On December 27, 2019 5:02:55 PM GMT+01:00, Richard Sandiford
<richard.sandif...@arm.com> wrote:
>The fold-left reduction code has a (rarely-used) fallback that handles
>cases in which the loop is fully-masked and the target has no native
>support for the reduction. The fallback includea a VEC_COND_EXPR
>between the reduction vector and a safe value, so we should check
>whether that VEC_COND_EXPR is supported.
>
>Tested on aarch64-linux-gnu and x86_64-linux-gnu. OK to install?
Ok.
Richard.
>Richard
>
>
>2019-12-27 Richard Sandiford <richard.sandif...@arm.com>
>
>gcc/
> * tree-vect-loop.c (vectorizable_reduction): Check whether the
> target supports the required VEC_COND_EXPR operation before
> allowing the fallback handling of masked fold-left reductions.
>
>gcc/testsuite/
> * gcc.target/aarch64/sve/mixed_size_10.c: New test.
>
>Index: gcc/tree-vect-loop.c
>===================================================================
>--- gcc/tree-vect-loop.c 2019-12-19 13:21:37.554633868 +0000
>+++ gcc/tree-vect-loop.c 2019-12-27 16:01:39.344247474 +0000
>@@ -6718,6 +6718,18 @@ vectorizable_reduction (stmt_vec_info st
> " conditional operation is available.\n");
> LOOP_VINFO_CAN_FULLY_MASK_P (loop_vinfo) = false;
> }
>+ else if (reduction_type == FOLD_LEFT_REDUCTION
>+ && reduc_fn == IFN_LAST
>+ && !expand_vec_cond_expr_p (vectype_in,
>+ truth_type_for (vectype_in),
>+ SSA_NAME))
>+ {
>+ if (dump_enabled_p ())
>+ dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
>+ "can't use a fully-masked loop because no"
>+ " conditional operation is available.\n");
>+ LOOP_VINFO_CAN_FULLY_MASK_P (loop_vinfo) = false;
>+ }
> else
> vect_record_loop_mask (loop_vinfo, masks, ncopies * vec_num,
> vectype_in, NULL);
>Index: gcc/testsuite/gcc.target/aarch64/sve/mixed_size_10.c
>===================================================================
>--- /dev/null 2019-09-17 11:41:18.176664108 +0100
>+++ gcc/testsuite/gcc.target/aarch64/sve/mixed_size_10.c 2019-12-27
>16:01:39.344247474 +0000
>@@ -0,0 +1,13 @@
>+/* { dg-options "-O3 -msve-vector-bits=256
>-fno-tree-loop-distribution" } */
>+
>+float
>+f (float *restrict x, double *restrict y)
>+{
>+ float res = 0.0;
>+ for (int i = 0; i < 100; ++i)
>+ {
>+ res += x[i];
>+ y[i] += y[i - 4] * 11;
>+ }
>+ return res;
>+}
