Added amd-smi partition as preliminary command.
new command includes following arguments:
- current - display the current partition information for the selected
gpu(s)
- memory - display memory partition information for the selected
gpu(s)
- accelerator - display accelerator partition information for the
selected gpu(s)
additional functionality will be added as more partition APIs are added.
Signed-off-by: gabrpham <Gabriel.Pham@amd.com>
Change-Id: Ica86160139002ef5213d6d4b0e390670aeef01c8
Этот коммит содержится в:
gabrpham
+24
-39
@@ -7,14 +7,15 @@ Full documentation for amd_smi_lib is available at [https://rocm.docs.amd.com/pr
|
||||
## amd_smi_lib for ROCm 6.3.0
|
||||
|
||||
### Changes
|
||||
- **Added support for GPU metrics 1.6 to `amdsmi_get_gpu_metrics_info()`**
|
||||
|
||||
- **Added support for GPU metrics 1.6 to `amdsmi_get_gpu_metrics_info()`**.
|
||||
Updated `amdsmi_get_gpu_metrics_info()` and structure `amdsmi_gpu_metrics_t` to include new fields for PVIOL / TVIOL, XCP (Graphics Compute Partitions) stats, and pcie_lc_perf_other_end_recovery:
|
||||
- `uint64_t accumulation_counter` - used for all throttled calculations
|
||||
- `uint64_t prochot_residency_acc` - Processor hot accumulator
|
||||
- `uint64_t ppt_residency_acc` - Package Power Tracking (PPT) accumulator (used in PVIOL calculations)
|
||||
- `uint64_t socket_thm_residency_acc` - Socket thermal accumulator - (used in TVIOL calculations)
|
||||
- `uint64_t vr_thm_residency_acc` - Voltage Rail (VR) thermal accumulator
|
||||
- `uint64_t hbm_thm_residency_acc` - High Bandwidth Memory (HBM) thermal accumulator
|
||||
- `uint64_t hbm_thm_residency_acc` - High Bandwidth Memory (HBM) thermal accumulator
|
||||
- `uint16_t num_partition` - corresponds to the current total number of partitions
|
||||
- `struct amdgpu_xcp_metrics_t xcp_stats[MAX_NUM_XCP]` - for each partition associated with current GPU, provides gfx busy & accumulators, jpeg, and decoder (VCN) engine utilizations
|
||||
- `uint32_t gfx_busy_inst[MAX_NUM_XCC]` - graphic engine utilization (%)
|
||||
@@ -23,11 +24,12 @@ Updated `amdsmi_get_gpu_metrics_info()` and structure `amdsmi_gpu_metrics_t` to
|
||||
- `uint64_t gfx_busy_acc[MAX_NUM_XCC]` - graphic engine utilization accumulated (%)
|
||||
- `uint32_t pcie_lc_perf_other_end_recovery` - corresponds to the pcie other end recovery counter
|
||||
|
||||
- **Added new violation status outputs and APIs: `amdsmi_status_t amdsmi_get_violation_status()`, `amd-smi metric --throttle`, and `amd-smi monitor --violation`**
|
||||
- **Added new violation status outputs and APIs: `amdsmi_status_t amdsmi_get_violation_status()`, `amd-smi metric --throttle`, and `amd-smi monitor --violation`**.
|
||||
***Only available for MI300+ ASICs.***
|
||||
Users can now retrieve violation status' through either our Python or C++ APIs. Additionally, we have
|
||||
added capability to view these outputs conviently through `amd-smi metric --throttle` and `amd-smi monitor --violation`.
|
||||
Example outputs are listed below (below is for reference, output is subject to change):
|
||||
|
||||
```shell
|
||||
$ amd-smi metric --throttle
|
||||
GPU: 0
|
||||
@@ -69,6 +71,7 @@ GPU: 1
|
||||
HBM_THERMAL_VIOLATION_PERCENT: 0 %
|
||||
...
|
||||
```
|
||||
|
||||
```shell
|
||||
$ amd-smi monitor --violation
|
||||
GPU PVIOL TVIOL PHOT_TVIOL VR_TVIOL HBM_TVIOL
|
||||
@@ -91,12 +94,12 @@ GPU PVIOL TVIOL PHOT_TVIOL VR_TVIOL HBM_TVIOL
|
||||
...
