This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH] Disable -fno-reorder-blocks-and-partition if no -fprofile-use to avoid unnecessary overhead
- From: Bernhard Reutner-Fischer <rep dot dot dot nop at gmail dot com>
- To: Andi Kleen <andi at firstfloor dot org>,Jeff Law <law at redhat dot com>
- Cc: Teresa Johnson <tejohnson at google dot com>,"gcc-patches at gcc dot gnu dot org" <gcc-patches at gcc dot gnu dot org>,Jan Hubicka <hubicka at ucw dot cz>,Xinliang David Li <davidxl at google dot com>
- Date: Sat, 26 Sep 2015 22:52:47 +0200
- Subject: Re: [PATCH] Disable -fno-reorder-blocks-and-partition if no -fprofile-use to avoid unnecessary overhead
- Authentication-results: sourceware.org; auth=none
- References: <CAAe5K+WwdWZyp96ODECAxPsu-nf97akiKzremnhPbiYGS2E_bw at mail dot gmail dot com> <87twqisaht dot fsf at tassilo dot jf dot intel dot com> <CAAe5K+UP9w=P6q5EKRTdJ2zgXmhTZgnxYgJbQfXse7nLDNWuag at mail dot gmail dot com> <56057DFB dot 2060703 at redhat dot com> <20150925175907 dot GM1747 at two dot firstfloor dot org>
On September 25, 2015 7:59:07 PM GMT+02:00, Andi Kleen <andi@firstfloor.org> wrote:
>On Fri, Sep 25, 2015 at 11:01:47AM -0600, Jeff Law wrote:
>> On 09/25/2015 10:58 AM, Teresa Johnson wrote:
>> >Woops, we crossed wires. I just committed this patch. Would you like
>> >me to revert it?
>> Leave it. If Andi can include a reversion if he can pound his work
>> around 66890 into submission.
>
>The patch is stable, was just gathering more data.
+@item partition-cold-min-freq
+When using doing function partitioning and there is no profile information
+consider edges below this frequency cold. Setting to zero disables
+any function splitting without profile information.
Trouble parsing "using doing", nice hack otherwise.
It's obviously a pity Google does its own thing and doesn't really push a balance between auto and such simple improvements. /me likes yours fwiw..
Cheers,
>
>> But I think it'd need some of the
>> varasm.c work Jan hinted at.
>
>The varasm work should be only needed if no function splitting is done,
>right? With my patch most functions do function splitting.
>
>-Andi