Profiling compilation time

Question

I have a C++ code which I am compiling using VC7 and it is taking a lot of time to build it. Is there any way I could profile it and find why it is taking time to build ?

Answer 1

In Visual Studio 2008, there's a setting for turning on build timing. It might be there in VC7 as well...

Tools / Options / Projects and Solutions / VC++ Project Settings / Build Timing: Yes

This applies to C++ projects, which (as of VS2008) don't use MSBuild. For MSBuild-based projects (such as C#), you want to increase the verbosity:

Tools / Options / Projects and Solutions / Build and Run / MSBuild project build output verbosity

By default, it's set to "Minimal".

Answer 2

If the code is template-intensive, then you could try doing the template instantiation profiling. Steven Watanabe came up with theprofiler and if I remember correctly it was supposed to work with VS (don't know the version).

Answer 3

Is the source code on a network? This sometimes slows the compilation a lot.

Answer 4

Recent Microsoft's C++ Build Insights SDK has introduced an option /timetrace to its vcperf which allows you to profile your build and visualise the build times of specific components in form of a flame graph inside of any chromium-based browser.

Assuming you have vcperf downloaded and installed, you need to:

1. Start its session by executing

vcperf /start SessionName

2. Run your build (the events are captured system-wide by vcperf)
3. Stop the session

vcperf /stop SessionName /timetrace output.json

You can now run your chromium-based browser, type in <browser_name>://tracing , (for example chrome://tracing ) and load up the output.json file for visualisation.

Example visualisation taken from here .

Answer 5

If your code makes extensive use of template, you might be interested in Templight , a tool developed by a hungarian research team for debugging and profiling C++ template metaprograms ( paper ). It seems very promising, but I'm not sure the tool is available for download...

Answer 6

My guess is that it would be difficult to get useful results from profiling. You could look at the create times of each .obj file and check if there are any files that are particularly slow, but I doubt this would be the case.

Have you gone through the compiler options such as pre-compiled headers to see what improvements ths provides? Similarly, turning off the optimizer where it is not required can speed the build up significantly. My advice would be to take some time to try out a few 'what if' scenarios.

Answer 7

如果可能，您可以在您的情况下尝试#include 所有 .cpp 文件到单个编译单元中的技巧，只是为了检查是否有很多文件和许多包含的开销。

Answer 8

使用预编译头