From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <gcc-patches-return-289767-listarch-gcc-patches=gcc.gnu.org@gcc.gnu.org>
Received: (qmail 27647 invoked by alias); 18 Apr 2011 11:10:24 -0000
Received: (qmail 27638 invoked by uid 22791); 18 Apr 2011 11:10:22 -0000
X-SWARE-Spam-Status: No, hits=-2.3 required=5.0	tests=AWL,BAYES_00,DKIM_SIGNED,DKIM_VALID,DKIM_VALID_AU,FREEMAIL_FROM,RCVD_IN_DNSWL_LOW,RFC_ABUSE_POST,T_TO_NO_BRKTS_FREEMAIL
X-Spam-Check-By: sourceware.org
Received: from mail-wy0-f175.google.com (HELO mail-wy0-f175.google.com) (74.125.82.175)    by sourceware.org (qpsmtpd/0.43rc1) with ESMTP; Mon, 18 Apr 2011 11:09:48 +0000
Received: by wye20 with SMTP id 20so4595360wye.20        for <gcc-patches@gcc.gnu.org>; Mon, 18 Apr 2011 04:09:46 -0700 (PDT)
MIME-Version: 1.0
Received: by 10.227.202.149 with SMTP id fe21mr4855653wbb.205.1303124986508; Mon, 18 Apr 2011 04:09:46 -0700 (PDT)
Received: by 10.227.0.140 with HTTP; Mon, 18 Apr 2011 04:09:46 -0700 (PDT)
In-Reply-To: <g44o63fu4r.fsf@linaro.org>
References: <g4pqorfvwp.fsf@linaro.org>	<g44o63fu4r.fsf@linaro.org>
Date: Mon, 18 Apr 2011 11:54:00 -0000
Message-ID: <BANLkTi=Q07DdYYiP_oWMbHtVud4z_xr9Bw@mail.gmail.com>
Subject: Re: [5/9] Main target-independent support for direct interleaving
From: Richard Guenther <richard.guenther@gmail.com>
To: gcc-patches@gcc.gnu.org, patches@linaro.org, richard.sandiford@linaro.org
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: quoted-printable
X-IsSubscribed: yes
Mailing-List: contact gcc-patches-help@gcc.gnu.org; run by ezmlm
Precedence: bulk
List-Id: <gcc-patches.gcc.gnu.org>
List-Archive: <http://gcc.gnu.org/ml/gcc-patches/>
List-Post: <mailto:gcc-patches@gcc.gnu.org>
List-Help: <mailto:gcc-patches-help@gcc.gnu.org>
Sender: gcc-patches-owner@gcc.gnu.org
X-SW-Source: 2011-04/txt/msg01371.txt.bz2

