MFC
Exascale flow solver
Loading...
Searching...
No Matches
m_amr_exchange.fpp.f90
Go to the documentation of this file.
1# 1 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
2!>
3!!@file
4!!@brief Contains module m_amr_exchange
5
6# 7 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
7# 8 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.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_exchange.fpp" 2
370
371!> @brief Parent/child data exchange: gather plans, pack/unpack and the fill waves.
373
376 use m_mpi_proxy, only: s_mpi_abort ! @:ASSERT expands to it
379 use m_amr_state
380 use m_amr_wave
382 use m_amr_ghost
384
385 implicit none
386
387 private
391
392contains
393
394 !> Parent value for a level >= 2 ghost fill: the current state, or its blend with the substep-start backup when th < 1.
395 pure real(wp) function f_amr_par_val(g1, g2, g3, i, qp, th) result(v)
396
397# 35 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
398#if MFC_OpenACC
399# 35 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
400!$acc routine seq
401# 35 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
402#elif MFC_OpenMP
403# 35 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
404
405# 35 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
406
407# 35 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
408!$omp declare target device_type(any)
409# 35 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
410#endif
411# 36 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
412 integer, intent(in) :: g1, g2, g3, i, qp
413 real(wp), intent(in) :: th
414
415 if (th < 1._wp) then
416 v = th*real(amr_cons_st(g1, g2, g3, i, qp), wp) + (1._wp - th)*real(amr_stor_st(g1, g2, g3, i, qp), wp)
417 else
418 v = real(amr_cons_st(g1, g2, g3, i, qp), wp)
419 end if
420
421 end function f_amr_par_val
422
423 !> Sub-box variants of the parent-patch pack/copy for the ring-clipped parent-fill wave; bl:bh are patch-local, the buffer
424 !! layout matches the full-patch kernels.
425 impure subroutine s_amr_pack_parent_box_device(qp, bl, bh, buf)
426
427 integer, intent(in) :: qp !< parent's flat-store slot
428 integer, intent(in) :: bl(3), bh(3)
429 real(wp), intent(inout), contiguous :: buf(:)
430 integer :: i, g1, g2, g3, o1, o2, o3, n1, n2, n3, l1, l2, l3, u1, u2, u3
431 real(wp) :: th
432
433 o1 = amr_cpat_off(1); o2 = amr_cpat_off(2); o3 = amr_cpat_off(3)
434 l1 = bl(1); l2 = bl(2); l3 = bl(3); u1 = bh(1); u2 = bh(2); u3 = bh(3)
435 n1 = u1 - l1 + 1; n2 = u2 - l2 + 1; n3 = u3 - l3 + 1
436 th = amr_par_theta
437
438# 61 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
439
440# 61 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
441#if defined(MFC_OpenACC)
442# 61 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
443!$acc parallel loop collapse(4) gang vector default(present) copyin(th) copyout(buf)
444# 61 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
445#elif defined(MFC_OpenMP)
446# 61 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
447
448# 61 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
449
450# 61 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
451
452# 61 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
453!$omp target teams loop defaultmap(firstprivate:scalar) bind(teams,parallel) collapse(4) defaultmap(tofrom:aggregate) defaultmap(tofrom:allocatable) defaultmap(tofrom:pointer) map(to:th) map(from:buf)
454# 61 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
455#endif
456# 62 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
457 do i = 1, sys_size
458 do g3 = l3, u3
459 do g2 = l2, u2
460 do g1 = l1, u1
461 buf(1 + (g1 - l1) + n1*((g2 - l2) + n2*((g3 - l3) + n3*(i - 1)))) = f_amr_par_val(g1 + o1, g2 + o2, &
462 & g3 + o3, i, qp, th)
463 end do
464 end do
465 end do
466 end do
467
468# 72 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
469#if defined(MFC_OpenACC)
470# 72 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
471!$acc end parallel loop
472# 72 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
473#elif defined(MFC_OpenMP)
474# 72 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
475
476# 72 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
477!$omp end target teams loop
478# 72 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
479#endif
480# 73 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
481
482 end subroutine s_amr_pack_parent_box_device
483
484 impure subroutine s_amr_copy_parent_box(qp, bl, bh)
485
486 integer, intent(in) :: qp !< parent's flat-store slot
487 integer, intent(in) :: bl(3), bh(3)
488 integer :: i, g1, g2, g3, o1, o2, o3, l1, l2, l3, u1, u2, u3
489 real(wp) :: th
490
491 o1 = amr_cpat_off(1); o2 = amr_cpat_off(2); o3 = amr_cpat_off(3)
492 l1 = bl(1); l2 = bl(2); l3 = bl(3); u1 = bh(1); u2 = bh(2); u3 = bh(3)
493 th = amr_par_theta
494
495# 86 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
496
497# 86 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
498#if defined(MFC_OpenACC)
499# 86 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
500!$acc parallel loop collapse(4) gang vector default(present) copyin(th)
501# 86 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
502#elif defined(MFC_OpenMP)
503# 86 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
504
505# 86 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
506
507# 86 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
508
509# 86 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
510!$omp target teams loop defaultmap(firstprivate:scalar) bind(teams,parallel) collapse(4) defaultmap(tofrom:aggregate) defaultmap(tofrom:allocatable) defaultmap(tofrom:pointer) map(to:th)
511# 86 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
512#endif
513# 87 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
514 do i = 1, sys_size
515 do g3 = l3, u3
516 do g2 = l2, u2
517 do g1 = l1, u1
518 amr_cg(i)%sf(g1, g2, g3) = f_amr_par_val(g1 + o1, g2 + o2, g3 + o3, i, qp, th)
519 end do
520 end do
521 end do
522 end do
523
524# 96 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
525#if defined(MFC_OpenACC)
526# 96 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
527!$acc end parallel loop
528# 96 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
529#elif defined(MFC_OpenMP)
530# 96 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
531
532# 96 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
533!$omp end target teams loop
534# 96 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
535#endif
536# 97 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
537
538 end subroutine s_amr_copy_parent_box
539
540 !> Pack (pack = T) the global box [bl:bh] of q_coarse into the wire buffer buf, or unpack it into amr_cg at patch-local index g
541 !! - o (pack = F; stp cast). One element order (i slowest, x fastest), the same wp cast as the host pack in
542 !! s_amr_gather_coarse_patch.
543# 104 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
544 impure subroutine s_amr_pack_box_device(q_coarse, bl, bh, o1, o2, o3, buf)
545# 105 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
546
547# 107 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
548 type(scalar_field), dimension(sys_size), intent(in) :: q_coarse
549 integer, intent(in) :: bl(3), bh(3), o1, o2, o3
550 real(wp), intent(inout), contiguous :: buf(:)
551# 114 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
552 integer :: i, g1, g2, g3, l1, l2, l3, u1, u2, u3, n1, n2, n3
553# 118 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
554
555 l1 = bl(1); l2 = bl(2); l3 = bl(3); u1 = bh(1); u2 = bh(2); u3 = bh(3)
556 n1 = u1 - l1 + 1; n2 = u2 - l2 + 1; n3 = u3 - l3 + 1
557# 124 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
558# 125 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
559
560# 125 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
561
562# 125 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
563#if defined(MFC_OpenACC)
564# 125 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
565!$acc parallel loop collapse(4) gang vector default(present) copyout(buf)
566# 125 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
567#elif defined(MFC_OpenMP)
568# 125 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
569
570# 125 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
571
572# 125 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
573
574# 125 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
575!$omp target teams loop defaultmap(firstprivate:scalar) bind(teams,parallel) collapse(4) defaultmap(tofrom:aggregate) defaultmap(tofrom:allocatable) defaultmap(tofrom:pointer) map(from:buf)
576# 125 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
577#endif
578# 126 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
579# 129 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
580 do i = 1, sys_size
581 do g3 = l3, u3
582 do g2 = l2, u2
583 do g1 = l1, u1
584# 134 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
585 buf(1 + (g1 - l1) + n1*((g2 - l2) + n2*((g3 - l3) + n3*(i - 1)))) = real(q_coarse(i)%sf(g1 - o1, &
586 & g2 - o2, g3 - o3), wp)
587# 140 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
588 end do
589 end do
590 end do
591 end do
592
593# 144 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
594#if defined(MFC_OpenACC)
595# 144 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
596!$acc end parallel loop
597# 144 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
598#elif defined(MFC_OpenMP)
599# 144 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
600
601# 144 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
602!$omp end target teams loop
603# 144 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
604#endif
605# 145 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
606
607 end subroutine s_amr_pack_box_device
608# 147 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
609# 104 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
610 impure subroutine s_amr_unpack_box_device(bl, bh, off, buf)
611# 105 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
612
613# 111 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
614 integer, intent(in) :: bl(3), bh(3), off(3)
615 real(wp), intent(in), contiguous :: buf(:)
616# 114 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
617 integer :: i, g1, g2, g3, l1, l2, l3, u1, u2, u3, n1, n2, n3
618# 116 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
619 integer :: o1, o2, o3
620# 118 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
621
622 l1 = bl(1); l2 = bl(2); l3 = bl(3); u1 = bh(1); u2 = bh(2); u3 = bh(3)
623 n1 = u1 - l1 + 1; n2 = u2 - l2 + 1; n3 = u3 - l3 + 1
624# 122 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
625 o1 = off(1); o2 = off(2); o3 = off(3)
626# 124 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
627# 127 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
628
629# 127 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
630
631# 127 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
632#if defined(MFC_OpenACC)
633# 127 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
634!$acc parallel loop collapse(4) gang vector default(present) copyin(buf)
635# 127 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
636#elif defined(MFC_OpenMP)
637# 127 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
638
639# 127 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
640
641# 127 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
642
643# 127 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
644!$omp target teams loop defaultmap(firstprivate:scalar) bind(teams,parallel) collapse(4) defaultmap(tofrom:aggregate) defaultmap(tofrom:allocatable) defaultmap(tofrom:pointer) map(to:buf)
645# 127 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
646#endif
647# 128 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
648# 129 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
649 do i = 1, sys_size
650 do g3 = l3, u3
651 do g2 = l2, u2
652 do g1 = l1, u1
653# 137 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
654 amr_cg(i)%sf(g1 - o1, g2 - o2, &
655 & g3 - o3) = real(buf(1 + (g1 - l1) + n1*((g2 - l2) + n2*((g3 - l3) + n3*(i - 1)))), stp)
656# 140 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
657 end do
658 end do
659 end do
660 end do
661
662# 144 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
663#if defined(MFC_OpenACC)
664# 144 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
665!$acc end parallel loop
666# 144 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
667#elif defined(MFC_OpenMP)
668# 144 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
669
670# 144 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
671!$omp end target teams loop
672# 144 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
673#endif
674# 145 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
675
676 end subroutine s_amr_unpack_box_device
677# 147 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
678# 148 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
679
680 !> Ring clip: split the hollow shell of patch [plo:phi] minus open core [clo:chi] into at most 6 disjoint slabs in fixed order
681 !! (x, y within core x, z within core x and y). Collapsed dims pass 0; the width-1 double cover is clamped. Both sides derive
682 !! the list from replicated metadata.
683 impure subroutine s_amr_shell_slabs(plo, phi, clo, chi, ns, sb, se, cells)
684
685 integer, intent(in) :: plo(3), phi(3), clo(3), chi(3)
686 integer, intent(out) :: ns, sb(3, 6), se(3, 6), cells
687 integer :: cb(3, 6), ce(3, 6), s, ss
688 integer(8) :: words, patchw, corew
689
690 cb(:,1) = plo; ce(:,1) = [clo(1) - 1, phi(2), phi(3)]
691 cb(:,2) = [max(chi(1) + 1, clo(1)), plo(2), plo(3)]; ce(:,2) = phi
692 cb(:,3) = [clo(1), plo(2), plo(3)]; ce(:,3) = [chi(1), clo(2) - 1, phi(3)]
693 cb(:,4) = [clo(1), max(chi(2) + 1, clo(2)), plo(3)]; ce(:,4) = [chi(1), phi(2), phi(3)]
694 cb(:,5) = [clo(1), clo(2), plo(3)]; ce(:,5) = [chi(1), chi(2), clo(3) - 1]
695 cb(:,6) = [clo(1), clo(2), max(chi(3) + 1, clo(3))]; ce(:,6) = [chi(1), chi(2), phi(3)]
696 ns = 0; words = 0
697 do s = 1, 6
698 if (any(cb(:,s) > ce(:,s))) cycle
699 ns = ns + 1
700 sb(:,ns) = cb(:,s); se(:,ns) = ce(:,s)
701 words = words + product(int(ce(:,s) - cb(:,s) + 1, 8))
702 end do
703 ! slabs must tile the shell exactly
704 do s = 1, ns - 1
705 do ss = s + 1, ns
706 if (.not. (any(max(sb(:, s), sb(:, ss)) > min(se(:, s), se(:, ss))))) then
707# 175 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
708 call s_mpi_abort("m_amr_exchange.fpp:175: " // "Assertion failed: any(max(sb(:, s), sb(:, ss)) > min(se(:, s), se(:, ss))). " &
709# 175 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
710 & // "shell slabs: overlap")
711# 175 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
712 end if
713# 176 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
714 end do
715 end do
716 patchw = int(phi(1) - plo(1) + 1, 8)*int(phi(2) - plo(2) + 1, 8)*int(phi(3) - plo(3) + 1, 8)
717 corew = int(max(chi(1) - clo(1) + 1, 0), 8)*int(max(chi(2) - clo(2) + 1, 0), 8)*int(max(chi(3) - clo(3) + 1, 0), 8)
718 if (.not. (words == patchw - corew)) then
719# 180 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
720 call s_mpi_abort("m_amr_exchange.fpp:180: " // "Assertion failed: words == patchw - corew. " &
721# 180 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
722 & // "shell slabs: coverage mismatch")
723# 180 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
724 end if
725# 181 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
726 cells = int(words)
727
728 end subroutine s_amr_shell_slabs
729
730 !> Intersect the shell slabs with box [bl:bh], keeping order; also returns the total cell count.
731 impure subroutine s_amr_shell_clip(ns, sb, se, bl, bh, ms, tb, te, cells)
732
733 integer, intent(in) :: ns, sb(3, 6), se(3, 6), bl(3), bh(3)
734 integer, intent(out) :: ms, tb(3, 6), te(3, 6), cells
735 integer :: s, l(3), u(3)
736
737 ms = 0; cells = 0
738 do s = 1, ns
739 l = max(sb(:,s), bl); u = min(se(:,s), bh)
740 if (any(l > u)) cycle
741 ms = ms + 1
742 tb(:,ms) = l; te(:,ms) = u
743 cells = cells + product(u - l + 1)
744 end do
745
746 end subroutine s_amr_shell_clip
747
748#ifdef MFC_DEBUG
749 !> Debug only: NaN-flood amr_cg before a clipped gather so any read of an unshipped cell NaNs the ghost fill.
750 impure subroutine s_amr_poison_patch_device(w1, w2, w3)
751
752 use ieee_arithmetic, only: ieee_value, ieee_quiet_nan
753 integer, intent(in) :: w1, w2, w3
754 integer :: i, g1, g2, g3
755 real(stp) :: nanv
756
757 nanv = real(ieee_value(0._wp, ieee_quiet_nan), stp)
758
759# 213 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
760
761# 213 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
762#if defined(MFC_OpenACC)
763# 213 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
764!$acc parallel loop collapse(4) gang vector default(present)
765# 213 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
766#elif defined(MFC_OpenMP)
767# 213 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
768
769# 213 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
770
771# 213 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
772
773# 213 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
774!$omp target teams loop defaultmap(firstprivate:scalar) bind(teams,parallel) collapse(4) defaultmap(tofrom:aggregate) defaultmap(tofrom:allocatable) defaultmap(tofrom:pointer)
775# 213 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
776#endif
777# 214 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
778 do i = 1, sys_size
779 do g3 = 0, w3
780 do g2 = 0, w2
781 do g1 = 0, w1
782 amr_cg(i)%sf(g1, g2, g3) = nanv
783 end do
784 end do
785 end do
786 end do
787
788# 223 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
789#if defined(MFC_OpenACC)
790# 223 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
791!$acc end parallel loop
792# 223 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
793#elif defined(MFC_OpenMP)
794# 223 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
795
796# 223 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
797!$omp end target teams loop
798# 223 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
799#endif
800# 224 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
801
802 end subroutine s_amr_poison_patch_device
803#endif
804
805 !> Ring-clipped own-box copy on device: queue shell-slab/own-box intersections as slab rows and copy q_coarse into amr_cg in one
806 !! fused kernel.
807 impure subroutine s_amr_gather_own_rows(q_coarse, cg, o1, o2, o3)
808
809 type(scalar_field), dimension(sys_size), intent(in) :: q_coarse
810 type(scalar_field), dimension(sys_size), intent(inout) :: cg !< amr_cg, passed so a regrown pool maps as an argument
811 integer, intent(in) :: o1, o2, o3
812 integer :: i, s, g, r, n1, n2, g1, g2, g3, stot, nsl, lo, hi, md
813
814 stot = f_amr_slab_push(nsl)
815 if (stot == 0) return ! no level-1 block on this rank: the store may be unallocated
816
817# 239 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
818
819# 239 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
820#if defined(MFC_OpenACC)
821# 239 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
822!$acc parallel loop collapse(2) gang vector default(present) private(s, lo, hi, md, r, n1, n2, g1, g2, g3)
823# 239 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
824#elif defined(MFC_OpenMP)
825# 239 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
826
827# 239 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
828
829# 239 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
830
831# 239 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
832!$omp target teams loop defaultmap(firstprivate:scalar) bind(teams,parallel) collapse(2) defaultmap(tofrom:aggregate) defaultmap(tofrom:allocatable) defaultmap(tofrom:pointer) &
833# 239 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
834!$omp& private(s, lo, hi, md, r, n1, n2, g1, g2, g3)
835# 239 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
836#endif
837# 240 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
838 do i = 1, sys_size
839 do g = 0, stot - 1
840 lo = 1; hi = nsl
841 do while (lo < hi)
842 md = (lo + hi + 1)/2
843 if (amr_slab_tab(7, md) <= g) then; lo = md; else; hi = md - 1; end if
844 end do
845 s = lo
846 r = g - amr_slab_tab(7, s)
847 n1 = amr_slab_tab(4, s) - amr_slab_tab(1, s) + 1; n2 = amr_slab_tab(5, s) - amr_slab_tab(2, s) + 1
848 g1 = amr_slab_tab(1, s) + mod(r, n1)
849 g2 = amr_slab_tab(2, s) + mod(r/n1, n2)
850 g3 = amr_slab_tab(3, s) + r/(n1*n2)
851 cg(i)%sf(g1 - amr_slab_tab(13, s), g2 - amr_slab_tab(14, s), g3 - amr_slab_tab(15, s)) = q_coarse(i)%sf(g1 - o1, &
852 & g2 - o2, g3 - o3)
853 end do
854 end do
855
856# 257 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
857#if defined(MFC_OpenACC)
858# 257 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
859!$acc end parallel loop
860# 257 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
861#elif defined(MFC_OpenMP)
862# 257 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
863
864# 257 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
865!$omp end target teams loop
866# 257 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
867#endif
868# 258 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
869
870 end subroutine s_amr_gather_own_rows
871
872 !> Fused-pack plan for transfers 1:nt: slab corner, extents, absolute pool payload offset and exclusive element prefix
873 !! amr_fx_pre. Rows 8:11 are left to the parent-store pack site.
874 impure subroutine s_amr_fx_plan(w)
875
876 type(t_amr_wave_side), intent(in) :: w
877 integer :: t, e, nt
878
879 nt = w%nx
880 if (allocated(amr_fx_pl)) then
881 if (size(amr_fx_pl, 2) < nt) deallocate (amr_fx_pl, amr_fx_pre)
882 end if
883 if (.not. allocated(amr_fx_pl)) allocate (amr_fx_pl(11, max(nt, 64)), amr_fx_pre(max(nt, 64) + 1))
884 e = 0
885 do t = 1, nt
886 amr_fx_pl(1:3,t) = w%bl(:,t)
887 amr_fx_pl(4:6,t) = w%bh(:,t) - w%bl(:,t) + 1
888 amr_fx_pl(7, t) = w%off(t)
889 amr_fx_pre(t) = e
890 e = e + sys_size*amr_fx_pl(4, t)*amr_fx_pl(5, t)*amr_fx_pl(6, t)
891 end do
892 amr_fx_pre(nt + 1) = e
893
894 end subroutine s_amr_fx_plan
895
896 !> Longest run t0:t1 of one box's recv transfers that is contiguous in the wire pool (unpacks in one launch).
897 pure subroutine s_amr_fx_run(k, blk, nx, t0, t1)
898
899 integer, intent(in) :: k, blk(:), nx, t0
900 integer, intent(out) :: t1
901
902 t1 = t0
903 do while (t1 < nx)
904 if (blk(t1 + 1) /= k) exit
905 if (amr_fx_pl(7, t1 + 1) /= amr_fx_pl(7, t1) + amr_fx_pre(t1 + 1) - amr_fx_pre(t1)) exit
906 t1 = t1 + 1
907 end do
908
909 end subroutine s_amr_fx_run
910
911 !> Fused coarse-field pack: the wave's whole send list in one launch; inverts the linear index s_amr_pack_box_device writes, so
912 !! wire bytes match the per-transfer packs.
