mirror of
https://github.com/zebrajr/pytorch.git
synced 2025-12-07 12:21:27 +01:00
93719440b8
Summary: Follow-up of https://github.com/pytorch/pytorch/issues/46461 with a similar goal Makes them more readable and possibly faster. Care has to be taken because `map` applies the function immediately while `(x for x in xs)` is a generator expression which gets evaluated later. This is a benefit in some cases where it is not required to actually create the list of values in memory (e.g. when passing to `tuple` or `extend` or `join`) Pull Request resolved: https://github.com/pytorch/pytorch/pull/46462 Reviewed By: zou3519 Differential Revision: D24422343 Pulled By: ezyang fbshipit-source-id: 252e33499c92ac0b15238f2df32681dbbda2b237
49 lines
1.2 KiB
Python
49 lines
1.2 KiB
Python
import tempfile
|
|
import contextlib
|
|
from . import cudart, check_error
|
|
|
|
|
|
DEFAULT_FLAGS = [
|
|
"gpustarttimestamp",
|
|
"gpuendtimestamp",
|
|
"gridsize3d",
|
|
"threadblocksize",
|
|
"streamid",
|
|
"enableonstart 0",
|
|
"conckerneltrace",
|
|
]
|
|
|
|
|
|
def init(output_file, flags=None, output_mode='key_value'):
|
|
rt = cudart()
|
|
if not hasattr(rt, 'cudaOutputMode'):
|
|
raise AssertionError("HIP does not support profiler initialization!")
|
|
flags = DEFAULT_FLAGS if flags is None else flags
|
|
if output_mode == 'key_value':
|
|
output_mode_enum = rt.cudaOutputMode.KeyValuePair
|
|
elif output_mode == 'csv':
|
|
output_mode_enum = rt.cudaOutputMode.CSV
|
|
else:
|
|
raise RuntimeError("supported CUDA profiler output modes are: key_value and csv")
|
|
with tempfile.NamedTemporaryFile(delete=True) as f:
|
|
f.write(b'\n'.join(f.encode('ascii') for f in flags))
|
|
f.flush()
|
|
check_error(rt.cudaProfilerInitialize(f.name, output_file, output_mode_enum))
|
|
|
|
|
|
def start():
|
|
check_error(cudart().cudaProfilerStart())
|
|
|
|
|
|
def stop():
|
|
check_error(cudart().cudaProfilerStop())
|
|
|
|
|
|
@contextlib.contextmanager
|
|
def profile():
|
|
try:
|
|
start()
|
|
yield
|
|
finally:
|
|
stop()
|