Re: [PATCH, rs6000] Fix PR80695 (vec_construct cost modeling issue)

2017-05-11 Thread Segher Boessenkool
On Thu, May 11, 2017 at 10:59:41AM -0500, Bill Schmidt wrote: > PR80695 identifies a case (similar to several others we've seen) where SLP > vectorization is too aggressive about vectorizing stores. The problem is > that we undervalue the cost of a vec_construct operation. vec_construct > is the

Re: [PATCH, rs6000] Fix PR80695 (vec_construct cost modeling issue)

2017-05-11 Thread Bill Schmidt
I meant to mention that I regstrapped this on both POWER8 and POWER9. Bill > On May 11, 2017, at 10:59 AM, Bill Schmidt > wrote: > > Hi, > > PR80695 identifies a case (similar to several others we've seen) where SLP > vectorization is too aggressive about vectorizing stores. The problem is >

[PATCH, rs6000] Fix PR80695 (vec_construct cost modeling issue)

2017-05-11 Thread Bill Schmidt
Hi, PR80695 identifies a case (similar to several others we've seen) where SLP vectorization is too aggressive about vectorizing stores. The problem is that we undervalue the cost of a vec_construct operation. vec_construct is the vectorizer's representation for building a vector from scalar ele