On Wed, 2025-05-28 at 18:17 +0100, Richard Sandiford wrote: > Sorry for the slow reply, had a few days off. > > Xi Ruoyao <xry...@xry111.site> writes: > > If we see a promoted subreg and TRULY_NOOP_TRUNCATION says the > > truncation is not a noop, then all bits of the inner reg are live. We > > cannot reduce the live mask to that of the mode of the subreg. > > > > gcc/ChangeLog: > > > > PR rtl-optimization/120050 > > * ext-dce.cc (ext_dce_process_uses): Break early if a SUBREG in > > rhs is promoted and the truncation from the inner mode to the > > outer mode is not a noop when handling SETs. > > --- > > > > Bootstrapped on mips64el-linux-gnuabi64. Ok for trunk? > > > > gcc/ext-dce.cc | 12 ++++++++++-- > > 1 file changed, 10 insertions(+), 2 deletions(-) > > > > diff --git a/gcc/ext-dce.cc b/gcc/ext-dce.cc > > index a0343950141..3b21e68b90c 100644 > > --- a/gcc/ext-dce.cc > > +++ b/gcc/ext-dce.cc > > @@ -35,6 +35,7 @@ along with GCC; see the file COPYING3. If not see > > #include "print-rtl.h" > > #include "dbgcnt.h" > > #include "diagnostic-core.h" > > +#include "target.h" > > > > /* These should probably move into a C++ class. */ > > static vec<bitmap_head> livein; > > @@ -764,13 +765,20 @@ ext_dce_process_uses (rtx_insn *insn, rtx obj, > > We don't want to mark those bits live unnecessarily > > as that inhibits extension elimination in important > > cases such as those in Coremark. So we need that > > - outer code. */ > > + outer code. > > + > > + But if !TRULY_NOOP_TRUNCATION_MODES_P, those bits > > + may be actually alive with any promoted subreg > > + regardless of the outer code. See PR 120050. */ > > How about expanding on this a bit: > > If !TRULY_NOOP_TRUNCATION_MODES_P holds true for > the subreg, then the mode change performed by Y > would normally need to be a TRUNCATE rather than > a SUBREG. It is probably the guarantee provided > by SUBREG_PROMOTED_VAR_P that allows the SUBREG > in Y as an exception. We must therefore preserve > that guarantee and treat the upper bits of the > inner register as live regardless of the outer code. > See PR 120050. */ > > OK with that change, thanks.
Pushed r16-1110 with the comment updated. Will wait for a few days before backporting to 15. -- Xi Ruoyao <xry...@xry111.site> School of Aerospace Science and Technology, Xidian University
