path: root/doc/ghdl.texi

\input texinfo   @c -*-texinfo-*-
@c %**start of header
@setfilename ghdl.info
@settitle GHDL guide
@c %**end of header

@direntry
* ghdl: (ghdl).         VHDL compiler.
@end direntry

@titlepage
@title GHDL guide
@subtitle GHDL, a VHDL compiler
@subtitle For GHDL version 0.32 (Dunoon edition)
@author Tristan Gingold
@c The following two commands start the copyright page.
@page
@vskip 0pt plus 1filll
Copyright @copyright{} 2002-2014 Tristan Gingold.

Permission is granted to copy, distribute and/or modify this document
under the terms of the GNU Free Documentation License, Version 1.1 or
any later version published by the Free Software Foundation.
@end titlepage

@ignore
Part I: User guide
1) Intro: what is vhdl, what is ghdl
2) starting with ghdl: a few examples
2.1) hello world
2.2) a nand gate
2.3) testsuite for a nand gate
2.4) a nand3 gate (using components)
2.5) testsuite for the nand3

Part II: Reference guide
1) command line options
1.1) filename extension.
2) Current standards
2.w) what is 93c
3) Linking with Ada or C code. FOREIGN use.
3) library organization
4) built-in libraries and pathes.
5) debugging your program.
6) report messages (runtime errors, boundary errors, assertion)
7) Error message, improve it.
8) current bugs, how to report a bug.
9) Copyright

done: ?) source representation
done: ?) copyright
done: ?) debugging
done: ?) executable options
done: ?) top entity characteristics
done: ?) work library
done: ?) ieee library
done: ?) file format (textio/not textio)

TODO:
XX: indexes
XXX: signals cannot be forced, only viewed in depth.
x: implementation dependant: files (see 4.3.1.4)

To check:
model vs modeling vs modelize
behaviour vs behavior
analyze vs analyse

Internal overview
 ortho
 grt subprograms
@end ignore

@contents

@ifnottex
@node Top, Introduction, (dir), (dir)
@top GHDL guide
GHDL, a VHDL compiler.

Copyright @copyright{} 2002-2014 Tristan Gingold.

Permission is granted to copy, distribute and/or modify this document
under the terms of the GNU Free Documentation License, Version 1.1
or any later version published by the Free Software Foundation.

@menu
* Introduction::                What is GHDL, what is VHDL
* Starting with GHDL::          Build a VHDL program with GHDL
* Invoking GHDL::               
* Simulation and runtime::      
* GHDL implementation of VHDL::  
* GHDL implementation of VITAL::  
* Flaws and bugs report::       
* Copyrights::                  
* Index::                       
@end menu

@end ifnottex

@node Introduction, Starting with GHDL, Top, Top
@comment  node-name,  next,  previous,  up
@chapter Introduction

@menu
* What is VHDL::                
* What is GHDL::                
@end menu

@section Content of this manual
This manual is the user and reference manual for GHDL.  It does not
contain an introduction to VHDL.  Thus, the reader should have at least
a basic knowledge of VHDL.  A good knowledge of VHDL language reference
manual (usually called LRM) is a plus.

@c FIXME: references: URL, LRM reference.

@node What is VHDL, What is GHDL, Introduction, Introduction
@comment  node-name,  next,  previous,  up
@section What is @code{VHDL}?
@dfn{VHDL} is an acronym for Very High Speed Integrated Circuit Hardware
Description Language which is a programming language used to describe a
logic circuit by function, data flow behaviour, or structure.

@code{VHDL} @emph{is} a programming language: although @code{VHDL} was
not designed for writing general purpose programs, you can write any
algorithm with the @code{VHDL} language.  If you are able to write
programs, you will find in @code{VHDL} features similar to those found
in procedural languages such as @code{C}, @code{Pascal} or @code{Ada}.
@code{VHDL} derives most of its syntax and semantics from @code{Ada}.
Knowing @code{Ada} is an advantage for learning @code{VHDL} (it is an
advantage in general as well).

However, @code{VHDL} was not designed as a general purpose language but as an
@code{HDL} (hardware description language).  As the name implies, @code{VHDL}
aims at modeling or documenting electronics systems.  Due to the nature
of hardware components which are always running, @code{VHDL} is a highly
concurrent language, built upon an event-based timing model.

Like a program written in any other language, a @code{VHDL} program
can be executed. Since @code{VHDL} is used to model designs, the term
@dfn{simulation} is often used instead of @dfn{execution}, with the
same meaning.

Like a program written in another hardware description language, a
@code{VHDL} program can be transformed with a @code{synthesis tool}
into a netlist, that is, a detailed gate-level implementation.

@node What is GHDL,  , What is VHDL, Introduction
@comment  node-name,  next,  previous,  up
@section What is @code{GHDL}?
@dfn{GHDL} is a shorthand for G Hardware Design Language.  Currently,
@code{G} has no meaning.

@dfn{GHDL} is a @code{VHDL} compiler that can execute (nearly) any
@code{VHDL} program. @code{GHDL} is @emph{not} a synthesis tool: you cannot
create a netlist with @code{GHDL}.

Unlike some other simulators, @code{GHDL} is a compiler: it directly
translates a @code{VHDL} file to machine code, using the @code{GCC}
back-end and without using an intermediary language such as @code{C}
or @code{C++}.  Therefore, the compiled code should be faster and
the analysis time should be shorter than with a compiler using an
intermediary language.

The Windows(TM) version of @code{GHDL} is not based on @code{GCC} but on
an internal code generator.

The current version of @code{GHDL} does not contain any graphical
viewer: you cannot see signal waves.  You can still check with a test
bench.  The current version can produce a @code{VCD} file which can be
viewed with a wave viewer, as well as @code{ghw} files to be viewed by
@samp{gtkwave}.

