Commit 4b956d8b by Arnaud Charlet

[multiple changes]

2009-07-23  Olivier Hainque  <hainque@adacore.com>

	* g-ssinty.ads: New unit. GNAT.SSE.Internal_Types. Factorize
	low level internal type definitions for distinct higher level
	binding development activities (user type definitions and
	operations).
	* gnat_rm.texi: Document it.
	* g-ssvety.ads: Use it.
	* gcc-interface/Makefile.in: (x86 32/64 linux, cygwin32 sections): Add
	g-ssinty.o to EXTRA_GNATRTL_NONTASKING_OBJS.
	* gcc-interface/utils.c (gnat_internal_attribute_table): Add entry
	for the "may_alias" attribute.

2009-07-23  Thomas Quinot  <quinot@adacore.com>

	* scos.ads: Minor typo fix
	* gcc-interface/decl.c (validate_alignment): For the case of an
	implicit array base type, look for alignment clause on first subtype.
	Code clean up.

2009-07-23  Ed Schonberg  <schonberg@adacore.com>

	* sem.adb (Walk_Library_Units): Handle properly the case where a unit
	in the context depends on the spec of the main unit, by delaying
	processing of the main unit body until all other units have been
	processed.

From-SVN: r149993
parent f8c6086b
2009-07-23 Olivier Hainque <hainque@adacore.com>
* g-ssinty.ads: New unit. GNAT.SSE.Internal_Types. Factorize
low level internal type definitions for distinct higher level
binding development activities (user type definitions and
operations).
* gnat_rm.texi: Document it.
* g-ssvety.ads: Use it.
* gcc-interface/Makefile.in: (x86 32/64 linux, cygwin32 sections): Add
g-ssinty.o to EXTRA_GNATRTL_NONTASKING_OBJS.
* gcc-interface/utils.c (gnat_internal_attribute_table): Add entry
for the "may_alias" attribute.
2009-07-23 Thomas Quinot <quinot@adacore.com>
* scos.ads: Minor typo fix
* gcc-interface/decl.c (validate_alignment): For the case of an
implicit array base type, look for alignment clause on first subtype.
Code clean up.
2009-07-23 Ed Schonberg <schonberg@adacore.com>
* sem.adb (Walk_Library_Units): Handle properly the case where a unit
in the context depends on the spec of the main unit, by delaying
processing of the main unit body until all other units have been
processed.
2009-07-23 Arnaud Charlet <charlet@adacore.com>
* a-convec.adb: Add comments about suspicious/subtle code.
......
------------------------------------------------------------------------------
-- --
-- GNAT COMPILER COMPONENTS --
-- --
-- G N A T . S S E . I N T E R N A L _ T Y P E S --
-- --
-- S p e c --
-- --
-- Copyright (C) 2009, Free Software Foundation, Inc. --
-- --
-- GNAT is free software; you can redistribute it and/or modify it under --
-- terms of the GNU General Public License as published by the Free Soft- --
-- ware Foundation; either version 3, or (at your option) any later ver- --
-- sion. GNAT is distributed in the hope that it will be useful, but WITH- --
-- OUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY --
-- or FITNESS FOR A PARTICULAR PURPOSE. --
-- --
-- As a special exception under Section 7 of GPL version 3, you are granted --
-- additional permissions described in the GCC Runtime Library Exception, --
-- version 3.1, as published by the Free Software Foundation. --
-- --
-- You should have received a copy of the GNU General Public License and --
-- a copy of the GCC Runtime Library Exception along with this program; --
-- see the files COPYING3 and COPYING.RUNTIME respectively. If not, see --
-- <http://www.gnu.org/licenses/>. --
-- --
-- GNAT was originally developed by the GNAT team at New York University. --
-- Extensive contributions were provided by Ada Core Technologies Inc. --
-- --
------------------------------------------------------------------------------
-- This unit exposes low level types to interface with the GCC vector
-- builtins directly. These are useful for the development of higher level
-- bindings to the reference Intel intrinsic operations.
-- See GNAT.SSE for the list of targets where this facility is supported.