On Tue, Apr 12, 2011 at 3:59 PM, Richard Sandiford
<richard.sandiford@linaro.org> wrote:
> This patch adds vec_load_lanes and vec_store_lanes optabs for instructions
> like NEON's vldN and vstN. =A0The optabs are defined this way because the
> vectors must be allocated to a block of consecutive registers.
>
> Tested on x86_64-linux-gnu and arm-linux-gnueabi. =A0OK to install?
>
> Richard
>
>
> gcc/
> =A0 =A0 =A0 =A0* doc/md.texi (vec_load_lanes, vec_store_lanes): Document.
> =A0 =A0 =A0 =A0* optabs.h (COI_vec_load_lanes, COI_vec_store_lanes): New
> =A0 =A0 =A0 =A0convert_optab_index values.
> =A0 =A0 =A0 =A0(vec_load_lanes_optab, vec_store_lanes_optab): New convert=
 optabs.
> =A0 =A0 =A0 =A0* genopinit.c (optabs): Initialize the new optabs.
> =A0 =A0 =A0 =A0* internal-fn.def (LOAD_LANES, STORE_LANES): New internal =
functions.
> =A0 =A0 =A0 =A0* internal-fn.c (get_multi_vector_move, expand_LOAD_LANES)
> =A0 =A0 =A0 =A0(expand_STORE_LANES): New functions.
> =A0 =A0 =A0 =A0* tree.h (build_simple_array_type): Declare.
> =A0 =A0 =A0 =A0* tree.c (build_simple_array_type): New function.
> =A0 =A0 =A0 =A0* tree-vectorizer.h (vect_model_store_cost): Add a bool ar=
gument.
> =A0 =A0 =A0 =A0(vect_model_load_cost): Likewise.
> =A0 =A0 =A0 =A0(vect_store_lanes_supported, vect_load_lanes_supported)
> =A0 =A0 =A0 =A0(vect_record_strided_load_vectors): Declare.
> =A0 =A0 =A0 =A0* tree-vect-data-refs.c (vect_lanes_optab_supported_p)
> =A0 =A0 =A0 =A0(vect_store_lanes_supported, vect_load_lanes_supported): N=
ew functions.
> =A0 =A0 =A0 =A0(vect_transform_strided_load): Split out statement recordi=
ng into...
> =A0 =A0 =A0 =A0(vect_record_strided_load_vectors): ...this new function.
> =A0 =A0 =A0 =A0* tree-vect-stmts.c (create_vector_array, read_vector_arra=
y)
> =A0 =A0 =A0 =A0(write_vector_array, create_array_ref): New functions.
> =A0 =A0 =A0 =A0(vect_model_store_cost): Add store_lanes_p argument.
> =A0 =A0 =A0 =A0(vect_model_load_cost): Add load_lanes_p argument.
> =A0 =A0 =A0 =A0(vectorizable_store): Try to use store-lanes functions for
> =A0 =A0 =A0 =A0interleaved stores.
> =A0 =A0 =A0 =A0(vectorizable_load): Likewise load-lanes and loads.
> =A0 =A0 =A0 =A0* tree-vect-slp.c (vect_get_and_check_slp_defs)
> =A0 =A0 =A0 =A0(vect_build_slp_tree):
>
> Index: gcc/doc/md.texi
> =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
> --- gcc/doc/md.texi =A0 =A0 2011-04-12 12:16:46.000000000 +0100
> +++ gcc/doc/md.texi =A0 =A0 2011-04-12 14:48:28.000000000 +0100
> @@ -3846,6 +3846,48 @@ into consecutive memory locations. =A0Oper
> =A0consecutive memory locations, operand 1 is the first register, and
> =A0operand 2 is a constant: the number of consecutive registers.
>
> +@cindex @code{vec_load_lanes@var{m}@var{n}} instruction pattern
> +@item @samp{vec_load_lanes@var{m}@var{n}}
> +Perform an interleaved load of several vectors from memory operand 1
> +into register operand 0. =A0Both operands have mode @var{m}. =A0The regi=
ster
> +operand is viewed as holding consecutive vectors of mode @var{n},
> +while the memory operand is a flat array that contains the same number
> +of elements. =A0The operation is equivalent to:
> +
> +@smallexample
> +int c =3D GET_MODE_SIZE (@var{m}) / GET_MODE_SIZE (@var{n});
> +for (j =3D 0; j < GET_MODE_NUNITS (@var{n}); j++)
> + =A0for (i =3D 0; i < c; i++)
> + =A0 =A0operand0[i][j] =3D operand1[j * c + i];
> +@end smallexample
> +
> +For example, @samp{vec_load_lanestiv4hi} loads 8 16-bit values
> +from memory into a register of mode @samp{TI}@. =A0The register
> +contains two consecutive vectors of mode @samp{V4HI}@.

So vec_load_lanestiv2qi would load ... ?  c =3D=3D 8 here.  Intuitively
such operation would have adjacent blocks of siv2qi memory.  But
maybe you want to constrain the mode size to GET_MODE_SIZE (@var{n})
* GET_MODE_NUNITS (@var{n})?  In which case the mode m is
redundant?  You could specify that we load NUNITS adjacent vectors into
an integer mode of appropriate size.

> +This pattern can only be used if:
> +@smallexample
> +TARGET_ARRAY_MODE_SUPPORTED_P (@var{n}, @var{c})
> +@end smallexample
> +is true. =A0GCC assumes that, if a target supports this kind of
> +instruction for some mode @var{n}, it also supports unaligned
> +loads for vectors of mode @var{n}.
> +
> +@cindex @code{vec_store_lanes@var{m}@var{n}} instruction pattern
> +@item @samp{vec_store_lanes@var{m}@var{n}}
> +Equivalent to @samp{vec_load_lanes@var{m}@var{n}}, with the memory
> +and register operands reversed. =A0That is, the instruction is
> +equivalent to:
> +
> +@smallexample
> +int c =3D GET_MODE_SIZE (@var{m}) / GET_MODE_SIZE (@var{n});
> +for (j =3D 0; j < GET_MODE_NUNITS (@var{n}); j++)
> + =A0for (i =3D 0; i < c; i++)
> + =A0 =A0operand0[j * c + i] =3D operand1[i][j];
> +@end smallexample
> +
> +for a memory operand 0 and register operand 1.
> +
> =A0@cindex @code{vec_set@var{m}} instruction pattern
> =A0@item @samp{vec_set@var{m}}
> =A0Set given field in the vector value. =A0Operand 0 is the vector to mod=
ify,
> Index: gcc/optabs.h
> =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
> --- gcc/optabs.h =A0 =A0 =A0 =A02011-04-12 12:16:46.000000000 +0100
> +++ gcc/optabs.h =A0 =A0 =A0 =A02011-04-12 14:48:28.000000000 +0100
> @@ -578,6 +578,9 @@ enum convert_optab_index
> =A0 COI_satfract,
> =A0 COI_satfractuns,
>
> + =A0COI_vec_load_lanes,
> + =A0COI_vec_store_lanes,
> +

