wehub-resource-sync
83 lines
3.1 KiB
Python
83 lines
3.1 KiB
Python
# Copyright (c) 2025 PaddlePaddle Authors. All Rights Reserved.
|
|
#
|
|
# Licensed under the Apache License, Version 2.0 (the "License");
|
|
# you may not use this file except in compliance with the License.
|
|
# You may obtain a copy of the License at
|
|
#
|
|
# http://www.apache.org/licenses/LICENSE-2.0
|
|
#
|
|
# Unless required by applicable law or agreed to in writing, software
|
|
# distributed under the License is distributed on an "AS IS" BASIS,
|
|
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
# See the License for the specific language governing permissions and
|
|
# limitations under the License.
|
|
|
|
import unittest
|
|
|
|
import numpy as np
|
|
|
|
import paddle
|
|
|
|
|
|
@unittest.skipIf(
|
|
(not paddle.is_compiled_with_cuda()) or paddle.is_compiled_with_rocm,
|
|
'should compile with cuda.',
|
|
)
|
|
class TestVMMAllocator(unittest.TestCase):
|
|
def setUp(self):
|
|
self.GB = 1000**3
|
|
self.MB = 1000**2
|
|
paddle.set_flags({'FLAGS_dump_vmm_allocation_info': True})
|
|
paddle.set_flags({'FLAGS_use_virtual_memory_auto_growth': True})
|
|
paddle.set_flags({'FLAGS_native_compact': True})
|
|
self.cmds = [
|
|
["Alloc", 6 * self.GB, "0x100000000"],
|
|
["Alloc", 11 * self.GB, "0x100000001"],
|
|
["Alloc", 1 * self.GB, "0x100000002"],
|
|
["Alloc", 10 * self.GB, "0x100000003"],
|
|
["Free", 6 * self.GB, "0x100000000"],
|
|
["Free", 10 * self.GB, "0x100000003"],
|
|
["Alloc", 17 * self.GB, "0x100000004"],
|
|
]
|
|
|
|
def test_vmm_allocator(self):
|
|
params = {}
|
|
old_tensor1, old_tensor1_ptr, new_tensor1, new_tensor1_ptr = 0, 0, 0, 0
|
|
|
|
for op, size, ptr in self.cmds:
|
|
paddle.device.synchronize()
|
|
paddle_reserved1 = paddle.device.cuda.memory_reserved() // self.MB
|
|
|
|
if op == "Alloc":
|
|
params[ptr] = paddle.randn(
|
|
[int(int(size) / 4)], dtype='float32'
|
|
)
|
|
if op == "Free" and ptr in params:
|
|
del params[ptr]
|
|
|
|
if ptr == '0x100000001':
|
|
old_tensor1 = params['0x100000001'].numpy()[0:100]
|
|
old_tensor1_ptr = hex(params['0x100000001'].data_ptr())
|
|
|
|
paddle.device.synchronize()
|
|
paddle_reserved2 = paddle.device.cuda.memory_reserved() // self.MB
|
|
paddle_allocated2 = paddle.device.cuda.memory_allocated() // self.MB
|
|
paddle_max_reserved = (
|
|
paddle.device.cuda.max_memory_reserved() // self.MB
|
|
)
|
|
paddle_max_allocated = (
|
|
paddle.device.cuda.max_memory_allocated() // self.MB
|
|
)
|
|
|
|
print(
|
|
f"reserved = {paddle_reserved2} allocated = {paddle_allocated2} auto growth = {paddle_reserved2 - paddle_reserved1} max_allocated = {paddle_max_allocated} max_reserved = {paddle_max_reserved}"
|
|
)
|
|
new_tensor1 = params['0x100000001'].numpy()[0:100]
|
|
new_tensor1_ptr = hex(params['0x100000001'].data_ptr())
|
|
np.testing.assert_array_equal(old_tensor1, new_tensor1)
|
|
assert old_tensor1_ptr != new_tensor1_ptr
|
|
|
|
|
|
if __name__ == '__main__':
|
|
unittest.main()
|