Micro perf

I used to do performance work professionally.  Some of it was really satisfying. 
The interconnecting pieces of a well optimized piece of code fit together like a sliding block puzzle.

Chips got faster and lots of the old styles of optimization became less important.
But then the intel SSE instruction set and its follow-ups were introduced.  The instruction set has a limited set of instructions that operate on very large (128bit) registers, fairly quickly.

But everything old is new again.  And a lot of the stuff that was on the bleeding performance edge two decades ago is now relevant and exciting again.

In particular I'm a big fan of the book "Hacker's Delight".  This is one of the only books mentioned in the text of Knuth volume 4.

But this stuff is only really important for small scale perf.  This is not algorithmic optimization.  Sometimes this is called pinhole optimization.  I call this micro perf.

I will be referencing a lot of instructions with deeper explanations in https://software.intel.com/sites/landingpage/IntrinsicsGuide/