|
||||
```
|
||||
|
||||
- **Added ability to view XCP (Graphics Compute Partition) activity within `amd-smi metric --usage`**
|
||||
- **Added ability to view XCP (Graphics Compute Partition) activity within `amd-smi metric --usage`**.
|
||||
***Partition specific features are only available on MI300+ ASICs***
|
||||
Users can now retrieve graphic utilization statistic on a per-XCP (per-partition) basis. Here all XCP activities will be listed,
|
||||
but the current XCP is the partition id listed under both `amd-smi list` and `amd-smi static --partition`.
|
||||
|
||||
but the current XCP is the partition id listed under both `amd-smi list` and `amd-smi static --partition`.
|
||||
Example outputs are listed below (below is for reference, output is subject to change):
|
||||
|
||||
```shell
|
||||
$ amd-smi metric --usage
|
||||
GPU: 0
|
||||
@@ -161,7 +164,6 @@ GPU: 0
|
||||
XCP_6: [N/A, N/A, N/A, N/A, N/A, N/A, N/A, N/A]
|
||||
XCP_7: [N/A, N/A, N/A, N/A, N/A, N/A, N/A, N/A]
|
||||
|
||||
|
||||
GPU: 1
|
||||
USAGE:
|
||||
GFX_ACTIVITY: 0 %
|
||||
@@ -227,9 +229,10 @@ GPU: 1
|
||||
...
|
||||
```
|
||||
|
||||
- **Added `LC_PERF_OTHER_END_RECOVERY` CLI output to `amd-smi metric --pcie` and updated `amdsmi_get_pcie_info()` to include this value**
|
||||
- **Added `LC_PERF_OTHER_END_RECOVERY` CLI output to `amd-smi metric --pcie` and updated `amdsmi_get_pcie_info()` to include this value**.
|
||||
***Feature is only available on MI300+ ASICs***
|
||||
Users can now retrieve both through `amdsmi_get_pcie_info()` which has an updated structure:
|
||||
|
||||
```C
|
||||
typedef struct {
|
||||
...
|
||||
@@ -247,9 +250,10 @@ typedef struct {
|
||||
} pcie_metric;
|
||||
uint64_t reserved[32];
|
||||
} amdsmi_pcie_info_t;
|
||||
```
|
||||
```
|
||||
|
||||
- Example outputs are listed below (below is for reference, output is subject to change):
|
||||
|
||||
Example outputs are listed below (below is for reference, output is subject to change):
|
||||
```shell
|
||||
$ amd-smi metric --pcie
|
||||
GPU: 0
|
||||
@@ -284,7 +288,7 @@ GPU: 1
|
||||
...
|
||||
```
|
||||
|
||||
- **Updated BDF commands to look use KFD SYSFS for BDF: `amdsmi_get_gpu_device_bdf()`**
|
||||
- **Updated BDF commands to look use KFD SYSFS for BDF: `amdsmi_get_gpu_device_bdf()`**.
|
||||
This aligns BDF output with ROCm SMI.
|
||||
See below for overview as seen from `rsmi_dev_pci_id_get()` now provides partition ID. See API for better detail. Previously these bits were reserved bits (right before domain) and partition id was within function.
|
||||
- bits [63:32] = domain
|
||||
@@ -292,7 +296,6 @@ See below for overview as seen from `rsmi_dev_pci_id_get()` now provides partiti
|
||||
- bits [27:16] = reserved
|
||||
- bits [15: 0] = pci bus/device/function
|
||||
|
||||
|
||||
- **Moved python tests directory path install location**.
|
||||
- `/opt/<rocm-path>/share/amd_smi/pytest/..` to `/opt/<rocm-path>/share/amd_smi/tests/python_unittest/..`
|
||||
- On amd-smi-lib-tests uninstall, the amd_smi tests folder is removed.
|
||||
@@ -306,9 +309,7 @@ See below for overview as seen from `rsmi_dev_pci_id_get()` now provides partiti
|
||||
- **Added `amd-smi set -L/--clk-limit ...` command**.
|
||||
Equivalent to rocm-smi's '--extremum' command which sets sclk's or mclk's soft minimum or soft maximum clock frequency.
