Re: Using __builtin_expect (likely/unlikely macros)

[Paul Eggert]

Óscar Fuentes wrote:
> I'm pretty sure that
> the speedup will vanish as code is modified and compilers and CPUs
> change.

I agree that if you disable GCC optimization or do lots of other 
plausible-but-low-performance things, that patch will not help performance. 
However, I'm skeptical that the patch would make performance worse in ordinary 
production builds on current Emacs platforms. And I wasn't surprised that the 
patch was a significant (albeit small) win on the platform I tested on, because 
I looked at the machine code that it generated and I knew something of my 
platform's architecture.

Of course I could be unpleasantly surprised about how well this performance win 
generalizes to other platforms - and as I wrote, I am often surprised. To become 
surprised, though, we need to see some numbers.

> I'm far from being a naive programmer.

I didn't think you were naive, and apologize if my comments could be interpreted 
that way. My comments were intended only to make it clear that I'm not a naive 
programmer.