https://gcc.gnu.org/bugzilla/show_bug.cgi?id=88713
--- Comment #50 from Uroš Bizjak <ubizjak at gmail dot com> --- (In reply to Thomas Koenig from comment #49) > (In reply to Uroš Bizjak from comment #48) > > (In reply to rguent...@suse.de from comment #47) > > > >But why don't we generate sqrtps for vector sqrtf? > > > > > > That's the default for - mrecip back in time we benchmarked it and scalar > > > recip miscompares sth. > > > > It was polyhedron benchmark, in one benchmark, the index was calculated from > > square root, and that was too sensitive for 2 ULP difference. > > Argh. Sacrificing performance for the sake of bugware... The details are in [1] and all the drama is documented in PR32352. [1] https://gcc.gnu.org/ml/gcc-patches/2007-06/msg01044.html