|
||||
|
||||
|
||||
|
||||
- **Added Pytest functionality to test amdsmi API calls in Python**.
|
||||
- **Added unittest functionality to test amdsmi API calls in Python**.
|
||||
|
||||
- **Changed the `power` parameter in `amdsmi_get_energy_count()` to `energy_accumulator`**.
|
||||
- Changes propagate forwards into the python interface as well, however we are maintaing backwards compatibility and keeping the `power` field in the python API until ROCm 6.4.
|
||||
@@ -341,7 +342,6 @@ Topology arguments:
|
||||
ID: 7 | BDF: 0000:df:00.0 | UUID: <redacted>
|
||||
all | Selects all devices
|
||||
|
||||
|
||||
-a, --access Displays link accessibility between GPUs
|
||||
-w, --weight Displays relative weight between GPUs
|
||||
-o, --hops Displays the number of hops between GPUs
|
||||
@@ -352,7 +352,6 @@ Topology arguments:
|
||||
-d, --dma Display P2P direct memory access (DMA) link capability between nodes
|
||||
-z, --bi-dir Display P2P bi-directional link capability between nodes
|
||||
|
||||
|
||||
Command Modifiers:
|
||||
--json Displays output in JSON format (human readable by default).
|
||||
--csv Displays output in CSV format (human readable by default).
|
||||
@@ -407,7 +406,6 @@ BI-DIRECTIONAL TABLE:
|
||||
0000:bf:00.0 F T T T F F SELF F
|
||||
0000:df:00.0 T T T F F T F SELF
|
||||
|
||||
|
||||
Legend:
|
||||
SELF = Current GPU
|
||||
ENABLED / DISABLED = Link is enabled or disabled
|
||||
@@ -504,10 +502,10 @@ GPU: 0
|
||||
TARGET_GRAPHICS_VERSION: gfx942
|
||||
```
|
||||
|
||||
- **Udpated Partition APIs and struct information and added and partition_id to `amd-smi static --partition` & `amd-smi list`**.
|
||||
- **Udpated Partition APIs and struct information and added and partition_id to `amd-smi static --partition`**.
|
||||
- As part of an overhaul to partition information, some partition information will be made available in the `amdsmi_accelerator_partition_profile_t`.
|
||||
- This struct will be filled out by a new API, `amdsmi_get_gpu_accelerator_partition_profile()`.
|
||||
- Future data from these APIs wil will eventually get added to `static --partition`.
|
||||
- Future data from these APIs wil will eventually get added to `amd-smi partition`.
|
||||
|
||||
```C
|
||||
#define AMDSMI_MAX_ACCELERATOR_PROFILE 32
|
||||
@@ -548,7 +546,6 @@ typedef union {
|
||||
uint32_t nps_cap_mask;
|
||||
} amdsmi_nps_caps_t;
|
||||
|
||||
|
||||
typedef struct {
|
||||
amdsmi_accelerator_partition_type_t profile_type; // SPX, DPX, QPX, CPX and so on
|
||||
uint32_t num_partitions; // On MI300X, SPX: 1, DPX: 2, QPX: 4, CPX: 8, length of resources array
|
||||
@@ -567,21 +564,6 @@ GPU: 0
|
||||
COMPUTE_PARTITION: CPX
|
||||
MEMORY_PARTITION: NPS4
|
||||
PARTITION_ID: 0
|
||||
|
||||
$ amd-smi list
|
||||
GPU: 0
|
||||
BDF: 0000:23:00.0
|
||||
UUID: <redacted>
|
||||
KFD_ID: 45412
|
||||
NODE_ID: 1
|
||||
PARTITION_ID: 0
|
||||
|
||||
GPU: 1
|
||||
BDF: 0000:26:00.0
|
||||
UUID: <redacted>
|
||||
KFD_ID: 59881
|
||||
NODE_ID: 2
|
||||
PARTITION_ID: 0
|
||||
```
|
||||
|
||||
### Removals
|
||||
@@ -610,7 +592,7 @@ plan to eventually remove partition ID from the function portion of the BDF (Bus
|
||||
- bits [7:3] = Device
|
||||
- bits [2:0] = Function (partition id maybe in bits [2:0]) <-- Fallback for non SPX modes
|
||||
|
||||
Previously in non-SPX modes (ex. CPX/TPX/DPX/etc) some MI3x ASICs would not report all logical GPU devices within AMD SMI.
