On Tuesday, 3 January 2012 at 18:49:35 UTC, Martin Nowak wrote:
For real performance bottlenecks one should always examine the
assembly. For most code inlining hardly ever matters for the
runtime of your program and focusing on efficient algorithms is
most important.
What really baffles me is that people want control over inlining
but nobody seems to ever have noticed that x64 switch doesn't
switch and x64 vector ops aren't vectorized. Both of which are
really important in performance sensitive code.
Quality of implementations' optimizations and a common syntax for
code compilation guarantees are orthogonal issues.