913 impure subroutine s_amr_fx_pack_box(q_coarse, t0, t1, o1, o2, o3, pl, pre, buf)
914
915 type(scalar_field), dimension(sys_size), intent(in) :: q_coarse
916 integer, intent(in) :: t0, t1, o1, o2, o3
917 integer, intent(in), contiguous :: pl(:,:), pre(:)
918 real(wp), intent(inout), contiguous :: buf(:)
919 integer :: e, t, lo, hi, md, r, n1, n2, n3, i, g1, g2, g3
920
921
922# 310 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
923
924# 310 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
925#if defined(MFC_OpenACC)
926# 310 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
927!$acc parallel loop gang vector default(present) private(t, lo, hi, md, r, n1, n2, n3, i, g1, g2, g3) copyin(pl, pre) copyout(buf)
928# 310 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
929#elif defined(MFC_OpenMP)
930# 310 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
931
932# 310 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
933
934# 310 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
935
936# 310 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
937!$omp target teams loop defaultmap(firstprivate:scalar) bind(teams,parallel) defaultmap(tofrom:aggregate) defaultmap(tofrom:allocatable) defaultmap(tofrom:pointer) &
938# 310 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
939!$omp& private(t, lo, hi, md, r, n1, n2, n3, i, g1, g2, g3) map(to:pl, pre) map(from:buf)
940# 310 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
941#endif
942# 311 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
943 do e = pre(t0) + 1, pre(t1 + 1)
944 lo = t0; hi = t1
945 do while (lo < hi)
946 md = (lo + hi)/2
947 if (pre(md + 1) < e) then; lo = md + 1; else; hi = md; end if
948 end do
949 t = lo
950 r = e - pre(t) - 1
951 n1 = pl(4, t); n2 = pl(5, t); n3 = pl(6, t)
952 g1 = pl(1, t) + mod(r, n1); g2 = pl(2, t) + mod(r/n1, n2); g3 = pl(3, t) + mod(r/(n1*n2), n3)
953 i = r/(n1*n2*n3) + 1
954 buf(pl(7, t) + 1 + r) = real(q_coarse(i)%sf(g1 - o1, g2 - o2, g3 - o3), wp)
955 end do
956
957# 324 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
958#if defined(MFC_OpenACC)
959# 324 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
960!$acc end parallel loop
961# 324 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
962#elif defined(MFC_OpenMP)
963# 324 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
964
965# 324 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
966!$omp end target teams loop
967# 324 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
968#endif
969# 325 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
970
971 end subroutine s_amr_fx_pack_box
972
973 !> Fused parent-store pack: as s_amr_fx_pack_box, with the source the flat store at parent slot pl(8,:) and child frame
974 !! pl(9:11,:).
975 impure subroutine s_amr_fx_pack_parent(t0, t1, pl, pre, buf)
976
977 integer, intent(in) :: t0, t1
978 integer, intent(in), contiguous :: pl(:,:), pre(:)
979 real(wp), intent(inout), contiguous :: buf(:)
980 integer :: e, t, lo, hi, md, r, n1, n2, n3, i, g1, g2, g3
981 real(wp) :: th
982
983 th = amr_par_theta
984
985# 339 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
986
987# 339 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
988#if defined(MFC_OpenACC)
989# 339 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
990!$acc parallel loop gang vector default(present) private(t, lo, hi, md, r, n1, n2, n3, i, g1, g2, g3) copyin(pl, pre, th) copyout(buf)
991# 339 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
992#elif defined(MFC_OpenMP)
993# 339 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
994
995# 339 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
996
997# 339 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
998
999# 339 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1000!$omp target teams loop defaultmap(firstprivate:scalar) bind(teams,parallel) defaultmap(tofrom:aggregate) defaultmap(tofrom:allocatable) defaultmap(tofrom:pointer) &
1001# 339 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1002!$omp& private(t, lo, hi, md, r, n1, n2, n3, i, g1, g2, g3) map(to:pl, pre, th) map(from:buf)
1003# 339 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1004#endif
1005# 340 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1006 do e = pre(t0) + 1, pre(t1 + 1)
1007 lo = t0; hi = t1
1008 do while (lo < hi)
1009 md = (lo + hi)/2
1010 if (pre(md + 1) < e) then; lo = md + 1; else; hi = md; end if
1011 end do
1012 t = lo
1013 r = e - pre(t) - 1
1014 n1 = pl(4, t); n2 = pl(5, t); n3 = pl(6, t)
1015 g1 = pl(1, t) + mod(r, n1); g2 = pl(2, t) + mod(r/n1, n2); g3 = pl(3, t) + mod(r/(n1*n2), n3)
1016 i = r/(n1*n2*n3) + 1
1017 buf(pl(7, t) + 1 + r) = f_amr_par_val(g1 + pl(9, t), g2 + pl(10, t), g3 + pl(11, t), i, pl(8, t), th)
1018 end do
1019
1020# 353 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1021#if defined(MFC_OpenACC)
1022# 353 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1023!$acc end parallel loop
1024# 353 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1025#elif defined(MFC_OpenMP)
1026# 353 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1027
1028# 353 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1029!$omp end target teams loop
1030# 353 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1031#endif
1032# 354 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1033
1034 end subroutine s_amr_fx_pack_parent
1035
1036 !> Fused unpack of the contiguous transfer run t0:t1 into amr_cg in one launch; serves both packs via the per-transfer frame in
1037 !! plan rows 9:11.
1038 impure subroutine s_amr_fx_unpack(t0, t1, base, pl, pre, buf, cg)
1039
1040 type(scalar_field), dimension(sys_size), intent(inout) :: cg !< amr_cg, passed so a regrown pool maps as an argument
1041 integer, intent(in) :: t0, t1, base
1042 integer, intent(in), contiguous :: pl(:,:), pre(:)
1043 real(wp), intent(in), contiguous :: buf(:)
1044 integer :: e, t, lo, hi, md, r, n1, n2, n3, i, g1, g2, g3
1045
1046
1047# 367 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1048
1049# 367 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1050#if defined(MFC_OpenACC)
1051# 367 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1052!$acc parallel loop gang vector default(present) private(t, lo, hi, md, r, n1, n2, n3, i, g1, g2, g3) copyin(pl, pre, buf)
1053# 367 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1054#elif defined(MFC_OpenMP)
1055# 367 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1056
1057# 367 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1058
1059# 367 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1060
1061# 367 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1062!$omp target teams loop defaultmap(firstprivate:scalar) bind(teams,parallel) defaultmap(tofrom:aggregate) defaultmap(tofrom:allocatable) defaultmap(tofrom:pointer) &
1063# 367 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1064!$omp& private(t, lo, hi, md, r, n1, n2, n3, i, g1, g2, g3) map(to:pl, pre, buf)
1065# 367 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1066#endif
1067# 368 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1068 do e = pre(t0) + 1, pre(t1 + 1)
1069 lo = t0; hi = t1
1070 do while (lo < hi)
1071 md = (lo + hi)/2
1072 if (pre(md + 1) < e) then; lo = md + 1; else; hi = md; end if
1073 end do
1074 t = lo
1075 r = e - pre(t) - 1
1076 n1 = pl(4, t); n2 = pl(5, t); n3 = pl(6, t)
1077 g1 = pl(1, t) + mod(r, n1); g2 = pl(2, t) + mod(r/n1, n2); g3 = pl(3, t) + mod(r/(n1*n2), n3)
1078 i = r/(n1*n2*n3) + 1
1079 cg(i)%sf(g1 - pl(9, t), g2 - pl(10, t), g3 - pl(11, t)) = real(buf(pl(7, t) - base + 1 + r), stp)
1080 end do
1081
1082# 381 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1083#if defined(MFC_OpenACC)
1084# 381 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1085!$acc end parallel loop
1086# 381 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1087#elif defined(MFC_OpenMP)
1088# 381 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1089
1090# 381 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1091!$omp end target teams loop
1092# 381 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1093#endif
1094# 382 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1095
1096 end subroutine s_amr_fx_unpack
1097
1098 !> Level-1 fill wave, exchange phase: each level-1 box's padded coarse patch (region +/- amr_cpat_mar) is assembled on its owner
1099 !! from the coarse owners, one message per peer; both sides derive the transfers from replicated metadata. Slices are
1100 !! ring-clipped to the hollow shell unless full (init, regrid). Consume with s_amr_l1_fill_consume per owned box ascending.
1101 impure subroutine s_amr_l1_fill_exchange(q_cons_coarse, full)
1102
1103 type(scalar_field), dimension(sys_size), intent(inout) :: q_cons_coarse
1104 logical, intent(in) :: full
1105 integer :: k, r, idx, ix, owner, o1, o2, o3, lo, hi, kk
1106 integer :: plo(3), phi(3), crlo(3), crhi(3), bl(3), bh(3), msl
1107 integer :: tb(3, 6), te(3, 6)
1108
1109 amr_wcur = 1
1110 if (amr_num_blocks <= 0) return
1111 o1 = amr_sidx(1); o2 = amr_sidx(2); o3 = amr_sidx(3)
1113
1116 ! the Lagrangian-overlap safety check must cover every level-1 block (mine included)
1117 if (bubbles_lagrange) then
1118 do k = 1, amr_num_blocks
1119 if (amr_block_level(k) /= 1) cycle
1120 call s_amr_select_slot(k)
1122 end do
1123 end if
1125 call s_amr_refresh_lists()
1126 do kk = 1, amr_n_l1p
1127 k = amr_l1p_blk(kk)
1128 owner = amr_block_owner(k)
1129 call s_amr_patch_box(k, plo, phi)
1130 call s_amr_rank_coarse_range(proc_rank, crlo, crhi)
1131 call s_amr_box_isect(plo, phi, crlo, crhi, bl, bh)
1132 if (any(bl > bh)) cycle
1133 call s_amr_l1_slice_slabs(k, plo, phi, bl, bh, full, msl, tb, te)
1134 call s_amr_wave_add_slabs(amr_wsend, owner, k, msl, tb, te)
1135 end do
1137
1140 do kk = 1, amr_n_my
1141 k = amr_my_blk(kk)
1142 if (amr_block_level(k) /= 1) cycle
1143 call s_amr_patch_box(k, plo, phi)
1144 do idx = 1, amr_ovl_gather_n(k)
1145 r = amr_ovl_gather(idx, k)
1146 if (r == proc_rank) cycle
1147 call s_amr_rank_coarse_range(r, crlo, crhi)
1148 call s_amr_box_isect(plo, phi, crlo, crhi, bl, bh)
1149 call s_amr_l1_slice_slabs(k, plo, phi, bl, bh, full, msl, tb, te)
1150 call s_amr_wave_add_slabs(amr_wrecv, r, k, msl, tb, te)
1151 end do
1152 end do
1154
1155 ! post recvs, pack sends, post sends, one wait
1157 if (amr_device_pack .and. amr_wsend%nx > 0) then
1158 ! the pool is exactly tiled by the side's words; padding would ship garbage
1160 call s_amr_fx_pack_box(q_cons_coarse, 1, amr_wsend%nx, o1, o2, o3, amr_fx_pl(:,1:amr_wsend%nx), &
1161 & amr_fx_pre(1:amr_wsend%nx + 1), amr_fw_sq(1:amr_wsend%words))
1162 else
1163 do ix = 1, amr_wsend%nx
1164 call s_amr_wave_slice(amr_wsend, ix, lo, hi)
1165 call s_amr_pack_box_device(q_cons_coarse, amr_wsend%bl(:,ix), amr_wsend%bh(:,ix), o1, o2, o3, amr_fw_sq(lo:hi))
1166 end do
1167 end if
1170 if (amr_device_pack .and. amr_wrecv%nx > 0) call s_amr_fx_plan(amr_wrecv)
1172
1173 end subroutine s_amr_l1_fill_exchange
1174
1175 !> A contributor's slice [bl, bh] of box k's patch as transfer slabs: the whole slice (full) or its hollow-shell intersection.
