aboutsummaryrefslogtreecommitdiffstats
path: root/gcc-4.8.3/gcc/ada/s-stausa.ads
diff options
context:
space:
mode:
Diffstat (limited to 'gcc-4.8.3/gcc/ada/s-stausa.ads')
-rw-r--r--gcc-4.8.3/gcc/ada/s-stausa.ads339
1 files changed, 339 insertions, 0 deletions
diff --git a/gcc-4.8.3/gcc/ada/s-stausa.ads b/gcc-4.8.3/gcc/ada/s-stausa.ads
new file mode 100644
index 000000000..c0449e8fb
--- /dev/null
+++ b/gcc-4.8.3/gcc/ada/s-stausa.ads
@@ -0,0 +1,339 @@
+------------------------------------------------------------------------------
+-- --
+-- GNU ADA RUN-TIME LIBRARY (GNARL) COMPONENTS --
+-- --
+-- S Y S T E M - S T A C K _ U S A G E --
+-- --
+-- S p e c --
+-- --
+-- Copyright (C) 2004-2011, Free Software Foundation, Inc. --
+-- --
+-- GNARL 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/>. --
+-- --
+-- GNARL was developed by the GNARL team at Florida State University. --
+-- Extensive contributions were provided by Ada Core Technologies, Inc. --
+-- --
+------------------------------------------------------------------------------
+
+with System;
+with System.Storage_Elements;
+with System.Address_To_Access_Conversions;
+with Interfaces;
+
+package System.Stack_Usage is
+ pragma Preelaborate;
+
+ package SSE renames System.Storage_Elements;
+
+ subtype Stack_Address is SSE.Integer_Address;
+ -- Address on the stack
+
+ function To_Stack_Address
+ (Value : System.Address) return Stack_Address
+ renames System.Storage_Elements.To_Integer;
+
+ Task_Name_Length : constant := 32;
+ -- The maximum length of task name displayed.
+ -- ??? Consider merging this variable with Max_Task_Image_Length.
+
+ type Task_Result is record
+ Task_Name : String (1 .. Task_Name_Length);
+
+ Value : Natural;
+ -- Amount of stack used. The value is calculated on the basis of the
+ -- mechanism used by GNAT to allocate it, and it is NOT a precise value.
+
+ Stack_Size : Natural;
+ -- Size of the stack
+ end record;
+
+ type Result_Array_Type is array (Positive range <>) of Task_Result;
+
+ type Stack_Analyzer is private;
+ -- Type of the stack analyzer tool. It is used to fill a portion of the
+ -- stack with Pattern, and to compute the stack used after some execution.
+
+ -- Usage:
+
+ -- A typical use of the package is something like:
+
+ -- A : Stack_Analyzer;
+
+ -- task T is
+ -- pragma Storage_Size (A_Storage_Size);
+ -- end T;
+
+ -- [...]
+
+ -- Bottom_Of_Stack : aliased Integer;
+ -- -- Bottom_Of_Stack'Address will be used as an approximation of
+ -- -- the bottom of stack. A good practise is to avoid allocating
+ -- -- other local variables on this stack, as it would degrade
+ -- -- the quality of this approximation.
+
+ -- begin
+ -- Initialize_Analyzer (A,
+ -- "Task t",
+ -- A_Storage_Size,
+ -- 0,
+ -- A_Storage_Size - A_Guard,
+ -- To_Stack_Address (Bottom_Of_Stack'Address));
+ -- Fill_Stack (A);
+ -- Some_User_Code;
+ -- Compute_Result (A);
+ -- Report_Result (A);
+ -- end T;
+
+ -- Errors:
+ --
+ -- We are instrumenting the code to measure the stack used by the user
+ -- code. This method has a number of systematic errors, but several methods
+ -- can be used to evaluate or reduce those errors. Here are those errors
+ -- and the strategy that we use to deal with them:
+
+ -- Bottom offset:
+
+ -- Description: The procedure used to fill the stack with a given
+ -- pattern will itself have a stack frame. The value of the stack
+ -- pointer in this procedure is, therefore, different from the value
+ -- before the call to the instrumentation procedure.
+
+ -- Strategy: The user of this package should measure the bottom of stack
+ -- before the call to Fill_Stack and pass it in parameter. The impact
+ -- is very minor unless the stack used is very small, but in this case
+ -- you aren't very interested by the figure.
+
+ -- Instrumentation threshold at writing:
+
+ -- Description: The procedure used to fill the stack with a given
+ -- pattern will itself have a stack frame. Therefore, it will
+ -- fill the stack after this stack frame. This part of the stack will
+ -- appear as used in the final measure.
