Re: How to force gcc to vectorize the loop with particular vectorization width

2017-10-21 Thread Denis Bakhvalov
Hello Richard, Thank you. I achieved vectorization with vf = 16, using #pragma GCC optimize ("no-unroll-loops") __attribute__ ((__target__ ("sse4.2"))) and options -march=core-avx2 -mprefer-avx-128 But now I have a question: Is it possible in gcc to have vectorization with vf < 16? On 20/10/2017,

Re: atomic_thread_fence() semantics

2017-10-21 Thread Torvald Riegel
On Fri, 2017-10-20 at 18:46 +0300, Alexander Monakov wrote: > On Fri, 20 Oct 2017, Torvald Riegel wrote: > > On Thu, 2017-10-19 at 15:31 +0300, Alexander Monakov wrote: > > > On Thu, 19 Oct 2017, Andrew Haley wrote: > > > > No, you did not. This looks like a bug. Please report it. > > > > > > Th

Re: How to force gcc to vectorize the loop with particular vectorization width

2017-10-21 Thread Richard Biener
On October 21, 2017 9:50:13 PM GMT+02:00, Denis Bakhvalov wrote: >Hello Richard, >Thank you. I achieved vectorization with vf = 16, using >#pragma GCC optimize ("no-unroll-loops") >__attribute__ ((__target__ ("sse4.2"))) >and options -march=core-avx2 -mprefer-avx-128 > >But now I have a question: