2016-04-18 11:53 GMT+03:00 Senthil Kumar Selvaraj
:
> Hi,
>
> While tracking down a performance regression for the AVR target from
> 4.9.x to trunk, I noticed that failing the SHRINK_WRAPPING_ENABLED
> check in ira.c led to noticeably worse code for the
Hi,
While tracking down a performance regression for the AVR target from
4.9.x to trunk, I noticed that failing the SHRINK_WRAPPING_ENABLED
check in ira.c led to noticeably worse code for the following
case. That check prevents live range splitting of pseudos containing
formal args, and