Intel AVX-512 compiler vectorization article

[tServo]

It's here:
http://https://www.codeproject.com/A...proved-Perform

Remember, the Xeon Phi system that AirSquirrels hosts for this forum ( thanks, David )
has this compiler and the Phi has AVX-512 registers.

[sdbardwick]

Corrected link: https://www.codeproject.com/Articles...proved-Perform

[ewmayer]

Originally Posted by tServo

It's here:
http://https://www.codeproject.com/A...proved-Perform

Remember, the Xeon Phi system that AirSquirrels hosts for this forum ( thanks, David )
has this compiler and the Phi has AVX-512 registers.

I find no advantage from using ICC over GCC on that system - in fact GCC gives me more reliable performance. I've also seen GCC doing lots of auto-vectorization, at least of simple snips of code which are obvious candidates for this kind of optimization.

Have not yet had an opportunity to play with the ICC advanced-performance-analysis tools - I still hold out hope those will prove useful once I complete the basic code-port/hand-optimization of my AVX2 stuff to AVX-512.

[Mysticial]

I'm gonna wait until Skylake Purley before jumping on the AVX512 train. Though based on what I know so far, I have a strong feeling that getting the right chip will require mortgaging my house.