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
|
# Copyright (C) 2021 - 2024 Advanced Micro Devices, Inc. All rights reserved.
#
# Permission is hereby granted, free of charge, to any person obtaining a copy
# of this software and associated documentation files (the "Software"), to deal
# in the Software without restriction, including without limitation the rights
# to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
# copies of the Software, and to permit persons to whom the Software is
# furnished to do so, subject to the following conditions:
#
# The above copyright notice and this permission notice shall be included in
# all copies or substantial portions of the Software.
#
# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
# AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
# OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
# THE SOFTWARE.
"""Timing utilities."""
import collections
import perflib
import logging
from dataclasses import dataclass, field
from pathlib import Path as path
from typing import List
from typing import Set
from numpy import ceil
@dataclass
class Timer:
bench: str = ""
accutest: str = ""
active_tests_tokens: Set[bytes] = field(default_factory=set)
lib: List[str] = field(default_factory=list)
out: List[str] = field(default_factory=list)
device: int = 0
ntrial: int = 10
mp_size: int = 1
mp_exec: str = ""
ingrid: List[int] = field(default_factory=list)
outgrid: List[int] = field(default_factory=list)
ngpus: int = 1
verbose: bool = False
timeout: float = 0
sequence: int = None
hipskip: bool = True
def run_cases(self, generator):
bench = path(self.bench)
if not bench.is_file():
raise RuntimeError(f"Unable to find (dyna-)bench: {self.bench}")
if self.mp_size > 1 and self.mp_exec == None:
raise RuntimeError(f"Unable to find mpi executable {self.bench}")
failed_tokens = []
# get a list of powers of two less or equal than the number of requested resources,
# to be used for scalability experiments
gpu_list_pow2 = lambda n: [
2**i for i in range(int(ceil(n**0.5)) + 1) if 2**i <= n
]
total_prob_count = 0
no_accutest_prob_count = 0
for prob in generator.generate_problems():
total_prob_count += 1
n_resources = 1
# scalability for single-proc multi-GPU:
if self.ngpus > 1 and self.mp_size == 1:
n_resources = self.ngpus
# scalability for single-proc using 1-GPU per MPI:
elif self.mp_size > 1 and self.ngpus == 1:
n_resources = self.mp_size
scaling = prob.meta.get('scaling')
if scaling != None:
list_of_gpus = gpu_list_pow2(n_resources)
else:
list_of_gpus = [n_resources]
ws_factor = 1
for g in list_of_gpus:
token, seconds, success, __, __ = perflib.bench.run(
self.bench,
tuple([ws_factor * l for l in prob.length]),
direction=prob.direction,
real=prob.real,
inplace=prob.inplace,
precision=prob.precision,
nbatch=prob.nbatch,
mp_size=g if self.mp_size > 1 else 1,
mp_exec=self.mp_exec,
ingrid=self.ingrid,
outgrid=self.outgrid,
ngpus=g if self.ngpus > 1 else 1,
ntrial=self.ntrial,
device=self.device,
libraries=self.lib,
verbose=self.verbose,
timeout=self.timeout,
sequence=self.sequence,
skiphip=self.hipskip,
scalability=(scaling != None))
if scaling == 'weak':
ws_factor *= 2
if success:
for idx, vals in enumerate(seconds):
out = path(self.out[idx])
logging.info("output: " + str(out))
meta = {'title': prob.tag}
meta.update(prob.meta)
perflib.utils.write_dat(out, token, seconds[idx], meta)
else:
failed_tokens.append(token)
if self.active_tests_tokens and token.encode(
) not in self.active_tests_tokens:
no_accutest_prob_count += 1
logging.info(f'No accuracy test coverage for: ' + token)
if no_accutest_prob_count > 0:
print('\t')
logging.warning(
str(no_accutest_prob_count) + f' out of ' +
str(total_prob_count) +
f' problems do not have accuracy coverage.' +
f' Refer to rocfft-perf.log for details.')
return failed_tokens
@dataclass
class GroupedTimer:
bench: str = ""
accutest: str = ""
active_tests_tokens: Set[bytes] = field(default_factory=set)
lib: List[str] = field(default_factory=list)
out: List[str] = field(default_factory=list)
device: int = 0
ntrial: int = 10
mp_size: int = 1
mp_exec: str = ""
ngpus: int = 1
verbose: bool = False
timeout: float = 0
def run_cases(self, generator):
failed_tokens = []
all_problems = collections.defaultdict(list)
for problem in generator.generate_problems():
all_problems[problem.tag].append(problem)
total_problems = sum([len(v) for v in all_problems.values()])
print(
f'Timing {total_problems} problems in {len(all_problems)} groups')
if self.accutest:
accutest = path(self.accutest)
if not accutest.is_file():
raise RuntimeError(
f'Unable to find accuracy test: {self.accutest}')
self.active_tests_tokens = perflib.accutest.get_active_tests_tokens(
accutest)
for i, (tag, problems) in enumerate(all_problems.items()):
print(
f'\n{tag} (group {i} of {len(all_problems)}): {len(problems)} problems'
)
timer = Timer(**self.__dict__)
timer.out = [path(x) / (tag + '.dat') for x in self.out]
failed_tokens += timer.run_cases(
perflib.generators.VerbatimGenerator(problems))
return failed_tokens
|
