https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78200
--- Comment #12 from Richard Biener <rguenth at gcc dot gnu.org> ---
Btw, I see with -mavx2
addq (%r9), %rax
jns .L90
.L90:
je .L92
cmpl $2, 24(%rdx)
je .L91
thus there is no extra cmpq $0, %rdi in the predecessor.
Note when I profile avx (base) vs. avx2 (peak) I see
18.17% 451662 mcf_base.amd64- mcf_base.amd64-m64-gcc42-nn [.]
refresh_potential
18.12% 424592 mcf_base.amd64- mcf_base.amd64-m64-gcc42-nn [.]
primal_bea_mpp
17.96% 465325 mcf_peak.amd64- mcf_peak.amd64-m64-gcc42-nn [.]
primal_bea_mpp
14.93% 373309 mcf_peak.amd64- mcf_peak.amd64-m64-gcc42-nn [.]
refresh_potential
plus a 3-run of avx (base) vs. avx2 (peak) gives me
429.mcf 9120 252 36.1 * 9120 264 34.6 S
429.mcf 9120 257 35.5 S 9120 253 36.0 S
429.mcf 9120 232 39.3 S 9120 258 35.4 *
which isn't really conclusive.
If you are trying to narrow down a regression GCC 6 vs. GCC 7 I wouldn't
look at flags but at profiling and what changed.
