
178# 10 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
179
180# 12 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
181# 1 "/home/runner/work/MFC/MFC/src/common/include/defaults/thermochem.fpp" 1
182! Default for Fypp runs outside the toolchain (e.g. documentation). Searched last, so the
183! build/include/<target>/thermochem.fpp the toolchain writes from m_thermochem takes precedence.
184# 4 "/home/runner/work/MFC/MFC/src/common/include/defaults/thermochem.fpp"
185# 5 "/home/runner/work/MFC/MFC/src/common/include/defaults/thermochem.fpp"
186# 13 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp" 2
187
188# 16 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
189# 17 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
190
191# 24 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
192
193# 53 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
194
195# 65 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
196
197# 75 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
198
199# 105 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
200
201# 117 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
202
203# 127 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
204
205# 174 "/home/runner/work/MFC/MFC/src/common/include/shared_parallel_macros.fpp"
206! New line at end of file is required for FYPP
207# 2 "/home/runner/work/MFC/MFC/src/common/include/acc_macros.fpp" 2
208
209# 7 "/home/runner/work/MFC/MFC/src/common/include/acc_macros.fpp"
210
211# 17 "/home/runner/work/MFC/MFC/src/common/include/acc_macros.fpp"
212
213# 22 "/home/runner/work/MFC/MFC/src/common/include/acc_macros.fpp"
214
215# 27 "/home/runner/work/MFC/MFC/src/common/include/acc_macros.fpp"
216
217# 32 "/home/runner/work/MFC/MFC/src/common/include/acc_macros.fpp"
218
219# 37 "/home/runner/work/MFC/MFC/src/common/include/acc_macros.fpp"
220
221# 42 "/home/runner/work/MFC/MFC/src/common/include/acc_macros.fpp"
222
223# 47 "/home/runner/work/MFC/MFC/src/common/include/acc_macros.fpp"
224
225# 52 "/home/runner/work/MFC/MFC/src/common/include/acc_macros.fpp"
226
227# 57 "/home/runner/work/MFC/MFC/src/common/include/acc_macros.fpp"
228
229# 62 "/home/runner/work/MFC/MFC/src/common/include/acc_macros.fpp"
230
231# 73 "/home/runner/work/MFC/MFC/src/common/include/acc_macros.fpp"
232
233# 78 "/home/runner/work/MFC/MFC/src/common/include/acc_macros.fpp"
234
235# 83 "/home/runner/work/MFC/MFC/src/common/include/acc_macros.fpp"
236
237# 88 "/home/runner/work/MFC/MFC/src/common/include/acc_macros.fpp"
238
239# 103 "/home/runner/work/MFC/MFC/src/common/include/acc_macros.fpp"
240
241# 131 "/home/runner/work/MFC/MFC/src/common/include/acc_macros.fpp"
242
243# 161 "/home/runner/work/MFC/MFC/src/common/include/acc_macros.fpp"
244
245# 176 "/home/runner/work/MFC/MFC/src/common/include/acc_macros.fpp"
246
247# 194 "/home/runner/work/MFC/MFC/src/common/include/acc_macros.fpp"
248
249# 216 "/home/runner/work/MFC/MFC/src/common/include/acc_macros.fpp"
250
251# 245 "/home/runner/work/MFC/MFC/src/common/include/acc_macros.fpp"
252
253# 260 "/home/runner/work/MFC/MFC/src/common/include/acc_macros.fpp"
254
255# 270 "/home/runner/work/MFC/MFC/src/common/include/acc_macros.fpp"
256
257# 279 "/home/runner/work/MFC/MFC/src/common/include/acc_macros.fpp"
258
259# 295 "/home/runner/work/MFC/MFC/src/common/include/acc_macros.fpp"
260
261# 305 "/home/runner/work/MFC/MFC/src/common/include/acc_macros.fpp"
262
263# 312 "/home/runner/work/MFC/MFC/src/common/include/acc_macros.fpp"
264! New line at end of file is required for FYPP
265# 4 "/home/runner/work/MFC/MFC/src/common/include/parallel_macros.fpp" 2
266
267! GPU parallel region (scalar reductions, maxval/minval)
268# 23 "/home/runner/work/MFC/MFC/src/common/include/parallel_macros.fpp"
269
270! GPU parallel loop over threads (most common GPU macro)
271# 44 "/home/runner/work/MFC/MFC/src/common/include/parallel_macros.fpp"
272
273! Required closing for GPU_PARALLEL_LOOP
274# 56 "/home/runner/work/MFC/MFC/src/common/include/parallel_macros.fpp"
275
276! Mark routine for device compilation
277# 113 "/home/runner/work/MFC/MFC/src/common/include/parallel_macros.fpp"
278
279! Declare device-resident data
280# 131 "/home/runner/work/MFC/MFC/src/common/include/parallel_macros.fpp"
281
282! Inner loop within a GPU parallel region
283# 146 "/home/runner/work/MFC/MFC/src/common/include/parallel_macros.fpp"
284
285! Scoped GPU data region
286# 165 "/home/runner/work/MFC/MFC/src/common/include/parallel_macros.fpp"
287
288! Host code with device pointers (for MPI with GPU buffers)
289# 191 "/home/runner/work/MFC/MFC/src/common/include/parallel_macros.fpp"
290
291! Allocate device memory (unscoped)
292# 205 "/home/runner/work/MFC/MFC/src/common/include/parallel_macros.fpp"
293
294! Free device memory
295# 217 "/home/runner/work/MFC/MFC/src/common/include/parallel_macros.fpp"
296
297! Atomic operation on device
298# 229 "/home/runner/work/MFC/MFC/src/common/include/parallel_macros.fpp"
299
300! End atomic capture block
301# 240 "/home/runner/work/MFC/MFC/src/common/include/parallel_macros.fpp"
302
303! Copy data between host and device
304# 252 "/home/runner/work/MFC/MFC/src/common/include/parallel_macros.fpp"
305
306! Synchronization barrier
307# 264 "/home/runner/work/MFC/MFC/src/common/include/parallel_macros.fpp"
308
309! Import GPU library module (openacc or omp_lib)
310# 273 "/home/runner/work/MFC/MFC/src/common/include/parallel_macros.fpp"
311
312! Emit code only for AMD compiler
313# 280 "/home/runner/work/MFC/MFC/src/common/include/parallel_macros.fpp"
314
315! Emit code for non-Cray compilers
316# 287 "/home/runner/work/MFC/MFC/src/common/include/parallel_macros.fpp"
317
318! Emit code only for Cray compiler
319# 294 "/home/runner/work/MFC/MFC/src/common/include/parallel_macros.fpp"
320
321! Cray compiler directive, emitted from Fypp so the formatter cannot turn it into a comment
322# 301 "/home/runner/work/MFC/MFC/src/common/include/parallel_macros.fpp"
323
324! Emit code for non-NVIDIA compilers
325# 308 "/home/runner/work/MFC/MFC/src/common/include/parallel_macros.fpp"
326
327# 310 "/home/runner/work/MFC/MFC/src/common/include/parallel_macros.fpp"
328# 311 "/home/runner/work/MFC/MFC/src/common/include/parallel_macros.fpp"
329! New line at end of file is required for FYPP
330# 2 "/home/runner/work/MFC/MFC/src/common/include/macros.fpp" 2
331
332# 14 "/home/runner/work/MFC/MFC/src/common/include/macros.fpp"
333
334! Caution: This macro requires the use of a binding script to set CUDA_VISIBLE_DEVICES, such that we have one GPU device per MPI
335! rank. That's because for both cudaMemAdvise (preferred location) and cudaMemPrefetchAsync we use location = device_id = 0. For an
336! example see misc/nvidia_uvm/bind.sh.
