wehub-resource-sync
141 lines
5.6 KiB
Python
141 lines
5.6 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 paddle
|
|
from paddle.device.cuda.memory_analyzer import MemoryAnalysisTool
|
|
|
|
|
|
@unittest.skipIf(
|
|
(not paddle.is_compiled_with_cuda()) or paddle.is_compiled_with_rocm(),
|
|
'should compile with cuda.',
|
|
)
|
|
class TestAllocatorVisitor(unittest.TestCase):
|
|
def setUp(self):
|
|
self.GB = 1000**3
|
|
self.MB = 1000**2
|
|
self.cmds = [
|
|
["Alloc", 1 * self.GB, "0x100000000"],
|
|
["Alloc", 2 * self.GB, "0x100000001"],
|
|
["Alloc", 1 * self.GB, "0x100000002"],
|
|
["Alloc", 2 * self.GB, "0x100000003"],
|
|
["Free", 1 * self.GB, "0x100000000"],
|
|
["Free", 2 * self.GB, "0x100000003"],
|
|
]
|
|
self.cmds2 = [
|
|
["Alloc", 1 * self.MB, "0x100000010"],
|
|
["Alloc", 2 * self.MB, "0x100000011"],
|
|
["Alloc", 1 * self.MB, "0x100000012"],
|
|
["Alloc", 2 * self.MB, "0x100000013"],
|
|
["Free", 1 * self.MB, "0x100000010"],
|
|
["Free", 2 * self.MB, "0x100000013"],
|
|
]
|
|
paddle.set_flags({'FLAGS_use_virtual_memory_auto_growth': True})
|
|
|
|
def allocate_cmds(self, cmds):
|
|
params = {}
|
|
for op, size, ptr in 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]
|
|
|
|
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}"
|
|
)
|
|
return params
|
|
|
|
def test_multi_scale_alloc_free(self):
|
|
params = self.allocate_cmds(self.cmds)
|
|
MemoryAnalysisTool.vmm_max_free_size()
|
|
|
|
def test_block_info(self):
|
|
paddle.device.cuda.memory_summary()
|
|
params = self.allocate_cmds(self.cmds)
|
|
x = MemoryAnalysisTool.vmm_free_block_info()
|
|
y = MemoryAnalysisTool.vmm_all_block_info()
|
|
self.assertEqual(x[0][0][0], 1000000000)
|
|
self.assertEqual(x[0][1][0], 2002049024)
|
|
self.assertEqual(len(y), 1) # 1 allocators
|
|
self.assertEqual(len(y[0]), 4) # 4 blocks
|
|
|
|
def test_pool_filter_block_info(self):
|
|
"""Test vmm_large_all_block_info and vmm_small_all_block_info."""
|
|
# Allocate in large pool (>= vmm_small_pool_size_in_mb)
|
|
params = self.allocate_cmds(self.cmds)
|
|
|
|
large_info = MemoryAnalysisTool.vmm_large_all_block_info()
|
|
small_info = MemoryAnalysisTool.vmm_small_all_block_info()
|
|
all_info = MemoryAnalysisTool.vmm_all_block_info()
|
|
|
|
# large_info should have blocks (our 1GB allocations go to large pool)
|
|
self.assertGreater(len(large_info), 0)
|
|
self.assertGreater(len(large_info[0]), 0)
|
|
|
|
# Total blocks from large + small should equal all
|
|
total_filtered = sum(len(g) for g in large_info) + sum(
|
|
len(g) for g in small_info
|
|
)
|
|
total_all = sum(len(g) for g in all_info)
|
|
self.assertEqual(total_filtered, total_all)
|
|
|
|
def test_memory_summary(self):
|
|
paddle.set_flags({'FLAGS_use_virtual_memory_auto_growth': True})
|
|
paddle.device.cuda.memory_summary()
|
|
params = self.allocate_cmds(self.cmds2)
|
|
# paddle.device.cuda.memory_summary()
|
|
|
|
def test_memory_record(self):
|
|
paddle.set_flags({'FLAGS_use_virtual_memory_auto_growth': True})
|
|
paddle.set_flags({'FLAGS_record_alloc_event': True})
|
|
params = self.allocate_cmds(self.cmds)
|
|
params2 = self.allocate_cmds(self.cmds2)
|
|
paddle.device.cuda.allocate_record_plot()
|
|
paddle.device.cuda.allocate_record_plot(save_path="ana.png")
|
|
paddle.device.cuda.allocate_record_table()
|
|
paddle.device.cuda.memory_summary()
|
|
|
|
def test_memory_record_with_guard(self):
|
|
paddle.set_flags({'FLAGS_use_virtual_memory_auto_growth': True})
|
|
for _ in range(2):
|
|
with paddle.device.cuda.allocate_record_guard(True):
|
|
params = self.allocate_cmds(self.cmds2)
|
|
paddle.set_flags({'FLAGS_record_alloc_event': True})
|
|
with paddle.device.cuda.allocate_record_guard(False):
|
|
params2 = self.allocate_cmds(self.cmds2)
|
|
paddle.device.cuda.allocate_record_plot()
|
|
paddle.device.cuda.allocate_record_plot(save_path="ana.png")
|
|
paddle.device.cuda.allocate_record_table()
|
|
paddle.device.cuda.memory_summary()
|
|
|
|
|
|
if __name__ == '__main__':
|
|
unittest.main()
|