1 ------------------------------------------------------------------------------
3 -- GNAT COMPILER COMPONENTS --
9 -- Copyright (C) 1992-2013, Free Software Foundation, Inc. --
11 -- GNAT is free software; you can redistribute it and/or modify it under --
12 -- terms of the GNU General Public License as published by the Free Soft- --
13 -- ware Foundation; either version 3, or (at your option) any later ver- --
14 -- sion. GNAT is distributed in the hope that it will be useful, but WITH- --
15 -- OUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY --
16 -- or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License --
17 -- for more details. You should have received a copy of the GNU General --
18 -- Public License distributed with GNAT; see file COPYING3. If not, go to --
19 -- http://www.gnu.org/licenses for a complete copy of the license. --
21 -- GNAT was originally developed by the GNAT team at New York University. --
22 -- Extensive contributions were provided by Ada Core Technologies Inc. --
24 ------------------------------------------------------------------------------
26 -- This package contains constants describing target properties
28 with Types
; use Types
;
34 ------------------------------
35 -- Host/Target Dependencies --
36 ------------------------------
38 -- It is vital to maintain a clear distinction between properties of
39 -- types on the host and types on the target, since in the general
40 -- case of a cross-compiler these will be different.
42 -- This package provides definitions of values that describe the properties
43 -- of the target types. All instances of target dependencies, including the
44 -- definitions of such packages as Standard and System depend directly or
45 -- indirectly on the definitions in the Ttypes packages.
47 -- In the source of the compiler, references to attributes such as
48 -- Integer'Size will give information regarding the host types (i.e.
49 -- the types within the compiler itself). Such references are therefore
50 -- almost always suspicious (it is hard for example to see that the
51 -- code in the compiler should even be using type Integer very much,
52 -- and certainly this code should not depend on the size of Integer).
54 -- On the other hand, it is perfectly reasonable for the compiler to
55 -- require access to the size of type Integer for the target machine,
56 -- e.g. in constructing the internal representation of package Standard.
57 -- For this purpose, instead of referencing the attribute Integer'Size,
58 -- a reference to Ttypes.Standard_Integer_Size will provide the needed
59 -- value for the target type.
61 -- Two approaches are used for handling target dependent values in the
62 -- standard library packages. Package Standard is handled specially,
63 -- being constructed internally (by package Stand). Target dependent
64 -- values needed in Stand are obtained by direct reference to Ttypes
67 -- For package System, the required constant values are obtained by
68 -- referencing appropriate attributes. Ada 95 already defines most of
69 -- the required attributes, and GNAT specific attributes have been
70 -- defined to cover the remaining cases (such as Storage_Unit). The
71 -- evaluation of these attributes obtains the required target dependent
72 -- values from Ttypes and Ttypef. The additional attributes that have
73 -- been added to GNAT (Address_Size, Storage_Unit, Word_Size, Max_Priority,
74 -- and Max_Interrupt_Priority) are for almost all purposes redundant with
75 -- respect to the corresponding references to System constants. For example
76 -- in a program, System.Address_Size and Standard'Address_Size yield the
77 -- same value. The critical use of the attribute is in writing the System
78 -- declaration of Address_Size which of course cannot refer to itself. By
79 -- this means we achieve complete target independence in the source code
80 -- of package System, i.e. there is only one copy of the source of System
83 -- Note that during compilation there are two versions of package System
84 -- around. The version that is directly with'ed by compiler packages
85 -- contains host-dependent definitions, which is what is needed in that
86 -- case (for example, System.Storage_Unit referenced in the source of the
87 -- compiler refers to the storage unit of the host, not the target). This
88 -- means that, like attribute references, any references to constants in
89 -- package System in the compiler code are suspicious, since it is strange
90 -- for the compiler to have such host dependencies. If the compiler needs
91 -- to access the target dependent values of such quantities as Storage_Unit
92 -- then it should reference the constants in this package (Ttypes), rather
93 -- than referencing System.Storage_Unit, or Standard'Storage_Unit, both of
94 -- which would yield the host value.
96 ---------------------------------------------------
97 -- Target-Dependent Values for Types in Standard --
98 ---------------------------------------------------
100 -- Note: GNAT always supplies all the following integer and float types,
101 -- but depending on the machine, some of the types may be identical. For
102 -- example, on some machines, Short_Float may be the same as Float, and
103 -- Long_Long_Float may be the same as Long_Float.
105 Standard_Short_Short_Integer_Size
: constant Pos
:=
107 Standard_Short_Short_Integer_Width
: constant Pos
:=
108 Get_Targ
.Width_From_Size
109 (Standard_Short_Short_Integer_Size
);
111 Standard_Short_Integer_Size
: constant Pos
:=
113 Standard_Short_Integer_Width
: constant Pos
:=
114 Get_Targ
.Width_From_Size
115 (Standard_Short_Integer_Size
);
117 Standard_Integer_Size
: constant Pos
:=
119 Standard_Integer_Width
: constant Pos
:=
120 Get_Targ
.Width_From_Size
121 (Standard_Integer_Size
);
123 Standard_Long_Integer_Size
: constant Pos
:=
125 Standard_Long_Integer_Width
: constant Pos
:=
126 Get_Targ
.Width_From_Size
127 (Standard_Long_Integer_Size
);
129 Standard_Long_Long_Integer_Size
: constant Pos
:=
130 Set_Targ
.Long_Long_Size
;
131 Standard_Long_Long_Integer_Width
: constant Pos
:=
132 Get_Targ
.Width_From_Size
133 (Standard_Long_Long_Integer_Size
);
135 Standard_Short_Float_Size
: constant Pos
:=
137 Standard_Short_Float_Digits
: constant Pos
:=
138 Get_Targ
.Digits_From_Size
139 (Standard_Short_Float_Size
);
141 Standard_Float_Size
: constant Pos
:=
143 Standard_Float_Digits
: constant Pos
:=
144 Get_Targ
.Digits_From_Size
145 (Standard_Float_Size
);
147 Standard_Long_Float_Size
: constant Pos
:=
148 Set_Targ
.Double_Size
;
149 Standard_Long_Float_Digits
: constant Pos
:=
150 Get_Targ
.Digits_From_Size
151 (Standard_Long_Float_Size
);
153 Standard_Long_Long_Float_Size
: constant Pos
:=
154 Set_Targ
.Long_Double_Size
;
155 Standard_Long_Long_Float_Digits
: constant Pos
:=
156 Get_Targ
.Digits_From_Size
157 (Standard_Long_Long_Float_Size
);
159 Standard_Character_Size
: constant Pos
:= Set_Targ
.Char_Size
;
161 Standard_Wide_Character_Size
: constant Pos
:= 16;
162 Standard_Wide_Wide_Character_Size
: constant Pos
:= 32;
163 -- Standard wide character sizes
165 -- Note: there is no specific control over the representation of
166 -- enumeration types. The convention used is that if an enumeration
167 -- type has fewer than 2**(Character'Size) elements, then the size
168 -- used is Character'Size, otherwise Integer'Size is used.
