Re: [PATCH] vect: Tweak vect_better_loop_vinfo_p handling of variable VFs

2020-04-20 Thread Richard Biener via Gcc-patches
On Fri, Apr 17, 2020 at 5:21 PM Richard Sandiford wrote: > > This patch fixes a large lmbench performance regression with > 128-bit SVE, compiled in length-agnostic mode. > > vect_better_loop_vinfo_p (new in GCC 10) tries to estimate whether > a new loop_vinfo is cheaper than a previous one, with

[PATCH] vect: Tweak vect_better_loop_vinfo_p handling of variable VFs

2020-04-17 Thread Richard Sandiford
This patch fixes a large lmbench performance regression with 128-bit SVE, compiled in length-agnostic mode. vect_better_loop_vinfo_p (new in GCC 10) tries to estimate whether a new loop_vinfo is cheaper than a previous one, with an in-built preference for the old one. For variable VF it prefers t