+
+ -- Strategy: As the user passes the value of the bottom of stack to
+ -- the instrumentation to deal with the bottom offset error, and as
+ -- the instrumentation procedure knows where the pattern filling start
+ -- on the stack, the difference between the two values is the minimum
+ -- stack usage that the method can measure. If, when the results are
+ -- computed, the pattern zone has been left untouched, we conclude
+ -- that the stack usage is inferior to this minimum stack usage.
+
+ -- Instrumentation threshold at reading:
+
+ -- Description: The procedure used to read the stack at the end of the
+ -- execution clobbers the stack by allocating its stack frame. If this
+ -- stack frame is bigger than the total stack used by the user code at
+ -- this point, it will increase the measured stack size.
+
+ -- Strategy: We could augment this stack frame and see if it changes the
+ -- measure. However, this error should be negligible.
+
+ -- Pattern zone overflow:
+
+ -- Description: The stack grows outer than the topmost bound of the
+ -- pattern zone. In that case, the topmost region modified in the
+ -- pattern is not the maximum value of the stack pointer during the
+ -- execution.
+
+ -- Strategy: At the end of the execution, the difference between the
+ -- topmost memory region modified in the pattern zone and the
+ -- topmost bound of the pattern zone can be understood as the
+ -- biggest allocation that the method could have detect, provided
+ -- that there is no "Untouched allocated zone" error and no "Pattern
+ -- usage in user code" error. If no object in the user code is likely
+ -- to have this size, this is not likely to happen.
+
+ -- Pattern usage in user code:
+
+ -- Description: The pattern can be found in the object of the user code.
+ -- Therefore, the address space where this object has been allocated
+ -- will appear as untouched.
+
+ -- Strategy: Choose a pattern that is uncommon. 16#0000_0000# is the
+ -- worst choice; 16#DEAD_BEEF# can be a good one. A good choice is an
+ -- address which is not a multiple of 2, and which is not in the
+ -- target address space. You can also change the pattern to see if it
+ -- changes the measure. Note that this error *very* rarely influence
+ -- the measure of the total stack usage: to have some influence, the
+ -- pattern has to be used in the object that has been allocated on the
+ -- topmost address of the used stack.
+
+ -- Stack overflow:
+
+ -- Description: The pattern zone does not fit on the stack. This may
+ -- lead to an erroneous execution.
+
+ -- Strategy: Specify a storage size that is bigger than the size of the
+ -- pattern. 2 times bigger should be enough.
+
+ -- Augmentation of the user stack frames:
+
+ -- Description: The use of instrumentation object or procedure may
+ -- augment the stack frame of the caller.
+
+ -- Strategy: Do *not* inline the instrumentation procedures. Do *not*
+ -- allocate the Stack_Analyzer object on the stack.
+
+ -- Untouched allocated zone:
+
+ -- Description: The user code may allocate objects that it will never
+ -- touch. In that case, the pattern will not be changed.
+
+ -- Strategy: There are no way to detect this error. Fortunately, this
+ -- error is really rare, and it is most probably a bug in the user
+ -- code, e.g. some uninitialized variable. It is (most of the time)
+ -- harmless: it influences the measure only if the untouched allocated
+ -- zone happens to be located at the topmost value of the stack
+ -- pointer for the whole execution.
+
+ procedure Initialize (Buffer_Size : Natural);
+ pragma Export (C, Initialize, "__gnat_stack_usage_initialize");
+ -- Initializes the size of the buffer that stores the results. Only the
+ -- first Buffer_Size results are stored. Any results that do not fit in
+ -- this buffer will be displayed on the fly.
+
+ procedure Fill_Stack (Analyzer : in out Stack_Analyzer);
+ -- Fill an area of the stack with the pattern Analyzer.Pattern. The size
+ -- of this area is Analyzer.Size. After the call to this procedure,
+ -- the memory will look like that:
+ --
+ -- Stack growing
+ -- ---------------------------------------------------------------------->
+ -- |<--------------------->|<----------------------------------->|
+ -- | Stack frames to | Memory filled with Analyzer.Pattern |
+ -- | Fill_Stack | |
+ -- ^ | ^
+ -- Analyzer.Stack_Base | Analyzer.Pattern_Limit
+ -- ^
+ -- Analyzer.Pattern_Limit +/- Analyzer.Pattern_Size
+ --
+
+ procedure Initialize_Analyzer
+ (Analyzer : in out Stack_Analyzer;
+ Task_Name : String;
+ Stack_Size : Natural;
+ Stack_Base : Stack_Address;
+ Pattern_Size : Natural;
+ Pattern : Interfaces.Unsigned_32 := 16#DEAD_BEEF#);
+ -- Should be called before any use of a Stack_Analyzer, to initialize it.