337# 52 "/home/runner/work/MFC/MFC/src/common/include/macros.fpp"
338
339! Allocate and create GPU device memory
340# 72 "/home/runner/work/MFC/MFC/src/common/include/macros.fpp"
341
342! Free GPU device memory and deallocate
343# 80 "/home/runner/work/MFC/MFC/src/common/include/macros.fpp"
344
345! Cray-specific GPU pointer setup for vector fields
346# 104 "/home/runner/work/MFC/MFC/src/common/include/macros.fpp"
347
348! Cray-specific GPU pointer setup for scalar fields
349# 120 "/home/runner/work/MFC/MFC/src/common/include/macros.fpp"
350
351! Cray-specific GPU pointer teardown for scalar fields - exact inverse of ACC_SETUP_SFs. Removes the
352! present-table entries ACC_SETUP_SFs added (the scalar_field descriptor and its %sf copyin). REQUIRED when
353! an ACC_SETUP'd field is freed MID-RUN (e.g. AMR slot free on regrid/restart): Cray 'exit data delete'
354! decrements the reference counter, so the lone @:DEALLOCATE(arg%sf) only undoes the @:ALLOCATE create ref
355! and leaves the descriptor + the ACC_SETUP %sf copyin dangling. Call BEFORE @:DEALLOCATE(arg%sf).
356# 140 "/home/runner/work/MFC/MFC/src/common/include/macros.fpp"
357
358! Cray-specific GPU pointer setup for acoustic source spatials
359# 165 "/home/runner/work/MFC/MFC/src/common/include/macros.fpp"
360
361! Whether stress index idx is a shear component. Spelled out because NVHPC 26.9 at -O2 vectorizes any(idx == shear_indices)
362! inside a loop to always true, which doubled the normal-stress elastic energy.
363# 171 "/home/runner/work/MFC/MFC/src/common/include/macros.fpp"
364
365# 177 "/home/runner/work/MFC/MFC/src/common/include/macros.fpp"
366
367# 184 "/home/runner/work/MFC/MFC/src/common/include/macros.fpp"
368! New line at end of file is required for FYPP
369# 9 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp" 2
370
371!> @brief Fine-block stage advance and the immersed-boundary/Lagrange fine services.
373
374 use m_derived_types ! scalar_field, t_box, int_bounds_info
376 use m_mpi_proxy, only: s_mpi_abort
377 use m_rhs, only: s_compute_rhs
384 use m_amr_state
386 use m_amr_store
388 use m_amr_frame
390
391 implicit none
392
393 private
395
396contains
397
398 !> Compute the fine-grid IB state (markers/ghost points/levelset) for every active block from the body geometry (static-body
399 !! AMR), once after the coarse IB setup: swap grid and IB globals to each owned block, run the fine IB pipeline, restore.
400 impure subroutine s_amr_setup_ib()
401
402 integer :: islot, save_cur
403 integer(kind=8), allocatable :: body_gps(:)
404
405 if (.not. amr .or. .not. ib) return
406 save_cur = amr_cur
407 allocate (body_gps(max(num_gbl_ibs, 1))); body_gps = 0_8
409 do islot = 1, amr_num_blocks
410 call s_amr_select_slot(islot)
411 if (.not. amr_rank_owns_block) cycle
412 call s_amr_swap_to_fine()
413 call s_ibm_swap_to_fine(islot, gps_on_device=.false.)
414 call s_ibm_setup_fine()
415 call s_ibm_tally_gps_by_body(body_gps)
416 call s_ibm_restore_from_fine(islot)
418 end do
419 call s_amr_select_slot(save_cur)
420
421 ! The fine-IB image-point stencil is not decomposition-exact across a rank seam: abort if one body's fine ghost points
422 ! appear on more than one rank.
423 call s_ibm_check_bodies_on_one_rank(body_gps)
424
425 end subroutine s_amr_setup_ib
426
427 !> Apply the IB state correction on the current fine block after its RK update, mirroring the coarse s_ibm_correct_state.
428 !! amr_cur is set by the caller.
429 impure subroutine s_amr_ib_correct_fine(q_prim_b, pb_in, mv_in)
430
431 !> the q_prim the block's RHS pass filled (pooled scratch)
432 type(scalar_field), dimension(1:sys_size), intent(inout) :: q_prim_b
433 real(stp), dimension(:,:,:,:,:), intent(inout) :: pb_in, mv_in !< unread: amr prohibits qbmm
434 integer :: loc_cur !< plain scalars for device routines (NVHPC -Minline)
435
436 if (.not. ib) return
437 if (.not. amr_rank_owns_block) return
438 ! the swap moves the whole marker field through the host, and a block with no ghost point needs no correction
439 if (f_ibm_fine_gps(amr_cur) == 0) return
440 call s_amr_swap_to_fine()
441 call s_ibm_swap_to_fine(amr_cur, gps_on_device=.true.)
442 loc_cur = amr_loc_of(amr_cur)
443 call s_amr_br_load(loc_cur)
444 call s_ibm_correct_state(amr_cons_br, q_prim_b, pb_in, mv_in, amr_block_level(amr_cur))
445 loc_cur = amr_loc_of(amr_cur)
446 call s_amr_br_store(loc_cur)
449
450 end subroutine s_amr_ib_correct_fine
451
452 !> The clamps after the IB correction on the current fine block; the species cleanup's report reads the block's own markers.