Um, they are not really conversion optabs.  Any reason they can't
use the direct_optab table and path?  What are the two modes
usually?  I don't see how you specify the kind of permutation that
is performed on the load - so, why not go the targetm.expand_builtin
path instead (well, targetm.expand_internal_fn, of course - or rather
targetm.expand_gimple_call which we need anyway for expanding
directly from gimple calls at some point).

> =A0 COI_MAX
> =A0};
>
> @@ -598,6 +601,8 @@ #define fract_optab (&convert_optab_tabl
> =A0#define fractuns_optab (&convert_optab_table[COI_fractuns])
> =A0#define satfract_optab (&convert_optab_table[COI_satfract])
> =A0#define satfractuns_optab (&convert_optab_table[COI_satfractuns])
> +#define vec_load_lanes_optab (&convert_optab_table[COI_vec_load_lanes])
> +#define vec_store_lanes_optab (&convert_optab_table[COI_vec_store_lanes])
>
> =A0/* Contains the optab used for each rtx code. =A0*/
> =A0extern optab code_to_optab[NUM_RTX_CODE + 1];
> Index: gcc/genopinit.c
> =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
> --- gcc/genopinit.c =A0 =A0 2011-04-12 12:16:46.000000000 +0100
> +++ gcc/genopinit.c =A0 =A0 2011-04-12 14:48:28.000000000 +0100
> @@ -74,6 +74,8 @@ static const char * const optabs[] =3D
> =A0 "set_convert_optab_handler (fractuns_optab, $B, $A, CODE_FOR_$(fractu=
ns$Q$a$I$b2$))",
> =A0 "set_convert_optab_handler (satfract_optab, $B, $A, CODE_FOR_$(satfra=
ct$a$Q$b2$))",
> =A0 "set_convert_optab_handler (satfractuns_optab, $B, $A, CODE_FOR_$(sat=
fractuns$I$a$Q$b2$))",
> + =A0"set_convert_optab_handler (vec_load_lanes_optab, $A, $B, CODE_FOR_$=
(vec_load_lanes$a$b$))",
> + =A0"set_convert_optab_handler (vec_store_lanes_optab, $A, $B, CODE_FOR_=
$(vec_store_lanes$a$b$))",
> =A0 "set_optab_handler (add_optab, $A, CODE_FOR_$(add$P$a3$))",
> =A0 "set_optab_handler (addv_optab, $A, CODE_FOR_$(add$F$a3$)),\n\
> =A0 =A0 set_optab_handler (add_optab, $A, CODE_FOR_$(add$F$a3$))",
> Index: gcc/internal-fn.def
> =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
> --- gcc/internal-fn.def 2011-04-12 14:10:42.000000000 +0100
> +++ gcc/internal-fn.def 2011-04-12 14:48:28.000000000 +0100
> @@ -32,3 +32,6 @@ along with GCC; see the file COPYING3.
>
> =A0 =A0where NAME is the name of the function and FLAGS is a set of
> =A0 =A0ECF_* flags. =A0*/
> +
> +DEF_INTERNAL_FN (LOAD_LANES, ECF_CONST | ECF_LEAF)
> +DEF_INTERNAL_FN (STORE_LANES, ECF_CONST | ECF_LEAF)
> Index: gcc/internal-fn.c
> =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
> --- gcc/internal-fn.c =A0 2011-04-12 14:10:42.000000000 +0100
> +++ gcc/internal-fn.c =A0 2011-04-12 14:48:28.000000000 +0100
> @@ -41,6 +41,69 @@ #define DEF_INTERNAL_FN(CODE, FLAGS) FLA
> =A0 0
> =A0};
>
> +/* ARRAY_TYPE is an array of vector modes. =A0Return the associated insn
> + =A0 for load-lanes-style optab OPTAB. =A0The insn must exist. =A0*/
> +
> +static enum insn_code
> +get_multi_vector_move (tree array_type, convert_optab optab)
> +{
> + =A0enum insn_code icode;
