Hello! > When changing reg_known_value to a VEC, Steven sneaked in a small > semantic change. That change has the advantage of improving compilation > time substantially on some testcases (including PR55489), but it is a > bit heavy-handed: it also makes set_known_reg_value a no-op, while > get_known_reg_value will always return NULL. > > This patch fixes the VEC usage. Bootstrap/regtest in progress, but I'm > going to commit this as obvious as soon as bootstrapping finishes. > > 2012-11-26 Paolo Bonzini <pbonz...@redhat.com> > > * alias.c (init_alias_analysis): Fix allocation of reg_known_value. > > Index: ../../gcc/alias.c > =================================================================== > --- ../../gcc/alias.c (revisione 193853) > +++ ../../gcc/alias.c (copia locale) > @@ -2808,7 +2808,7 @@ init_alias_analysis (void) > > timevar_push (TV_ALIAS_ANALYSIS); > > - vec_alloc (reg_known_value, maxreg - FIRST_PSEUDO_REGISTER); > + vec_safe_grow_cleared (reg_known_value, maxreg - FIRST_PSEUDO_REGISTER); > reg_known_equiv_p = sbitmap_alloc (maxreg - FIRST_PSEUDO_REGISTER); > > /* If we have memory allocated from the previous run, use it. */
This one-liner causes following runtime test failure [1] for alphaev68-linux-gnu: FAIL: gfortran.fortran-torture/execute/save_1.f90 execution, -O2 FAIL: gfortran.fortran-torture/execute/save_1.f90 execution, -O2 -fomit-frame-pointer -finline-functions FAIL: gfortran.fortran-torture/execute/save_1.f90 execution, -O2 -fomit-frame-pointer -finline-functions -funroll-loops FAIL: gfortran.fortran-torture/execute/save_1.f90 execution, -O2 -fbounds-check The patch miscompiles libgfortran library. I will provide more info tomorrow, any hint what/where should I look for differences? [1] http://gcc.gnu.org/ml/gcc-testresults/2012-11/msg02492.html Uros.
