Use SCALAR_TYPE_MODE in vect_create_epilog_for_reduction

[Richard Sandiford]

This follows on from similar changes a couple of months ago and
is needed when general modes have variable size.

Tested on aarch64-linux-gnu, x86_64-linux-gnu and powerpc64le-linux-gnu.
Applied as obvious.

Richard


2017-10-23  Richard Sandiford  <richard.sandif...@linaro.org>

gcc/
        * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
        SCALAR_TYPE_MODE instead of TYPE_MODE.

Index: gcc/tree-vect-loop.c
===================================================================
--- gcc/tree-vect-loop.c        2017-10-22 21:04:50.136830154 +0100
+++ gcc/tree-vect-loop.c        2017-10-23 10:39:37.711243373 +0100
@@ -4487,7 +4487,7 @@ vect_create_epilog_for_reduction (vec<tr
       gcc_assert (gimple_assign_rhs_code (vec_stmt) == VEC_COND_EXPR);
 
       int scalar_precision
-       = GET_MODE_PRECISION (TYPE_MODE (TREE_TYPE (vectype)));
+       = GET_MODE_PRECISION (SCALAR_TYPE_MODE (TREE_TYPE (vectype)));
       tree cr_index_scalar_type = make_unsigned_type (scalar_precision);
       tree cr_index_vector_type = build_vector_type
        (cr_index_scalar_type, TYPE_VECTOR_SUBPARTS (vectype));