[Bug tree-optimization/126371] New: [17 Regression] 26% slowdown of s122 from TSVC benchmark suite on Aarch64 (Neoverse N1)

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=126371

            Bug ID: 126371
           Summary: [17 Regression] 26% slowdown of s122 from TSVC
                    benchmark suite on Aarch64 (Neoverse N1)
           Product: gcc
           Version: 17.0
            Status: UNCONFIRMED
          Keywords: missed-optimization, needs-bisection
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: pheeck at gcc dot gnu.org
  Target Milestone: ---
              Host: aarch64-gnu-linux
            Target: aarch64-gnu-linux

Created attachment 65116
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=65116&action=edit
s122 tsvc testcase

Testcase s122 from TEST SUITE FOR VECTORIZING COMPILERS slowed down by 26% when
compiled with -Ofast -march=armv8.2-a+crypto+fp16+rcpc+dotprod+ssbs.  I've seen
this on Neoverse N1 machines.

This happened somewhere between

r17-2271-ge5af677005c7ab
r17-2297-g1f949c2f1ab60a

A graph showing the slowdown:
https://lnt.opensuse.org/db_default/v4/CPP/graph?plot.0=353.820.0


To reproduce, build the attached testcase:

gcc *.c -Ofast -march=armv8.2-a+crypto+fp16+rcpc+dotprod+ssbs -o tsvc -lm

and run the testcase

$ ./tsvc
Loop    Time(sec)       Checksum
 s122        0.683      196490.531250

$ ./tsvc
Loop    Time(sec)       Checksum
 s122        0.861      196490.531250