OpenE2K
/
gcc
2
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

[multiple changes] 

2010-09-10  Jose Ruiz  <ruiz@adacore.com>

	* exp_cg.adb (Is_Predefined_Dispatching_Operation): Add the "__" scope
	separator when trying the pattern matching to detect predefined
	primitive operations.

2010-09-10  Robert Dewar  <dewar@adacore.com>

	* bindgen.adb, atree.adb: Minor reformatting.

2010-09-10  Ben Brosgol  <brosgol@adacore.com>

	* ug_words, gnat_ugn.texi: Revised "Transitioning to 64-Bit GNAT for
	OpenVMS" section.

From-SVN: r164175
This commit is contained in:
Arnaud Charlet 2010-09-10 15:48:24 +02:00
parent 8c4353b7c5
commit 0856403616
6 changed files with 158 additions and 83 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

15
gcc/ada/ChangeLog
View File

@ -1,3 +1,18 @@
2010-09-10 Jose Ruiz <ruiz@adacore.com>
* exp_cg.adb (Is_Predefined_Dispatching_Operation): Add the "__" scope
separator when trying the pattern matching to detect predefined
primitive operations.
2010-09-10 Robert Dewar <dewar@adacore.com>
* bindgen.adb, atree.adb: Minor reformatting.
2010-09-10 Ben Brosgol <brosgol@adacore.com>
* ug_words, gnat_ugn.texi: Revised "Transitioning to 64-Bit GNAT for
OpenVMS" section.
2010-09-10 Doug Rupp <rupp@adacore.com>
* bindgen.adb: Minor comment fix for -H switch.

8
gcc/ada/atree.adb
View File

@ -1261,7 +1261,7 @@ package body Atree is
-- New_Node_Breakpoint --
-------------------------
procedure nn is -- New_Node_Breakpoint
procedure nn is
begin
Write_Str ("Watched node ");
Write_Int (Int (Watch_Node));
@ -1273,7 +1273,7 @@ package body Atree is
-- New_Node_Debugging_Output --
-------------------------------
procedure nnd (N : Node_Id) is -- New_Node_Debugging_Output
procedure nnd (N : Node_Id) is
Node_Is_Watched : constant Boolean := N = Watch_Node;
begin
@ -1666,7 +1666,7 @@ package body Atree is
-- Rewrite_Breakpoint --
-------------------------
procedure rr is -- Rewrite_Breakpoint
procedure rr is
begin
Write_Str ("Watched node ");
Write_Int (Int (Watch_Node));
@ -1678,7 +1678,7 @@ package body Atree is
-- Rewrite_Debugging_Output --
------------------------------
procedure rrd (Old_Node, New_Node : Node_Id) is -- Rewrite_Debugging_Output
procedure rrd (Old_Node, New_Node : Node_Id) is
Node_Is_Watched : constant Boolean := Old_Node = Watch_Node;
begin

5
gcc/ada/bindgen.adb
View File

@ -792,8 +792,9 @@ package body Bindgen is
Write_Statement_Buffer;
-- Generate call to Install_Handler
-- In .NET, when binding with -z, we don't install the signal
-- handler to let the caller handle the last exception handler.
-- In .NET, when binding with -z, we don't install the signal handler
-- to let the caller handle the last exception handler.
if VM_Target /= CLI_Target
or else Bind_Main_Program

21
gcc/ada/exp_cg.adb
View File

@ -271,12 +271,21 @@ package body Exp_CG is
for J in Predef_Names_95'Range loop
Get_Name_String (Predef_Names_95 (J));
if Full_Name'Last - Suffix_Length > Name_Len
-- The predefined primitive operations are identified by the
-- names "_size", "_alignment", etc. If we try a pattern
-- matching against this string, we can wrongly match other
-- primitive operations like "get_size". To avoid this, we
-- add the "__" scope separator, which can only prepend
-- predefined primitive operations because other primitive
-- operations can neither start with an underline nor
-- contain two consecutive underlines in its name.
if Full_Name'Last - Suffix_Length > Name_Len + 2
and then
Full_Name
(Full_Name'Last - Name_Len - Suffix_Length + 1
(Full_Name'Last - Name_Len - 2 - Suffix_Length + 1
.. Full_Name'Last - Suffix_Length) =
Name_Buffer (1 .. Name_Len)
"__" & Name_Buffer (1 .. Name_Len)
then
-- For the equality operator the type of the two operands
-- must also match.
@ -291,12 +300,12 @@ package body Exp_CG is
for J in Predef_Names_05'Range loop
Get_Name_String (Predef_Names_05 (J));
if Full_Name'Last - Suffix_Length > Name_Len
if Full_Name'Last - Suffix_Length > Name_Len + 2
and then
Full_Name
(Full_Name'Last - Name_Len - Suffix_Length + 1
(Full_Name'Last - Name_Len - 2 - Suffix_Length + 1
.. Full_Name'Last - Suffix_Length) =
Name_Buffer (1 .. Name_Len)
"__" & Name_Buffer (1 .. Name_Len)
then
return True;
end if;

