Re: [PATCH] tree-optimization/104010 - fix SLP scalar costing with patterns

On Thu, 14 Apr 2022, Richard Sandiford wrote: > Richard Biener writes: > > On Thu, 14 Apr 2022, Richard Sandiford wrote: > > > >> Richard Biener writes: > >> > When doing BB vectorization the scalar cost compute is derailed > >> > by patterns, causing lanes to be considered live and thus not > >

Re: [PATCH] tree-optimization/104010 - fix SLP scalar costing with patterns

Richard Biener writes: > On Thu, 14 Apr 2022, Richard Sandiford wrote: > >> Richard Biener writes: >> > When doing BB vectorization the scalar cost compute is derailed >> > by patterns, causing lanes to be considered live and thus not >> > costed on the scalar side. For the testcase in PR104010

Re: [PATCH] tree-optimization/104010 - fix SLP scalar costing with patterns

On Thu, 14 Apr 2022, Richard Sandiford wrote: > Richard Biener writes: > > When doing BB vectorization the scalar cost compute is derailed > > by patterns, causing lanes to be considered live and thus not > > costed on the scalar side. For the testcase in PR104010 this > > prevents vectorization

Re: [PATCH] tree-optimization/104010 - fix SLP scalar costing with patterns

Richard Biener writes: > When doing BB vectorization the scalar cost compute is derailed > by patterns, causing lanes to be considered live and thus not > costed on the scalar side. For the testcase in PR104010 this > prevents vectorization which was done by GCC 11. PR103941 > shows similar case

Re: [PATCH] tree-optimization/104010 - fix SLP scalar costing with patterns

On Wed, 13 Apr 2022, Richard Biener wrote: > When doing BB vectorization the scalar cost compute is derailed > by patterns, causing lanes to be considered live and thus not > costed on the scalar side. For the testcase in PR104010 this > prevents vectorization which was done by GCC 11. PR103941

Re: [PATCH] tree-optimization/104010 - fix SLP scalar costing with patterns

On 4/13/22 14:57, Richard Biener via Gcc-patches wrote: When doing BB vectorization the scalar cost compute is derailed by patterns, causing lanes to be considered live and thus not costed on the scalar side. For the testcase in PR104010 this prevents vectorization which was done by GCC 11.

[PATCH] tree-optimization/104010 - fix SLP scalar costing with patterns

When doing BB vectorization the scalar cost compute is derailed by patterns, causing lanes to be considered live and thus not costed on the scalar side. For the testcase in PR104010 this prevents vectorization which was done by GCC 11. PR103941 shows similar cases of missed optimizations that are