1176 impure subroutine s_amr_l1_slice_slabs(k, plo, phi, bl, bh, full, msl, tb, te)
1177
1178 integer, intent(in) :: k, plo(3), phi(3), bl(3), bh(3)
1179 logical, intent(in) :: full
1180 integer, intent(out) :: msl, tb(3, 6), te(3, 6)
1181 integer :: clo(3), chi(3), nsh, scells, sb(3, 6), se(3, 6)
1182
1183 if (full) then
1184 msl = 1; tb(:,1) = bl; te(:,1) = bh
1185 return
1186 end if
1187 call s_amr_patch_core(k, clo, chi)
1188 call s_amr_shell_slabs(plo, phi, clo, chi, nsh, sb, se, scells)
1189 call s_amr_shell_clip(nsh, sb, se, bl, bh, msl, tb, te, scells)
1190
1191 end subroutine s_amr_l1_slice_slabs
1192
1193 !> Drain block k's received transfers at the wave cursor into amr_cg (patch origin off): one fused unpack per contiguous run
1194 !! under amr_device_pack, else one unpack per transfer.
1195 impure subroutine s_amr_fill_drain(k, off)
1196
1197 integer, intent(in) :: k, off(3)
1198 integer :: lo, hi, ie, jx, boff
1199
1200 do while (amr_wcur <= amr_wrecv%nx)
1201 if (amr_wrecv%blk(amr_wcur) /= k) exit
1202 if (amr_device_pack) then
1203 call s_amr_fx_run(k, amr_wrecv%blk, amr_wrecv%nx, amr_wcur, ie)
1204 boff = amr_fx_pl(7, amr_wcur)
1205 do jx = amr_wcur, ie
1206 amr_fx_pl(9:11,jx) = off
1207 end do
1208 call s_amr_fx_unpack(amr_wcur, ie, boff, amr_fx_pl(:,1:amr_wrecv%nx), amr_fx_pre(1:amr_wrecv%nx + 1), &
1209 & amr_fw_rq(boff + 1:amr_fx_pl(7, ie) + amr_fx_pre(ie + 1) - amr_fx_pre(ie)), amr_cg)
1210 amr_wcur = ie + 1
1211 cycle
1212 end if
1213 call s_amr_wave_slice(amr_wrecv, amr_wcur, lo, hi)
1215 amr_wcur = amr_wcur + 1
1216 end do
1217
1218 end subroutine s_amr_fill_drain
1219
1220 !> Level-1 fill wave, consume for owned box k: set the patch frame, copy the own slice, unpack k's transfers. Consume in
1221 !! ascending order: transfers are box-major, so k's are the next run at the cursor.
1222 impure subroutine s_amr_l1_fill_consume(q_cons_coarse, k, full)
1223
1224 type(scalar_field), dimension(sys_size), intent(in) :: q_cons_coarse
1225 integer, intent(in) :: k
1226 logical, intent(in) :: full
1227 integer :: o1, o2, o3, msl, isl
1228 integer :: plo(3), phi(3), crlo(3), crhi(3), bl(3), bh(3)
1229 integer :: tb(3, 6), te(3, 6)
1230
1232 o1 = amr_sidx(1); o2 = amr_sidx(2); o3 = amr_sidx(3)
1233 call s_amr_patch_box(k, plo, phi)
1234 amr_cpat_off = plo
1235#ifdef MFC_DEBUG
1236 ! NaN-flood before the writes land
1237 call s_amr_poison_patch_device(phi(1) - plo(1), phi(2) - plo(2), phi(3) - plo(3))
1238#endif
1239 call s_amr_rank_coarse_range(proc_rank, crlo, crhi)
1240 call s_amr_box_isect(plo, phi, crlo, crhi, bl, bh)
1241 call s_amr_l1_slice_slabs(k, plo, phi, bl, bh, full, msl, tb, te)
1242 if (msl > 0 .and. all(bl <= bh)) then
1243 do isl = 1, msl
1244 call s_amr_slab_add(tb(:,isl), te(:,isl), 0, plo, plo)
1245 end do
1246 call s_amr_gather_own_rows(q_cons_coarse, amr_cg, o1, o2, o3)
1247 end if
1250
1251 end subroutine s_amr_l1_fill_consume
1252
1253 !> Every receive transfer of the open fill wave has been consumed.
1254 impure subroutine s_amr_fill_wave_done()
1255
1256 if (.not. (amr_wcur == amr_wrecv%nx + 1)) then
1257# 543 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1258 call s_mpi_abort("m_amr_exchange.fpp:543: " // "Assertion failed: amr_wcur == amr_wrecv%nx + 1. " &
1259# 543 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1260 & // "fill wave: unconsumed recv transfers")
1261# 543 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1262 end if
1263# 544 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1264
1265 end subroutine s_amr_fill_wave_done
1266
1267 !> Per-stage level-1 fill: the wave, then every owned level-1 box's consume and ghost fill at once.
1268 impure subroutine s_amr_stage_fill_wave(q_cons_coarse)
1269
1270 type(scalar_field), dimension(sys_size), intent(inout) :: q_cons_coarse
1271
1272 call s_amr_l1_fill_exchange(q_cons_coarse, .false.)
1273 call s_amr_l1_fill_batched(q_cons_coarse)
1274
1275 end subroutine s_amr_stage_fill_wave
1276
1277 !> Unpack the queued slab rows (row 9: payload offset in buf) into the pooled amr_cg patches: a whole fill wave in one launch.
1278 impure subroutine s_amr_unpack_rows(buf, cg)
1279
1280 real(wp), intent(in), contiguous :: buf(:)
1281 type(scalar_field), dimension(sys_size), intent(inout) :: cg
1282 integer :: i, s, g, r, n1, n2, g1, g2, g3, stot, nsl, lo, hi, md
1283
1284 stot = f_amr_slab_push(nsl)
1285 if (stot == 0) return ! no level-1 block on this rank: the store may be unallocated
1286
1287# 566 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1288
1289# 566 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1290#if defined(MFC_OpenACC)
1291# 566 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1292!$acc parallel loop collapse(2) gang vector default(present) private(s, lo, hi, md, r, n1, n2, g1, g2, g3) copyin(buf)
1293# 566 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1294#elif defined(MFC_OpenMP)
1295# 566 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1296
1297# 566 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1298
1299# 566 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1300
1301# 566 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1302!$omp target teams loop defaultmap(firstprivate:scalar) bind(teams,parallel) collapse(2) defaultmap(tofrom:aggregate) defaultmap(tofrom:allocatable) defaultmap(tofrom:pointer) &
1303# 566 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1304!$omp& private(s, lo, hi, md, r, n1, n2, g1, g2, g3) map(to:buf)
1305# 566 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1306#endif
1307# 567 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1308 do i = 1, sys_size
1309 do g = 0, stot - 1
1310 lo = 1; hi = nsl
1311 do while (lo < hi)
1312 md = (lo + hi + 1)/2
1313 if (amr_slab_tab(7, md) <= g) then; lo = md; else; hi = md - 1; end if
1314 end do
1315 s = lo
1316 r = g - amr_slab_tab(7, s)
1317 n1 = amr_slab_tab(4, s) - amr_slab_tab(1, s) + 1; n2 = amr_slab_tab(5, s) - amr_slab_tab(2, s) + 1
1318 g1 = amr_slab_tab(1, s) + mod(r, n1)
1319 g2 = amr_slab_tab(2, s) + mod(r/n1, n2)
1320 g3 = amr_slab_tab(3, s) + r/(n1*n2)
1321 cg(i)%sf(g1 - amr_slab_tab(13, s), g2 - amr_slab_tab(14, s), g3 - amr_slab_tab(15, s)) = real(buf(amr_slab_tab(9, &
1322 & s) + 1 + r + amr_slab_tab(8, s)*(i - 1)), stp)
1323 end do
1324 end do
1325
1326# 584 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1327#if defined(MFC_OpenACC)
1328# 584 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1329!$acc end parallel loop
1330# 584 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1331#elif defined(MFC_OpenMP)
1332# 584 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1333
1334# 584 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1335!$omp end target teams loop
1336# 584 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1337#endif
1338# 585 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1339
1340 end subroutine s_amr_unpack_rows
1341
1342 !> Restriction over the queued slab rows (row 9 store column, 10:12 region lo, 13 wire payload offset): children average into
1343 !! the coarse field (sf, frame o) or into buf (wire layout of s_amr_unpack_rows).
