]> gcc.gnu.org Git - gcc.git/commit
[PR tree-optimization/107394] Canonicalize global franges as they are read back.
authorAldy Hernandez <aldyh@redhat.com>
Tue, 25 Oct 2022 20:44:51 +0000 (22:44 +0200)
committerAldy Hernandez <aldyh@redhat.com>
Thu, 27 Oct 2022 09:40:16 +0000 (11:40 +0200)
commit2b1fb720818a85d5c893ce65d140add40debf2ff
treee89fbf13fcc8de28fa6e8d83c6bec7489ca9c972
parent9119431bc1563217c7c770035b0456d1e2bc596d
[PR tree-optimization/107394] Canonicalize global franges as they are read back.

The problem here is that we're inlining a global range with NANs into
a function that has been tagged with __attribute__((optimize
("-ffinite-math-only"))).  As the global range is copied from
SSA_NAME_RANGE_INFO, its NAN bits are copied, which then cause
frange::verify_range() to fail a sanity check making sure no NANs
creep in when !HONOR_NANS.

I think what we should do is nuke the NAN bits as we're restoring the
global range.  For that matter, if we use the frange constructor,
everything except that NAN sign will be done automatically, including
dropping INFs to the min/max representable range when appropriate.

PR tree-optimization/107394

gcc/ChangeLog:

* value-range-storage.cc (frange_storage_slot::get_frange): Use
frange constructor.

gcc/testsuite/ChangeLog:

* gcc.dg/tree-ssa/pr107394.c: New test.
gcc/testsuite/gcc.dg/tree-ssa/pr107394.c [new file with mode: 0644]
gcc/value-range-storage.cc
This page took 0.073322 seconds and 6 git commands to generate.