derek.holloway
7502bf64a4
CodeQL / Analyze (actions) (push) Canceled after 0s
CodeQL / Analyze (python) (push) Canceled after 0s
pre-commit / pre-commit (push) Canceled after 0s
Multi-Arch CI / setup (push) Canceled after 0s
Publish build_manylinux_x86_64 images / publish_build_manylinux_x86_64 (push) Canceled after 0s
Setup (Multi-Arch) / setup (push) Canceled after 0s
Multi-Arch CI / Linux::${{ fromJSON(needs.setup.outputs.linux_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish no_rocm_image_ubi10 image / publish_no_rocm_image_ubi10 (push) Canceled after 0s
Multi-Arch CI / Windows::${{ fromJSON(needs.setup.outputs.windows_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Multi-Arch CI / Manifest Diff (push) Canceled after 0s
TheRock Manifest Diff Report / Generate Manifest Diff Report (push) Canceled after 0s
Multi-Arch CI / CI Summary (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04 (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04_ocl_rt images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04_rocgdb images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Unit Tests / Unit Tests :: ubuntu-24.04 (push) Canceled after 0s
Unit Tests / Unit Tests :: windows-2022 (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_media (push) Canceled after 0s
Unit Tests / Unit Tests Summary (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_openmpi (push) Canceled after 0s
421 lines
18 KiB
Python
Executable File
421 lines
18 KiB
Python
Executable File
#!/usr/bin/env python3
|
|
#
|
|
# Copyright (C) Advanced Micro Devices. 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.
|
|
"""GPU partition: compute partition set, XGMI PLPD."""
|
|
|
|
import unittest
|
|
|
|
import common.common as common
|
|
from common.common import amdsmi
|
|
|
|
|
|
class TestGpuPartition(unittest.TestCase):
|
|
@classmethod
|
|
def setUpClass(cls):
|
|
cls.common = common.Common(common.verbose)
|
|
|
|
@classmethod
|
|
def tearDownClass(cls):
|
|
try:
|
|
amdsmi.amdsmi_shut_down()
|
|
except amdsmi.AmdSmiLibraryException:
|
|
pass
|
|
|
|
def setUp(self):
|
|
self.raise_exception = None
|
|
self.common.amdsmi_smart_init()
|
|
self.common.processors = amdsmi.amdsmi_get_processor_handles()
|
|
|
|
def tearDown(self):
|
|
amdsmi.amdsmi_shut_down()
|
|
|
|
def test_set_gpu_compute_partition(self):
|
|
self.common.print_func_name("")
|
|
|
|
if self.common.TODO_SKIP_FAIL:
|
|
msg = "\tSkipping test_set_gpu_compute_partition as it fails on MI300."
|
|
self.common.print(msg)
|
|
self.skipTest(msg)
|
|
|
|
for i, gpu in enumerate(self.common.processors):
|
|
self.common.print_device_header(i)
|
|
default_compute_partition_type = common.COMPUTE_PARTITION_TYPES[0][1]
|
|
msg = f"\t### amdsmi_get_gpu_compute_partition(gpu={i}):"
|
|
try:
|
|
default_compute_partition_name = amdsmi.amdsmi_get_gpu_compute_partition(gpu)
|
|
self.common.print(msg, default_compute_partition_name)
|
|
self.common.check_ret("", "", self.common.PASS)
|
|
except (amdsmi.AmdSmiLibraryException, amdsmi.AmdSmiParameterException) as e:
|
|
if self.common.check_ret(msg, e, self.common.PASS):
|
|
self.raise_exception = e
|
|
continue
|
|
|
|
for (
|
|
compute_partition_type_name,
|
|
compute_partition_type,
|
|
compute_partition_type_cond,
|
|
) in common.COMPUTE_PARTITION_TYPES:
|
|
if default_compute_partition_name == compute_partition_type_name:
|
|
default_compute_partition_type = compute_partition_type
|
|
msg = f"\t### amdsmi_set_gpu_compute_partition(gpu={i}, compute_partition_type={compute_partition_type_name}):"
|
|
try:
|
|
ret = amdsmi.amdsmi_set_gpu_compute_partition(gpu, compute_partition_type)
|
|
self.common.print(msg, ret)
|
|
self.common.check_ret("", "", self.common.PASS)
|
|
except (amdsmi.AmdSmiLibraryException, amdsmi.AmdSmiParameterException) as e:
|
|
if self.common.check_ret(msg, e, compute_partition_type_cond):
|
|
self.raise_exception = e
|
|
|
|
msg = f"\t### amdsmi_set_gpu_compute_partition(gpu={i}, default_compute_partition={default_compute_partition_name}):"
|
|
try:
|
|
ret = amdsmi.amdsmi_set_gpu_compute_partition(gpu, default_compute_partition_type)
|
|
self.common.print(msg, ret)
|
|
self.common.check_ret("", "", self.common.PASS)
|
|
except (amdsmi.AmdSmiLibraryException, amdsmi.AmdSmiParameterException) as e:
|
|
if self.common.check_ret(msg, e, self.common.PASS):
|
|
self.raise_exception = e
|
|
continue
|
|
|
|
if self.raise_exception:
|
|
raise self.raise_exception
|
|
return
|
|
|
|
# integration
|
|
|
|
def test_xgmi_plpd(self):
|
|
self.common.print_func_name("")
|
|
|
|
if self.common.TODO_SKIP_FAIL:
|
|
msg = "\tSkipping test_set_xgmi_plpd as it fails on MI300."