|
||||
- Previously in non-SPX modes (ex. CPX/TPX/DPX/etc) some MI3x ASICs would not report all logical GPU devices within AMD SMI.
|
||||
|
||||
```shell
|
||||
$ amd-smi monitor -p -t -v
|
||||
@@ -650,9 +632,8 @@ GPU POWER GPU_TEMP MEM_TEMP VRAM_USED VRAM_TOTAL
|
||||
```
|
||||
|
||||
- **Fixed incorrect implementation of the Python API `amdsmi_get_gpu_metrics_header_info()`**.
|
||||
- **`amdsmitst` TestGpuMetricsRead now prints metric in correct units**
|
||||
|
||||
- **`amd-smi static --partition` will have updates with additional partition information from `amdsmi_get_gpu_accelerator_partition_profile()`**.
|
||||
- **`amdsmitst` TestGpuMetricsRead now prints metric in correct units**.
|
||||
|
||||
### Known issues
|
||||
|
||||
@@ -662,6 +643,10 @@ GPU POWER GPU_TEMP MEM_TEMP VRAM_USED VRAM_TOTAL
|
||||
|
||||
- **Python API for `amdsmi_get_energy_count()` will deprecate the `power` field in ROCm 6.4 and use `energy_accumulator` field instead**.
|
||||
|
||||
- **Added preliminary `amd-smi partition` command**.
|
||||
- The new partition command can be used to display GPU information, including memory and accelerator partition information.
|
||||
- The command will be at full functionality once additional partition information from `amdsmi_get_gpu_accelerator_partition_profile()` has been implemented.
|
||||
|
||||
## amd_smi_lib for ROCm 6.2.1
|
||||
|
||||
### Additions
|
||||
|
||||
@@ -94,7 +94,8 @@ if __name__ == "__main__":
|
||||
amd_smi_commands.reset,
|
||||
amd_smi_commands.monitor,
|
||||
amd_smi_commands.rocm_smi,
|
||||
amd_smi_commands.xgmi)
|
||||
amd_smi_commands.xgmi,
|
||||
amd_smi_commands.partition)
|
||||
try:
|
||||
try:
|
||||
argcomplete.autocomplete(amd_smi_parser)
|
||||
@@ -128,7 +129,6 @@ if __name__ == "__main__":
|
||||
sys.tracebacklimit = 10
|
||||
else:
|
||||
sys.tracebacklimit = -1
|
||||
|
||||
# Execute subcommands
|
||||
args.func(args)
|
||||
except amdsmi_cli_exceptions.AmdSmiException as e:
|
||||
|
||||
@@ -5043,6 +5043,8 @@ class AMDSMICommands():
|
||||
bitrate = pcie_speed_GTs_value
|
||||
max_bandwidth = bitrate * pcie_static['max_pcie_width']
|
||||
except amdsmi_exception.AmdSmiLibraryException as e:
|
||||
bitrate = "N/A"
|
||||
max_bandwidth = "N/A"
|
||||
logging.debug("Failed to get bitrate and bandwidth for GPU %s | %s", src_gpu_id,
|
||||
e.get_error_info())
|
||||
|
||||
@@ -5084,6 +5086,8 @@ class AMDSMICommands():
|
||||
read = metrics_info['xgmi_read_data_acc'][dest_gpu_id]
|
||||
write = metrics_info['xgmi_write_data_acc'][dest_gpu_id]
|
||||
except amdsmi_exception.AmdSmiLibraryException as e:
|
||||
read = "N/A"
|
||||
write = "N/A"
|
||||
logging.debug("Failed to get read data for %s to %s | %s",
|
||||
self.helpers.get_gpu_id_from_device_handle(src_gpu),
|
||||
self.helpers.get_gpu_id_from_device_handle(dest_gpu),
|
||||
@@ -5172,6 +5176,207 @@ class AMDSMICommands():
|
||||
self.logger.print_output(multiple_device_enabled=True)
|
||||
|
||||
|
||||
def partition(self, args, multiple_devices=False, gpu=None, current=None, memory=None, accelerator=None):
|
||||
""" Display parition information for the target GPU
|
||||
param:
|
||||
args - argparser args to pass to subcommand
|
||||
multiple_devices (bool) - True if checking for multiple devices
|
||||
gpu (device_handle) - device_handle for target device
|
||||