170 -- Similarly, the size of fixed-point types depends on the size of the
171 -- corresponding integer type, which is the smallest predefined integer
172 -- type capable of representing the required range of values.
174 -------------------------------------------------
175 -- Target-Dependent Values for Types in System --
176 -------------------------------------------------
178 System_Address_Size
: constant Pos
:= Set_Targ
.Pointer_Size
;
179 -- System.Address'Size (also size of all thin pointers)
181 System_Max_Binary_Modulus_Power
: constant Pos
:=
182 Standard_Long_Long_Integer_Size
;
184 System_Max_Nonbinary_Modulus_Power
: constant Pos
:= Standard_Integer_Size
;
186 System_Storage_Unit
: constant Pos
:= Set_Targ
.Bits_Per_Unit
;
187 System_Word_Size
: constant Pos
:= Set_Targ
.Bits_Per_Word
;
189 System_Tick_Nanoseconds
: constant Pos
:= 1_000_000_000
;
190 -- Value of System.Tick in nanoseconds. At the moment, this is a fixed
191 -- constant (with value of 1.0 seconds), but later we should add this
192 -- value to the GCC configuration file so that its value can be made
193 -- configuration dependent.
195 -----------------------------------------------------
196 -- Target-Dependent Values for Types in Interfaces --
197 -----------------------------------------------------
199 Interfaces_Wchar_T_Size
: constant Pos
:= Set_Targ
.Wchar_T_Size
;
201 ----------------------------------------
202 -- Other Target-Dependent Definitions --
203 ----------------------------------------
205 Maximum_Alignment
: constant Pos
:= Set_Targ
.Maximum_Alignment
;
206 -- The maximum alignment, in storage units, that an object or type may
207 -- require on the target machine.
209 System_Allocator_Alignment
: constant Pos
:=
210 Set_Targ
.System_Allocator_Alignment
;
211 -- The alignment in storage units of addresses returned by malloc
213 Max_Unaligned_Field
: constant Pos
:= Set_Targ
.Max_Unaligned_Field
;
214 -- The maximum supported size in bits for a field that is not aligned
215 -- on a storage unit boundary.
217 Bytes_Big_Endian
: Boolean := Set_Targ
.Bytes_BE
/= 0;
218 -- Important note: for Ada purposes, the important setting is the bytes
219 -- endianness (Bytes_Big_Endian), not the bits value (Bits_Big_Endian).
220 -- This is because Ada bit addressing must be compatible with the byte
221 -- ordering (otherwise we would end up with non-contiguous fields). It
222 -- is rare for the two to be different, but if they are, Bits_Big_Endian
223 -- is relevant only for the generation of instructions with bit numbers,
224 -- and thus relevant only to the back end. Note that this is a variable
225 -- rather than a constant, since it can be modified (flipped) by -gnatd8.
227 Target_Short_Enums
: constant Boolean := Set_Targ
.Short_Enums
/= 0;
228 -- True if we are in short enums mode, where foreign convention
229 -- (in particular C and C++) enumeration types will be sized as in Ada,
230 -- using the shortest possibility from 8,16,32 bits, signed or unsigned.
231 -- A zero value means Short_Enums are not in use, and in this case all
232 -- foreign convention enumeration types are given the same size as c int.
234 Target_Strict_Alignment
: Boolean :=
235 Set_Targ
.Strict_Alignment
/= 0;
236 -- True if instructions will fail if data is misaligned. Note that this
237 -- is a variable rather than a constant since it can be modified (set to
238 -- True) if the debug flag -gnatd.A is used.
240 Target_Double_Float_Alignment
: constant Nat
:=
241 Set_Targ
.Double_Float_Alignment
;
242 -- The default alignment of "double" floating-point types, i.e. floating
243 -- point types whose size is equal to 64 bits, or 0 if this alignment is
244 -- not lower than the largest power of 2 multiple of System.Storage_Unit
245 -- that does not exceed either the object size of the type or the maximum
246 -- allowed alignment.
248 Target_Double_Scalar_Alignment
: constant Nat
:=
249 Set_Targ
.Double_Scalar_Alignment
;
250 -- The default alignment of "double" or larger scalar types, i.e. scalar
251 -- types whose size is greater or equal to 64 bits, or 0 if this alignment
252 -- is not lower than the largest power of 2 multiple of System.Storage_Unit
253 -- that does not exceed either the object size of the type or the maximum
254 -- allowed alignment.