|
|
self.common.print(msg)
|
|
self.skipTest(msg)
|
|
|
|
for i, gpu in enumerate(self.common.processors):
|
|
self.common.print_device_header(i)
|
|
msg = f"gpu({i}):"
|
|
|
|
# Get current policy info
|
|
msg = f"\t### amdsmi_get_xgmi_plpd(gpu={i}):"
|
|
try:
|
|
policy_info = amdsmi.amdsmi_get_xgmi_plpd(gpu)
|
|
self.common.print(msg, "")
|
|
self.common.check_ret("", "", self.common.PASS)
|
|
|
|
num_supported = policy_info["num_supported"]
|
|
if not isinstance(num_supported, int):
|
|
self.common.print(f"Cannot determine num_supported={num_supported}", "")
|
|
continue
|
|
policy_id_current = policy_info["current_id"]
|
|
if not isinstance(policy_id_current, int):
|
|
self.common.print(f"Cannot determine policy_id_current={policy_id_current}", "")
|
|
continue
|
|
policy_id_orig = policy_info["policies"][policy_id_current]["policy_id"]
|
|
if not isinstance(policy_id_orig, int):
|
|
self.common.print(f"Cannot determine orig policy_id={policy_id_orig}", "")
|
|
continue
|
|
index = 0
|
|
if num_supported >= 2:
|
|
if policy_id_current != 0:
|
|
index = 1
|
|
policy_id = policy_info["policies"][index]["policy_id"]
|
|
if not isinstance(policy_id, int):
|
|
self.common.print(f"Cannot determine policy_id={policy_id}", "")
|
|
continue
|
|
except (amdsmi.AmdSmiLibraryException, amdsmi.AmdSmiParameterException) as e:
|
|
if self.common.check_ret(msg, e, self.common.PASS):
|
|
self.raise_exception = e
|
|
continue
|
|
|
|
# Set policy
|
|
msg = f"\t### amdsmi_set_xgmi_plpd(gpu={i}, policy_id={policy_id}):"
|
|
try:
|
|
ret = amdsmi.amdsmi_set_xgmi_plpd(gpu, policy_id)
|
|
self.common.print(msg, ret)
|
|
self.common.check_ret("", "", self.common.PASS)
|
|
except (amdsmi.AmdSmiLibraryException, amdsmi.AmdSmiParameterException) as e:
|
|
if self.common.check_ret(msg, e, self.common.PASS):
|
|
self.raise_exception = e
|
|
|
|
# Set back to original policy
|
|
msg = f"\t### amdsmi_set_xgmi_plpd(gpu={i}, policy_id={policy_id_orig}):"
|
|
try:
|
|
ret = amdsmi.amdsmi_set_xgmi_plpd(gpu, policy_id_orig)
|
|
self.common.check_ret("", "", self.common.PASS)
|
|
except (amdsmi.AmdSmiLibraryException, amdsmi.AmdSmiParameterException) as e:
|
|
if self.common.check_ret(msg, e, self.common.PASS):
|
|
self.raise_exception = e
|
|
|
|
if self.raise_exception:
|
|
raise self.raise_exception
|
|
return
|
|
|
|
# import multiprocessing
|
|
# Unstable on workstation cards
|
|
# def test_walkthrough_multiprocess(self):
|
|
# print("\n\n========> test_walkthrough_multiprocess start <========\n")
|
|
# processors = amdsmi.amdsmi_get_processor_handles()
|
|
# self.assertGreaterEqual(len(processors), 1)
|
|
# self.assertLessEqual(len(processors), self.common.max_num_physical_devices)
|
|
# p0 = multiprocessing.Process(target=walk_through, args=[self])
|
|
# p1 = multiprocessing.Process(target=walk_through, args=[self])
|
|
# p2 = multiprocessing.Process(target=walk_through, args=[self])
|
|