current - boolean which dictates whether the current partition information is shown
|
||||
memory - boolean which dictates whether the memory partition information is shown
|
||||
accelerator - boolean which dictates whether the accelerator partition information is shown
|
||||
returns:
|
||||
nothing
|
||||
"""
|
||||
|
||||
if gpu:
|
||||
args.gpu = gpu
|
||||
if args.gpu == None:
|
||||
args.gpu = self.device_handles
|
||||
if not isinstance(args.gpu, list):
|
||||
args.gpu = [args.gpu]
|
||||
if current:
|
||||
args.current = current
|
||||
if memory:
|
||||
args.memory = memory
|
||||
if accelerator:
|
||||
args.accelerator = accelerator
|
||||
|
||||
# if no args are present, then everything should be displayed
|
||||
if not args.current and not args.memory and not args.accelerator:
|
||||
args.current = True
|
||||
args.memory = True
|
||||
args.accelerator = True
|
||||
|
||||
if args.current:
|
||||
self.logger.table_header = ''.rjust(7)
|
||||
current_header = "GPU_ID".ljust(13) + \
|
||||
"MEMORY".ljust(8) + \
|
||||
"ACCELERATOR_TYPE".ljust(18) + \
|
||||
"ACCELERATOR_PROFILE_INDEX".ljust(27) + \
|
||||
"PARTITION_ID".ljust(14)
|
||||
self.logger.table_header = current_header + self.logger.table_header.strip()
|
||||
|
||||
tabular_output = []
|
||||
for gpu in args.gpu:
|
||||
gpu_id = self.helpers.get_gpu_id_from_device_handle(gpu)
|
||||
try:
|
||||
partition_dict = amdsmi_interface.amdsmi_get_gpu_accelerator_partition_profile(gpu)
|
||||
profile_type = partition_dict['partition_profile']['profile_type']
|
||||
profile_index = partition_dict['partition_profile']['profile_index']
|
||||
partition_id = partition_dict['partition_id']
|
||||
except amdsmi_exception.AmdSmiLibraryException as e:
|
||||
profile_type = "N/A"
|
||||
profile_index = "N/A"
|
||||
partition_id = "N/A"
|
||||
logging.debug("Failed to get accelerator partition profile for GPU %s | %s", gpu_id, e.get_error_info())
|
||||
try:
|
||||
current_mem_cap = amdsmi_interface.amdsmi_get_gpu_memory_partition(gpu)
|
||||
except amdsmi_exception.AmdSmiLibraryException as e:
|
||||
current_mem_cap = "N/A"
|
||||
logging.debug("Failed to get current memory partition capabilties for GPU %s | %s", gpu_id, e.get_error_info())
|
||||
|
||||
tabular_output_dict = {"gpu_id": gpu_id,
|
||||
"memory": current_mem_cap,
|
||||
"accelerator_type": profile_type,
|
||||
"accelerator_profile_index": profile_index,
|
||||
"partition_id": partition_id}
|
||||
tabular_output.append(tabular_output_dict)
|
||||
|
||||
self.logger.multiple_device_output = tabular_output
|
||||
self.logger.table_title = "CURRENT_PARTITION"
|
||||
self.logger.print_output(multiple_device_enabled=True, tabular=True)
|
||||
self.logger.clear_multiple_devices_ouput()
|
||||
|
||||
if args.memory:
|
||||
for gpu in args.gpu:
|
||||
gpu_id = self.helpers.get_gpu_id_from_device_handle(gpu)
|
||||
try:
|
||||
memory_partition = amdsmi_interface.amdsmi_get_gpu_memory_partition(gpu) # this info likely actually comes from different apis than used here
|
||||
except amdsmi_exception.AmdSmiLibraryException as e:
|
||||
memory_partition = "N/A"
|
||||
logging.debug("Failed to get current memory partition for GPU %s | %s", gpu_id, e.get_error_info())
|
||||
try:
|
||||
partition_dict = amdsmi_interface.amdsmi_get_gpu_accelerator_partition_profile(gpu)
|
||||
temp_mem_caps = partition_dict['partition_profile']['memory_caps']
