[PATCH] D19995: Optimize access to global variable references in PIE mode when linker supports copy relocations for PIE

[Sriraman Tallam via llvm-commits]

On Mon, May 9, 2016 at 4:04 PM, Rafael Espíndola
<rafael.espindola at gmail.com> wrote:
> On 8 May 2016 at 20:26, John McCall <rjmccall at gmail.com> wrote:
>> rjmccall added a comment.
>>
>> It seems reasonable to me to have an option enabling the use of copy relocations for globals, and I agree with David that that option should logically be applicable to PIE.  I would also tend to agree with Joerg that that option should default to off, even in non-PIC modes, but it's not my call.
>>
>> RE: the performance impact of relying on the linker to optimize accesses: it certainly can increase register pressure, but it shouldn't in the most common cases where the access sequence ends in putting a new value in a GPR anyway (usually either the address of the variable or something loaded from it).  More importantly, I don't really understand how you can avoid this without compiling code assuming that all globals can be copy-relocated, since the *compiler* doesn't know whether an external symbol is defined as protected or default.  (Unless that's exactly what we're doing — are people really this fast-and-loose in ELF land?)
>>
>> RE: the semantic problems with copy relocations exposing the layout of globals: yes, binary compatibility problems usually don't exist when you recompile all the dependent code.  I can understand why system maintainers wouldn't be thrilled with the idea that certain kinds of updates to system libraries can't be made without recompiling the entire user space, though.
>
> I agree with all remarks.
>
> Using this email to note that I think there is a point that was missed
> in some messages in this thread:
>
> Using copy relocations is *not* what is making the program faster.
>
> What is making the program faster is that it is not accessing its own
> variables with a got. The way copy relocations come into the picture
> is that they allow the compiler guess to be wrong in the case of the
> main program. If it is, the dynamic linker produces a copy.

Yes, I completely agree, no confusion here :)

>
> So you could get the same (even a bit better maybe) performance by either
>
> * Marking every decl in the main program protected (or hidden if
> doesn't take plugins)

This seems impractical solution for us, changing the source to mark
every declaration  as a protected symbol.


> * Marking every library decl with dllimport and keeping the default of
> assuming externals are local to this DSO.
>
> Copy relocations are just an easy enabler. Another easy one that might
> produce close results is building the main program with
> -fvisibility=protected. That will mark every definition protected and
> avoid some got uses. The linker will hopefully optimize the others.
>
> Cheers,
> Rafael