Statistics View

To view the API Statistic, AscendCL API, OS Runtime API, Runtime API, OPs, and Op Info invocation data, click Statistics in the data pane at the bottom and select a data type from the drop-down list in the upper left corner.

If the current version is used to collect profile data, no AscendCL API, OS Runtime API, and Runtime API data is displayed. Instead, the related data is summarized and displayed in API Statistic.

API Statistic

Figure 1 API Statistic

**Table 1** Field description
Field	Description
Level	Level of an API, including AscendCL, Runtime, Node, Model, and HCCL.
API Name	API name
Time(us)	Time (μs) taken by the API You can rearrange the information in ascending or descending order by clicking the triangle icon next to the field name.
Count	Number of API calls
Avg(us)	Average duration (μs) of a single API call
Min(us)	Minimum duration (μs) of an API call
Max(us)	Maximum duration (μs) of an API call
Variance	Time consumption variance of an API call.

AscendCL APIs and OS Runtime APIs

Figure 2 AscendCL APIs

Figure 3 OS Runtime APIs

**Table 2** Field description
Field	Description
Name	API name
Type (AscendCL API only)	Type of the AscendCL API
Time(%)	Percentage of time taken by the API
Time(us)	Time (μs) taken by the API You can rearrange the information in ascending or descending order by clicking the triangle icon next to the field name.
Count	Number of API calls
Avg(us)	Average duration (μs) of a single API call
Min(us)	Minimum duration (μs) of an API call
Max(us)	Maximum duration (μs) of an API call
Process ID	Process ID of the API
Thread ID	Thread ID of the API

Runtime API

Figure 4 Runtime API

**Table 3** Field description
Field	Description
Name	API name
Stream ID	Stream ID
Time(%)	Percentage of time taken by the API
Time(ns)	Time (ns) taken by the API You can rearrange the information in ascending or descending order by clicking the triangle icon next to the field name.
Calls	Runtime API calls
Avg(ns)	Average duration (ns) of a single API call
Min(ns)	Minimum duration (ns) of an API call
Max(ns)	Maximum duration (ns) of an API call
Process ID	Process ID of the API
Thread ID	Thread ID of the API

OPs

Figure 5 OPs (Ascend 310 AI Processor and Ascend 310B AI Processor)

**Table 4** Field description
Field	Description
Time(%)	Percentage of time taken by the API
Time(us)	Time (μs) taken by the API You can rearrange the information in ascending or descending order by clicking the triangle icon next to the field name.
Count	Number of times that a task is executed
Avg(us)	Average execution time of a task, in μs
Min(us)	Minimum execution time of a task, in μs
Max(us)	Maximum execution time of a task, in μs
Waiting(us)	Total waiting time of a task, in μs
Running(us)	Total running time of a task, in μs. If a task has been running for a long time, the operator implementation may be incorrect.
Pending(us)	Total pending time of a task, in μs
Type	Task type
API	Called API
Task ID	Task ID
Op Name	Operator name
Stream ID	Stream ID

Figure 6 OPs (Ascend 310P AI Processor, Atlas A2 Training Series Product/Atlas 800I A2 Inference Product, and Ascend 910 AI Processor)

**Table 5** Field description
Field	Description
Kernel Name	Kernel name
Kernel Type	Kernel type
Stream Id	Stream ID
Task Id	Task ID
Task Time(us)	Task duration (μs)
Task Start(us	Task start time (μs).
Task Stop(us)	Task end time (μs).

Op Info

Figure 7 OP Info

**Table 6** Field description
Field	Description
Model Name	Model name This parameter is displayed only when offline model calling (ACL function API: aclModelExecute) is involved in an inference application.
Task ID	Task ID
Stream ID	Stream ID
Op Name	Operator name
Op Type	Operator type
Task Start Time	Time when the task starts to run
Task Duration(us)	Task running duration (us)
Task Wait Time(us)	Interval between tasks (μs)
Block Dim	Index of the core that executes the task
Input Shapes	Input shapes
Input Data Types	Input data types
Input Formats	Input formats
Output Shapes	Output shapes
Output Data Types	Output data types
Output Formats	Output formats
Aicore Time(us) (displayed only when the collection mode is Task-Based)	AI Core running duration (μs)
Total Cycles (displayed only when the collection mode is Task-Based)	Number of cycles taken to execute all task instructions

Parent topic: Timeline Views