llvm dev - Mar 2015 - [LLVMdev] [RFC] AArch64: Should we disable GlobalMerge?

> On Feb 27, 2015, at 2:15 PM, Ahmed Bougacha <ahmed.bougacha at
gmail.com> wrote:
> 
> On Fri, Feb 27, 2015 at 2:01 PM, Renato Golin <renato.golin at
linaro.org> wrote:
>> On 27 February 2015 at 21:26, Ahmed Bougacha <ahmed.bougacha at
gmail.com> wrote:
>>> Which brings us to my fallback proposal:  what about disabling the
>>> pass on darwin only?
>> 
>> That's a decision for Jim/Evan. I'm ok if they are.
> 
> Jim, thoughts?
I would prefer Darwin not differ in this regard, but I don’t feel incredibly
strongly about it. Just a general preference to keeping platform dependencies
and differences to a minimum. Whatever y’all decide is fine with me.
> 
>> 
>>> As for other targets, as a first step, making the pass run under
-O3
>>> rather than -O1 is hopefully agreeable to everyone?
>> 
>> Sounds reasonable.
> 
> Great!
> 
>> Even though it conflicts with LTO, that's what O3 means, as you
said,
>> instability. People at O3 might want to fiddle with the passes
>> (on/off) to get the best performance for their own code/workload.
> 
> By the way, I'm not convinced LTO being either -O3 or -O0 is sensible.
> But that's a discussion for another day =)
> 
> -Ahmed
> 
>> cheers,
>> --renato

On Fri, Feb 27, 2015 at 3:06 PM, Jim Grosbach <grosbach at apple.com>
wrote:
> I would prefer Darwin not differ in this regard, but I don’t feel
incredibly strongly about it. Just a general preference to keeping platform
dependencies and differences to a minimum.
Same.
> Whatever y’all decide is fine with me.
We might not need this after all =)   With a module-level flag, we
could only enable it under -O3 even for LTO, which is fine no matter
the platform.

-Ahmed

On Fri, Feb 27, 2015 at 3:20 PM, Ahmed Bougacha
<ahmed.bougacha at gmail.com> wrote:
> We might not need this after all =)   With a module-level flag, we
> could only enable it under -O3 even for LTO, which is fine no matter
> the platform.
Done, in:
clang: http://reviews.llvm.org/D7968
LLVM: http://reviews.llvm.org/D7969

The clang patch also only enables the pass with -O3 on AArch64.  I
kept -O1 for ARM because I need to do some sanity checks, but the goal
is to have it enabled on both AArch64 and ARM only with -O3.

-Ahmed

Alright, I switched both AArch64 and ARM (which looks about the same)
to enable GlobalMerge with -O3 rather than -O1, in r233024.  There are
recent LTO changes that should take care of the opt-level issue, so
that part should be covered as well.

As for actual improvements, I submitted some patches (though not for
the module pass issue yet), we'll see how that goes.

Thanks all!

-Ahmed

Nice! Thanks, Ahmed.

> On Mar 23, 2015, at 2:29 PM, Ahmed Bougacha <ahmed.bougacha at
gmail.com> wrote:
> 
> Alright, I switched both AArch64 and ARM (which looks about the same)
> to enable GlobalMerge with -O3 rather than -O1, in r233024.  There are
> recent LTO changes that should take care of the opt-level issue, so
> that part should be covered as well.
> 
> As for actual improvements, I submitted some patches (though not for
> the module pass issue yet), we'll see how that goes.
> 
> Thanks all!
> 
> -Ahmed