https://gcc.gnu.org/bugzilla/show_bug.cgi?id=69657
--- Comment #5 from Wilco <wdijkstr at arm dot com> --- (In reply to Andrew Pinski from comment #4) > (In reply to Jonathan Wakely from comment #3) > > Recategorising as component=c++, and removing the regression marker (because > > the change in libstdc++ that reveals this issue is required for > > conformance). > > Adding back regression as it is an user visible regression even though only > the library change, the user does not understand that. In fact I suspect > this comes from SPEC CPU 2006 or some other heavily used benchmarking code. Correct, I found it while investigating major regressions in a few benchmarks. Note I see the long, long long, int128, float, double, long double overloads, but where is the plain int overload defined? I bet that adding an int overload that redirects to __builtin_abs similar to the others will fix the issue.