|
||||
|
||||
if temp_mem_caps.amdsmi_nps_flags_t == None:
|
||||
mem_caps = temp_mem_caps.nps_cap_mask
|
||||
mem_caps_list = []
|
||||
if mem_caps & 1 == 1:
|
||||
mem_caps_list.append("NPS1")
|
||||
if mem_caps & 2 == 2:
|
||||
mem_caps_list.append("NPS2")
|
||||
if mem_caps & 4 == 4:
|
||||
mem_caps_list.append("NPS4")
|
||||
if mem_caps & 8 == 8:
|
||||
mem_caps_list.append("NPS8")
|
||||
mem_caps_str = str(mem_caps_list).replace("]", "").replace("[", "")
|
||||
else:
|
||||
mem_caps = temp_mem_caps.amdsmi_nps_flags_t
|
||||
mem_caps_list = []
|
||||
if mem_caps.nps1_cap == 1:
|
||||
mem_caps_list.append("NPS1")
|
||||
if mem_caps.nps2_cap == 1:
|
||||
mem_caps_list.append("NPS2")
|
||||
if mem_caps.nps4_cap == 1:
|
||||
mem_caps_list.append("NPS4")
|
||||
if mem_caps.nps8_cap == 1:
|
||||
mem_caps_list.append("NPS8")
|
||||
mem_caps_str = str(mem_caps_list).replace("]", "").replace("[", "")
|
||||
if mem_caps_str == "":
|
||||
mem_caps_str = "N/A"
|
||||
except amdsmi_exception.AmdSmiLibraryException as e:
|
||||
mem_caps_str = "N/A"
|
||||
logging.debug("Failed to get accelerator partition profile for GPU %s | %s", gpu_id, e.get_error_info())
|
||||
|
||||
memory_dict = {'caps': mem_caps_str, 'current': memory_partition}
|
||||
self.logger.store_output(gpu, 'memory_partition', memory_dict)
|
||||
self.logger.store_multiple_device_output()
|
||||
self.logger.print_output(multiple_device_enabled=True)
|
||||
self.logger.clear_multiple_devices_ouput()
|
||||
if args.accelerator:
|
||||
self.logger.table_header = ''.rjust(7)
|
||||
current_header = "GPU_ID".ljust(13) + \
|
||||
"PROFILE_INDEX".ljust(15) + \
|
||||
"MEMORY_PARTITION_CAPS".ljust(23) + \
|
||||
"ACCELERATOR_TYPE".ljust(18) + \
|
||||
"PARTITION_ID".ljust(14) + \
|
||||
"NUM_PARTITIONS".ljust(16) + \
|
||||
"NUM_RESOURCES".ljust(15) + \
|
||||
"RESOURCE_INDEX".ljust(16) + \
|
||||
"RESOURCE_TYPE".ljust(15) + \
|
||||
"RESOURCE_INSTANCES".ljust(20) + \
|
||||
"RESOURCES_SHARED".ljust(18)
|
||||
self.logger.table_header = current_header + self.logger.table_header.strip()
|
||||
|
||||
tabular_output = []
|
||||
for gpu in args.gpu:
|
||||
gpu_id = self.helpers.get_gpu_id_from_device_handle(gpu)
|
||||
try:
|
||||
partition_dict = amdsmi_interface.amdsmi_get_gpu_accelerator_partition_profile(gpu)
|
||||
profile_type = partition_dict['partition_profile']['profile_type']
|
||||
profile_index = partition_dict['partition_profile']['profile_index']
|
||||
temp_mem_caps = partition_dict['partition_profile']['memory_caps']
|
||||
parition_id = partition_dict['partition_id']
|
||||
num_resources = partition_dict['partition_profile']['num_resources']
|
||||
resources = partition_dict['partition_profile']['resources']
|
||||
|
||||
if temp_mem_caps.amdsmi_nps_flags_t == None:
|
||||
mem_caps = temp_mem_caps.nps_cap_mask
|
||||
mem_caps_list = []
|
||||
if mem_caps & 1 == 1:
|
||||
mem_caps_list.append("NPS1")
|
||||
if mem_caps & 2 == 2:
|
||||
mem_caps_list.append("NPS2")
|
||||
if mem_caps & 4 == 4:
|
||||
mem_caps_list.append("NPS4")
|
||||
if mem_caps & 8 == 8:
|
||||
mem_caps_list.append("NPS8")
|
||||
mem_caps_str = str(mem_caps_list).replace("]", "").replace("[", "")
|
||||
else:
|
||||
mem_caps = temp_mem_caps.amdsmi_nps_flags_t
|
||||
mem_caps_list = []
|
||||
if mem_caps.nps1_cap == 1:
|
||||
mem_caps_list.append("NPS1")
|
||||
if mem_caps.nps2_cap == 1:
|
||||
mem_caps_list.append("NPS2")
|
||||
if mem_caps.nps4_cap == 1:
|
||||
mem_caps_list.append("NPS4")
|
||||
if mem_caps.nps8_cap == 1:
|
||||
mem_caps_list.append("NPS8")
|
||||
mem_caps_str = str(mem_caps_list).replace("]", "").replace("[", "")
|
||||
if mem_caps_str == "":
|
||||
mem_caps_str = "N/A"
|
||||
except amdsmi_exception.AmdSmiLibraryException as e:
|
||||
profile_type = "N/A"
|
||||
profile_index = "N/A"
|
||||
temp_mem_caps = "N/A"
|
||||
parition_id = "N/A"
|
||||
num_resources = "N/A"
|
||||
resources = "N/A"
|
||||
mem_caps_str = "N/A"
|
||||
logging.debug("Failed to get accelerator partition profile for GPU %s | %s", gpu_id, e.get_error_info())
|
||||
|
||||
tabular_output_dict = {"gpu_id": gpu_id,
|
||||
"profile_index": profile_index,
|
||||
"memory_partition_caps": mem_caps_str,
|
||||
"accelerator_type": profile_type,
|
||||
"partition_id": parition_id,
|
||||
"num_partitions": 0,
|
||||
"num_resources": num_resources,
|
||||
"resource_index": resources,
|
||||
"resource_type": resources,
|
||||
"resource_instances": resources,
|
||||
"resources_shared": resources}
|
||||
tabular_output.append(tabular_output_dict)
|
||||
|
||||
self.logger.multiple_device_output = tabular_output
|
||||
self.logger.table_title = "ACCELERATOR_PARTITION_PROFILES"
|
||||
self.logger.print_output(multiple_device_enabled=True, tabular=True)
|
||||
self.logger.clear_multiple_devices_ouput()
|
||||
|
||||
|
||||
def _event_thread(self, commands, i):
|
||||
devices = commands.device_handles
|
||||
if len(devices) == 0:
|
||||
|
||||
@@ -150,8 +150,32 @@ class AMDSMILogger():
|
||||
table_values += string_value.ljust(14)
|
||||
elif key == "link_type":
|
||||
table_values += string_value.ljust(10)
|
||||
elif key == "memory":
|
||||
table_values += string_value.ljust(8)
|
||||
elif key == "accelerator_type":
|
||||
table_values += string_value.ljust(18)
|
||||
elif key == "partition_id":
|
||||
table_values += string_value.ljust(14)
|
||||
elif key == "accelerator_profile_index":
|
||||
table_values += string_value.ljust(27)
|
||||
elif key == "profile_index":
|
||||
table_values += string_value.ljust(15)
|
||||
elif key == "memory_partition_caps":
|
||||
table_values += string_value.ljust(23)
|
||||
elif key == "num_partitions":
|
||||
table_values += string_value.ljust(16)
|
||||
elif key == "num_resources":
|
||||
table_values += string_value.ljust(15)
|
||||
elif key == "resource_index":
|
||||
table_values += string_value.ljust(16)
|
||||
elif key == "resource_type":
|
||||
table_values += string_value.ljust(15)
|
||||
elif key == "resource_instances":
|
||||
table_values += string_value.ljust(20)
|
||||
elif key == "resources_shared":
|
||||
table_values += string_value.ljust(18)
|
||||
elif key == "RW":
|
||||
table_values += " " + string_value.ljust(52)
|
||||
table_values += string_value.ljust(52)
|
||||
elif key == "process_list":
|
||||
#Add an additional padding between the first instance of GPU and NAME
|
||||
table_values += ' '
|
||||
|
||||
@@ -71,7 +71,7 @@ class AMDSMIParser(argparse.ArgumentParser):
|
||||
"""
|
||||
def __init__(self, version, list, static, firmware, bad_pages, metric,
|
||||
process, profile, event, topology, set_value, reset, monitor,
|
||||
rocmsmi, xgmi):
|
||||
rocmsmi, xgmi, partition):
|
||||
|
||||
# Helper variables
|
||||
self.helpers = AMDSMIHelpers()
|
||||
