Re: [boost] non-branching min()

Subject: Re: [boost] non-branching min()
From: DE (satan66613_at_[hidden])
Date: 2009-09-11 11:11:23

on 11.09.2009 at 3:19
 Cory Nelson wrote :
> I'm curious, have you actually tried to look at the generated
> instructions?
of course

> Because on a 686+ this should compile this into two
> non-branching instructions for min():
> cmp a, b
> cmova a, b
> And the same thing but with cmovb for max().
neither icc11 nor msvc80 generate such instructions
msvc generates a conditional jump and icc generates code "without
branches but with the sbb asm instruction" (c) Joel Falcou

> You solution should generate at least 5 instructions.
yep

i got an interesting result
msvc happily optimizes a lut implementation but generates branch for
traditional one (?:)
however icc eliminates branch for ?: but generates slow code for lut
in the end we have too competitors: msvc_lut and icc_?:
and the record holds msvc which is some percent faster than icc code

and also branches were faster than lut on amd athlon and athlon xp+
so i guess such implementation is not a good idea in general

-- 
Pavel