-
Notifications
You must be signed in to change notification settings - Fork 2
/
loop.py
354 lines (281 loc) · 12.3 KB
/
loop.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
import numpy as np
import type_, kernel, adapter
import adapter
from adapter import gpu_config
import numba
from numba import objmode, njit, jit, prange, cuda
from constant import *
simulation = None
# =============================================================================
# History-based
# =============================================================================
#@jit(nopython=True)
def HISTORY_simulation(mcdc, hostco):
# =========================================================================
# Simulation loop
# =========================================================================
for i_history in prange(mcdc['N_history']):
# =====================================================================
# Initialize history
# =====================================================================
# Create particle
P = kernel.create(type_.particle)
# Set RNG seed
P['seed'] = mcdc['seed']
kernel.rng_skip_ahead(i_history*mcdc['history_stride'], P, mcdc)
# Initialize particle
kernel.source(P, mcdc)
# "Push" to the bank
mcdc['bank']['content'][0] = kernel.record_particle(P)
mcdc['bank']['size'] = 1
# Reset main seed
mcdc['seed'] = P['seed']
# =====================================================================
# History loop
# =====================================================================
while mcdc['bank']['size'] > 0:
# =================================================================
# Initialize particle
# =================================================================
# "Pop" particle from bank
mcdc['bank']['size'] -= 1
idx = mcdc['bank']['size']
P = kernel.read_particle(mcdc['bank']['content'][idx])
# Set particle seed
P['seed'] = mcdc['seed']
# =================================================================
# Particle loop
# =================================================================
# Particle loop
while P['alive']:
# Move to event
kernel.move(P, mcdc)
# Event
event = P['event']
# Collision
if event == EVENT_SCATTERING:
kernel.scattering(P, mcdc)
elif event == EVENT_FISSION:
kernel.fission(P, mcdc)
elif event == EVENT_LEAKAGE:
kernel.leakage(P, mcdc)
elif event == EVENT_BRANCHLESS_COLLISION:
kernel.branchless_collision(P, mcdc)
# Update main seed
mcdc['seed'] = P['seed']
# =============================================================================
# Event-based
# =============================================================================
#init_stack = None
def EVENT_simulation(mcdc, hostco):
# =========================================================================
# Initialize simulation
# =========================================================================
#init_stack = None
#if mcdc['gpu']:
#print('Location A')
#kernel.initialize_stack(mcdc,hostco)
#else:
#!kernel.initialize_stack(mcdc, hostco)
#b,t = adapter.gpu_config(mcdc['N_particle'], hostco)
b,t = adapter.gpu_config(int(1E6), hostco)
gpu_hostco = cuda.to_device(hostco)
gpu_mcdc = cuda.to_device(mcdc)
kernel.initialize_stack[b,t](gpu_mcdc, gpu_hostco)
# =========================================================================
# Simulation loop
# =========================================================================
#print('To simulation')
it = 0
while np.max(hostco['stack_size'][1:]) > 0:
it += 1
#print(it)
# =====================================================================
# Initialize event
# =====================================================================
# Determine next event executed based on the longest stack
#gpu_hostco.copy_to_host(hostco)
stack = np.argmax(hostco['stack_size'][1:]) + 1 # Offset for EVENT_NONE
event = hostco['event_idx'][stack]
#print(event)
# =================================================================
# Event loop
# =================================================================
if event == EVENT_SOURCE:
#print('Source! {}'.format(event))
kernel.source(mcdc, gpu_mcdc, hostco, gpu_hostco)
elif event == EVENT_MOVE:
#print('Move! {}'.format(event))
kernel.move(mcdc, gpu_mcdc, hostco, gpu_hostco)
elif event == EVENT_SCATTERING:
#print('Scattering! {}'.format(event))
kernel.scattering(mcdc, gpu_mcdc, hostco, gpu_hostco)
elif event == EVENT_FISSION:
#print('Fission! {}'.format(event))
kernel.fission(mcdc, gpu_mcdc, hostco, gpu_hostco)
elif event == EVENT_LEAKAGE:
#print('Leak! {}'.format(event))
kernel.leakage(mcdc, gpu_mcdc, hostco, gpu_hostco)
elif event == EVENT_BRANCHLESS_COLLISION:
#print('Branchless Collision!', event)
kernel.branchless_collision(mcdc, gpu_mcdc, hostco, gpu_hostco)
'''
print(hostco['stack_size'])
print(mcdc['stack_']['size'])
for i in range(hostco['stack_size'].shape[0]):
size = mcdc['stack_'][i]['size']
if size > 0:
print(i, size, mcdc['stack_'][i]['content'][:size])
print(mcdc['bank'])
print('\n\n')
'''
gpu_mcdc.copy_to_host(mcdc)
path_to_harmonize='../harmonize'
import sys
sys.path.append(path_to_harmonize)
import harmonize as harm
def ASYNC_simulation_factory(single_fn=True, asynchronous=True):
