Re: [PATCH 00/10] jump label: introduce very_[un]likely + cleanups+ docs

[H. Peter Anvin]

On 02/21/2012 11:25 PM, Ingo Molnar wrote:
> 
> There is a fundamental assymetry, and intentionally so. You 
> *really* have to think what the common case is, and make sure 
> the build defaults to that. It's not the end of the world to 
> have it flipped over, but there's costs and those costs are 
> higher even in the branch path than a regular 
> likely()/unlikely().
> 

No, not really -- it's still an unconditional branch, which means you
will not tax the branch predictor in any way and which can be followed
by the front end without taking a speculation hit.  So although there is
an out-of-line penalty -- which you hit for any conditional, after all
you can only have *one* piece of code which is straight line -- it
should be less than for a normal conditional branch.

> So you are rather wrong about your expectations - I think that 
> is one more piece of evidence that the naming was less than 
> optimal.
> 
>> So the key aspect of this is the staticness of the 
>> conditional, NOT the degree of bias of the branch.  Hence my 
>> past insistence on the "static_branch" name (rather than 
>> "jump_label")... the branch part can be omitted, as an 
>> implementation detail, but the staticness of it is its 
>> absolutely key defining characteristic.
> 
> I don't think you understand this facility as well as you think 
> you do.

Uh, no, I do... see the above, but combine that of course with the sheer
astronomical cost of flipping the conditional.

	-hpa


-- 
H. Peter Anvin, Intel Open Source Technology Center
I work for Intel.  I don't speak on their behalf.

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/