@@ -117,7 +117,7 @@ class AMDSMIParser(argparse.ArgumentParser):
|
||||
# Store possible subcommands & aliases for later errors
|
||||
self.possible_commands = ['version', 'list', 'static', 'firmware', 'ucode', 'bad-pages',
|
||||
'metric', 'process', 'profile', 'event', 'topology', 'set',
|
||||
'reset', 'monitor', 'dmon', 'xgmi']
|
||||
'reset', 'monitor', 'dmon', 'xgmi', 'partition']
|
||||
|
||||
# Add all subparsers
|
||||
self._add_version_parser(self.subparsers, version)
|
||||
@@ -135,6 +135,7 @@ class AMDSMIParser(argparse.ArgumentParser):
|
||||
self._add_monitor_parser(self.subparsers, monitor)
|
||||
self._add_rocm_smi_parser(self.subparsers, rocmsmi)
|
||||
self._add_xgmi_parser(self.subparsers, xgmi)
|
||||
self._add_partition_parser(self.subparsers, partition)
|
||||
|
||||
|
||||
def _not_negative_int(self, int_value):
|
||||
@@ -1286,6 +1287,40 @@ class AMDSMIParser(argparse.ArgumentParser):
|
||||
xgmi_parser.add_argument('-m', '--metric', action='store_true', required=False, help=metrics_help)
|
||||
|
||||
|
||||
def _add_partition_parser(self, subparsers, func):
|
||||
if not self.helpers.is_amdgpu_initialized():
|
||||
# The partition subcommand is only applicable to systems with amdgpu initialized
|
||||
return
|
||||
|
||||
# Subparser help text
|
||||
partition_help = "Displays partition information of the devices"
|
||||
partition_subcommand_help = "If no GPU is specified, returns information for all GPUs on the system.\
|
||||
\nIf no partition argument is provided all partition information will be displayed."
|
||||
partition_optionals_title = "partition arguments"
|
||||
|
||||
# Options help text
|
||||
current_help = "display the current partition information"
|
||||
memory_help = "display the current memory partition mode and capabilities"
|
||||
accelerator_help = "display accelerator partition information"
|
||||
|
||||
# Create partition subparser
|
||||
partition_parser = subparsers.add_parser('partition', help=partition_help, description=partition_subcommand_help)
|
||||
partition_parser._optionals.title = partition_optionals_title
|
||||
partition_parser.formatter_class=lambda prog: AMDSMISubparserHelpFormatter(prog)
|
||||
partition_parser.set_defaults(func=func)
|
||||
|
||||
# Add Universal Arguments
|
||||
self._add_device_arguments(partition_parser, required=False)
|
||||
|
||||
# Handle GPU Options
|
||||
partition_parser.add_argument('-c', '--current', action='store_true', required=False, help=current_help)
|
||||
partition_parser.add_argument('-m', '--memory', action='store_true', required=False, help=memory_help)
|
||||
partition_parser.add_argument('-a', '--accelerator', action='store_true', required=False, help=accelerator_help)
|
||||
|
||||
# Add command modifiers to the bottom
|
||||
self._add_command_modifiers(partition_parser)
|
||||
|
||||
|
||||
def error(self, message):
|
||||
outputformat = self.helpers.get_output_format()
|
||||
|
||||
|
||||
@@ -2788,7 +2788,7 @@ def amdsmi_get_gpu_accelerator_partition_profile(
|
||||
"profile_type" : profile.profile_type,
|
||||
"num_partitions" : profile.num_partitions,
|
||||
"profile_index" : profile.profile_index,
|
||||
"memory_caps" : "N/A",
|
||||
"memory_caps" : profile.memory_caps,
|
||||
"num_resources" : profile.num_resources,
|
||||
"resources" : "N/A"
|
||||
}
|
||||
|
||||
Ссылка в новой задаче
Block a user