453 impure subroutine s_amr_fine_hooks_post_ib()
454
455 call s_amr_fine_enter()
456 if (chemistry .and. ib) call s_ibm_load_fine_markers(1, [amr_cur], reshape([m, n, p], [3, 1]), 1, 0)
458 call s_amr_fine_leave()
459
460 end subroutine s_amr_fine_hooks_post_ib
461
462 !> Rebuild the current fine block's IB state from the moving body's position (prescribed motion, moving_ibm==1) by reusing the
463 !! coarse s_update_mib on the swapped-in slot. Must precede s_amr_ib_correct_fine.
464 impure subroutine s_amr_update_mib_fine()
465
466 integer :: i, blo(3), bhi(3)
467 logical :: inside, moving_here
468
469 if (.not. ib) return
470 if (.not. amr_rank_owns_block) return
471 moving_here = .false.
472 ! A moving body must stay inside its block (an edge overlap would silently clip the forcing). Under dynamic regrid the
473 ! expansion left margin max(amr_buf,4) and body + image-point stencil reach (2 coarse cells) must stay contained between
474 ! regrids; on a static block only the body bbox must stay inside. Consecutive contained positions keep every sub-time
475 ! interpolate contained (boxes are convex in the linearly-moving corners).
476 if (any(patch_ib(1:num_ibs)%moving_ibm /= 0)) then
477 do i = 1, num_ibs
478 if (patch_ib(i)%moving_ibm == 0) cycle
479 call s_amr_body_bbox(i, merge(2, 0, amr_regrid_int > 0), blo, bhi)
480 if (.not. f_amr_boxes_overlap(blo, bhi, amr_slots(amr_cur)%region%lo, amr_slots(amr_cur)%region%hi)) cycle
481 inside = all((blo >= amr_slots(amr_cur)%region%lo .and. bhi <= amr_slots(amr_cur)%region%hi) .or. .not. amr_dim)
482 if (.not. inside) then
483 write (0, '(A,I0,A,3ES11.3,A,3ES11.3,A,6I6,A,6I6)') ' [amr] body ', i, ' at ', patch_ib(i)%x_centroid, &
484 & patch_ib(i)%y_centroid, patch_ib(i)%z_centroid, ' vel ', patch_ib(i)%vel(1:3), ' bbox(lo,hi) ', blo, &
485 & bhi, ' block(lo,hi) ', amr_slots(amr_cur)%region%lo, amr_slots(amr_cur)%region%hi
486 call s_mpi_abort('amr with moving ib: the body reached the fine-block boundary; ' &
487 & // 'under dynamic regrid reduce amr_regrid_int or increase amr_buf, ' &
488 & // 'for a static block enlarge it to contain the trajectory')
489 end if
490 moving_here = .true.
491 end do
492 end if
493 ! the recompute rebuilds the whole slot for every body, so skip blocks with no moving body over them
494 if (.not. moving_here) return
495 call s_amr_swap_to_fine()
496 call s_ibm_swap_to_fine(amr_cur, gps_on_device=.true.)
497 call s_update_mib(num_ibs)
500
501 end subroutine s_amr_update_mib_fine
502
503 !> Batched fine advance: owned fine blocks are advanced in batches of up to amr_bat_lim blocks of the same level and (padded)
504 !! extent, stacked two ghost shells apart along the last active dimension, so one s_compute_rhs call and one RK kernel cover a
505 !! batch. Blocks are independent, so grouping order is free. Members read the leader's coordinate arrays in the non-stacked
506 !! dimensions (see s_amr_init_report).
507 impure subroutine s_amr_fine_stage_advance_batched(s, coefs, bc_type, q_T_sf, pb_in, rhs_pb, mv_in, rhs_mv, t_step)
508
509 integer, intent(in) :: s, t_step
510 real(wp), intent(in) :: coefs(4)
511 type(integer_field), dimension(1:num_dims,1:2), intent(in) :: bc_type
512 type(scalar_field), intent(inout) :: q_t_sf
513 real(stp), dimension(:,:,:,:,:), intent(inout) :: pb_in, mv_in
514 real(wp), dimension(:,:,:,:,:), intent(inout) :: rhs_pb, rhs_mv
515 integer :: i, g, ibm
516 integer, allocatable :: my(:)
517 logical, allocatable :: done(:)
518 logical :: last_batch
519 integer :: h_bat_blk_ibm !< plain scalars for device routines (NVHPC -Minline)
520 real(wp) :: h_coefs_1, h_coefs_2, h_coefs_3, h_coefs_4, h_merge_1_wp_dt_igr
521
523 allocate (done(amr_n_my)); done = .false.
524 my = amr_my_blk(1:amr_n_my)
525 call s_amr_ibp_reserve(my)
526 do i = 1, amr_n_my
527 if (done(i)) cycle
528 done(i) = .true.
529 g = my(i)
530 if (amr_adv_level > 0 .and. amr_block_level(g) /= amr_adv_level) cycle ! subcycling: one level at a time
531 call s_amr_form_batch(i, my, done)
532 ! no fine block left undone: this batch's restore must push the coarse grid state (the coarse stage reads it)
533 last_batch = .not. any(.not. done .and. (amr_adv_level == 0 .or. amr_block_level(my) == amr_adv_level))
534 ! leader selected (swap, capture and RK read amr_cur / its extents); members' store columns
535 call s_amr_select_slot(g)
536 amr_bat_ext = [amr_slots(g)%m, amr_slots(g)%n, amr_slots(g)%p]
537 call s_amr_bat_layout()
538 do ibm = 1, amr_bat_n
540 amr_bat_mext(:,ibm) = [amr_slots(amr_bat_blk(ibm))%m, amr_slots(amr_bat_blk(ibm))%n, amr_slots(amr_bat_blk(ibm))%p]
541 end do
542
543# 181 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
544#if defined(MFC_OpenACC)
545# 181 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
546!$acc update device(amr_bat_loc, amr_bat_mext, amr_bat_off)
547# 181 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
548#elif defined(MFC_OpenMP)
549# 181 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
550!$omp target update to(amr_bat_loc, amr_bat_mext, amr_bat_off)
551# 181 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
552#endif
553# 182 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
554 ! step-entry backup for the SSP-RK combination, per member
555 if (s == 1) then
556 do ibm = 1, amr_bat_n
557 h_bat_blk_ibm = amr_bat_blk(ibm)
558 call s_amr_copy_fine_fields(h_bat_blk_ibm)
559 end do
560 end if
561 amr_in_fine_advance = .true.
