Re: [PATCH v5] tree-ssa-sink: Improve code sinking pass

[Ajit Agarwal via Gcc-patches]

On 23/06/23 7:44 am, Peter Bergner wrote:
> On 6/1/23 11:54 PM, Ajit Agarwal via Gcc-patches wrote:
>>
>>
>> On 01/06/23 2:06 pm, Bernhard Reutner-Fischer wrote:
>>> On 1 June 2023 09:20:08 CEST, Ajit Agarwal <aagar...@linux.ibm.com> wrote:
>>>> Hello All:
>>>>
>>>> This patch improves code sinking pass to sink statements before call to 
>>>> reduce
>>>> register pressure.
>>>> Review comments are incorporated.
>>>
>>> Hi Ajit!
>>>
>>> I had two comments for v4 that you did not address in v5 or followed up.
>>> thanks,
>>
>> Which comments I didn't address. Please let me know.
> 
> I believe he's referring to these two comments:
> 
>   > +   && dominated_by_p (CDI_DOMINATORS, new_best_bb, early_bb))
>   > + {
>   > +   if (def_use_same_block (use))
>   > +     return best_bb;
>   > +
>   > +   return new_best_bb;
>   > + }
>   > + return best_bb;
>   > +    }
>   >  
> 
>   Many returns.
>   I'd have said
>         && !def_use_same_block (use)
>       return new_best_bb;
>       else
>       return best_bb;
> 
>   and rephrase the comment above list of Things to consider accordingly.
> 
> 
> I agree with Bernhard's comment that it could be rewritten to be clearer.
> Although, the "else" isn't really required.  So Bernhard's version would
> look like:
> 
>   if (new_best_bb
>       && use
>       && new_best_bb != best_bb
>       && new_best_bb != early_bb
>       && !is_gimple_call (stmt)
>       && gsi_end_p (gsi_start_phis (new_best_bb))
>       && gimple_bb (use) != early_bb
>       && !is_gimple_call (use)
>       && dominated_by_p (CDI_POST_DOMINATORS, new_best_bb, gimple_bb (use))
>       && dominated_by_p (CDI_DOMINATORS, new_best_bb, early_bb)
>       && !def_use_same_block (use))
>     return new_best_bb;
>   else
>     return best_bb;
> 
> ...or just:
> 
>   if (new_best_bb
>       && use
>       && new_best_bb != best_bb
>       && new_best_bb != early_bb
>       && !is_gimple_call (stmt)
>       && gsi_end_p (gsi_start_phis (new_best_bb))
>       && gimple_bb (use) != early_bb
>       && !is_gimple_call (use)
>       && dominated_by_p (CDI_POST_DOMINATORS, new_best_bb, gimple_bb (use))
>       && dominated_by_p (CDI_DOMINATORS, new_best_bb, early_bb)
>       && !def_use_same_block (use))
>     return new_best_bb;
> 
>   return best_bb;
> 
> 
> Either works.

Thanks Peter. I will incorporate and send the new version of the patch.

> 
> 
> Peter
>