Utilize your CPU power – Cache optimizations and SIMD instructions

Comments

Comments are closed.

Thanks a lot Mario!

I think that your talk is very important for C++ community because explain very well the power of C++ in numerical computation.

Very interesting talk, the concepts were explained excellently thanks to the use of diagrams and graphs.

Optimization is something that I find really interesting and fun to think about - being able to see how much cache optimizations and special SIMD instructions affect the run-time performance of a real-world physics problem solver was an eye-opener.

Boost.SIMD looks extremely easy to use (yet powerful), as well.

Thanks!

I think I should convince some people at work to try Boost.SIMD :) Great talk, really clear and inspiring.

A fundamental talk about cache-oriented optimization techniques and SIMD operations in C++. These arguments are particularly important in a community that really cares about performance!

Anonymous at 10:33 on 27 Jun 2015

Caches, caches, caches. As C++ programmers, performance of our code is always in our hearts. This talk shows nicely that by taking into account the memory hierarchy you can "easily" reach a 3x speedup.

The SIMD part using Boost.SIMD was also very nice and full of interesting ideas.