This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [GOOGLE] Refactor the profile propagation for AutoFDO
- From: Dehao Chen <dehao at google dot com>
- To: Diego Novillo <dnovillo at google dot com>
- Cc: Xinliang David Li <davidxl at google dot com>, GCC Patches <gcc-patches at gcc dot gnu dot org>
- Date: Mon, 25 Nov 2013 10:30:55 -0800
- Subject: Re: [GOOGLE] Refactor the profile propagation for AutoFDO
- Authentication-results: sourceware.org; auth=none
- References: <CAO2gOZU+3A0g1O4isju=7N6_NFW+iaCZJfY4-eBf6Vm7bwWY8w at mail dot gmail dot com> <CAD_=9DSFjktrsR0nBfQHp0t2a6bMZcf5GBXt1KSvYZuFDcyAhQ at mail dot gmail dot com> <CAAkRFZLuWfLu3rrTjY0xVOr-JiNPn_1QJy_Hp_ZD7wfNbDS-Qg at mail dot gmail dot com> <CAD_=9DQ0XrRjbATPsQ9kBub-Q0c7OTJcKapUePgwWRDbCO=CKQ at mail dot gmail dot com>
On Mon, Nov 25, 2013 at 10:26 AM, Diego Novillo <dnovillo@google.com> wrote:
> On Mon, Nov 25, 2013 at 1:22 PM, Xinliang David Li <davidxl@google.com> wrote:
>> In this case the backedge will be a critical edge, which will be split by GCC.
>
> Right. So, if I split it, I will reach essentially the same
> conclusion, I think. The new block will get the original block's
> weight, which (in turn) will translate into the (now only outgoing)
> edge.
Why do you want to set the back edge count as the BB count? I think
the right formula is: count(back_edge) = count(BB) -
count(entry_edge), in which entry_edge is the edge that enters the
loop.
Dehao
>
>
> Diego.