> + =A0enum machine_mode imode;
> + =A0enum machine_mode vmode;
> +
> + =A0gcc_assert (TREE_CODE (array_type) =3D=3D ARRAY_TYPE);
> + =A0imode =3D TYPE_MODE (array_type);
> + =A0vmode =3D TYPE_MODE (TREE_TYPE (array_type));
> +
> + =A0icode =3D convert_optab_handler (optab, imode, vmode);
> + =A0gcc_assert (icode !=3D CODE_FOR_nothing);
> + =A0return icode;
> +}
> +
> +/* Expand: LHS =3D LOAD_LANES (ARGS[0]). =A0*/
> +
> +static void
> +expand_LOAD_LANES (tree lhs, tree *args)
> +{
> + =A0struct expand_operand ops[2];
> + =A0tree type;
> + =A0rtx target, mem;
> +
> + =A0type =3D TREE_TYPE (lhs);
> +
> + =A0target =3D expand_expr (lhs, NULL_RTX, VOIDmode, EXPAND_WRITE);
> + =A0mem =3D expand_normal (args[0]);
> +
> + =A0gcc_assert (MEM_P (mem));
> + =A0PUT_MODE (mem, TYPE_MODE (type));
> +
> + =A0create_output_operand (&ops[0], target, TYPE_MODE (type));
> + =A0create_fixed_operand (&ops[1], mem);
> + =A0expand_insn (get_multi_vector_move (type, vec_load_lanes_optab), 2, =
ops);
> +}
> +
> +/* Expand: LHS =3D STORE_LANES (ARGS[0]). =A0*/
> +
> +static void
> +expand_STORE_LANES (tree lhs, tree *args)
> +{
> + =A0struct expand_operand ops[2];
> + =A0tree type;
> + =A0rtx target, rhs;
> +
> + =A0type =3D TREE_TYPE (args[0]);
> +
> + =A0target =3D expand_expr (lhs, NULL_RTX, VOIDmode, EXPAND_WRITE);
> + =A0rhs =3D expand_normal (args[0]);
> +
> + =A0gcc_assert (MEM_P (target));
> + =A0PUT_MODE (target, TYPE_MODE (type));
> +
> + =A0create_fixed_operand (&ops[0], target);
> + =A0create_input_operand (&ops[1], rhs, TYPE_MODE (type));
> + =A0expand_insn (get_multi_vector_move (type, vec_store_lanes_optab), 2,=
 ops);
> +}
> +
> =A0/* Routines to expand each internal function, indexed by function numb=
er.
> =A0 =A0Each routine has the prototype:
>
> Index: gcc/tree.h
> =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
> --- gcc/tree.h =A02011-04-12 12:16:46.000000000 +0100
> +++ gcc/tree.h =A02011-04-12 14:48:28.000000000 +0100
> @@ -4198,6 +4198,7 @@ extern tree build_type_no_quals (tree);
> =A0extern tree build_index_type (tree);
> =A0extern tree build_array_type (tree, tree);
> =A0extern tree build_nonshared_array_type (tree, tree);
> +extern tree build_simple_array_type (tree, unsigned HOST_WIDE_INT);
> =A0extern tree build_function_type (tree, tree);
> =A0extern tree build_function_type_list (tree, ...);
> =A0extern tree build_function_type_skip_args (tree, bitmap);
> Index: gcc/tree.c
> =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
> --- gcc/tree.c =A02011-04-12 12:16:46.000000000 +0100
> +++ gcc/tree.c =A02011-04-12 14:48:28.000000000 +0100
> @@ -7385,6 +7385,15 @@ build_nonshared_array_type (tree elt_typ
> =A0 return build_array_type_1 (elt_type, index_type, false);
> =A0}
>
> +/* Return a representation of ELT_TYPE[NELTS], using indices of type
> + =A0 sizetype. =A0*/
> +
> +tree
> +build_simple_array_type (tree elt_type, unsigned HOST_WIDE_INT nelts)

build_array_type_nelts

The rest looks ok to me.

Richard.