The following makes sure to perform the vector element extraction
using the element type and convert to the original, possibly
bit-precision, result afterwards. I've also used gimple_build
for the BIT_FIELD_REF since that simplifies the code.
Bootstrapped on x86_64-unknown-linux-gnu, testing in progress.
PR tree-optimization/122158
* tree-vect-loop.cc (vect_create_epilog_for_reduction): Handle
bit-precision result.
* gcc.dg/vect/pr122158.c: New testcase.
---
gcc/testsuite/gcc.dg/vect/pr122158.c | 15 +++++++++++++++
gcc/tree-vect-loop.cc | 14 +++++---------
2 files changed, 20 insertions(+), 9 deletions(-)
create mode 100644 gcc/testsuite/gcc.dg/vect/pr122158.c
diff --git a/gcc/testsuite/gcc.dg/vect/pr122158.c
b/gcc/testsuite/gcc.dg/vect/pr122158.c
new file mode 100644
index 00000000000..5d0f7dac7dc
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/vect/pr122158.c
@@ -0,0 +1,15 @@
+/* { dg-do compile } */
+
+int g_3, g_181, g_7;
+void func_54(unsigned int, int *);
+int func_24()
+{
+ int p_25;
+ int *l_49 = &g_3;
+ func_54(g_7, l_49);
+}
+void func_54(unsigned int p_55, int *p_56)
+{
+ for (g_181 = 27; g_181 != 9; --g_181)
+ p_55 || (*p_56 = 0);
+}
diff --git a/gcc/tree-vect-loop.cc b/gcc/tree-vect-loop.cc
index 160cc37a0d6..003dc734c01 100644
--- a/gcc/tree-vect-loop.cc
+++ b/gcc/tree-vect-loop.cc
@@ -6055,8 +6055,6 @@ vect_create_epilog_for_reduction (loop_vec_info
loop_vinfo,
Create: va = vop <va, va'>
} */
- tree rhs;
-
if (dump_enabled_p ())
dump_printf_loc (MSG_NOTE, vect_location,
"Reduce using vector shifts\n");
@@ -6075,7 +6073,6 @@ vect_create_epilog_for_reduction (loop_vec_info
loop_vinfo,
new_temp = gimple_build (&stmts, code,
vectype1, new_name, new_temp);
}
- gsi_insert_seq_before (&exit_gsi, stmts, GSI_SAME_STMT);
/* 2.4 Extract the final scalar result. Create:
s_out3 = extract_field <v_out2, bitpos> */
@@ -6084,12 +6081,11 @@ vect_create_epilog_for_reduction (loop_vec_info
loop_vinfo,
dump_printf_loc (MSG_NOTE, vect_location,
"extract scalar result\n");
- rhs = build3 (BIT_FIELD_REF, scalar_type, new_temp,
- bitsize, bitsize_zero_node);
- epilog_stmt = gimple_build_assign (new_scalar_dest, rhs);
- new_temp = make_ssa_name (new_scalar_dest, epilog_stmt);
- gimple_assign_set_lhs (epilog_stmt, new_temp);
- gsi_insert_before (&exit_gsi, epilog_stmt, GSI_SAME_STMT);
+ new_temp = gimple_build (&stmts, BIT_FIELD_REF, TREE_TYPE (vectype1),
+ new_temp, bitsize, bitsize_zero_node);
+ new_temp = gimple_build (&stmts, VIEW_CONVERT_EXPR,
+ scalar_type, new_temp);
+ gsi_insert_seq_before (&exit_gsi, stmts, GSI_SAME_STMT);
scalar_results.safe_push (new_temp);
}
else
--
2.51.0