190
gcc/ada/gnat_ugn.texi
View File

@ -8057,6 +8057,7 @@ be presented in subsequent sections.
* Binder Error Message Control::
* Elaboration Control::
* Output Control::
* Dynamic Allocation Control::
* Binding with Non-Ada Main Programs::
* Binding Programs with No Main Subprogram::
@end menu
@ -8168,6 +8169,17 @@ flag checks are generated.
@cindex @option{^-h^/HELP^} (@command{gnatbind})
Output usage (help) information
@item ^-H32^/32_MALLOC^
@cindex @option{^-H32^/32_MALLOC^} (@command{gnatbind})
Use 32-bit allocations for @code{__gnat_malloc} (and thus for access types).
For further details see @ref{Dynamic Allocation Control}.
@item ^-H64^/64_MALLOC^
@cindex @option{^-H32^/32_MALLOC^} (@command{gnatbind})
Use 64-bit allocations for @code{__gnat_malloc} (and thus for access types).
@cindex @code{__gnat_malloc}
For further details see @ref{Dynamic Allocation Control}.
@item ^-I^/SEARCH^
@cindex @option{^-I^/SEARCH^} (@command{gnatbind})
Specify directory to be searched for source and ALI files.
@ -8596,6 +8608,35 @@ be used to improve code generation in some cases.
@end table
@node Dynamic Allocation Control
@subsection Dynamic Allocation Control
@noindent
The heap control switches -- @option{-H32} and @option{-H64} --
determine whether dynamic allocation uses 32-bit or 64-bit memory.
They only affect compiler-generated allocations via @code{__gnat_malloc};
explicit calls to @code{malloc} and related functions from the C
run-time library are unaffected.
@table @option
@item -H32
Allocate memory on 32-bit heap
@item -H64
Allocate memory on 64-bit heap. This is the default
unless explicitly overridden by a @code{'Size} clause on the access type.
@end table
@ifset vms
@noindent
See also @ref{Access types and 32/64-bit allocation}.
@end ifset
@ifclear vms
@noindent
These switches are only effective on VMS platforms.
@end ifclear
@node Binding with Non-Ada Main Programs
@subsection Binding with Non-Ada Main Programs
@ -26509,10 +26550,11 @@ Such code will be referred to below as @emph{64-bit code}.
@menu
* Address types::
* Access types::
* Access types and 32/64-bit allocation::
* Unchecked conversions::
* Predefined constants::
* Interfacing with C::
* 32/64-bit descriptors::
* Experience with source compatibility::
@end menu
@ -26527,9 +26569,13 @@ approach has been taken:
@itemize @bullet
@item
@code{System.Address} always has a size of 64 bits
@cindex @code{System.Address} size
@cindex @code{Address} size
@item
@code{System.Short_Address} is a 32-bit subtype of @code{System.Address}
@cindex @code{System.Short_Address} size
@cindex @code{Short_Address} size
@end itemize
@noindent
@ -26568,31 +26614,64 @@ required in any code setting or accessing the field; the compiler will
automatically perform any needed conversions between address
formats.
@node Access types
@subsubsection Access types
@node Access types and 32/64-bit allocation
@subsubsection Access types and 32/64-bit allocation
@cindex 32-bit allocation
@cindex 64-bit allocation
@noindent
By default, objects designated by access values are always
allocated in the 32-bit
address space. Thus legacy code will never contain
any objects that are not addressable with 32-bit addresses, and
the compiler will never raise exceptions as result of mixing
32-bit and 64-bit addresses.
By default, objects designated by access values are always allocated in
the 64-bit address space, and access values themselves are represented
in 64 bits. If these defaults are not appropriate, and 32-bit allocation
is required (for example if the address of an allocated object is assigned
to a @code{Short_Address} variable), then several alternatives are available:
However, the access values themselves are represented in 64 bits, for optimum
performance and future compatibility with 64-bit code. As was
the case with @code{System.Address}, the compiler will give an error message
if an object or record component has a representation clause that
requires the access value to fit in 32 bits. In such a situation,
an explicit size clause for the access type, specifying 32 bits,
will have the desired effect.
@itemize @bullet
@item
A pool-specific access type (ie, an @w{Ada 83} access type, whose
definition is @code{access T} versus @code{access all T} or
@code{access constant T}), may be declared with a @code{'Size} representation
clause that establishes the size as 32 bits.
In such circumstances allocations for that type will
be from the 32-bit heap. Such a clause is not permitted
for a general access type (declared with @code{access all} or
@code{access constant}) as values of such types must be able to refer
to any object of the designated type, including objects residing outside
the 32-bit address range. Existing @w{Ada 83} code will not contain such
type definitions, however, since general access types were introduced
in @w{Ada 95}.
@item
Switches for @command{GNAT BIND} control whether the internal GNAT
allocation routine @code{__gnat_malloc} uses 64-bit or 32-bit allocations.
@cindex @code{__gnat_malloc}
The switches are respectively @option{-H64} (the default) and
@option{-H32}.
@cindex @option{-H32} (@command{gnatbind})
@cindex @option{-H64} (@command{gnatbind})
@item
The environment variable (logical name) @code{GNAT$NO_MALLOC_64}
@cindex @code{GNAT$NO_MALLOC_64} environment variable
may be used to force @code{__gnat_malloc} to use 32-bit allocation.
If this variable is left
undefined, or defined as @code{"DISABLE"}, @code{"FALSE"}, or @code{"0"},
then the default (64-bit) allocation is used.
If defined as @code{"ENABLE"}, @code{"TRUE"}, or @code{"1"},
then 32-bit allocation is used. The gnatbind qualifiers described above
override this logical name.
@item
A ^gcc switch^gcc switch^ for OpenVMS, @option{-mno-malloc64}, operates
@cindex @option{-mno-malloc64} (^gcc^gcc^)
at a low level to convert explicit calls to @code{malloc} and related
functions from the C run-time library so that they perform allocations
in the 32-bit heap.
Since all internal allocations from GNAT use @code{__gnat_malloc},
this switch is not required unless the program makes explicit calls on
@code{malloc} (or related functions) from interfaced C code.
@end itemize
General access types (declared with @code{access all}) can never be
32 bits, as values of such types must be able to refer to any object
of the designated type,
including objects residing outside the 32-bit address range.
Existing Ada 83 code will not contain such type definitions,
however, since general access types were introduced in Ada 95.
@node Unchecked conversions
@subsubsection Unchecked conversions
@ -26665,6 +26744,20 @@ pragma Convention(C, int_star);
for int_star'Size use 64; -- Necessary to get 64 and not 32 bits
@end smallexample
@node 32/64-bit descriptors
@subsubsection 32/64-bit descriptors
@noindent
By default, GNAT uses a 64-bit descriptor mechanism. For an imported
subprogram (i.e., a subprogram identified by pragma @code{Import_Function},
@code{Import_Procedure}, or @code{Import_Valued_Procedure}) that specifies
@code{Short_Descriptor} as its mechanism, a 32-bit descriptor is used.
@cindex @code{Short_Descriptor} mechanism for imported subprograms
If the configuration pragma @code{Short_Descriptors} is supplied, then
all descriptors will be 32 bits.
@cindex pragma @code{Short_Descriptors}
@node Experience with source compatibility
@subsubsection Experience with source compatibility
@ -26697,8 +26790,6 @@ these sorts of potential source code porting problems.
* Making code 64 bit clean::
* Allocating memory from the 64 bit storage pool::
* Restrictions on use of 64 bit objects::
* Using 64 bit storage pools by default::
* General access types::
* STARLET and other predefined libraries::
@end menu
@ -26742,13 +26833,10 @@ Any attempt to do this will raise @code{Constraint_Error}.
@subsubsection Allocating memory from the 64-bit storage pool
@noindent
For any access type @code{T} that potentially requires memory allocations
beyond the 32-bit address space,
use the following representation clause:
@smallexample @c ada
for T'Storage_Pool use System.Pool_64;
@end smallexample
By default, all allocations -- for both pool-specific and general
access types -- use the 64-bit storage pool. To override
this default, for an individual access type or globally, see
@ref{Access types and 32/64-bit allocation}.
@node Restrictions on use of 64 bit objects
@subsubsection Restrictions on use of 64-bit objects
@ -26763,46 +26851,6 @@ or assigning it to a variable of type @code{Short_Address}, will cause
no exception is raised and execution
will become erroneous.
@node Using 64 bit storage pools by default
@subsubsection Using 64-bit storage pools by default
@noindent
In some cases it may be desirable to have the compiler allocate
from 64-bit storage pools by default. This may be the case for
libraries that are 64-bit clean, but may be used in both 32-bit
and 64-bit contexts. For these cases the following configuration
pragma may be specified:
@smallexample @c ada
pragma Pool_64_Default;
@end smallexample
@noindent
Any code compiled in the context of this pragma will by default
use the @code{System.Pool_64} storage pool. This default may be overridden
for a specific access type @code{T} by the representation clause:
@smallexample @c ada
for T'Storage_Pool use System.Pool_32;
@end smallexample
@noindent
Any object whose address may be passed to a subprogram with a
@code{Short_Address} argument, or assigned to a variable of type
@code{Short_Address}, needs to be allocated from this pool.
@node General access types
@subsubsection General access types
@noindent
Objects designated by access values from a
general access type (declared with @code{access all}) are never allocated
from a 64-bit storage pool. Code that uses general access types will
accept objects allocated in either 32-bit or 64-bit address spaces,
but never allocate objects outside the 32-bit address space.
Using general access types ensures maximum compatibility with both
32-bit and 64-bit code.
@node STARLET and other predefined libraries
@subsubsection STARLET and other predefined libraries

2
gcc/ada/ug_words
View File

@ -218,3 +218,5 @@ stderr ^ SYS$ERROR
-O1 ^ /OPTIMIZE=SOME
-O2 ^ /OPTIMIZE=ALL
-O3 ^ /OPTIMIZE=INLINING
-H32 ^ /32_MALLOC
-H64 ^ /64_MALLOC