https://gcc.gnu.org/bugzilla/show_bug.cgi?id=128348
Bug ID: 128348
Summary: [17 Regression] ~70-80% exec time slowdown of
Polyhedron induct2 benchmark since
r17-4390-gd528b3e5314fde
Product: gcc
Version: 17.0
Status: UNCONFIRMED
Keywords: missed-optimization, needs-reduction
Severity: normal
Priority: P3
Component: tree-optimization
Assignee: unassigned at gcc dot gnu.org
Reporter: pheeck at gcc dot gnu.org
CC: rguenth at gcc dot gnu.org
Target Milestone: ---
Host: x86_64-pc-linux-gnu
Target: x86_64-pc-linux-gnu
As seen here
https://lnt.opensuse.org/db_default/v4/CPP/graph?plot.0=176.773.0
https://lnt.opensuse.org/db_default/v4/CPP/graph?plot.0=466.773.0
https://lnt.opensuse.org/db_default/v4/CPP/graph?plot.0=286.773.0
https://lnt.opensuse.org/db_default/v4/CPP/graph?plot.0=556.773.0
https://lnt.opensuse.org/db_default/v4/CPP/graph?plot.0=994.773.0
there was a ~70-80% slowdown of the Polyhedron Fortran Benchmarks
(www.fortran.uk) induct2. This bisects to
commit d528b3e5314fde594080406c95ac39c77881c770
Author: Richard Biener <[email protected]>
AuthorDate: Thu Jul 23 15:51:31 2026 +0200
Commit: Richard Biener <[email protected]>
CommitDate: Thu Sep 17 14:49:51 2026 +0200
Make vect_build_slp_tree_1 compute a set of matches[]
Where to get the benchmarks:
https://fortran.uk/pb11.zip
You can get to this download link from:
https://fortran.uk/fortran-compiler-comparisons/the-polyhedron-solutions-benchmark-suite/
I would include it as an attachment but I'm not sure how the benchmark is
licensed.
To reproduce:
Compile the fortran source induct2.f90 with -O2 with GCC before and after the
commit, put the binary into the same folder as the attached input induct2.in
and run the binary.