dev_state_type = numba.from_dtype(type_.global_)
grp_state_type = numba.from_dtype(np.dtype([ ]))
thd_state_type = numba.from_dtype(np.dtype([ ]))
particle = numba.from_dtype(type_.particle)
def initialize(prog: numba.uintp):
pass
def finalize(prog: numba.uintp):
pass
def continuation(prog: numba.uintp, P: particle):
if P['event'] == EVENT_SOURCE:
source_async(prog,P)
elif P['event'] == EVENT_MOVE:
move_async(prog,P)
elif P['event'] == EVENT_SCATTERING:
scattering_async(prog,P)
elif P['event'] == EVENT_FISSION:
fission_async(prog,P)
elif P['event'] == EVENT_LEAKAGE:
leakage_async(prog,P)
elif P['event'] == EVENT_BRANCHLESS_COLLISION:
bcollision_async(prog,P)
def source(prog: numba.uintp, P: particle):
kernel.source(P, device(prog))
continuation(prog,P)
def move(prog: numba.uintp, P: particle):
kernel.move(P, device(prog))
continuation(prog,P)
def scattering(prog: numba.uintp, P: particle):
kernel.scattering(P, device(prog))
continuation(prog,P)
def fission(prog: numba.uintp, P: particle):
n = kernel.fission(P, device(prog))
for i in range(n):
P_new = numba.cuda.local.array(1,particle)[0]
P_new['x'] = P['x']
P_new['ux'] = -1.0 + 2.0*kernel.rng(P, device(prog))
P_new['w'] = P['w']
P_new['seed'] = P['seed']
P_new['event'] = EVENT_MOVE
P_new['alive'] = True
continuation(prog,P_new)
kernel.terminate_particle(P)
def leakage(prog: numba.uintp, P: particle):
kernel.leakage(P, device(prog))
continuation(prog,P)
def bcollision(prog: numba.uintp, P: particle):
kernel.branchless_collision(P, device(prog))
continuation(prog,P)
def iterate(prog: numba.uintp, P: particle):
if P['event'] == EVENT_SOURCE:
kernel.source(P, device(prog))
elif P['event'] == EVENT_MOVE:
kernel.move(P, device(prog))
elif P['event'] == EVENT_SCATTERING:
kernel.scattering(P, device(prog))
elif P['event'] == EVENT_FISSION:
n = kernel.fission(P, device(prog))
for i in range(n):
P_new = numba.cuda.local.array(1,particle)[0]
P_new['x'] = P['x']
P_new['ux'] = -1.0 + 2.0*kernel.rng(P, device(prog))
P_new['w'] = P['w']
P_new['seed'] = P['seed']
P_new['event'] = EVENT_MOVE
P_new['alive'] = True
iterate_async(prog,P_new)
kernel.terminate_particle(P)
elif P['event'] == EVENT_LEAKAGE:
kernel.leakage(P, device(prog))
elif P['event'] == EVENT_BRANCHLESS_COLLISION:
kernel.branchless_collision(P, device(prog))
if P['event'] != EVENT_NONE:
iterate_async(prog,P)
state_spec = (dev_state_type,grp_state_type,thd_state_type)
one_event_fns = [iterate]
multi_event_fns = [source,move,scattering,fission,leakage,bcollision]
device, group, thread = harm.RuntimeSpec.access_fns(state_spec)
iterate_async, = harm.RuntimeSpec.async_dispatch(iterate)
source_async, move_async, scattering_async, fission_async, leakage_async, bcollision_async = \
harm.RuntimeSpec.async_dispatch(source,move,scattering,fission,leakage,bcollision)
continuation = adapter.compiler(continuation,"gpu_device")
program_spec = None
if single_fn:
def make_work(prog: numba.uintp) -> numba.boolean:
N_particle = device(prog)['N_particle']
WARP_SIZE = 32
shared_offset = numba.cuda.shared.array(1,numba.uint64)
if numba.cuda.threadIdx.x == 0:
shared_offset[0] = numba.cuda.atomic.add(device(prog)['source_counter'],0,WARP_SIZE)
index = numba.cuda.atomic.add(shared_offset,0,1)
if index >= N_particle:
return False
new_particle = numba.cuda.local.array(1,particle)[0]
new_particle['event'] = EVENT_SOURCE
new_particle['seed'] = index
iterate_async(prog,new_particle)
return True
base_fns = (initialize,finalize,make_work)
program_spec = harm.RuntimeSpec("mcdc",state_spec,base_fns,one_event_fns,WORK_ARENA_SIZE=655360)
else:
def make_work(prog: numba.uintp) -> numba.boolean:
N_particle = device(prog)['N_particle']
index = numba.cuda.atomic.add(device(prog)['source_counter'],0,1)
if index >= N_particle:
return False
new_particle = numba.cuda.local.array(1,particle)[0]
new_particle['event'] = EVENT_SOURCE
new_particle['seed'] = index
source_async(prog,new_particle)
return True
base_fns = (initialize,finalize,make_work)
program_spec = harm.RuntimeSpec("mcdc",state_spec,base_fns,multi_event_fns,WORK_ARENA_SIZE=655360)
if asynchronous:
runtime = program_spec.harmonize_instance()
else:
runtime = program_spec.event_instance(io_capacity=65536*4,load_margin=1024)
def runner(mcdc,hostco):
runtime.init(256)
runtime.store_state(mcdc)
if asynchronous:
runtime.exec(655360,256)
else:
runtime.exec(4,1024)
runtime.load_state(mcdc)
return runner
# =============================================================================
# Factory
# =============================================================================
def make_loops(alg, target):
global simulation
if alg == 'history':
simulation = adapter.loop(HISTORY_simulation, target)
elif alg == 'event':
simulation = adapter.loop(EVENT_simulation, target)
elif alg == 'async':
simulation = ASYNC_simulation_factory(True,True)
elif alg == 'async-multi':
simulation = ASYNC_simulation_factory(False,True)
elif alg == 'new-event':
simulation = ASYNC_simulation_factory(True,False)
elif alg == 'new-event-multi':
simulation = ASYNC_simulation_factory(False,False)
else:
print(f"[ERROR] Unrecognized algorithm type '{alg}'")