https://gcc.gnu.org/bugzilla/show_bug.cgi?id=119367
Thomas Schwinge <tschwinge at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Summary|[15/16 Regression][gcn] |[15 Regression][gcn]
|libgomp.fortran/target1.f90 |libgomp.fortran/target1.f90
|with '-O2 -g': '.2byte |with '-O2 -g': '.2byte
|.LM6-.LM5' – error: ... out |.LM6-.LM5' – error: ... out
|of range – since |of range – since
|r15-8047-gadb14c7625178b |r15-8047-gadb14c7625178b
--- Comment #10 from Thomas Schwinge <tschwinge at gcc dot gnu.org> ---
GCN target 'gfortran.dg/class_allocate_22.f90',
'gfortran.dg/inline_matmul_6.f90' had already recovered recently (in the past
few months), and the remaining ones: GCN target 'gfortran.dg/associate_18.f08'
as well as GCN offloading 'libgomp.fortran/target1.f90',
'libgomp.fortran/target6.f90' have now recovered via commit
r16-2377-gbdc4062a0796788e44d5e6ecd753268a8b453cc7 "amdgcn: add more insn
patterns using vec_duplicate", and/or commit
r16-2378-gf194924984c4eb9c8be5310f78b191b35e576ab8 "amdgcn: Add ashlvNm, mulvNm
macros", and/or commit r16-2379-g351fa55c58a036f148d13bca972e687a0bacd113
"amdgcn: add DImode offsets for gather/scatter".
However, as far as I understand, the underlying problem (location views) has
not yet been resolved; is now latent.
