将配置文件数据导入通用分析和可视化工具

发布于 2024-09-09 01:48:52 字数 700 浏览 9 评论 0原文

我有一种生成几种不同类型的配置文件数据的机制。是否有任何工具可以从外部源导入配置文件数据并生成有用的交互式可视化效果?似乎有许多分析工具都实现了自己的渲染引擎,但驱动它们的原始数据之间确实没有太大区别。

到目前为止,我只是模仿了 gprof 的文本输出,这比人们想象的更有用。但这并不能说明什么。

示例:用于 Windows 的 xperf、用于 Mac 的 Shark(尽管我没有使用过)、用于 Google Chrome 的 SpeedTracer、oprofileui 等等。

gprof2dot 是我正在寻找的一个示例,因为它接受来自多个配置文件数据生成器的输出,但是(1)我发现它的输出不是非常有用,特别是在大型配置文件上; (2) 最理想的是,我想要一个基于 Web 的交互式用户界面。

更好的是,是否有一个标准的配置文件数据格式可以用来提供多个工具?

首先,我有基本的动态调用图数据,包括挂钟计时和调用计数,但我还想用 CPU 计数器、缓存未命中计数、页面错误计数等代替计时信息。我还想将多个来源的配置文件数据混合在一起(例如,来自 C++ 和脚本分析工具的单独配置文件数据,通过时间戳或显式标记进行关联。)

更新:我进一步研究了 SpeedTracer。它定义了数据的导入/导出格式。因此,如果不出意外,我可以序列化为该格式并将数据加载到 SpeedTracer 中。我并不热衷于与 Chrome 和 GWT 结合,但这是一个好的开始。

I have a mechanism for generating profile data of a couple of different types. Are there any tools that will import profile data from external sources and produce useful interactive visualizations? It seems like there are many profiling tools out there that all implement their own rendering engines, but there really isn't much difference between the raw data that is driving them.

So far, I just mimicked gprof's text output, which is more useful than people give it credit for. But that isn't saying much.

Examples: xperf for Windows, Shark for Mac (though I haven't used it), SpeedTracer for Google Chrome, oprofileui, and many others.

gprof2dot is an example of what I'm looking for, in that it accepts output from multiple profile data generators, but (1) I don't find its output terribly useful, especially on large profiles; and (2) optimally, I would like an interactive web-based UI.

Better yet, is there a standard format for profile data that I can use to feed multiple tools?

To start with, I have basic dynamic callgraph data with wallclock timings and call counts, but I would also like to substitute CPU counters, cache miss counts, page fault counts, etc. for the timing information. I would also like to mix together profile data from multiple sources (eg separate profile data from C++ and script profiling tools, related via timestamps or explicit markers.)

Update: I looked a bit further into SpeedTracer. It defines an import/export format for its data. So if nothing else, I can serialize into that format and load my data into SpeedTracer. I'm not crazy about being bonded to Chrome and GWT, but it's a good start.

如果你对这篇内容有疑问,欢迎到本站社区发帖提问 参与讨论,获取更多帮助,或者扫码二维码加入 Web 技术交流群。

扫码二维码加入Web技术交流群

发布评论

需要 登录 才能够评论, 你可以免费 注册 一个本站的账号。

评论(3

牵你手 2024-09-16 01:48:52

您是否考虑过仅从工具导出并通过像 SSRS< /a>?这甚至可以自动化,没有太多麻烦......

Have you considered just exporting from tools and importing to a reporting system with drilldown like SSRS? This can even be automated with not much fuss...

安穩 2024-09-16 01:48:52

如果你愿意花一些钱TableauSpotfire 是两个选项。对于开源的东西,您也许可以破解 Geovista。这些都是通用的可视化平台,而不是专门针对分析的。当然,你也可以使用Excel。

If you are willing to spend some money Tableau and Spotfire are two options. For something open source, you might be able to hack Geovista. These are all general visualization platforms, not specific to profiling. Of course, you could also use Excel.

落在眉间の轻吻 2024-09-16 01:48:52

Spotfire 刚刚宣布对其主要可视化工具进行一年免费试用,并免费发布到网络 - 在 Silver Spotfire 上搜索

Spotfire just announced a free one-year trial of their main visualization tool with free publishing to the Web - search on Silver Spotfire

~没有更多了~
我们使用 Cookies 和其他技术来定制您的体验包括您的登录状态等。通过阅读我们的 隐私政策 了解更多相关信息。 单击 接受 或继续使用网站,即表示您同意使用 Cookies 和您的相关数据。
原文