562 call s_phase_tic(ph_swap)
563 call s_amr_swap_to_fine(widen=.true.) ! the leader's grid, extended into the slab (amr_bat_n > 1)
564 call s_phase_toc(ph_swap)
565 call s_phase_tic(ph_rhs)
567 ! each member's fine markers at its slab offset, for the RHS body-cell zeroing and the IB face forces
568 if (ib) then
570 & amr_bat_w)
572 end if
573 call s_compute_rhs(amr_cons_br, q_t_sf, amr_scr_prim, bc_type, amr_scr_rhs, pb_in, rhs_pb, mv_in, rhs_mv, t_step, s)
574 call s_phase_toc(ph_rhs)
575 call s_phase_tic(ph_swap)
576 call s_amr_restore_coarse(sync_device=last_batch)
577 call s_phase_toc(ph_swap)
578 amr_in_fine_advance = .false.
579 call s_phase_tic(ph_rk)
580 ! IGR folds dt into its RHS, so the update multiplies by 1 there
581 h_coefs_1 = coefs(1)
582 h_coefs_2 = coefs(2)
583 h_coefs_3 = coefs(3)
584 h_coefs_4 = coefs(4)
585 h_merge_1_wp_dt_igr = merge(1._wp, dt, igr)
586 call s_amr_fine_rk_update_batch(amr_bat_n, amr_scr_rhs, h_coefs_1, h_coefs_2, h_coefs_3, h_coefs_4, h_merge_1_wp_dt_igr)
588 call s_amr_batch_post_rk_hooks(coefs, pb_in, mv_in)
589 end if
590 call s_phase_toc(ph_rk)
591 end do
592 amr_bat_n = 0
593 deallocate (done)
594
596
597 !> Start a batch at leader my(i): add later same-level blocks that fit the leader's extent, wasting <= amr_bat_pad of the
598 !! member's cells; marks members done.
599 impure subroutine s_amr_form_batch(i, my, done)
600
601 integer, intent(in) :: i, my(:)
602 logical, intent(inout) :: done(:)
603 integer :: j, g, h
604
605 g = my(i)
606 amr_bat_n = 1; amr_bat_blk(1) = g
607 do j = i + 1, amr_n_my
608 if (amr_bat_n == amr_bat_lim) exit
609 h = my(j)
610 if (done(j) .or. amr_block_level(h) /= amr_block_level(g)) cycle
611 ! padded membership: no larger than the leader in any dim, wasting <= amr_bat_pad of the member's cells
612 if (amr_slots(h)%m > amr_slots(g)%m .or. amr_slots(h)%n > amr_slots(g)%n .or. amr_slots(h)%p > amr_slots(g)%p) cycle
613 if (real((amr_slots(g)%m + 1)*(amr_slots(g)%n + 1)*(amr_slots(g)%p + 1) - (amr_slots(h)%m + 1)*(amr_slots(h)%n + 1) &
614 & *(amr_slots(h)%p + 1), wp) > amr_bat_pad*real((amr_slots(h)%m + 1)*(amr_slots(h)%n + 1)*(amr_slots(h)%p + 1), &
615 & wp)) cycle
616 amr_bat_n = amr_bat_n + 1; amr_bat_blk(amr_bat_n) = h; done(j) = .true.
617 end do
618
619 end subroutine s_amr_form_batch
620
621 !> For the IB face forces on this batch (s_accumulate_ib_face_fluxes): each member's origin relative to the leader's, and the
622 !! boxes its level+1 children cover, in slab indices.
623 impure subroutine s_amr_bat_ib_tables()
624
625 integer :: ibm, kc, d, lead, blk, plo(3), phi(3)
626
627 lead = amr_bat_blk(1)
628 amr_cov_n = 0
629 do ibm = 1, amr_bat_n
630 blk = amr_bat_blk(ibm)
631 amr_bat_xsh(:,ibm) = 0._wp
632 amr_bat_xsh(1, ibm) = amr_slots(blk)%x_cb(-1) - amr_slots(lead)%x_cb(-1)
633 if (n_glb > 0) amr_bat_xsh(2, ibm) = amr_slots(blk)%y_cb(-1) - amr_slots(lead)%y_cb(-1)
634 if (p_glb > 0) amr_bat_xsh(3, ibm) = amr_slots(blk)%z_cb(-1) - amr_slots(lead)%z_cb(-1)
635 if (amr_block_level(blk) >= amr_max_level) cycle
636 do kc = 1, amr_num_blocks
637 if (.not. f_amr_is_child(kc, blk)) cycle
638 call s_amr_parent_foot(kc, blk, plo, phi)
639 amr_cov_n = amr_cov_n + 1
640 do d = 1, 3
641 amr_cov_box(d, amr_cov_n) = plo(d) + amr_bat_off(d, ibm)
642 amr_cov_box(d + 3, amr_cov_n) = phi(d) + amr_bat_off(d, ibm)
643 end do
644 end do
645 end do
646
647# 274 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
648#if defined(MFC_OpenACC)
649# 274 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
650!$acc update device(amr_bat_xsh, amr_cov_box)
651# 274 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
652#elif defined(MFC_OpenMP)
653# 274 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
654!$omp target update to(amr_bat_xsh, amr_cov_box)
655# 274 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
656#endif
657# 275 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
658
659 end subroutine s_amr_bat_ib_tables
660
661 !> Per-member post-RK hooks of the coarse stage (q_cons_ts(1), m_time_steppers).
662 impure subroutine s_amr_batch_post_rk_hooks(coefs, pb_in, mv_in)
663
664 real(wp), intent(in) :: coefs(4)
665 real(stp), dimension(:,:,:,:,:), intent(inout) :: pb_in, mv_in
666 integer :: ibm, nb
667 logical :: pre, post
668
669 ! the coarse stage's post-RK hooks (m_time_steppers) run here per member in its own frame, from the same lists
670 ! (m_stage_hooks) around the same IB correction. Each reads only the member's cells, so member order is free.
671 ! amr_bat_n = 1 while visiting members: s_amr_swap_to_fine would otherwise extend the grid into the slab, and the hooks
672 ! need the member's extents.
