AI/rocm-systems
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

[rocprofiler-compute] Improve amdsmi interface (#2245)

Browse Source 
* Improve amdsmi interface

* Fix issue where max mem clock was being set as max gfx clock

* Handle the case when all device handles might not be usable due to
  devices being hidden by ROCR and HIP environment variables

* Fix get gpu vram size to return str in KB

* Improve testing of amdsmi interface functions
This commit is contained in:
vedithal-amd
2025-12-12 09:02:37 -05:00
committed by GitHub
parent ca0f3a6b5a
commit 793732a04e
3 changed files with 116 additions and 86 deletions
Download Patch File Download Diff File Expand all files Collapse all files
projects/rocprofiler-compute/tests/test_utils.py
+21 -14
View File
@@ -7719,21 +7719,22 @@ def test_amdsmi_ctx():
amdsmi_shutdown_mock.assert_called_once()
def test_amdsmi_get_device_handle():
from utils.amdsmi_interface import get_device_handle, import_amdsmi_module
def test_amdsmi_get_device_handles():
from utils.amdsmi_interface import get_device_handles, import_amdsmi_module
_ = import_amdsmi_module()
with mock.patch("amdsmi.amdsmi_get_processor_handles") as device_handles_mock:
device_handles_mock.return_value = [12345]
get_device_handle()
handles = get_device_handles()
assert handles[0] == 12345
device_handles_mock.assert_called_once()
with mock.patch(
"amdsmi.amdsmi_get_processor_handles", side_effect=Exception("Mock exception")
) as device_handles_mock:
handle = get_device_handle()
assert handle is None
handle = get_device_handles()
assert len(handle) == 0
def test_amdsmi_get_mem_max_clock():
@@ -7741,12 +7742,18 @@ def test_amdsmi_get_mem_max_clock():
_ = import_amdsmi_module()
with mock.patch("amdsmi.amdsmi_get_processor_handles") as device_handles_mock:
device_handles_mock.return_value = [12345]
with mock.patch("utils.amdsmi_interface.get_device_handles") as device_handles_mock:
device_handles_mock.return_value = [0, 4567]
with mock.patch("amdsmi.amdsmi_get_clock_info") as mem_max_clock_mock:
mem_max_clock_mock.return_value = {"max_clk": 100}
def side_effect(handle, *args, **kwargs):
if handle == 0:
raise Exception("Invalid handle: 0")
return {"max_clk": 100}
mem_max_clock_mock.side_effect = side_effect
clk = get_mem_max_clock()
mem_max_clock_mock.assert_called_once()
assert mem_max_clock_mock.call_count == 2
assert clk == 100
@@ -7755,7 +7762,7 @@ def test_amdsmi_get_gpu_model():
_ = import_amdsmi_module()
with mock.patch("amdsmi.amdsmi_get_processor_handles") as device_handles_mock:
with mock.patch("utils.amdsmi_interface.get_device_handles") as device_handles_mock:
device_handles_mock.return_value = [12345]
with mock.patch("amdsmi.amdsmi_get_gpu_board_info") as device_name_mock:
with mock.patch("amdsmi.amdsmi_get_gpu_asic_info") as asic_name_mock:
@@ -7771,7 +7778,7 @@ def test_amdsmi_get_gpu_model():
"amdsmi.amdsmi_get_gpu_board_info", side_effect=Exception("Mock exception")
):
model = get_gpu_model()
assert model == "N/A"
assert model == ("N/A", "N/A", "N/A")
def test_amdsmi_get_gpu_vbios_part_number():
@@ -7779,7 +7786,7 @@ def test_amdsmi_get_gpu_vbios_part_number():
_ = import_amdsmi_module()
with mock.patch("amdsmi.amdsmi_get_processor_handles") as device_handles_mock:
with mock.patch("utils.amdsmi_interface.get_device_handles") as device_handles_mock:
device_handles_mock.return_value = [12345]
with mock.patch("amdsmi.amdsmi_get_gpu_vbios_info") as vbios_part_number_mock:
vbios_part_number_mock.return_value = {
@@ -7801,7 +7808,7 @@ def test_amdsmi_get_gpu_compute_partition():
_ = import_amdsmi_module()
with mock.patch("amdsmi.amdsmi_get_processor_handles") as device_handles_mock:
with mock.patch("utils.amdsmi_interface.get_device_handles") as device_handles_mock:
device_handles_mock.return_value = [12345]
with mock.patch(
"amdsmi.amdsmi_get_gpu_compute_partition"
@@ -7824,7 +7831,7 @@ def test_amdsmi_get_gpu_memory_partition():
_ = import_amdsmi_module()
with mock.patch("amdsmi.amdsmi_get_processor_handles") as device_handles_mock:
with mock.patch("utils.amdsmi_interface.get_device_handles") as device_handles_mock:
device_handles_mock.return_value = [12345]
with mock.patch(
"amdsmi.amdsmi_get_gpu_memory_partition"