+ -- Max_Pattern_Size is the size of the pattern zone, might be smaller than
+ -- the full stack size Stack_Size in order to take into account e.g. the
+ -- secondary stack and a guard against overflow. The actual size taken
+ -- will be readjusted with data already used at the time the stack is
+ -- actually filled.
+
+ Is_Enabled : Boolean := False;
+ -- When this flag is true, then stack analysis is enabled
+
+ procedure Compute_Result (Analyzer : in out Stack_Analyzer);
+ -- Read the pattern zone and deduce the stack usage. It should be called
+ -- from the same frame as Fill_Stack. If Analyzer.Probe is not null, an
+ -- array of Unsigned_32 with Analyzer.Probe elements is allocated on
+ -- Compute_Result's stack frame. Probe can be used to detect the error:
+ -- "instrumentation threshold at reading". See above. After the call
+ -- to this procedure, the memory will look like:
+ --
+ -- Stack growing
+ -- ----------------------------------------------------------------------->
+ -- |<---------------------->|<-------------->|<--------->|<--------->|
+ -- | Stack frames | Array of | used | Memory |
+ -- | to Compute_Result | Analyzer.Probe | during | filled |
+ -- | | elements | the | with |
+ -- | | | execution | pattern |
+ -- | | |
+ -- |<----------------------------------------------------> |
+ -- Stack used ^
+ -- Pattern_Limit
+
+ procedure Report_Result (Analyzer : Stack_Analyzer);
+ -- Store the results of the computation in memory, at the address
+ -- corresponding to the symbol __gnat_stack_usage_results. This is not
+ -- done inside Compute_Result in order to use as less stack as possible
+ -- within a task.
+
+ procedure Output_Results;
+ -- Print the results computed so far on the standard output. Should be
+ -- called when all tasks are dead.
+
+ pragma Export (C, Output_Results, "__gnat_stack_usage_output_results");
+
+private
+
+ package Unsigned_32_Addr is
+ new System.Address_To_Access_Conversions (Interfaces.Unsigned_32);
+
+ subtype Pattern_Type is Interfaces.Unsigned_32;
+ Bytes_Per_Pattern : constant := Pattern_Type'Object_Size / Storage_Unit;
+
+ type Stack_Analyzer is record
+ Task_Name : String (1 .. Task_Name_Length);
+ -- Name of the task
+
+ Stack_Base : Stack_Address;
+ -- Address of the base of the stack, as given by the caller of
+ -- Initialize_Analyzer.
+
+ Stack_Size : Natural;
+ -- Entire size of the analyzed stack
+
+ Pattern_Size : Natural;
+ -- Size of the pattern zone
+
+ Pattern : Pattern_Type;
+ -- Pattern used to recognize untouched memory
+
+ Pattern_Limit : Stack_Address;
+ -- Bound of the pattern area farthest to the base
+
+ Topmost_Touched_Mark : Stack_Address;
+ -- Topmost address of the pattern area whose value it is pointing
+ -- at has been modified during execution. If the systematic error are
+ -- compensated, it is the topmost value of the stack pointer during
+ -- the execution.
+
+ Pattern_Overlay_Address : System.Address;
+ -- Address of the stack abstraction object we overlay over a
+ -- task's real stack, typically a pattern-initialized array.
+
+ Result_Id : Positive;
+ -- Id of the result. If less than value given to gnatbind -u corresponds
+ -- to the location in the result array of result for the current task.
+ end record;
+
+ Environment_Task_Analyzer : Stack_Analyzer;
+
+ Compute_Environment_Task : Boolean;
+
+ type Result_Array_Ptr is access all Result_Array_Type;
+
+ Result_Array : Result_Array_Ptr;
+ pragma Export (C, Result_Array, "__gnat_stack_usage_results");
+ -- Exported in order to have an easy accessible symbol in when debugging
+
+ Next_Id : Positive := 1;
+ -- Id of the next stack analyzer
+
+ function Stack_Size
+ (SP_Low : Stack_Address;
+ SP_High : Stack_Address) return Natural;
+ pragma Inline (Stack_Size);
+ -- Return the size of a portion of stack delimited by SP_High and SP_Low
+ -- (), i.e. the difference between SP_High and SP_Low. The storage element
+ -- pointed by SP_Low is not included in the size. Inlined to reduce the
+ -- size of the stack used by the instrumentation code.
+
+end System.Stack_Usage;