---
title: 使用Roofline分析
description: "Roofline分析可帮助用户在给定硬件平台上，分析出应用程序的瓶颈点位置，从而有针对性的进行优化。"
url: https://www.hikunpeng.com/document/detail/zh/kunpengdevkithistory/devkit_24_rc2/cliuserguide/KunpengDevKitCli_0037.html
sourcePath: /source/zh/kunpengdevkithistory/devkit_24_rc2/cliuserguide/KunpengDevKitCli_0037.html
indexId: c0861585ee4b930c07b77dff789b5ffd48c82e1efea9b316fe6bd7a32db81f7285
---
# 使用Roofline分析

#### 命令功能

Roofline分析可帮助用户在给定硬件平台上，分析出应用程序的瓶颈点位置，从而有针对性的进行优化。

- 仅支持基于鲲鹏920的服务器，不支持容器环境。
- Roofline使用DBI技术，待分析的应用必须为ELF格式的二进制文件。
- Roofline采集时需确保应用已完成运行；Roofline采集时长大概为应用运行时长三倍。


#### 命令格式

```
devkit tuner roofline [-h] [-l {0,1,2,3}] [-m {total,region}] [-o <file>] workload ...
```


#### 参数说明


**表1 参数说明**

| 参数 | 参数选项 | 说明 |
| --- | --- | --- |
| \-h/\-\-help | \- | 获取帮助信息。 |
| \-l/\-\-log\-level | 0/1/2/3 | 配置日志等级。默认值为2(warning)。 0(debug) 1(info) 2(warning) 3(error) |
| \-m/\-\-mode | total/region | 配置分析范围为整个二进制应用或用户选择的regions；默认为total。 total：采集整个应用的roofline数据。 region：采集应用中各个region的roofline数据，程序的region划分需要用户按需进行插桩。 |
| \-o/\-\-outpath | \- | 配置生成数据包名称。默认为当前所在目录生成，默认文件名为roofline\-YYYYMMDD\-HMS。 |


#### 使用示例

采集已划分好region的应用：

```
devkit tuner roofline -m region /mysharedir/devkit/SystemProfilerBackend/tuner_cli/docs/matrix_multiply_c
```

返回信息如下：

```
Note:
1. Roofline task is currently only supported on the 920 platform.
2. The application must be a binary file in ELF format.
3. Roofline task collection needs to ensure the application has finished running.
4. The estimated time of roofline collection is about 3 * application estimated time.
RFCOLLECT: Start collection for /mysharedir/devkit/SystemProfilerBackend/tuner_cli/docs/matrix_multiply_c
RFCOLLECT: Launch application to collect performance metrics of /mysharedir/devkit/SystemProfilerBackend/tuner_cli/docs/matrix_multiply_c
ROOFLINE_EVENTS are initialized.
Initialization time: 0.070167 seconds
Calculation time: 0.206211 seconds
The dimension of the matrices is too large to print.
RFCOLLECT: Launch application to do binary instrumentation of /mysharedir/devkit/SystemProfilerBackend/tuner_cli/docs/matrix_multiply_c
Initialization time: 0.168616 seconds
Calculation time: 2.243492 seconds
The dimension of the matrices is too large to print.
RFCOLLECT: Launch benchmarks for measuring roofs
RFCOLLECT: Processing all collected data
RFCOLLECT: Result is captured at /mysharedir/devkit/SystemProfilerBackend/sys_perf/components/sys_tools/rfcollect-20240424-143840.json
RFCOLLECT: Run "rfreport /mysharedir/devkit/SystemProfilerBackend/sys_perf/components/sys_tools/rfcollect-20240424-143840.json" to get report.
Get roofline report ...
The roofline json report: /mysharedir/devkit/SystemProfilerBackend/sys_perf/components/sys_tools/roofline-20240424-143840.json
The roofline html report: /mysharedir/devkit/SystemProfilerBackend/sys_perf/components/sys_tools/roofline-20240424-143840.html
```

任务将生成JSON文件和HTML文件，如需进行数据分析可直接使用JSON文件，查看性能数据可在浏览器中打开HTML文件。

图1 Roofline HTML文件