673
675 nb = amr_bat_n; amr_bat_n = 1
676 do ibm = 1, nb
678 if (pre .and. amr_rank_owns_block) then
679 ! same ldt as the coarse stage; the synthetic turbulence forcing reads the member's own primitive state
680 if (synthetic_turbulence) call s_amr_bat_member_prim(ibm, amr_scr_prim, amr_scr_prim_blk)
681 call s_amr_fine_enter()
682 if (synthetic_turbulence) then
683 call s_stage_hooks_pre_ib(amr_cons_br, amr_scr_prim_blk, amr_scr_rhs, coefs(3)*dt/coefs(4), .true.)
684 else
685 call s_stage_hooks_pre_ib(amr_cons_br, amr_scr_prim, amr_scr_rhs, coefs(3)*dt/coefs(4), .true.)
686 end if
687 call s_amr_fine_leave()
688 end if
689 if (ib) then
690 ! a moving body's block: the correction and the clamps after it follow the bodies' move (s_amr_fine_stage_ib)
691 if (f_amr_ib_waits(amr_bat_blk(ibm))) then
692 call s_amr_ibp_save(ibm)
693 cycle
694 end if
696 call s_amr_ib_correct_fine(amr_scr_prim_blk, pb_in, mv_in)
697 end if
698 if (post .and. amr_rank_owns_block) call s_amr_fine_hooks_post_ib()
699 end do
700 amr_bat_n = nb
701
702 end subroutine s_amr_batch_post_rk_hooks
703
704 !> Whether block blk's IB correction waits for the bodies to move: it holds ghost points, or a moving body (one coarse cell
705 !! beyond the margin s_amr_update_mib_fine keeps it inside, for this stage's displacement) overlaps it.
706 impure logical function f_amr_ib_waits(blk) result(waits)
707
708 integer, intent(in) :: blk
709 integer :: i, blo(3), bhi(3)
710
711 waits = .false.
712 if (.not. (ib .and. moving_immersed_boundary_flag)) return
713 if (amr_block_owner(blk) /= proc_rank) return
714 waits = f_ibm_fine_gps(blk) > 0
715 do i = 1, num_ibs
716 if (waits) return
717 if (patch_ib(i)%moving_ibm == 0) cycle
718 call s_amr_body_bbox(i, merge(2, 0, amr_regrid_int > 0) + 1, blo, bhi)
719 waits = f_amr_boxes_overlap(blo, bhi, amr_slots(blk)%region%lo, amr_slots(blk)%region%hi)
720 end do
721
722 end function f_amr_ib_waits
723
724 !> Size amr_ibp for the owned blocks my(:) that wait (their columns need not survive a stage) and start the stage's list.
725 impure subroutine s_amr_ibp_reserve(my)
726
727 integer, intent(in) :: my(:)
728 integer :: i, nw
729
730 amr_ibp_n = 0
731 if (.not. (ib .and. moving_immersed_boundary_flag)) return
732 nw = 0
733 do i = 1, size(my)
734 if (f_amr_ib_waits(my(i))) nw = nw + 1
735 end do
736 if (allocated(amr_ibp)) then
737 if (size(amr_ibp, 5) >= nw) return
738#ifdef MFC_DEBUG
739# 355 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
740 block
741# 355 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
742 use iso_fortran_env, only: output_unit
743# 355 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
744
745# 355 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
746 print *, 'm_amr_advance.fpp:355: ', '@:DEALLOCATE(amr_ibp)'
747# 355 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
748
749# 355 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
750 call flush (output_unit)
751# 355 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
752 end block
753# 355 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
754#endif
755# 355 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
756
757# 355 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
758#if defined(MFC_OpenACC)
759# 355 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
760!$acc exit data delete(amr_ibp)
761# 355 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
762#elif defined(MFC_OpenMP)
763# 355 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
764!$omp target exit data map(release:amr_ibp)
765# 355 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
766#endif
767# 355 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
768 deallocate (amr_ibp)
769# 356 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
770 deallocate (amr_ibp_blk)
771 end if
772 if (nw == 0) return
773#ifdef MFC_DEBUG
774# 359 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
775 block
776# 359 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
777 use iso_fortran_env, only: output_unit
778# 359 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
779
780# 359 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
781 print *, 'm_amr_advance.fpp:359: ', '@:ALLOCATE(amr_ibp(mbuf_lo(1):mbuf_hi(1), mbuf_lo(2):mbuf_hi(2), mbuf_lo(3):mbuf_hi(3), 1:sys_size, 1:nw))'
782# 359 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
783
784# 359 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
785 call flush (output_unit)
786# 359 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
787 end block
788# 359 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
789#endif
790# 359 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
791 allocate (amr_ibp(mbuf_lo(1):mbuf_hi(1), mbuf_lo(2):mbuf_hi(2), mbuf_lo(3):mbuf_hi(3), 1:sys_size, 1:nw))
792# 359 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
793
794# 359 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
795
796# 359 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
797#if defined(MFC_OpenACC)
798# 359 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
799!$acc enter data create(amr_ibp)
800# 359 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
801#elif defined(MFC_OpenMP)
802# 359 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
803!$omp target enter data map(always,alloc:amr_ibp)
804# 359 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
805#endif
806# 360 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
807 allocate (amr_ibp_blk(nw))
808
809 end subroutine s_amr_ibp_reserve
810
811 !> Keep batch member ibm's primitive state from the RHS pass (amr_scr_prim, at its slab offset) in the next amr_ibp column.
