On December 28, 2019 5:29:20 PM GMT+01:00, Richard Sandiford
<richard.sandif...@arm.com> wrote:
>The EXTRACT_LAST_REDUCTION handling needs to generate a separate
>comparison instruction that feeds the vector mask argument of the
>IFN_EXTRACT_LAST call. We weren't checking whether that comparison
>was supported, leading to an ICE on the testcase.
>
>Tested on aarch64-linux-gnu and x86_64-linux-gnu. OK to install?
Ok.
Richard.
>Richard
>
>
>2019-12-28 Richard Sandiford <richard.sandif...@arm.com>
>
>gcc/
> * tree-vect-stmts.c (vectorizable_condition): For extract-last
> reductions, check that the target supports the required comparison
> operation.
>
>gcc/testsuite/
> * gcc.dg/vect/vect-cond-12.c: New test.
>
>Index: gcc/tree-vect-stmts.c
>===================================================================
>--- gcc/tree-vect-stmts.c 2019-12-27 16:53:46.000000000 +0000
>+++ gcc/tree-vect-stmts.c 2019-12-28 16:28:16.344143831 +0000
>@@ -10081,6 +10081,16 @@ vectorizable_condition (stmt_vec_info st
> cond_code = SSA_NAME;
> }
>
>+ if (TREE_CODE_CLASS (cond_code) == tcc_comparison
>+ && reduction_type == EXTRACT_LAST_REDUCTION
>+ && !expand_vec_cmp_expr_p (comp_vectype, vec_cmp_type,
>cond_code))
>+ {
>+ if (dump_enabled_p ())
>+ dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
>+ "reduction comparison operation not supported.\n");
>+ return false;
>+ }
>+
> if (!vec_stmt)
> {
> if (bitop1 != NOP_EXPR)
>Index: gcc/testsuite/gcc.dg/vect/vect-cond-12.c
>===================================================================
>--- /dev/null 2019-09-17 11:41:18.176664108 +0100
>+++ gcc/testsuite/gcc.dg/vect/vect-cond-12.c 2019-12-28
>16:28:16.344143831 +0000
>@@ -0,0 +1,14 @@
>+/* { dg-do compile } */
>+
>+int
>+f (int *x, short *y)
>+{
>+ int res = 100;
>+ for (int i = 0; i < 40; ++i)
>+ {
>+ if (y[i] > 1)
>+ res = x[i];
>+ x[i] += y[i];
>+ }
>+ return res;
>+}
