Re: [PATCH] Microoptimization of lib/poll.c

[Paolo Bonzini]

> > 3) access to thread-local errno is slow, it's also better to cache it.
>
> Pheww! That is micro-optimization indeed.

Well, thread-local errno might involve a function call on some systems. 
 If the function call is not CSE'd by the compiler, it means half a 
dozen function calls.

> Do you have an objection against this further patch? Should save 2 instructions
> in the case that you were not profiling.

*That* is micro-optimization! :-)  It actually makes the code clearer, 
so go on.  But I guess a division is more expensive than those 2 
instructions...

(I used Shark to get the profile.  rpl_poll was responsible for 6% 
before my optimizations and 3% afterwards.  So not a big change from 
Amdahl's point of view :-) but a big change for this specific function).

Paolo