812 impure subroutine s_amr_ibp_save(ibm)
813
814 integer, intent(in) :: ibm
815 integer :: i, j, k, l, h, q, o1, o2, o3, b1l, b1h, b2l, b2h, b3l, b3h
816
817 h = amr_bat_blk(ibm)
818 amr_ibp_n = amr_ibp_n + 1; q = amr_ibp_n
819 amr_ibp_blk(q) = h
820 o1 = amr_bat_off(1, ibm); o2 = amr_bat_off(2, ibm); o3 = amr_bat_off(3, ibm)
821 b1l = amr_slots(h)%idwbuff(1)%beg; b1h = amr_slots(h)%idwbuff(1)%end
822 b2l = amr_slots(h)%idwbuff(2)%beg; b2h = amr_slots(h)%idwbuff(2)%end
823 b3l = amr_slots(h)%idwbuff(3)%beg; b3h = amr_slots(h)%idwbuff(3)%end
824
825# 377 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
826
827# 377 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
828#if defined(MFC_OpenACC)
829# 377 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
830!$acc parallel loop collapse(4) gang vector default(present) copyin(q, o1, o2, o3, b1l, b1h, b2l, b2h, b3l, b3h)
831# 377 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
832#elif defined(MFC_OpenMP)
833# 377 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
834
835# 377 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
836
837# 377 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
838
839# 377 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
840!$omp target teams loop defaultmap(firstprivate:scalar) bind(teams,parallel) collapse(4) defaultmap(tofrom:aggregate) defaultmap(tofrom:allocatable) defaultmap(tofrom:pointer) &
841# 377 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
842!$omp& map(to:q, o1, o2, o3, b1l, b1h, b2l, b2h, b3l, b3h)
843# 377 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
844#endif
845# 378 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
846 do i = 1, sys_size
847 do l = b3l, b3h
848 do k = b2l, b2h
849 do j = b1l, b1h
850 amr_ibp(j, k, l, i, q) = amr_scr_prim(i)%sf(j + o1, k + o2, l + o3)
851 end do
852 end do
853 end do
854 end do
855
856# 387 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
857#if defined(MFC_OpenACC)
858# 387 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
859!$acc end parallel loop
860# 387 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
861#elif defined(MFC_OpenMP)
862# 387 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
863
864# 387 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
865!$omp end target teams loop
866# 387 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
867#endif
868# 388 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
869
870 end subroutine s_amr_ibp_save
871
872 !> After the bodies have moved on the stage's complete force: the IB rebuild, the correction and the clamps after it, on each
873 !! block s_amr_batch_post_rk_hooks left waiting, from its saved primitive state.
874 impure subroutine s_amr_fine_stage_ib(pb_in, mv_in)
875
876 real(stp), dimension(:,:,:,:,:), intent(inout) :: pb_in, mv_in
877 integer :: i, j, k, l, h, q, b1l, b1h, b2l, b2h, b3l, b3h
878 logical :: post
879
880 if (amr_ibp_n == 0) return
881 post = f_stage_hooks_post_ib()
882 amr_bat_n = 1 ! as s_amr_batch_post_rk_hooks: each block in its own frame
883 do q = 1, amr_ibp_n
884 h = amr_ibp_blk(q)
885 call s_amr_select_slot(h)
887 b1l = amr_slots(h)%idwbuff(1)%beg; b1h = amr_slots(h)%idwbuff(1)%end
888 b2l = amr_slots(h)%idwbuff(2)%beg; b2h = amr_slots(h)%idwbuff(2)%end
889 b3l = amr_slots(h)%idwbuff(3)%beg; b3h = amr_slots(h)%idwbuff(3)%end
890
891# 409 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
892
893# 409 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
894#if defined(MFC_OpenACC)
895# 409 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
896!$acc parallel loop collapse(4) gang vector default(present) copyin(q, b1l, b1h, b2l, b2h, b3l, b3h)
897# 409 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
898#elif defined(MFC_OpenMP)
899# 409 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
900
901# 409 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
902
903# 409 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
904
905# 409 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
906!$omp target teams loop defaultmap(firstprivate:scalar) bind(teams,parallel) collapse(4) defaultmap(tofrom:aggregate) defaultmap(tofrom:allocatable) defaultmap(tofrom:pointer) &
907# 409 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
908!$omp& map(to:q, b1l, b1h, b2l, b2h, b3l, b3h)
909# 409 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
910#endif
911# 410 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
912 do i = 1, sys_size
913 do l = b3l, b3h
914 do k = b2l, b2h
915 do j = b1l, b1h
916 amr_scr_prim_blk(i)%sf(j, k, l) = amr_ibp(j, k, l, i, q)
917 end do
918 end do
919 end do
920 end do
921
922# 419 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
923#if defined(MFC_OpenACC)
924# 419 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
925!$acc end parallel loop
926# 419 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
927#elif defined(MFC_OpenMP)
928# 419 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
929
930# 419 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
931!$omp end target teams loop
932# 419 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
933#endif
934# 420 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
935 call s_amr_ib_correct_fine(amr_scr_prim_blk, pb_in, mv_in)
936 if (post) call s_amr_fine_hooks_post_ib()
937 end do
938 amr_bat_n = 0
939 amr_ibp_n = 0
940 call s_amr_select_slot(1) ! the next stage's coarse RHS captures creg into slot 1
941
942 end subroutine s_amr_fine_stage_ib
943
944 !> The batch's slab layout: members stacked along the last active dim, stride amr_bat_w.
945 impure subroutine s_amr_bat_layout()
946
947 integer :: ibm
948
949 amr_bat_sd = num_dims
950 amr_bat_w = amr_bat_ext(amr_bat_sd) + 2*buff_size + 1
951 do ibm = 1, amr_bat_n
952 amr_bat_off(:,ibm) = 0
953 amr_bat_off(amr_bat_sd, ibm) = (ibm - 1)*amr_bat_w
954 end do
955
956 end subroutine s_amr_bat_layout
957
958 !> Device RK stage update over the batch: q = (c1*q + c2*q_stor + c3*dt_in*rhs)/c4 (compute in wp, store stp), reading each
959 !! member's rhs at its slab offset. Keep in lockstep with the non-IGR rk_coef form in s_tvd_rk (m_time_steppers).
960 impure subroutine s_amr_fine_rk_update_batch(nb, q_rhs, c1, c2, c3, c4, dt_in)
961
962 integer, intent(in) :: nb
963 type(scalar_field), dimension(sys_size), intent(in) :: q_rhs
964 real(wp), intent(in) :: c1, c2, c3, c4, dt_in
965 integer :: ibm, i, fi, fj, fk, fm, fn, fp, loc
966
967 fm = amr_bat_ext(1); fn = amr_bat_ext(2); fp = amr_bat_ext(3)
968
969# 453 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
970
971# 453 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
972#if defined(MFC_OpenACC)
973# 453 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
974!$acc parallel loop collapse(5) gang vector default(present) private(loc) copyin(nb, fm, fn, fp)
975# 453 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
976#elif defined(MFC_OpenMP)
977# 453 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
978
979# 453 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
980
981# 453 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
982
983# 453 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
984!$omp target teams loop defaultmap(firstprivate:scalar) bind(teams,parallel) collapse(5) defaultmap(tofrom:aggregate) defaultmap(tofrom:allocatable) defaultmap(tofrom:pointer) private(loc) &
985# 453 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
986!$omp& map(to:nb, fm, fn, fp)
987# 453 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
988#endif
989# 454 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
990 do ibm = 1, nb
991 do i = 1, sys_size
992 do fk = 0, fp
993 do fj = 0, fn
994 do fi = 0, fm
995 ! padding
996 if (fi > amr_bat_mext(1, ibm) .or. fj > amr_bat_mext(2, ibm) .or. fk > amr_bat_mext(3, ibm)) cycle
997 loc = amr_bat_loc(ibm)
998 amr_cons_st(fi, fj, fk, i, loc) = (c1*real(amr_cons_st(fi, fj, fk, i, loc), &
999 & wp) + c2*real(amr_stor_st(fi, fj, fk, i, loc), &
1000 & wp) + c3*dt_in*real(q_rhs(i)%sf(fi + amr_bat_off(1, ibm), fj + amr_bat_off(2, ibm), &
1001 & fk + amr_bat_off(3, ibm)), wp))/c4
1002 end do
1003 end do
1004 end do
1005 end do
1006 end do
1007
1008# 471 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
1009#if defined(MFC_OpenACC)
1010# 471 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
1011!$acc end parallel loop
1012# 471 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
1013#elif defined(MFC_OpenMP)
1014# 471 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
1015
1016# 471 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
1017!$omp end target teams loop
1018# 471 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
1019#endif
1020# 472 "/home/runner/work/MFC/MFC/src/simulation/m_amr_advance.fpp"
1021
1022 end subroutine s_amr_fine_rk_update_batch
1023
1024end module m_amr_advance
integer, intent(in) k
integer, intent(in) j
integer, intent(in) l
Fine-block stage advance and the immersed-boundary/Lagrange fine services.
impure subroutine s_amr_fine_hooks_post_ib()
The clamps after the IB correction on the current fine block; the species cleanup's report reads the ...
impure subroutine s_amr_form_batch(i, my, done)
Start a batch at leader my(i): add later same-level blocks that fit the leader's extent,...
impure subroutine s_amr_bat_layout()
The batch's slab layout: members stacked along the last active dim, stride amr_bat_w.
impure subroutine, public s_amr_fine_stage_ib(pb_in, mv_in)
After the bodies have moved on the stage's complete force: the IB rebuild, the correction and the cla...
impure subroutine s_amr_ibp_reserve(my)
Size amr_ibp for the owned blocks my(:) that wait (their columns need not survive a stage) and start ...
impure logical function f_amr_ib_waits(blk)
Whether block blk's IB correction waits for the bodies to move: it holds ghost points,...
impure subroutine s_amr_update_mib_fine()
Rebuild the current fine block's IB state from the moving body's position (prescribed motion,...
impure subroutine s_amr_fine_rk_update_batch(nb, q_rhs, c1, c2, c3, c4, dt_in)
Device RK stage update over the batch: q = (c1*q + c2*q_stor + c3*dt_in*rhs)/c4 (compute in wp,...
impure subroutine s_amr_ibp_save(ibm)
Keep batch member ibm's primitive state from the RHS pass (amr_scr_prim, at its slab offset) in the n...
impure subroutine, public s_amr_setup_ib()
Compute the fine-grid IB state (markers/ghost points/levelset) for every active block from the body g...
impure subroutine, public s_amr_fine_stage_advance_batched(s, coefs, bc_type, q_t_sf, pb_in, rhs_pb, mv_in, rhs_mv, t_step)
Batched fine advance: owned fine blocks are advanced in batches of up to amr_bat_lim blocks of the sa...
impure subroutine s_amr_batch_post_rk_hooks(coefs, pb_in, mv_in)
Per-member post-RK hooks of the coarse stage (q_cons_ts(1), m_time_steppers).
impure subroutine s_amr_ib_correct_fine(q_prim_b, pb_in, mv_in)
Apply the IB state correction on the current fine block after its RK update, mirroring the coarse s_i...
impure subroutine s_amr_bat_ib_tables()
For the IB face forces on this batch (s_accumulate_ib_face_fluxes): each member's origin relative to ...
Rank decomposition, box arithmetic and block ownership (SFC cut, owner lookups, owned-block lists).
impure subroutine, public s_amr_refresh_my_blocks()
Refresh the owned-block list if any owner write has happened since the last refresh....
impure subroutine, public s_amr_body_bbox(i, mrg, blo, bhi)
Margin-padded global coarse-index bounding box of immersed body i (supported analytic geometries only...
Parent/child data exchange: gather plans, pack/unpack and the fill waves.
Block-frame swap (fine grid state in/out of the shared solver).
impure subroutine, public s_amr_fine_enter()
Swap to the current fine block and load its solver frame amr_cons_br; paired with s_amr_fine_leave.
impure subroutine, public s_amr_swap_to_fine(widen)
Swap the global grid state to the fine block. Must be paired with s_amr_restore_coarse.
impure subroutine, public s_amr_restore_coarse(sync_device)
Restore the global grid state saved by s_amr_swap_to_fine.
impure subroutine, public s_amr_fine_leave()
Store the solver frame back to the fine block and restore the coarse grid.
AMR flux registers: per-RK-stage refluxing at the coarse/fine block boundary. Depends only on m_deriv...
pure subroutine, public s_amr_parent_foot(k, pblk, plo, phi)
Parent-fine footprint of block k inside its parent pblk from replicated metadata, so every rank gets ...
pure logical function, public f_amr_is_child(kc, g)
Block kc is a level+1 child of block g: one level finer with an overlapping region box.
Module-scope AMR state (block pool, flat field store, wave buffers) and the leaf helpers every AMR mo...
real(stp), dimension(:,:,:,:,:), allocatable amr_ibp
Moving bodies: the primitive state of the RHS pass of each block whose IB correction waits for the bo...
integer amr_ibp_n
columns in use this stage
type(scalar_field), dimension(:), allocatable amr_scr_prim_blk
block-frame prim scratch for the batched per-member IB correction (ib only); the slab prim stacks mem...
type(t_level), dimension(:), allocatable amr_slots
Fixed pool of refined-block slots; the working slot amr_cur selects which one per-block routines oper...
type(scalar_field), dimension(:), allocatable amr_cons_br
Copy bridge to the shared solver, whose routines take type(scalar_field) and serve the monolithic pat...
type(scalar_field), dimension(:), allocatable amr_scr_rhs
integer, dimension(:), allocatable amr_loc_of
Dense local index for live slots: amr_slots is sparse over 1:amr_max_blocks, the field store is dense...
pure logical function f_amr_boxes_overlap(alo, ahi, blo, bhi)
Do two coarse-index boxes [alo:ahi] and [blo:bhi] overlap? Collapsed dims never disqualify.
integer, dimension(amr_bat_max) amr_bat_loc
Batched advance: the bridge spans amr_br_batch blocks along the last active dimension,...
integer, dimension(:), allocatable amr_ibp_blk
the block of each amr_ibp column
integer, dimension(:), allocatable amr_my_blk
Block indices this rank owns, ascending (the paired MPI_SENDRECVs depend on the order); built in s_am...
type(scalar_field), dimension(:), allocatable amr_scr_prim
Pooled advance scratch shared by every fine block (the fused advance leaves no cross-block lifetime);...
Flat per-block field store: dense slot indexing, reservation, alloc/free and the prim/bridge loaders.
impure subroutine, public s_amr_br_load_batch(nb)
Batched bridge load: every batch member (amr_bat_blk/amr_bat_loc, extents amr_bat_ext) lands at offse...
impure subroutine, public s_amr_bat_member_prim(ibm, src, dst)
Copy batch member ibm's primitive state from the slab scratch into the block-frame scratch over its b...
impure subroutine, public s_amr_br_store(loc)
impure subroutine, public s_amr_copy_fine_fields(k)
Device copy amr_cons_st -> amr_stor_st over block k's buffered box, all sys_size fields (RK step-entr...
impure subroutine, public s_amr_br_load(loc)
Move block loc's conserved state between the flat store and the bridge over the whole buffered box (a...
Prolongation, restriction and Berger-Colella reflux, including the restrict/reflux/freg waves.
Shared derived types for field data, patch geometry, bubble dynamics, and MPI I/O structures.
Global parameters for the computational domain, fluid properties, and simulation algorithm configurat...
real(wp), parameter amr_bat_pad
A block may join a batch led by a larger block when padding it to the leader's extent wastes at most ...
integer proc_rank
Rank of the local processor.
integer, dimension(amr_bat_max) amr_bat_blk
integer, dimension(:), allocatable amr_block_level
Refinement level of each active block (1..amr_max_level) and the deepest level populated....
integer, dimension(3) amr_bat_ext
real(wp), dimension(3, amr_bat_max) amr_bat_xsh
For the IB face forces on a fine frame: each batch member's origin minus the leader's (members read t...
integer amr_adv_level
the level the batched fine advance is restricted to (0: every refined level)
logical, dimension(3) amr_dim
The active dimensions, and this rank's coarse subdomain origin and extent padded to 3-vectors (0 in c...
integer, dimension(:,:), allocatable amr_cov_box
integer, dimension(3, amr_bat_max) amr_bat_mext
integer, dimension(:), allocatable amr_block_owner
Fine-level distribution map: SFC/work-balanced single-owner rank per active block,...
integer amr_num_blocks
Number of active AMR fine-block slots (>= 1) and the working slot index selecting which one the per-b...
integer, dimension(3, amr_bat_max) amr_bat_off
logical amr_rank_owns_block
true on the current block's single owner rank: amr_block_owner(amr_cur) == proc_rank; kept by s_set_a...
subroutine s_amr_select_slot(islot)
Make block slot islot the working slot: set amr_cur and copy its stored mirrors (region,...
Ghost-node immersed boundary method: locates ghost/image points, computes interpolation coefficients,...
impure subroutine s_update_mib(num_ibs)
impure subroutine, public s_ibm_load_fine_markers(nb, slots, mext, sd, w)
Load ib_markers_fine for the next fine RHS pass: member ibm's fine markers (interior 0:mext(:,...
impure subroutine, public s_ibm_setup_fine()
Compute the fine-grid IB state (markers, ghost points, levelset, image points, interpolation coeffs) ...
logical moving_immersed_boundary_flag
impure subroutine, public s_ibm_tally_gps_by_body(cnt)
Add to cnt(1:num_gbl_ibs) the ghost points each body has in the active list (the fine slot's when swa...
impure integer function, public f_ibm_fine_gps(islot)
Ghost points stored for fine slot islot: zero for a block that holds no immersed body.
subroutine, public s_ibm_correct_state(q_cons_vf, q_prim_vf, pb_in, mv_in, lvl)
Update the conservative variables at the ghost points.
integer, public num_gps
Number of ghost points.
impure subroutine, public s_ibm_check_bodies_on_one_rank(cnt)
Abort if any one body has ghost points on more than one rank: the fine image-point stencil is exact o...
impure subroutine, public s_ibm_reset_fine_park()
Forget every slot's parked fine list (before the fine lists are rebuilt for the current block set).
impure subroutine, public s_ibm_swap_to_fine(islot, gps_on_device)
Swap the module IB globals (ib_markers/ghost_points/num_gps) to fine slot islot's state; the coarse s...
impure subroutine, public s_ibm_restore_from_fine(islot)
Restore the coarse IB globals saved by s_ibm_swap_to_fine, parking the (possibly updated) fine state ...
MPI halo exchange, domain decomposition, and buffer packing/unpacking for the simulation solver.
Per-phase wall-time budget of the AMR step, reported when rank_time_wrt is on. Brackets are disjoint ...
integer, parameter, public ph_rk
fine-block RK update + IB
integer, parameter, public ph_rhs
fine-block RHS
impure subroutine, public s_phase_tic(id)
integer, parameter, public ph_swap
per-batch grid-state swap and restore
impure subroutine, public s_phase_toc(id)
Assemble the right-hand side of the governing equations using finite-volume flux differencing,...
impure subroutine, public s_compute_rhs(q_cons_vf, q_t_sf, q_prim_vf, bc_type, rhs_vf, pb_in, rhs_pb, mv_in, rhs_mv, t_step, stage)
Compute the right-hand side of the semi-discrete governing equations for a single time stage.
The hooks applied to the conservative state after each RK stage update, as one ordered list shared by...
impure subroutine, public s_stage_hooks_pre_ib(q_cons, q_prim, rhs, ldt, on_block, q_t_sf)
Sources and closures, before the IB correction. on_block: q_cons is an AMR fine block in its own fram...
impure subroutine, public s_stage_hooks_post_ib(q_cons, lvl)
Clamps, after the IB correction. lvl: the AMR level of the frame (0 = the coarse grid),...
pure logical function, public f_stage_hooks_post_ib()
Whether s_stage_hooks_post_ib does anything.
pure logical function, public f_stage_hooks_pre_ib_on_block()
Whether s_stage_hooks_pre_ib does anything on a fine block (it costs a frame swap there).
Derived type annexing an integer scalar field (SF).
Derived type annexing a scalar field (SF).