# p3 = multiprocessing.Process(target=walk_through, args=[self])
|
|
# p0.start()
|
|
# p1.start()
|
|
# p2.start()
|
|
# p3.start()
|
|
# p0.join()
|
|
# p1.join()
|
|
# p2.join()
|
|
# p3.join()
|
|
# print("\n========> test_walkthrough_multiprocess end <========\n")
|
|
|
|
# import threading
|
|
# Unstable on workstation cards
|
|
# def test_walkthrough_multithread(self):
|
|
# print("\n\n========> test_walkthrough_multithread start <========\n")
|
|
# processors = amdsmi.amdsmi_get_processor_handles()
|
|
# self.assertGreaterEqual(len(processors), 1)
|
|
# self.assertLessEqual(len(processors), self.common.max_num_physical_devices)
|
|
# t0 = threading.Thread(target=walk_through, args=[self])
|
|
# t1 = threading.Thread(target=walk_through, args=[self])
|
|
# t2 = threading.Thread(target=walk_through, args=[self])
|
|
# t3 = threading.Thread(target=walk_through, args=[self])
|
|
# t0.start()
|
|
# t1.start()
|
|
# t2.start()
|
|
# t3.start()
|
|
# t0.join()
|
|
# t1.join()
|
|
# t2.join()
|
|
# t3.join()
|
|
# print("\n========> test_walkthrough_multithread end <========\n")
|
|
|
|
# # Unstable - do not run
|
|
# def test_z_gpureset_asicinfo_multithread(self):
|
|
# def get_asic_info(processor):
|
|
# try:
|
|
# print("\n###Test amdsmi_get_gpu_asic_info \n")
|
|
# asic_info = amdsmi.amdsmi_get_gpu_asic_info(processor)
|
|
# except amdsmi.AmdSmiLibraryException as e:
|
|
# self.common.check_exception(e)
|
|
# continue
|
|
# print(" asic_info['market_name'] is: {}".format(
|
|
# asic_info['market_name']))
|
|
# print(" asic_info['vendor_id'] is: {}".format(
|
|
# asic_info['vendor_id']))
|
|
# print(" asic_info['vendor_name'] is: {}".format(
|
|
# asic_info['vendor_name']))
|
|
# print(" asic_info['device_id'] is: {}".format(
|
|
# asic_info['device_id']))
|
|
# print(" asic_info['rev_id'] is: {}".format(
|
|
# asic_info['rev_id']))
|
|
# print(" asic_info['asic_serial'] is: {}".format(
|
|
# asic_info['asic_serial']))
|
|
# print(" asic_info['oam_id'] is: {}\n".format(
|
|
# asic_info['oam_id']))
|
|
# def gpu_reset(processor):
|
|
# print("\n###Test amdsmi_reset_gpu \n")
|
|
# amdsmi.amdsmi_reset_gpu(processor)
|
|
# print(" GPU reset completed.\n")
|
|
# print("\n\n========> test_z_gpureset_asicinfo_multithread start <========\n")
|
|
# processors = amdsmi.amdsmi_get_processor_handles()
|
|
# self.assertGreaterEqual(len(processors), 1)
|
|
# self.assertLessEqual(len(processors), self.common.max_num_physical_devices)
|
|
# for i in range(0, len(processors)):
|
|
# bdf = amdsmi.amdsmi_get_gpu_device_bdf(processors[i])
|
|
# print("\n\n###Test Processor {}, bdf: {}".format(i, bdf))
|
|
# t0 = threading.Thread(target=get_asic_info, args=[processors[i]])
|
|
# t1 = threading.Thread(target=gpu_reset, args=[processors[i]])
|
|
# # t2 = threading.Thread(target=walk_through, args=[self])
|
|
# # t3 = threading.Thread(target=walk_through, args=[self])
|
|
# t0.start()
|
|
# t1.start()
|
|
# # t2.start()
|
|
# # t3.start()
|
