Re: [PATCH] Allow FSM to thread single block cases too

[Richard Biener <richard.guenther@gmail.com>]

On Wed, Oct 14, 2015 at 2:42 PM, Jeff Law <law@redhat.com> wrote:
> On 10/14/2015 04:16 AM, Richard Biener wrote:
>>
>> On Tue, Oct 13, 2015 at 2:52 PM, Richard Biener
>> <richard.guenther@gmail.com> wrote:
>>>
>>> On Tue, Oct 13, 2015 at 2:21 PM, Jeff Law <law@redhat.com> wrote:
>>>>
>>>>
>>>> One of the cases that was missing in the FSM support is threading when
>>>> the
>>>> path is a single block.  ie, a control statement's output can be
>>>> statically
>>>> determined just by looking at PHIs in the control statement's block for
>>>> one
>>>> or incoming edges.
>>>>
>>>> This is necessary to fix a regression if I turn off the old jump
>>>> threader's
>>>> backedge support.  Just as important, Jan has in the past asked about a
>>>> trivial jump threader to be run during early optimizations.  Limiting
>>>> the
>>>> FSM bits to this case would likely satisfy that need in the future.
>>>
>>>
>>> I think he asked for trivial forward threads though due to repeated
>>> tests.
>>> I hacked FRE to do this (I think), but maybe some trivial cleanup
>>> opportunities
>>> are still left here.  Honza?
>>
>>
>> This or other related patches in the range r228731:228774 has caused a
>> quite
>> big jump in SPEC CPU 2000 binary sizes (notably 176.gcc - so maybe
>> affecting
>> bootstrap as well, at -O3).  Are you sure this doesn't re-introduce DOM
>> effectively peeling all loops once?
>
> It's possible.  I've actually got a patch in overnight testing that
> introduces some of the heuristics to avoid mucking up loops to the FSM bits.

Like never threading a loop exit test to the loop header (but only to the exit).
At least if it is the only exit in the loop (but maybe better for all exits).

Richard.

> jeff
>