[PATCH] Add a shrink-wrapping pass to improve the placement of prologue and epilogue.

[Quentin Colombet]

Hi Kristof,

> On Apr 23, 2015, at 10:47 AM, Kristof Beyls <kristof.beyls at arm.com> wrote:
> 
> 
> 
>> -----Original Message-----
>> From: Quentin Colombet [mailto:qcolombet at apple.com]
>> Sent: 23 April 2015 17:52
>> To: qcolombet at apple.com; grosbach at apple.com
>> Cc: kparzysz at codeaurora.org; Kristof Beyls; dberlin at dberlin.org; Amara
>> Emerson; justin.holewinski at gmail.com; llvm-commits at cs.uiuc.edu
>> Subject: Re: [PATCH] Add a shrink-wrapping pass to improve the placement
>> of prologue and epilogue.
>> 
>> Hi Kristof,
>> 
>> Thanks for your feedbacks.
>> 
>>> Overall, I guess that this needs support in at least one backend, so
>> regression tests can be added before this is in a state ready to commit?
>> 
>> 
>> Whatever people prefers. I have a patch to add the support for shrink-
>> wrapping for AArch64, I can just merge it with this one with the path
>> disabled by default.
> 
> I guess I'm missing something obvious. I'm assuming that for the functions
> where the shrink-wrapping optimization triggers, it's always, or almost always
> a win? I am assuming that the effect of the patch as shown, when triggering,
> moves a single prologue and a single epilogue sequence from a path that's
> executed once per function call to a path that's executed at most once per
> function call.
> 
> If that assumption were true, there would be no reason to disable this pass
> by default, right? So, how is my assumption incorrect?

No, you are right.

The thing is I haven’t give it as much testing as I want and I would like other people to try it out on their test suite before we make it the default.
The bottom line is I want to ascertain the correctness of the AArch64 changes :).

Cheers,
-Quentin
> 
> Thanks,
> 
> Kristof
> 
> 
> 
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits