This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: Another fix for single-element permutes (PR 84265)
- From: Richard Biener <richard dot guenther at gmail dot com>
- To: GCC Patches <gcc-patches at gcc dot gnu dot org>, Richard Sandiford <richard dot sandiford at linaro dot org>
- Date: Thu, 8 Feb 2018 16:10:47 +0100
- Subject: Re: Another fix for single-element permutes (PR 84265)
- Authentication-results: sourceware.org; auth=none
- References: <878tc3y5h0.fsf@linaro.org>
On Thu, Feb 8, 2018 at 2:29 PM, Richard Sandiford
<richard.sandiford@linaro.org> wrote:
> PR83753 was about a case in which we ended up trying to "vectorise"
> a group of loads ore stores using single-element vectors. The problem
> was that we were classifying the load or store as VMAT_CONTIGUOUS_PERMUTE
> rather than VMAT_CONTIGUOUS, even though it doesn't make sense to permute
> a single-element vector.
>
> In that PR it was enough to change get_group_load_store_type,
> because vectorisation ended up being unprofitable and so we didn't
> take things further. But when vectorisation is profitable, the same
> fix is needed in vectorizable_load and vectorizable_store.
>
> Tested on aarch64-linux-gnu, aarch64_be-elf and x86_64-linux-gnu.
> OK to install?
OK.
Richard.
> Richard
>
>
> 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
>
> gcc/
> PR tree-optimization/84265
> * tree-vect-stmts.c (vectorizable_store): Don't treat
> VMAT_CONTIGUOUS accesses as grouped.
> (vectorizable_load): Likewise.
>
> gcc/testsuite/
> PR tree-optimization/84265
> * gcc.dg/vect/pr84265.c: New test.
>
> Index: gcc/tree-vect-stmts.c
> ===================================================================
> --- gcc/tree-vect-stmts.c 2018-01-30 09:45:27.710764075 +0000
> +++ gcc/tree-vect-stmts.c 2018-02-08 13:26:39.242566948 +0000
> @@ -6214,7 +6214,8 @@ vectorizable_store (gimple *stmt, gimple
> }
>
> grouped_store = (STMT_VINFO_GROUPED_ACCESS (stmt_info)
> - && memory_access_type != VMAT_GATHER_SCATTER);
> + && memory_access_type != VMAT_GATHER_SCATTER
> + && (slp || memory_access_type != VMAT_CONTIGUOUS));
> if (grouped_store)
> {
> first_stmt = GROUP_FIRST_ELEMENT (stmt_info);
> @@ -7696,7 +7697,8 @@ vectorizable_load (gimple *stmt, gimple_
> return true;
> }
>
> - if (memory_access_type == VMAT_GATHER_SCATTER)
> + if (memory_access_type == VMAT_GATHER_SCATTER
> + || (!slp && memory_access_type == VMAT_CONTIGUOUS))
> grouped_load = false;
>
> if (grouped_load)
> Index: gcc/testsuite/gcc.dg/vect/pr84265.c
> ===================================================================
> --- /dev/null 2018-02-08 11:17:10.862716283 +0000
> +++ gcc/testsuite/gcc.dg/vect/pr84265.c 2018-02-08 13:26:39.240567025 +0000
> @@ -0,0 +1,23 @@
> +/* { dg-do compile } */
> +
> +struct a
> +{
> + unsigned long b;
> + unsigned long c;
> + int d;
> + int *e;
> + char f;
> +};
> +
> +struct
> +{
> + int g;
> + struct a h[];
> +} i;
> +
> +int j, k;
> +void l ()
> +{
> + for (; k; k++)
> + j += (int) (i.h[k].c - i.h[k].b);
> +}