package GNAT.SSE.Internal_Types is
type v4sf is private;
type v2df is private;
type v2di is private;
private
-- GCC'wise, vector operations operate on objects of vector modes,
-- conveyed through vector types obtained in C by setting an attribute on
-- what looks like a component typedef. For example, in xmmintrin.h:
--
-- typedef float __v4sf __attribute__ ((__vector_size__ (16)));
-- Applying a 'vector_size' machine attribute in Ada, as in
--
-- type Vf is new Float;
-- pragma Machine_Attribute (Vf, "vector_size", 16);
--
-- makes Vf a 16bytes long V4SFmode GCC type but the effect on the type
-- layout is not conveyed to the front-end. The latter still sees "Vf"
-- as a 4bytes long single float, with numerous potential pitfalls.
-- We devised a 'vector_type' alternate machine attribute, which applies
-- to array types of the proper size and alignment from the front-end
-- perspective:
type v4sf is array (1 .. 4) of GNAT.SSE.Float32;
for v4sf'Alignment use GNAT.SSE.VECTOR_ALIGN;
pragma Machine_Attribute (v4sf, "vector_type");
type v2di is array (1 .. 2) of GNAT.SSE.Integer64;
for v2di'Alignment use GNAT.SSE.VECTOR_ALIGN;
pragma Machine_Attribute (v2di, "vector_type");
type v2df is array (1 .. 2) of GNAT.SSE.Float64;
for v2df'Alignment use GNAT.SSE.VECTOR_ALIGN;
pragma Machine_Attribute (v2df, "vector_type");
end GNAT.SSE.Internal_Types;
......@@ -30,7 +30,9 @@
------------------------------------------------------------------------------
-- This unit exposes the Ada __m128 like data types to represent the contents
-- of SSE registers, for use by the SSE intrinsics.
-- of SSE registers, for use by bindings to the SSE intrinsic operations.
-- See GNAT.SSE for the list of targets where this facility is supported.
package GNAT.SSE.Vector_Types is
......@@ -49,32 +51,23 @@ package GNAT.SSE.Vector_Types is
-- * Use new data types only with the respective intrinsics described
-- in this documentation. >>
type M128 is private; -- SSE >= 1
type M128d is private; -- SSE >= 2
type M128i is private; -- SSE >= 2
type m128 is private; -- SSE >= 1
type m128d is private; -- SSE >= 2
type m128i is private; -- SSE >= 2
private
-- GCC'wise, vector operations operate on objects of vector modes,
-- conveyed through vector types obtained by setting an attribute on what
-- looks like a component typedef. For example, in C (xmmintrin.h):
--
-- typedef float __v4sf __attribute__ ((__vector_size__ (16)));
-- We can obtain the same low level GCC effect in Ada with
-- Machine_Attribute pragmas, as in
-- Each of the m128 types maps to a specific vector_type with
-- an extra "may_alias" attribute as in GCC's definitions for C,
-- for instance in xmmintrin.h:
--
-- type Vf is new Float;
-- pragma Machine_Attribute (Vf, "vector_size", 16);
-- /* The Intel API is flexible enough that we must allow aliasing
-- with other vector types, and their scalar components. */
-- typedef float __m128
-- __attribute__ ((__vector_size__ (16), __may_alias__));
--
-- which makes Vf a 16bytes long V4SFmode type for GCC. The effect on the
-- type layout is not conveyed to the front-end, however, so the latter
-- still sees "Vf" as a 4bytes long single float. This leads to numerous
-- potential pitfalls if this type is directly exposed user land, so we
-- add wrapper records with rep clauses to compensate.
-- The wrapper records all have a single component of the twisted low
-- level type, so they inherit the mode while the rep clauses convey the
-- size and alignment information to the front-end.
-- /* Internal data types for implementing the intrinsics. */
-- typedef float __v4sf __attribute__ ((__vector_size__ (16)));
------------
-- M128 --
......@@ -82,44 +75,32 @@ private
-- << The __m128 data type can hold four 32-bit floating-point values. >>
type V4sf is new Float32;
pragma Machine_Attribute (V4sf, "vector_size", VECTOR_BYTES);
type M128 is record
Value : V4sf;
end record;
for M128'Size use VECTOR_BYTES * 8;
