On December 10, 2019 12:39:53 PM GMT+01:00, Richard Sandiford <richard.sandif...@arm.com> wrote: >The direct_slp_reduc code in vect_create_epilog_for_reduction was >still assuming that all types involved in a reduction are the same >(up to types_compatible_p), whereas we now support differences in >sign. This was causing an ICE in gcc.dg/vect/pr92324-4.c for SVE. > >Tested on aarch64-linux-gnu and x86_64-linux-gnu. OK to install?
Ok. Richard. >Richard > > >2019-12-10 Richard Sandiford <richard.sandif...@arm.com> > >gcc/ > * tree-vect-loop.c (vect_create_epilog_for_reduction): When > handling direct_slp_reduc, allow the PHI arguments to have > a different type from the vector elements. > >Index: gcc/tree-vect-loop.c >=================================================================== >--- gcc/tree-vect-loop.c 2019-12-10 11:30:23.506138368 +0000 >+++ gcc/tree-vect-loop.c 2019-12-10 11:38:37.322774821 +0000 >@@ -5054,6 +5054,8 @@ vect_create_epilog_for_reduction (stmt_v > tree scalar_value > = PHI_ARG_DEF_FROM_EDGE (orig_phis[i]->stmt, > loop_preheader_edge (loop)); >+ scalar_value = gimple_convert (&seq, TREE_TYPE (vectype), >+ scalar_value); > vector_identity = gimple_build_vector_from_val (&seq, vectype, > scalar_value); > }
