[Bug target/82666] [13/14/15/16 regression]: sum += (x>128 ? x : 0) puts the cmov on the critical path (at -O2)

[rguenth at gcc dot gnu.org via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82666

--- Comment #19 from Richard Biener <rguenth at gcc dot gnu.org> ---
Elsewhere we discussed that both such case and the FMA "defer/disable" we do on
GIMPLE should ideally be done on RTL (as a loop pass?) to decouple the actual
reduction from long latency addends there.