for M128'Alignment use VECTOR_ALIGN;
type m128 is array (1 .. 4) of Float32;
for m128'Alignment use VECTOR_ALIGN;
pragma Machine_Attribute (m128, "vector_type");
pragma Machine_Attribute (m128, "may_alias");
-------------
-- M128d --
-- m128d --
-------------
-- << The __m128d data type can hold two 64-bit floating-point values. >>
type V2df is new Float64;
pragma Machine_Attribute (V2df, "vector_size", VECTOR_BYTES);
type M128d is record
Value : V2df;
end record;
for M128d'Size use VECTOR_BYTES * 8;
for M128d'Alignment use VECTOR_ALIGN;
type m128d is array (1 .. 2) of Float64;
for m128d'Alignment use VECTOR_ALIGN;
pragma Machine_Attribute (m128d, "vector_type");
pragma Machine_Attribute (m128d, "may_alias");
-------------
-- M128i --
-- m128i --
-------------
-- << The __m128i data type can hold sixteen 8-bit, eight 16-bit, four
-- 32-bit, or two 64-bit integer values. >>
type V2di is new Integer64;
pragma Machine_Attribute (V2di, "vector_size", VECTOR_BYTES);
type M128i is record
Value : V2di;
end record;
for M128i'Size use VECTOR_BYTES * 8;
for M128i'Alignment use VECTOR_ALIGN;
type m128i is array (1 .. 2) of Integer64;
for m128i'Alignment use VECTOR_ALIGN;
pragma Machine_Attribute (m128i, "vector_type");
pragma Machine_Attribute (m128i, "may_alias");
end GNAT.SSE.Vector_Types;
......@@ -1052,7 +1052,7 @@ ifeq ($(strip $(filter-out %86 linux%,$(arch) $(osys))),)
endif
THREADSLIB = -lpthread
EXTRA_GNATRTL_NONTASKING_OBJS=g-sse.o g-ssvety.o
EXTRA_GNATRTL_NONTASKING_OBJS=g-sse.o g-ssvety.o g-ssinty.o
EXTRA_GNATRTL_TASKING_OBJS=s-linux.o
endif
......@@ -1593,7 +1593,7 @@ ifeq ($(strip $(filter-out cygwin32% mingw32% pe,$(osys))),)
endif
EXTRA_GNATRTL_NONTASKING_OBJS = \
s-win32.o s-winext.o g-regist.o g-sse.o g-ssvety.o
s-win32.o s-winext.o g-regist.o g-sse.o g-ssvety.o g-ssinty.o
EXTRA_GNATRTL_TASKING_OBJS = a-exetim.o
MISCLIB = -lws2_32
......@@ -2003,7 +2003,7 @@ ifeq ($(strip $(filter-out %x86_64 linux%,$(arch) $(osys))),)
mlib-tgt-specific.adb<mlib-tgt-specific-linux.adb \
indepsw.adb<indepsw-gnu.adb
EXTRA_GNATRTL_NONTASKING_OBJS=g-sse.o g-ssvety.o
EXTRA_GNATRTL_NONTASKING_OBJS=g-sse.o g-ssvety.o g-ssinty.o
EXTRA_GNATRTL_TASKING_OBJS=s-linux.o
EH_MECHANISM=-gcc
THREADSLIB=-lpthread
......
......@@ -122,6 +122,7 @@ const struct attribute_spec gnat_internal_attribute_table[] =
{ "type generic", 0, 0, false, true, true, handle_type_generic_attribute },
{ "vector_size", 1, 1, false, true, false, handle_vector_size_attribute },
{ "may_alias", 0, 0, false, true, false, NULL },
/* ??? format and format_arg are heavy and not supported, which actually
prevents support for stdio builtins, which we however declare as part
......
......@@ -382,6 +382,7 @@ The GNAT Library
* GNAT.Spitbol.Table_Integer (g-sptain.ads)::
* GNAT.Spitbol.Table_VString (g-sptavs.ads)::
* GNAT.SSE (g-sse.ads)::
* GNAT.SSE.Internal_Types (g-ssinty.ads)::
* GNAT.SSE.Vector_Types (g-ssvety.ads)::
* GNAT.Strings (g-string.ads)::
* GNAT.String_Split (g-strspl.ads)::
......@@ -13571,6 +13572,7 @@ of GNAT, and will generate a warning message.
* GNAT.Spitbol.Table_Integer (g-sptain.ads)::
* GNAT.Spitbol.Table_VString (g-sptavs.ads)::
* GNAT.SSE (g-sse.ads)::
* GNAT.SSE.Internal_Types (g-ssinty.ads)::
* GNAT.SSE.Vector_Types (g-ssvety.ads)::
* GNAT.Strings (g-string.ads)::
* GNAT.String_Split (g-strspl.ads)::
......@@ -14641,6 +14643,15 @@ the Intel(r) Streaming SIMD Extensions with GNAT on the x86 family of
targets. It exposes vector component types together with a general
introduction to the binding contents and use.
@node GNAT.SSE.Internal_Types (g-ssinty.ads)
@section @code{GNAT.SSE.Internal_Types} (@file{g-ssinty.ads})
@cindex @code{GNAT.SSE.Internal_Types} (@file{g-ssinty.ads})
@noindent
Low level GCC vector types for direct use of the vector related
builtins, required for the development of higher level bindings to SSE
intrinsic operations.
@node GNAT.SSE.Vector_Types (g-ssvety.ads)