1344# 591 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1345 impure subroutine s_amr_restrict_rows_sf(coarse_tgt, o1, o2, o3)
1346# 592 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1347
1348# 594 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1349 type(scalar_field), dimension(sys_size), intent(inout) :: coarse_tgt
1350 integer, intent(in) :: o1, o2, o3
1351# 599 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1352 integer :: i, s, g, r, n1, n2, ci, cj, ck, stot, nsl, lo, hi, md, rr, fi0, fj0, fk0, ddi, ddj, ddk, dj_hi, dk_hi, nchild
1353 real(wp) :: acc
1354
1355 stot = f_amr_slab_push(nsl)
1356 if (stot == 0) return
1357 rr = amr_ref_ratio
1358 dj_hi = merge(rr - 1, 0, amr_dim(2)); dk_hi = merge(rr - 1, 0, amr_dim(3)); nchild = rr**num_dims
1359# 610 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1360
1361# 610 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1362
1363# 610 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1364#if defined(MFC_OpenACC)
1365# 610 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1366!$acc parallel loop collapse(2) gang vector default(present) private(s, lo, hi, md, r, n1, n2, ci, cj, ck, fi0, fj0, fk0, ddi, ddj, ddk, acc)
1367# 610 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1368#elif defined(MFC_OpenMP)
1369# 610 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1370
1371# 610 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1372
1373# 610 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1374
1375# 610 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1376!$omp target teams loop defaultmap(firstprivate:scalar) bind(teams,parallel) collapse(2) defaultmap(tofrom:aggregate) defaultmap(tofrom:allocatable) defaultmap(tofrom:pointer) &
1377# 610 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1378!$omp& private(s, lo, hi, md, r, n1, n2, ci, cj, ck, fi0, fj0, fk0, ddi, ddj, ddk, acc)
1379# 610 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1380#endif
1381# 611 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1382# 612 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1383 do i = 1, sys_size
1384 do g = 0, stot - 1
1385 lo = 1; hi = nsl
1386 do while (lo < hi)
1387 md = (lo + hi + 1)/2
1388 if (amr_slab_tab(7, md) <= g) then; lo = md; else; hi = md - 1; end if
1389 end do
1390 s = lo
1391 r = g - amr_slab_tab(7, s)
1392 n1 = amr_slab_tab(4, s) - amr_slab_tab(1, s) + 1; n2 = amr_slab_tab(5, s) - amr_slab_tab(2, s) + 1
1393 ci = amr_slab_tab(1, s) + mod(r, n1)
1394 cj = amr_slab_tab(2, s) + mod(r/n1, n2)
1395 ck = amr_slab_tab(3, s) + r/(n1*n2)
1396 fi0 = (ci - amr_slab_tab(10, s))*rr; fj0 = (cj - amr_slab_tab(11, s))*rr; fk0 = (ck - amr_slab_tab(12, s))*rr
1397 acc = 0._wp
1398 do ddk = 0, dk_hi
1399 do ddj = 0, dj_hi
1400 do ddi = 0, rr - 1
1401 acc = acc + real(amr_cons_st(fi0 + ddi, fj0 + ddj, fk0 + ddk, i, amr_slab_tab(9, s)), wp)
1402 end do
1403 end do
1404 end do
1405# 635 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1406 coarse_tgt(i)%sf(ci - o1, cj - o2, ck - o3) = real(acc/real(nchild, wp), stp)
1407# 639 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1408 end do
1409 end do
1410
1411# 641 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1412#if defined(MFC_OpenACC)
1413# 641 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1414!$acc end parallel loop
1415# 641 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1416#elif defined(MFC_OpenMP)
1417# 641 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1418
1419# 641 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1420!$omp end target teams loop
1421# 641 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1422#endif
1423# 642 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1424
1425 end subroutine s_amr_restrict_rows_sf
1426# 644 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1427# 591 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1428 impure subroutine s_amr_restrict_rows_wire(buf)
1429# 592 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1430
1431# 597 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1432 real(wp), intent(inout), contiguous :: buf(:)
1433# 599 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1434 integer :: i, s, g, r, n1, n2, ci, cj, ck, stot, nsl, lo, hi, md, rr, fi0, fj0, fk0, ddi, ddj, ddk, dj_hi, dk_hi, nchild
1435 real(wp) :: acc
1436
1437 stot = f_amr_slab_push(nsl)
1438 if (stot == 0) return
1439 rr = amr_ref_ratio
1440 dj_hi = merge(rr - 1, 0, amr_dim(2)); dk_hi = merge(rr - 1, 0, amr_dim(3)); nchild = rr**num_dims
1441# 607 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1442
1443# 607 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1444
1445# 607 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1446#if defined(MFC_OpenACC)
1447# 607 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1448!$acc parallel loop collapse(2) gang vector default(present) private(s, lo, hi, md, r, n1, n2, ci, cj, ck, fi0, fj0, fk0, ddi, ddj, ddk, acc) copyout(buf)
1449# 607 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1450#elif defined(MFC_OpenMP)
1451# 607 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1452
1453# 607 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1454
1455# 607 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1456
1457# 607 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1458!$omp target teams loop defaultmap(firstprivate:scalar) bind(teams,parallel) collapse(2) defaultmap(tofrom:aggregate) defaultmap(tofrom:allocatable) defaultmap(tofrom:pointer) &
1459# 607 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1460!$omp& private(s, lo, hi, md, r, n1, n2, ci, cj, ck, fi0, fj0, fk0, ddi, ddj, ddk, acc) map(from:buf)
1461# 607 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1462#endif
1463# 609 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1464# 612 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1465 do i = 1, sys_size
1466 do g = 0, stot - 1
1467 lo = 1; hi = nsl
1468 do while (lo < hi)
1469 md = (lo + hi + 1)/2
1470 if (amr_slab_tab(7, md) <= g) then; lo = md; else; hi = md - 1; end if
1471 end do
1472 s = lo
1473 r = g - amr_slab_tab(7, s)
1474 n1 = amr_slab_tab(4, s) - amr_slab_tab(1, s) + 1; n2 = amr_slab_tab(5, s) - amr_slab_tab(2, s) + 1
1475 ci = amr_slab_tab(1, s) + mod(r, n1)
1476 cj = amr_slab_tab(2, s) + mod(r/n1, n2)
1477 ck = amr_slab_tab(3, s) + r/(n1*n2)
1478 fi0 = (ci - amr_slab_tab(10, s))*rr; fj0 = (cj - amr_slab_tab(11, s))*rr; fk0 = (ck - amr_slab_tab(12, s))*rr
1479 acc = 0._wp
1480 do ddk = 0, dk_hi
1481 do ddj = 0, dj_hi
1482 do ddi = 0, rr - 1
1483 acc = acc + real(amr_cons_st(fi0 + ddi, fj0 + ddj, fk0 + ddk, i, amr_slab_tab(9, s)), wp)
1484 end do
1485 end do
1486 end do
1487# 637 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1488 buf(amr_slab_tab(13, s) + 1 + r + amr_slab_tab(8, s)*(i - 1)) = acc/real(nchild, wp)
1489# 639 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1490 end do
1491 end do
1492
1493# 641 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1494#if defined(MFC_OpenACC)
1495# 641 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1496!$acc end parallel loop
1497# 641 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1498#elif defined(MFC_OpenMP)
1499# 641 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1500
1501# 641 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1502!$omp end target teams loop
1503# 641 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1504#endif
1505# 642 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1506
1507 end subroutine s_amr_restrict_rows_wire
1508# 644 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1509# 645 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1510
1511 !> Hold np coarse patches in amr_cg stacked along dim 3 (patch p at p*(amr_cpat_hi(3) + 1)); grows only, contents do not survive
1512 !! a fill.
1513 impure subroutine s_amr_cg_reserve(np)
1514
1515 integer, intent(in) :: np
1516 integer :: i
1517
1518 if (np <= amr_cg_np) return
1519 amr_cg_np = max(np, 2*amr_cg_np)
1520 do i = 1, sys_size
1521#ifdef _CRAYFTN
1522# 656 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1523 block
1524# 656 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1525#ifdef MFC_DEBUG
1526# 656 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1527 block
1528# 656 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1529 use iso_fortran_env, only: output_unit
1530# 656 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1531
1532# 656 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1533 print *, 'm_amr_exchange.fpp:656: ', '@:ACC_TEARDOWN_SFs(amr_cg(i))'
1534# 656 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1535
1536# 656 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1537 call flush (output_unit)
1538# 656 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1539 end block
1540# 656 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1541#endif
1542# 656 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1543
1544# 656 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1545 if (associated(amr_cg(i)%sf)) then
1546# 656 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1547
1548# 656 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1549#if defined(MFC_OpenACC)
1550# 656 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1551!$acc exit data delete(amr_cg(i)%sf)
1552# 656 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1553#elif defined(MFC_OpenMP)
1554# 656 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1555!$omp target exit data map(release:amr_cg(i)%sf)
1556# 656 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1557#endif
1558# 656 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1559 end if
1560# 656 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1561
1562# 656 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1563#if defined(MFC_OpenACC)
1564# 656 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1565!$acc exit data delete(amr_cg(i))
1566# 656 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1567#elif defined(MFC_OpenMP)
1568# 656 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1569!$omp target exit data map(release:amr_cg(i))
1570# 656 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1571#endif
1572# 656 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1573 end block
1574# 656 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1575#endif
1576# 657 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1577#ifdef MFC_DEBUG
1578# 657 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1579 block
1580# 657 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1581 use iso_fortran_env, only: output_unit
1582# 657 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1583
1584# 657 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1585 print *, 'm_amr_exchange.fpp:657: ', '@:DEALLOCATE(amr_cg(i)%sf)'
1586# 657 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1587
1588# 657 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1589 call flush (output_unit)
1590# 657 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1591 end block
1592# 657 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1593#endif
1594# 657 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1595
1596# 657 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1597#if defined(MFC_OpenACC)
1598# 657 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1599!$acc exit data delete(amr_cg(i)%sf)
1600# 657 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1601#elif defined(MFC_OpenMP)
1602# 657 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1603!$omp target exit data map(release:amr_cg(i)%sf)
1604# 657 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1605#endif
1606# 657 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1607 deallocate (amr_cg(i)%sf)
1608# 658 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1609#ifdef MFC_DEBUG
1610# 658 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1611 block
1612# 658 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1613 use iso_fortran_env, only: output_unit
1614# 658 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1615
1616# 658 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1617 print *, 'm_amr_exchange.fpp:658: ', '@:ALLOCATE(amr_cg(i)%sf(0:amr_cpat_hi(1), 0:amr_cpat_hi(2), 0:amr_cg_np*(amr_cpat_hi(3) + 1) - 1))'
1618# 658 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1619
1620# 658 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1621 call flush (output_unit)
1622# 658 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1623 end block
1624# 658 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1625#endif
1626# 658 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1627 allocate (amr_cg(i)%sf(0:amr_cpat_hi(1), 0:amr_cpat_hi(2), 0:amr_cg_np*(amr_cpat_hi(3) + 1) - 1))
1628# 658 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1629
1630# 658 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1631
1632# 658 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1633#if defined(MFC_OpenACC)
1634# 658 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1635!$acc enter data create(amr_cg(i)%sf)
1636# 658 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1637#elif defined(MFC_OpenMP)
1638# 658 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1639!$omp target enter data map(always,alloc:amr_cg(i)%sf)
1640# 658 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1641#endif
1642# 659 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1643 amr_cg(i)%sf = 0._stp ! finite padding for the device copy
1644#ifdef _CRAYFTN
1645# 660 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1646 block
1647# 660 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1648#ifdef MFC_DEBUG
1649# 660 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1650 block
1651# 660 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1652 use iso_fortran_env, only: output_unit
1653# 660 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1654
1655# 660 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1656 print *, 'm_amr_exchange.fpp:660: ', '@:ACC_SETUP_SFs(amr_cg(i))'
1657# 660 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1658
1659# 660 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1660 call flush (output_unit)
1661# 660 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1662 end block
1663# 660 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1664#endif
1665# 660 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1666
1667# 660 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1668
1669# 660 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1670#if defined(MFC_OpenACC)
1671# 660 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1672!$acc enter data copyin(amr_cg(i))
1673# 660 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1674#elif defined(MFC_OpenMP)
1675# 660 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1676!$omp target enter data map(to:amr_cg(i))
1677# 660 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1678#endif
1679# 660 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1680 if (associated(amr_cg(i)%sf)) then
1681# 660 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1682
1683# 660 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1684#if defined(MFC_OpenACC)
1685# 660 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1686!$acc enter data copyin(amr_cg(i)%sf)
1687# 660 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1688#elif defined(MFC_OpenMP)
1689# 660 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1690!$omp target enter data map(to:amr_cg(i)%sf)
1691# 660 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1692#endif
1693# 660 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1694 end if
1695# 660 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1696 end block
1697# 660 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1698#endif
1699# 661 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1700 end do
1701
1702 end subroutine s_amr_cg_reserve
1703
1704 !> Level-1 consume and ghost fill for all owned boxes at once: patches pooled in amr_cg (s_amr_cg_reserve), one kernel each.
1705 impure subroutine s_amr_l1_fill_batched(q_cons_coarse)
1706
1707 type(scalar_field), dimension(sys_size), intent(in) :: q_cons_coarse
1708
1709 call s_amr_l1_pool_gather(q_cons_coarse)
1710 call s_amr_l1_pool_fill()
1711
1712 end subroutine s_amr_l1_fill_batched
1713
1714 !> Consume the open level-1 fill wave into the pooled amr_cg: box k at dim-3 offset amr_cg_pz(k), amr_cg_nuse patches.
1715 impure subroutine s_amr_l1_pool_gather(q_cons_coarse)
1716
1717 type(scalar_field), dimension(sys_size), intent(in) :: q_cons_coarse
1718 integer :: k, kk, isl, msl, t
1719 integer :: plo(3), phi(3), crlo(3), crhi(3), bl(3), bh(3), tb(3, 6), te(3, 6)
1720
1721 call s_phase_tic(ph_gather)
1722 if (.not. allocated(amr_cg_pz)) allocate (amr_cg_pz(amr_max_blocks))
1723 amr_cg_nuse = 0
1724 do kk = 1, amr_n_my
1725 k = amr_my_blk(kk)
1726 if (amr_block_level(k) /= 1) cycle
1727 amr_cg_pz(k) = amr_cg_nuse*(amr_cpat_hi(3) + 1); amr_cg_nuse = amr_cg_nuse + 1
1728 end do
1729 call s_amr_cg_reserve(amr_cg_nuse)
1730 call s_amr_rank_coarse_range(proc_rank, crlo, crhi)
1731 do kk = 1, amr_n_my
1732 k = amr_my_blk(kk)
1733 if (amr_block_level(k) /= 1) cycle
1734 call s_amr_patch_box(k, plo, phi)
1735 call s_amr_box_isect(plo, phi, crlo, crhi, bl, bh)
1736 if (any(bl > bh)) cycle
1737 call s_amr_l1_slice_slabs(k, plo, phi, bl, bh, .false., msl, tb, te)
1738 do isl = 1, msl
1739 call s_amr_slab_add(tb(:,isl), te(:,isl), 0, plo, plo - [0, 0, amr_cg_pz(k)])
1740 end do
1741 end do
1742 call s_amr_gather_own_rows(q_cons_coarse, amr_cg, amr_sidx(1), amr_sidx(2), amr_sidx(3))
1743 if (amr_wrecv%nx > 0) then
1744 do t = 1, amr_wrecv%nx
1745 k = amr_wrecv%blk(t)
1746 call s_amr_patch_box(k, plo, phi)
1747 call s_amr_slab_add(amr_wrecv%bl(:,t), amr_wrecv%bh(:,t), amr_wrecv%off(t), plo, plo - [0, 0, amr_cg_pz(k)])
1748 end do
1749 call s_amr_unpack_rows(amr_fw_rq(1:amr_wrecv%words), amr_cg)
1750 end if
1751 amr_wcur = amr_wrecv%nx + 1
1752 call s_phase_toc(ph_gather)
1753
1754 end subroutine s_amr_l1_pool_gather
1755
1756 !> Ghost-fill every owned level-1 box from its pooled patch (s_amr_l1_pool_gather's layout) in one kernel.
1757 impure subroutine s_amr_l1_pool_fill()
1758
1759 integer :: k, kk, plo(3), phi(3)
1760
1761 call s_phase_tic(ph_gfill)
1762 do kk = 1, amr_n_my
1763 k = amr_my_blk(kk)
1764 if (amr_block_level(k) /= 1) cycle
1765 call s_amr_select_slot(k)
1766 call s_amr_patch_box(k, plo, phi)
1767 amr_cpat_off = plo
1768 call s_amr_queue_ghost_slabs(amr_loc_of(k), amr_cg_pz(k))
1769 end do
1770 call s_amr_fill_ghost_rows(amr_cg)
1771 call s_phase_toc(ph_gfill)
1772
1773 end subroutine s_amr_l1_pool_fill
1774
1775 !> Padded parent-patch extents of a child with parent footprint [plo, phi] (amr_cpat_mar coarse cells each side).
1776 pure subroutine s_amr_patch_width(plo, phi, w1, w2, w3)
1777
1778 integer, intent(in) :: plo(3), phi(3)
1779 integer, intent(out) :: w1, w2, w3
1780
1781 w1 = phi(1) - plo(1) + 2*amr_cpat_mar
1782 w2 = merge(phi(2) - plo(2) + 2*amr_cpat_mar, 0, amr_dim(2)); w3 = merge(phi(3) - plo(3) + 2*amr_cpat_mar, 0, amr_dim(3))
1783
1784 end subroutine s_amr_patch_width
1785
1786 !> Parent-fill transfer list of a child with padded patch (w1, w2, w3), patch-local: the whole patch (full) or its hollow shell.
1787 !! Send, receive and consume share it.
1788 impure subroutine s_amr_parent_slabs(w1, w2, w3, full, msl, tb, te)
1789
1790 integer, intent(in) :: w1, w2, w3
1791 logical, intent(in) :: full
1792 integer, intent(out) :: msl, tb(3, 6), te(3, 6)
1793 integer :: clo(3), chi(3), scells
1794
1795 if (full) then
1796 msl = 1; tb(:,1) = 0; te(:,1) = [w1, w2, w3]
1797 return
1798 end if
1799 clo = merge(amr_cpat_mar + f_amr_prolong_reach(), 0, amr_dim)
1800 chi = merge([w1, w2, w3] - amr_cpat_mar - f_amr_prolong_reach(), 0, amr_dim)
1801 call s_amr_shell_slabs([0, 0, 0], [w1, w2, w3], clo, chi, msl, tb, te, scells)
1802
1803 end subroutine s_amr_parent_slabs
1804
1805 !> A child's patch frame: the global parent-fine origin of its padded parent patch (region footprint - amr_cpat_mar).
1806 impure subroutine s_amr_parent_frame(k, plo, phi, w1, w2, w3)
1807
1808 integer, intent(in) :: k
1809 integer, intent(out) :: plo(3), phi(3), w1, w2, w3
1810
1811 call s_amr_parent_foot(k, amr_parent_blk(k), plo, phi)
1812 call s_amr_patch_width(plo, phi, w1, w2, w3)
1813 amr_cpat_off = merge(plo - amr_cpat_mar, 0, amr_dim)
1814
1815 end subroutine s_amr_parent_frame
1816
1817 !> Level-lev parent-fill wave, exchange phase: each block's parent patch from its parent's fine array (parent-fine frame); split
1818 !! children ship from the parent's owner, co-located parents copy at consume. Called per level ascending so parents are
1819 !! complete. Consume with s_amr_parent_fill_consume per owned box.
1820 impure subroutine s_amr_parent_fill_exchange(lev, full)
1821
1822 integer, intent(in) :: lev
1823 logical, intent(in) :: full
1824 integer :: k, ix, pblk, powner, cowner, lo, hi, kk
1825 integer :: w1, w2, w3, plo(3), phi(3), msl
1826 integer :: tb(3, 6), te(3, 6)
1827 integer :: loc_parent_blk_k !< plain scalar for device routines (NVHPC -Minline)
1828
1829 amr_wcur = 1
1830 if (amr_num_blocks <= 0) return
1831 call s_amr_wave_open(amr_wave, amr_band_parent_fill)
1832
1833 call s_phase_tic(ph_gather)
1834 ! the Lagrangian-overlap safety check must cover every level-lev block (mine included)
1835 if (bubbles_lagrange) then
1836 do k = 1, amr_num_blocks
1837 if (amr_block_level(k) /= lev) cycle
1838 call s_amr_select_slot(k)
1839 call s_amr_check_lag_clear()
1840 end do
1841 end if
1842 call s_amr_refresh_lists()
1843 ! send side: every level-lev block whose parent I own but whose owner is another rank
1844 call s_amr_wave_reset(amr_wsend)
1845 do kk = 1, amr_n_fch
1846 k = amr_fch_blk(kk)
1847 if (amr_block_level(k) /= lev) cycle
1848 pblk = amr_parent_blk(k)
1849 powner = amr_block_owner(pblk); cowner = amr_block_owner(k)
1850 if (powner == cowner .or. powner /= proc_rank) cycle
1851 call s_amr_parent_frame(k, plo, phi, w1, w2, w3)
1852 call s_amr_parent_slabs(w1, w2, w3, full, msl, tb, te)
1853 call s_amr_wave_add_slabs(amr_wsend, cowner, k, msl, tb, te)
1854 end do
1855 call s_amr_wave_close(amr_wsend, amr_fw_sq, amr_fw_dev)
1856 ! receive side: every level-lev block I own whose parent lives on another rank
1857 call s_amr_wave_reset(amr_wrecv)
1858 call s_amr_refresh_my_blocks()
1859 do kk = 1, amr_n_my
1860 k = amr_my_blk(kk)
1861 if (amr_block_level(k) /= lev) cycle
1862 pblk = amr_parent_blk(k)
1863 powner = amr_block_owner(pblk)
1864 if (powner == proc_rank) cycle
1865 call s_amr_parent_frame(k, plo, phi, w1, w2, w3)
1866 call s_amr_parent_slabs(w1, w2, w3, full, msl, tb, te)
1867 call s_amr_wave_add_slabs(amr_wrecv, powner, k, msl, tb, te)
1868 end do
1869 call s_amr_wave_close(amr_wrecv, amr_fw_rq, amr_fw_dev)
1870 call s_amr_wave_post(amr_wave, amr_wrecv, amr_fw_rq, amr_fw_dev)
1871 ! packs: one fused launch over the whole send list under amr_device_pack, else one launch per transfer
1872 if (amr_device_pack .and. amr_wsend%nx > 0) then
1873 call s_amr_fx_plan(amr_wsend)
1874 do ix = 1, amr_wsend%nx
1875 k = amr_wsend%blk(ix)
1876 call s_amr_parent_frame(k, plo, phi, w1, w2, w3)
1877 amr_fx_pl(8, ix) = amr_loc_of(amr_parent_blk(k))
1878 amr_fx_pl(9:11,ix) = amr_cpat_off
1879 end do
1880 call s_amr_fx_pack_parent(1, amr_wsend%nx, amr_fx_pl(:,1:amr_wsend%nx), amr_fx_pre(1:amr_wsend%nx + 1), &
1881 & amr_fw_sq(1:amr_wsend%words))
1882 else
1883 do ix = 1, amr_wsend%nx
1884 k = amr_wsend%blk(ix)
1885 call s_amr_parent_frame(k, plo, phi, w1, w2, w3)
1886 call s_amr_wave_slice(amr_wsend, ix, lo, hi)
1887 loc_parent_blk_k = amr_loc_of(amr_parent_blk(k))
1888 call s_amr_pack_parent_box_device(loc_parent_blk_k, amr_wsend%bl(:,ix), amr_wsend%bh(:,ix), amr_fw_sq(lo:hi))
1889 end do
1890 end if
1891 call s_amr_wave_send(amr_wave, amr_wsend, amr_fw_sq, amr_fw_dev)
1892 call s_amr_wave_wait(amr_wave)
1893 if (amr_device_pack .and. amr_wrecv%nx > 0) call s_amr_fx_plan(amr_wrecv)
1894 call s_phase_toc(ph_gather)
1895
1896 end subroutine s_amr_parent_fill_exchange
1897
1898 !> The parent-fill wave, consume phase for owned box k (the current slot): set the patch frame, then a co-located parent's
1899 !! own-box device copies or k's received transfers (box-major: the next contiguous run at the cursor).
1900 impure subroutine s_amr_parent_fill_consume(k, full)
1901
1902 integer, intent(in) :: k
1903 logical, intent(in) :: full
1904 integer :: pblk, isl, msl
1905 integer :: w1, w2, w3, plo(3), phi(3)
1906 integer :: tb(3, 6), te(3, 6)
1907 integer :: loc_pblk !< plain scalar for device routines (NVHPC -Minline)
1908
1909 call s_phase_tic(ph_gather)
1910 pblk = amr_parent_blk(k)
1911 call s_amr_parent_frame(k, plo, phi, w1, w2, w3)
1912#ifdef MFC_DEBUG
1913 call s_amr_poison_patch_device(w1, w2, w3)
1914#endif
1915 if (amr_block_owner(pblk) == proc_rank) then
1916 call s_amr_parent_slabs(w1, w2, w3, full, msl, tb, te)
1917 do isl = 1, msl
1918 loc_pblk = amr_loc_of(pblk)
1919 call s_amr_copy_parent_box(loc_pblk, tb(:,isl), te(:,isl))
1920 end do
1921 else
1922 if (.not. (amr_wcur <= amr_wrecv%nx .and. amr_wrecv%blk(amr_wcur) == k)) then
1923# 883 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1924 call s_mpi_abort("m_amr_exchange.fpp:883: " // .and."Assertion failed: amr_wcur <= amr_wrecv%nx amr_wrecv%blk(amr_wcur) == k. " &
1925# 883 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1926 & // "parent-fill wave: missing recv transfer")
1927# 883 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1928 end if
1929# 884 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1930 call s_amr_fill_drain(k, [0, 0, 0])
1931 end if
1932 call s_phase_toc(ph_gather)
1933
1934 end subroutine s_amr_parent_fill_consume
1935
1936 !> Per-stage level-lev fill: the wave, then per owned level-lev box the consume and the ghost fill.
1937 impure subroutine s_amr_parent_fill_wave(lev)
1938
1939 integer, intent(in) :: lev
1940 integer :: k, kk
1941 integer :: loc_cur !< plain scalar for device routines (NVHPC -Minline)
1942
1943 call s_amr_parent_fill_exchange(lev, .false.)
1944 do kk = 1, amr_n_own
1945 k = amr_own_blk(kk)
1946 if (amr_block_level(k) /= lev) cycle
1947 call s_amr_select_slot(k)
1948 if (.not. amr_rank_owns_block) cycle
1949 call s_amr_parent_fill_consume(k, .false.)
1950 call s_phase_tic(ph_gfill)
1951 loc_cur = amr_loc_of(amr_cur)
1952 call s_amr_fill_fine_ghosts(amr_cg, loc_cur)
1953 call s_phase_toc(ph_gfill)
1954 end do
1956
1957 end subroutine s_amr_parent_fill_wave
1958
1959 !> Device pack (to_buf=T) / unpack (F) of the interior block [o+0:o+fm] of the flat store at column loc <-> contiguous MPI
1960 !! buffer buf for P2P migration; wp buffer cast to/from stp. Only buf is mapped (flang miscomputes strided %sf sections in map
1961 !! clauses).
1962 impure subroutine s_amr_pack_unpack_block_st(loc, o1, o2, o3, fm1, fm2, fm3, buf, to_buf)
1963
1964 integer, intent(in) :: loc, o1, o2, o3, fm1, fm2, fm3
1965 real(wp), intent(inout), contiguous :: buf(:)
1966 logical, intent(in) :: to_buf
1967 integer :: i, j, k, l
1968
1969# 924 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1970# 925 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1971 if (to_buf) then
1972
1973# 926 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1974
1975# 926 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1976#if defined(MFC_OpenACC)
1977# 926 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1978!$acc parallel loop collapse(4) gang vector default(present) copyout(buf)
1979# 926 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1980#elif defined(MFC_OpenMP)
1981# 926 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1982
1983# 926 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1984
1985# 926 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1986
1987# 926 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1988!$omp target teams loop defaultmap(firstprivate:scalar) bind(teams,parallel) collapse(4) defaultmap(tofrom:aggregate) defaultmap(tofrom:allocatable) defaultmap(tofrom:pointer) map(from:buf)
1989# 926 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1990#endif
1991# 927 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1992# 931 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1993 do i = 1, sys_size
1994 do l = 0, fm3
1995 do k = 0, fm2
1996 do j = 0, fm1
1997# 936 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
1998 buf(1 + j + (fm1 + 1)*(k + (fm2 + 1)*(l + (fm3 + 1)*(i - 1)))) = real(amr_cons_st(o1 + j, &
1999 & o2 + k, o3 + l, i, loc), wp)
2000# 942 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
2001 end do
2002 end do
2003 end do
2004 end do
2005
2006# 946 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
2007#if defined(MFC_OpenACC)
2008# 946 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
2009!$acc end parallel loop
2010# 946 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
2011#elif defined(MFC_OpenMP)
2012# 946 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
2013
2014# 946 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
2015!$omp end target teams loop
2016# 946 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
2017#endif
2018# 947 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
2019# 924 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
2020# 928 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
2021 else
2022
2023# 929 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
2024
2025# 929 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
2026#if defined(MFC_OpenACC)
2027# 929 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
2028!$acc parallel loop collapse(4) gang vector default(present) copyin(buf)
2029# 929 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
2030#elif defined(MFC_OpenMP)
2031# 929 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
2032
2033# 929 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
2034
2035# 929 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
2036
2037# 929 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
2038!$omp target teams loop defaultmap(firstprivate:scalar) bind(teams,parallel) collapse(4) defaultmap(tofrom:aggregate) defaultmap(tofrom:allocatable) defaultmap(tofrom:pointer) map(to:buf)
2039# 929 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
2040#endif
2041# 930 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
2042# 931 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
2043 do i = 1, sys_size
2044 do l = 0, fm3
2045 do k = 0, fm2
2046 do j = 0, fm1
2047# 939 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
2048 amr_cons_st(o1 + j, o2 + k, o3 + l, i, &
2049 & loc) = real(buf(1 + j + (fm1 + 1)*(k + (fm2 + 1)*(l + (fm3 + 1)*(i - 1)))), stp)
2050# 942 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
2051 end do
2052 end do
2053 end do
2054 end do
2055
2056# 946 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
2057#if defined(MFC_OpenACC)
2058# 946 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
2059!$acc end parallel loop
2060# 946 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
2061#elif defined(MFC_OpenMP)
2062# 946 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
2063
2064# 946 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
2065!$omp end target teams loop
2066# 946 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
2067#endif
2068# 947 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
2069# 948 "/home/runner/work/MFC/MFC/src/simulation/m_amr_exchange.fpp"
2070 end if
2071
2072 end subroutine s_amr_pack_unpack_block_st
2073
2074end module m_amr_exchange
integer, intent(in) k
integer, intent(in) j
integer, intent(in) l
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_refresh_lists()
Rebuild the epoch-keyed block lists when the mesh epoch has moved (one O(global blocks) walk per regr...
pure subroutine, public s_amr_rank_coarse_range(r, crlo, crhi)
Rank r's contiguous owned coarse-cell range per dim: the interior plus buff_size ghosts where the sub...
pure subroutine, public s_amr_box_isect(alo, ahi, blo, bhi, olo, ohi)
Per-dim intersection of two global boxes [alo:ahi] and [blo:bhi] -> [olo:ohi] (empty when olo > ohi i...
Parent/child data exchange: gather plans, pack/unpack and the fill waves.
impure subroutine s_amr_l1_fill_batched(q_cons_coarse)
Level-1 consume and ghost fill for all owned boxes at once: patches pooled in amr_cg (s_amr_cg_reserv...
impure subroutine s_amr_fx_plan(w)
Fused-pack plan for transfers 1:nt: slab corner, extents, absolute pool payload offset and exclusive ...
impure subroutine s_amr_fx_pack_box(q_coarse, t0, t1, o1, o2, o3, pl, pre, buf)
Fused coarse-field pack: the wave's whole send list in one launch; inverts the linear index s_amr_pac...
impure subroutine s_amr_shell_slabs(plo, phi, clo, chi, ns, sb, se, cells)
Ring clip: split the hollow shell of patch [plo:phi] minus open core [clo:chi] into at most 6 disjoin...
impure subroutine s_amr_pack_parent_box_device(qp, bl, bh, buf)
Sub-box variants of the parent-patch pack/copy for the ring-clipped parent-fill wave; bl:bh are patch...
impure subroutine s_amr_gather_own_rows(q_coarse, cg, o1, o2, o3)
Ring-clipped own-box copy on device: queue shell-slab/own-box intersections as slab rows and copy q_c...
impure subroutine s_amr_fx_pack_parent(t0, t1, pl, pre, buf)
Fused parent-store pack: as s_amr_fx_pack_box, with the source the flat store at parent slot pl(8,...
impure subroutine, public s_amr_l1_pool_gather(q_cons_coarse)
Consume the open level-1 fill wave into the pooled amr_cg: box k at dim-3 offset amr_cg_pz(k),...
impure subroutine s_amr_fx_unpack(t0, t1, base, pl, pre, buf, cg)
Fused unpack of the contiguous transfer run t0:t1 into amr_cg in one launch; serves both packs via th...
impure subroutine, public s_amr_l1_fill_consume(q_cons_coarse, k, full)
Level-1 fill wave, consume for owned box k: set the patch frame, copy the own slice,...
pure subroutine s_amr_fx_run(k, blk, nx, t0, t1)
Longest run t0:t1 of one box's recv transfers that is contiguous in the wire pool (unpacks in one lau...
impure subroutine, public s_amr_restrict_rows_sf(coarse_tgt, o1, o2, o3)
Restriction over the queued slab rows (row 9 store column, 10:12 region lo, 13 wire payload offset): ...
impure subroutine s_amr_parent_slabs(w1, w2, w3, full, msl, tb, te)
Parent-fill transfer list of a child with padded patch (w1, w2, w3), patch-local: the whole patch (fu...
impure subroutine s_amr_parent_frame(k, plo, phi, w1, w2, w3)
A child's patch frame: the global parent-fine origin of its padded parent patch (region footprint - a...
impure subroutine s_amr_pack_box_device(q_coarse, bl, bh, o1, o2, o3, buf)
Pack (pack = T) the global box [bl:bh] of q_coarse into the wire buffer buf, or unpack it into amr_cg...
impure subroutine, public s_amr_pack_unpack_block_st(loc, o1, o2, o3, fm1, fm2, fm3, buf, to_buf)
Device pack (to_buf=T) / unpack (F) of the interior block [o+0:o+fm] of the flat store at column loc ...
impure subroutine s_amr_fill_drain(k, off)
Drain block k's received transfers at the wave cursor into amr_cg (patch origin off): one fused unpac...
impure subroutine s_amr_copy_parent_box(qp, bl, bh)
impure subroutine, public s_amr_unpack_rows(buf, cg)
Unpack the queued slab rows (row 9: payload offset in buf) into the pooled amr_cg patches: a whole fi...
impure subroutine s_amr_l1_slice_slabs(k, plo, phi, bl, bh, full, msl, tb, te)
A contributor's slice [bl, bh] of box k's patch as transfer slabs: the whole slice (full) or its holl...
impure subroutine, public s_amr_parent_fill_exchange(lev, full)
Level-lev parent-fill wave, exchange phase: each block's parent patch from its parent's fine array (p...
pure real(wp) function f_amr_par_val(g1, g2, g3, i, qp, th)
Parent value for a level >= 2 ghost fill: the current state, or its blend with the substep-start back...
impure subroutine s_amr_unpack_box_device(bl, bh, off, buf)
impure subroutine, public s_amr_l1_fill_exchange(q_cons_coarse, full)
Level-1 fill wave, exchange phase: each level-1 box's padded coarse patch (region +/- amr_cpat_mar) i...
impure subroutine, public s_amr_parent_fill_wave(lev)
Per-stage level-lev fill: the wave, then per owned level-lev box the consume and the ghost fill.
impure subroutine s_amr_cg_reserve(np)
Hold np coarse patches in amr_cg stacked along dim 3 (patch p at p*(amr_cpat_hi(3) + 1)); grows only,...
impure subroutine, public s_amr_l1_pool_fill()
Ghost-fill every owned level-1 box from its pooled patch (s_amr_l1_pool_gather's layout) in one kerne...
impure subroutine, public s_amr_restrict_rows_wire(buf)
pure subroutine s_amr_patch_width(plo, phi, w1, w2, w3)
Padded parent-patch extents of a child with parent footprint [plo, phi] (amr_cpat_mar coarse cells ea...
impure subroutine, public s_amr_stage_fill_wave(q_cons_coarse)
Per-stage level-1 fill: the wave, then every owned level-1 box's consume and ghost fill at once.
impure subroutine, public s_amr_fill_wave_done()
Every receive transfer of the open fill wave has been consumed.
impure subroutine, public s_amr_parent_fill_consume(k, full)
The parent-fill wave, consume phase for owned box k (the current slot): set the patch frame,...
impure subroutine s_amr_poison_patch_device(w1, w2, w3)
Debug only: NaN-flood amr_cg before a clipped gather so any read of an unshipped cell NaNs the ghost ...
impure subroutine s_amr_shell_clip(ns, sb, se, bl, bh, ms, tb, te, cells)
Intersect the shell slabs with box [bl:bh], keeping order; also returns the total cell count.
Coarse-to-fine ghost fill: slab table, limited-linear or quartic fill, quartic positivity guard,...
impure integer function, public f_amr_slab_push(nsl)
Push the slab table to the device and empty it for the next build; returns the total cell count.
impure subroutine, public s_amr_slab_add(lo, hi, loc, ilo, coff)
Append slab [lo, hi] to amr_slab_tab (rows: 1:3 lo, 4:6 hi, 7 flat offset, 8 cells,...
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 ...
Same-level fine-fine seams and the coarse conservative halo.
impure subroutine, public s_amr_build_seam_pairs()
Rebuild the same-level seam-pair list once per regrid/restart in identical (xb, yb) order on all rank...
Module-scope AMR state (block pool, flat field store, wave buffers) and the leaf helpers every AMR mo...
integer, dimension(:,:), allocatable amr_fx_pl
Fused exchange packs (amr_device_pack): one row per wave transfer (slab corner (1:3),...
integer, dimension(:), allocatable amr_fx_pre
integer amr_wcur
consume cursor into the open fill wave's receive transfers (box-major)
integer, parameter amr_band_stage_fill
real(wp), dimension(:), allocatable amr_fw_sq
Wire pools of the pooled waves (m_amr_wave lays them out; contents never survive a wave)....
pure subroutine s_amr_patch_box(k, plo, phi)
Block k's padded coarse patch box [plo, phi] (region +/- amr_cpat_mar; collapsed dims 0:0).
integer, dimension(:), allocatable amr_ovl_gather_n
integer amr_seam_pairs_nblk
type(scalar_field), dimension(:), allocatable amr_cg
logical amr_fw_dev
Device-resident wire pools: with rdma_mpi MPI sends/receives them by device address,...
integer amr_cg_np
patches amr_cg holds, stacked along dim 3 (s_amr_cg_reserve)
logical amr_seam_pairs_dirty
real(stp), dimension(:,:,:,:,:), allocatable amr_cons_st
Flat per-block field store (x, y, z, var, local slot) so one kernel can run over every live block; al...
real(stp), dimension(:,:,:,:,:), allocatable amr_stor_st
real(wp), dimension(:), allocatable amr_fw_rq
integer, dimension(:), allocatable amr_l1p_blk
padded variant (region +/- amr_cpat_mar vs my coarse range): stage-fill gather
integer, dimension(:,:), allocatable amr_slab_tab
Per-block gathered coarse patch: the owner may not hold the coarse cells its block refines,...
integer, dimension(:), allocatable amr_my_blk
Block indices this rank owns, ascending (the paired MPI_SENDRECVs depend on the order); built in s_am...
integer, dimension(:,:), allocatable amr_ovl_gather
cached per-block P2P overlap-rank lists, rebuilt with the seam list: amr_ovl_gather(:,...
integer, dimension(3) amr_cpat_off
global coarse index of amr_cg local cell 0 (region_lo - amr_cpat_mar)
impure subroutine s_amr_check_lag_clear()
Rank-local per-stage guard: the local bubbles' padded bbox must stay clear of the current block (init...
pure subroutine s_amr_patch_core(k, clo, chi)
Core [clo, chi] of the patch box that the ghost fill never reads, so the fill waves ship only the she...
MPI point-to-point waves for AMR exchange: post all receives, pack, post all sends,...
impure subroutine, public s_amr_wave_send(w, s, pool, dev)
One send per peer (the caller has packed every slice).
impure subroutine, public s_amr_wave_post(w, s, pool, dev)
One receive per peer into the side's pool runs.
type(t_amr_wave_side), public amr_wsend
Fill and restrict waves share amr_wsend/amr_wrecv; the seam wave stays open across the fills,...
impure subroutine, public s_amr_wave_add_slabs(s, peer, blk, msl, tb, te)
Add the msl slabs [tb:te] of block blk as transfers with peer.
impure subroutine, public s_amr_wave_close(s, pool, dev)
Lay the side out in its pool (peers in first-appearance order, transfers contiguous per peer); grows ...
impure subroutine, public s_amr_wave_wait(w)
Wait for all requests; under MFC_DEBUG each receive length must equal its plan.
impure subroutine, public s_amr_wave_open(w, b)
Open a wave on tag band b.
type(t_amr_wave), target, public amr_wave
pure subroutine, public s_amr_wave_slice(s, t, lo, hi)
Pool bounds [lo, hi] of transfer t's payload.
impure subroutine, public s_amr_wave_reset(s)
Start enumerating a side's transfers.
type(t_amr_wave_side), public amr_wrecv
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...
integer, dimension(3) amr_sidx
integer proc_rank
Rank of the local processor.
integer, dimension(:), allocatable amr_block_level
Refinement level of each active block (1..amr_max_level) and the deepest level populated....
logical, dimension(3) amr_dim
The active dimensions, and this rank's coarse subdomain origin and extent padded to 3-vectors (0 in c...
real(wp) amr_par_theta
subcycled level>=2 fills read (1-theta)*parent backup + theta*parent (1: current only)
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...
subroutine s_amr_select_slot(islot)
Make block slot islot the working slot: set amr_cur and copy its stored mirrors (region,...
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_gather
fill waves: coarse-patch gather into every block
real(wp), dimension(ph_n) acc
impure subroutine, public s_phase_tic(id)
impure subroutine, public s_phase_toc(id)
One side (send or receive) of a pooled wave.
Derived type annexing a scalar field (SF).