NDK Performance Tips in C++

[Andreas]

Hi everyone,

I would like to know if anyone can give me any tips on how to improve
the performance of a critical part of my application (C++) which in
performed in native code. The particular part contains a lot of matrix
transformations.. So any tips like what kind of
arrays(float,double, ...) I should use or any other tips like for
example avoid setters and getters are welcome. I was trying to figure
out if there is a way to profile the native code part with something
like traceview in order to find performance bottlenecks but I had no
luck. TraceView sees ndk code as a black box...If anyone knows any
tips or uses any tool that could help with performance on a native
code part written completely in C++ please let me know.

Thanks in advance,

Andreas

[mic _]

If you're targeting processors with NEON extensions you could perhaps make use of this: http://code.google.com/p/math-neon/source/browse/#svn%2Ftrunk   (see math_mat4.c for example)

You can read more about the concepts here: http://blogs.arm.com/software-enablement/241-coding-for-neon-part-3-matrix-multiplication/


/Michael

--
You received this message because you are subscribed to the Google Groups "android-ndk" group.
To post to this group, send email to andro...@googlegroups.com.
To unsubscribe from this group, send email to android-ndk...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/android-ndk?hl=en.

[Andreas]

Michael thank you for your answer.

I am looking for something similar but not only for processors with
NEON extensions!

On Mar 9, 9:46 am, mic _ <micol...@gmail.com> wrote:
> If you're targeting processors with NEON extensions you could perhaps make
> use of this:http://code.google.com/p/math-neon/source/browse/#svn%2Ftrunk
> (see math_mat4.c for example)
>

> You can read more about the concepts here:http://blogs.arm.com/software-enablement/241-coding-for-neon-part-3-m...
>
> /Michael