@section @code{GNAT.SSE.Vector_Types} (@file{g-ssvety.ads})
@cindex @code{GNAT.SSE.Vector_Types} (@file{g-ssvety.ads})
......
......@@ -103,7 +103,7 @@ package SCOs is
-- Statement lines
-- These lines correspond to a sequence of one or more statements which
-- are always exeecuted in sequence, The first statement may be an entry
-- are always executed in sequence, The first statement may be an entry
-- point (e.g. statement after a label), and the last statement may be
-- an exit point (e.g. an exit statement), but no other entry or exit
-- points may occur within the sequence of statements. The idea is that
......
......@@ -107,7 +107,6 @@ package body Sem is
procedure Analyze (N : Node_Id) is
begin
Debug_A_Entry ("analyzing ", N);
-- Immediate return if already analyzed
if Analyzed (N) then
......@@ -1510,6 +1509,12 @@ package body Sem is
-- after we have fully processed X, and is used only for debugging
-- printouts and assertions.
Do_Main : Boolean := False;
-- Flag to delay processing the main body until after all other units.
-- This is needed because the spec of the main unit may appear in the
-- context of some other unit. We do not want this to force processing
-- of the main body before all other units have been processed.
procedure Do_Action (CU : Node_Id; Item : Node_Id);
-- Calls Action, with some validity checks
......@@ -1712,7 +1717,8 @@ package body Sem is
if not Nkind_In (Item, N_Package_Body, N_Subprogram_Body)
or else Acts_As_Spec (CU)
or else CU = Cunit (Main_Unit)
or else (CU = Cunit (Main_Unit) and then Do_Main)
then
Do_Action (CU, Item);
......@@ -1733,14 +1739,47 @@ package body Sem is
-- be possible to restrict the list to those bodies that are used
-- in the main unit. Possible optimization ???
-- Such bodies can also appear in a circular dependency list, where
-- spec A depends on spec B and the body of B depends on spec A.
-- This is not an elaboration issue, but body B must be excluded
-- from the processing.
if Nkind (Item) = N_Package_Declaration then
declare
Body_Unit : constant Node_Id := Library_Unit (CU);
function Circular_Dependence (B : Node_Id) return Boolean;
-- Check whether this body depends on a spec that is pending,
-- that is to say has been seen but not processed yet.
function Circular_Dependence (B : Node_Id) return Boolean is
Item : Node_Id;
UN : Unit_Number_Type;
begin
Item := First (Context_Items (B));
while Present (Item) loop
if Nkind (Item) = N_With_Clause then
UN := Get_Cunit_Unit_Number (Library_Unit (Item));
if Seen (UN)
and then not Done (UN)
then
return True;
end if;
end if;
Next (Item);
end loop;
return False;
end Circular_Dependence;
begin
if Present (Body_Unit)
and then Body_Unit /= Cunit (Main_Unit)
and then Unit_Num /= Get_Source_Unit (System_Aux_Id)
and then not Circular_Dependence (Body_Unit)
then
Do_Unit_And_Dependents (Body_Unit, Unit (Body_Unit));
Do_Action (Body_Unit, Unit (Body_Unit));
......@@ -1801,16 +1840,13 @@ package body Sem is
case Nkind (N) is
-- If it's a body, then ignore it, unless it's the main unit
-- Otherwise bodies appear in the list because of inlining or
-- instantiations, and they are processed immediately after
-- the corresponding specs.
-- If it's a body, ignore it. Bodies appear in the list only
-- because of inlining/instantiations, and they are processed
-- immediately after the corresponding specs.
-- The main unit is processed separately after all other units.
when N_Package_Body | N_Subprogram_Body =>
if CU = Cunit (Main_Unit) then
Do_Unit_And_Dependents (CU, N);
end if;
null;
-- It's a spec, so just do it
......@@ -1822,6 +1858,11 @@ package body Sem is
Next_Elmt (Cur);
end loop;
if not Done (Main_Unit) then
Do_Main := True;
Do_Unit_And_Dependents (Cunit (Main_Unit), Unit (Cunit (Main_Unit)));
end if;
if Debug_Unit_Walk then
if Done /= (Done'Range => True) then
Write_Eol;
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment