Hi, thanks for reading this post first.

I'm new to use this awesome library, and I have a question regarding performance.

I've tested with several methods(accelerate framework, math-neon, and OpenCV),
compute some dot product, and result shows Eigen is faster than OpenCV but is slower than others.

Eigen is very simple to implement. If anyone has experience with these or got any answers, please enlighten me:) Is Eigen faster than those two methods?

To make Eigen fast on ARM-NEON you have to make sure you use a very recent version of GCC (4.5) and of course that you enabled support for neon instructions (-mfpu=neon). Then I don't know about Accelerate nor math-neon.

ggael wrote:To make Eigen fast on ARM-NEON you have to make sure you use a very recent version of GCC (4.5) and of course that you enabled support for neon instructions (-mfpu=neon). Then I don't know about Accelerate nor math-neon.

Thanks for the comment! I was looking for fast method to calculate some matrices..
Actually, I already enabled neon with cflags. For GCC, the version is 4.2.1.. may be I need to update it and try again.