[Bug rtl-optimization/117239] [12/13/14/15 Regression] wrong code at -O{s, 2} with "-fno-inline -fschedule-insns" on x86_64-linux-gnu since r12-7472-g609e8c492d62d9

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=117239

--- Comment #9 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Created attachment 60362
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=60362&action=edit
gcc15-pr117239-stats.patch

To guide me on which cases commonly occur during compilations, I've
bootstrapped/regtested this patch to print statistics (results included in the
first line) on how many mems to be potentially invalidated are
SP_DERIVED_VALUE_P or that + constant, how many are other values, how many are
non-VALUES (none) and which of those pass or fail canon_anti_dependence.