|
# t0.join()
|
|
# t1.join()
|
|
# # t2.join()
|
|
# # t3.join()
|
|
# print("\n========> test_z_gpureset_asicinfo_multithread end <========\n")
|
|
|
|
def test_get_gpu_accelerator_partition_profile(self):
|
|
self.common.print_func_name("")
|
|
self.common.Test_API_Per_GPU(
|
|
amdsmi_get_gpu_accelerator_partition_profile=amdsmi.amdsmi_get_gpu_accelerator_partition_profile
|
|
)
|
|
return
|
|
|
|
def test_get_gpu_accelerator_partition_profile_config(self):
|
|
self.common.print_func_name("")
|
|
self.common.Test_API_Per_GPU(
|
|
amdsmi_get_gpu_accelerator_partition_profile_config=amdsmi.amdsmi_get_gpu_accelerator_partition_profile_config
|
|
)
|
|
return
|
|
|
|
def test_get_gpu_compute_partition(self):
|
|
self.common.print_func_name("")
|
|
self.common.Test_API_Per_GPU(
|
|
amdsmi_get_gpu_compute_partition=amdsmi.amdsmi_get_gpu_compute_partition
|
|
)
|
|
return
|
|
|
|
def test_get_gpu_accelerator_partition_mem_alloc_mode(self):
|
|
self.common.print_func_name("")
|
|
self.common.Test_API_Per_GPU(
|
|
amdsmi_get_gpu_accelerator_partition_mem_alloc_mode=amdsmi.amdsmi_get_gpu_accelerator_partition_mem_alloc_mode
|
|
)
|
|
return
|
|
|
|
def test_get_gpu_memory_partition(self):
|
|
self.common.print_func_name("")
|
|
self.common.Test_API_Per_GPU(
|
|
amdsmi_get_gpu_memory_partition=amdsmi.amdsmi_get_gpu_memory_partition
|
|
)
|
|
return
|
|
|
|
def test_get_gpu_memory_partition_config(self):
|
|
self.common.print_func_name("")
|
|
|
|
if self.common.TODO_SKIP_FAIL:
|
|
msg = "\tSkipping test_get_gpu_memory_partition_config as it fails on MI300."
|
|
self.common.print(msg)
|
|
self.skipTest(msg)
|
|
|
|
self.common.Test_API_Per_GPU(
|
|
amdsmi_get_gpu_memory_partition_config=amdsmi.amdsmi_get_gpu_memory_partition_config
|
|
)
|
|
return
|
|
|
|
def test_set_gpu_accelerator_partition_profile(self):
|
|
self.common.print_func_name("")
|
|
|
|
if self.common.TODO_SKIP_NOT_COMPLETE:
|
|
msg = "\tSkipping test_set_gpu_accelerator_partition_profile as it is not complete."
|
|
self.common.print(msg)
|
|
self.skipTest(msg)
|
|
|
|
# TODO profile_index = 0
|
|
profile_index = 0
|
|
|
|
self.common.Test_API_Per_GPU(
|
|
amdsmi_set_gpu_accelerator_partition_profile=amdsmi.amdsmi_set_gpu_accelerator_partition_profile,
|
|
profile_index=profile_index,
|
|
)
|
|
return
|
|
|
|
# Uses clk_type_name instead of clk_type
|
|
# Uses clk_limit_type_name instead of clk_limit_type
|
|
|
|
def test_set_gpu_memory_partition_return_codes(self):
|
|
"""Verify that set accepts supported modes and rejects unsupported ones."""
|
|
self.common.print_func_name("")
|
|
|
|
self.common.Test_Per_GPU_With_One_Enum(
|
|
amdsmi_set_gpu_memory_partition_mode=amdsmi.amdsmi_set_gpu_memory_partition_mode,
|
|
memory_partition_type=common.MEMORY_PARTITION_TYPES,
|
|
)
|
|
return
|
|
|
|
def test_set_gpu_memory_partition_idempotent(self):
|
|
"""Setting the current partition mode must succeed and leave the read-back unchanged."""
