[Bug tree-optimization/84114] global reassociation pass prevents fma usage, generates slower code

[wilco at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84114

--- Comment #7 from Wilco <wilco at gcc dot gnu.org> ---
Author: wilco
Date: Mon Mar  5 14:40:55 2018
New Revision: 258248

URL: https://gcc.gnu.org/viewcvs?rev=258248&root=gcc&view=rev
Log:
[AArch64] PR84114: Avoid reassociating FMA

As discussed in the PR, the reassociation phase runs before FMAs are formed
and so can significantly reduce FMA opportunities.  Although reassociation
could be switched off, it helps in many cases, so a better alternative is to
only avoid reassociation of floating point additions.  This fixes the testcase
and gives 1% speedup on SPECFP2017, fixing the performance regression.

    gcc/
        PR tree-optimization/84114
        * config/aarch64/aarch64.c (aarch64_reassociation_width)
        Avoid reassociation of FLOAT_MODE addition.

Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/config/aarch64/aarch64.c