[PATCH] PowerPC: CTR shouldn't fire if a TLS call is in the loop

[David Majnemer]

On Sat, Dec 27, 2014 at 4:34 AM, Joerg Sonnenberger <joerg at britannica.bec.de
> wrote:

> On Sat, Dec 27, 2014 at 09:12:04AM +0000, David Majnemer wrote:
> > Determining the address of a TLS variable results in a function call in
> > certain TLS models.  This means that a simple ICmpInst might actually
> > result in invalidating the CTR register.
>
> Why is %load not moved out of the loop?
>

It could be if a middle-end pass decided to LICM it, there is no reason why
it must be hoisted; we have the same issue with instructions which cannot
be hoisted like volatile loads/stores/etc.


>
> Joerg
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20141227/01a6bfd8/attachment.html>