|
|
self.common.print_func_name("")
|
|
|
|
NPS_NAME_TO_TYPE = {
|
|
"NPS1": amdsmi.AmdSmiMemoryPartitionType.NPS1,
|
|
"NPS2": amdsmi.AmdSmiMemoryPartitionType.NPS2,
|
|
"NPS4": amdsmi.AmdSmiMemoryPartitionType.NPS4,
|
|
"NPS8": amdsmi.AmdSmiMemoryPartitionType.NPS8,
|
|
}
|
|
|
|
raise_exception = None
|
|
for i, gpu in enumerate(self.common.processors):
|
|
self.common.print_device_header(i)
|
|
|
|
# Read current partition.
|
|
current_partition = None
|
|
msg = f"\t### amdsmi_get_gpu_memory_partition(gpu={i}):"
|
|
try:
|
|
current_partition = amdsmi.amdsmi_get_gpu_memory_partition(gpu)
|
|
self.common.print(msg, current_partition)
|
|
except (amdsmi.AmdSmiLibraryException, amdsmi.AmdSmiParameterException) as e:
|
|
if self.common.check_ret(msg, e, self.common.PASS):
|
|
raise_exception = e
|
|
continue
|
|
|
|
if current_partition not in NPS_NAME_TO_TYPE:
|
|
self.common.print(f"\t [skip] Unknown partition mode: {current_partition}", "")
|
|
continue
|
|
|
|
# Set the same mode again — must succeed.
|
|
msg = f"\t### amdsmi_set_gpu_memory_partition idempotent(gpu={i}, mode={current_partition}):"
|
|
try:
|
|
amdsmi.amdsmi_set_gpu_memory_partition(gpu, NPS_NAME_TO_TYPE[current_partition])
|
|
self.common.print(msg, "SUCCESS")
|
|
self.common.check_ret("", "", self.common.PASS)
|
|
except (amdsmi.AmdSmiLibraryException, amdsmi.AmdSmiParameterException) as e:
|
|
if self.common.check_ret(msg, e, self.common.PASS):
|
|
raise_exception = e
|
|
continue
|
|
|
|
# Read back — value must be unchanged (no reload occurred).
|
|
msg = f"\t### amdsmi_get_gpu_memory_partition post-idempotent-set(gpu={i}):"
|
|
try:
|
|
post_partition = amdsmi.amdsmi_get_gpu_memory_partition(gpu)
|
|
self.common.print(msg, post_partition)
|
|
if post_partition != current_partition:
|
|
self.common.print(
|
|
f"\t TEST FAILURE: partition changed from {current_partition} "
|
|
f"to {post_partition} without a driver reload",
|
|
"",
|
|
)
|
|
raise_exception = amdsmi.AmdSmiLibraryException(0)
|
|
except (amdsmi.AmdSmiLibraryException, amdsmi.AmdSmiParameterException) as e:
|
|
if self.common.check_ret(msg, e, self.common.PASS):
|
|
raise_exception = e
|
|
|
|
if raise_exception:
|
|
raise raise_exception
|
|
|
|
def test_set_gpu_memory_partition_mode(self):
|
|
self.common.print_func_name("")
|
|
|
|
if self.common.TODO_SKIP_FAIL:
|
|
msg = "\tSkipping test_set_gpu_memory_partition_mode as it fails (MI350X, AMDSMI_STATUS_UNEXPECTED_DATA)."
|
|
self.common.print(msg)
|
|
self.skipTest(msg)
|
|
|
|
self.common.Test_Per_GPU_With_One_Enum(
|
|
amdsmi_set_gpu_memory_partition_mode=amdsmi.amdsmi_set_gpu_memory_partition_mode,
|
|
memory_partition_mode=common.MEMORY_PARTITION_TYPES,
|
|
)
|
|
return
|
|
|
|
def test_set_gpu_accelerator_partition_mem_alloc_mode(self):
|
|
self.common.print_func_name("")
|
|
|
|
self.common.Test_Per_GPU_With_One_Enum(
|
|
amdsmi_set_gpu_accelerator_partition_mem_alloc_mode=amdsmi.amdsmi_set_gpu_accelerator_partition_mem_alloc_mode,
|
|
accelerator_partition_mem_alloc_mode=common.ACCELERATOR_PARTITION_MEM_ALLOC_MODE_TYPES,
|
|
)
|
|
return
|
|
|
|
# out of order freq_ind then value then clk_type
|