@code{GHDL} aims at implementing @code{VHDL} as defined by IEEE 1076.
It supports most of the 1987 standard and most features added by the
1993 standard.


@node Starting with GHDL, Invoking GHDL, Introduction, Top
@comment  node-name,  next,  previous,  up
@chapter Starting with GHDL
In this chapter, you will learn how to use the GHDL compiler by
working on two examples.

@menu
* The hello word program::      
* A full adder::                
* Starting with a design::      
@end menu

@node The hello word program, A full adder, Starting with GHDL, Starting with GHDL
@comment  node-name,  next,  previous,  up
@section The hello world program
To illustrate the large purpose of VHDL, here is a commented VHDL
"Hello world" program.

@example
--  @r{Hello world program.}
use std.textio.all; --  @r{Imports the standard textio package.}

--  @r{Defines a design entity, without any ports.}
entity hello_world is
end hello_world;

architecture behaviour of hello_world is
begin
   process
      variable l : line;
   begin
      write (l, String'("Hello world!"));
      writeline (output, l);
      wait;
   end process;
end behaviour;
@end example

Suppose this program is contained in the file @file{hello.vhdl}.
First, you have to compile the file; this is called @dfn{analysis} of a design
file in VHDL terms.
@smallexample
$ ghdl -a hello.vhdl
@end smallexample
This command creates or updates a file @file{work-obj93.cf}, which
describes the library @samp{work}.  On GNU/Linux, this command generates a
file @file{hello.o}, which is the object file corresponding to your
VHDL program.  The object file is not created on Windows.

Then, you have to build an executable file.
@smallexample
$ ghdl -e hello_world
@end smallexample
The @samp{-e} option means @dfn{elaborate}.  With this option, @code{GHDL}
creates code in order to elaborate a design, with the @samp{hello}
entity at the top of the hierarchy.

On GNU/Linux, the result is an executable program called @file{hello}
which can be run:
@smallexample
$ ghdl -r hello_world
@end smallexample
or directly:
@smallexample
$ ./hello_world
@end smallexample

On Windows, no file is created.  The simulation is launched using this command:
@smallexample
> ghdl -r hello_world
@end smallexample

The result of the simulation appears on the screen:
@smallexample
Hello world!
@end smallexample

@node A full adder, Starting with a design, The hello word program, Starting with GHDL
@comment  node-name,  next,  previous,  up
@section A full adder
VHDL is generally used for hardware design.  This example starts with
a full adder described in the @file{adder.vhdl} file:

@example
entity adder is
  -- @r{@var{i0}, @var{i1} and the carry-in @var{ci} are inputs of the adder.}
  -- @r{@var{s} is the sum output, @var{co} is the carry-out.}
  port (i0, i1 : in bit; ci : in bit; s : out bit; co : out bit);
end adder;

architecture rtl of adder is
begin
   --  @r{This full-adder architecture contains two concurrent assignment.}
   --  @r{Compute the sum.}
   s <= i0 xor i1 xor ci;
   --  @r{Compute the carry.}
   co <= (i0 and i1) or (i0 and ci) or (i1 and ci);
end rtl;
@end example

You can analyze this design file:
@smallexample
$ ghdl -a adder.vhdl
@end smallexample

You can try to execute the @samp{adder} design, but this is useless,
since nothing externally visible will happen.  In order to
check this full adder, a testbench has to be run.  This testbench is
very simple, since the adder is also simple: it checks exhaustively all
inputs.  Note that only the behaviour is tested, timing constraints are
not checked.  The file @file{adder_tb.vhdl} contains the testbench for
the adder:
@example
--  @r{A testbench has no ports.}
entity adder_tb is
end adder_tb;

architecture behav of adder_tb is
   --  @r{Declaration of the component that will be instantiated.}
   component adder
     port (i0, i1 : in bit; ci : in bit; s : out bit; co : out bit);
   end component;
   --  @r{Specifies which entity is bound with the component.}
   for adder_0: adder use entity work.adder;
   signal i0, i1, ci, s, co : bit;
begin
   --  @r{Component instantiation.}
   adder_0: adder port map (i0 => i0, i1 => i1, ci => ci,
                            s => s, co => co);

   --  @r{This process does the real job.}
   process
      type pattern_type is record
         --  @r{The inputs of the adder.}
         i0, i1, ci : bit;
         --  @r{The expected outputs of the adder.}
         s, co : bit;
      end record;
      --  @r{The patterns to apply.}
      type pattern_array is array (natural range <>) of pattern_type;
      constant patterns : pattern_array :=
        (('0', '0', '0', '0', '0'),
         ('0', '0', '1', '1', '0'),
         ('0', '1', '0', '1', '0'),
         ('0', '1', '1', '0', '1'),
         ('1', '0', '0', '1', '0'),
         ('1', '0', '1', '0', '1'),
         ('1', '1', '0', '0', '1'),
         ('1', '1', '1', '1', '1'));
   begin
      --  @r{Check each pattern.}
      for i in patterns'range loop
         --  @r{Set the inputs.}
         i0 <= patterns(i).i0;
         i1 <= patterns(i).i1;
         ci <= patterns(i).ci;
         --  @r{Wait for the results.}
         wait for 1 ns;
         --  @r{Check the outputs.}
         assert s = patterns(i).s
            report "bad sum value" severity error;
         assert co = patterns(i).co
            report "bad carray out value" severity error;
      end loop;
      assert false report "end of test" severity note;
      --  @r{Wait forever; this will finish the simulation.}
      wait;
   end process;
end behav;
@end example

As usual, you should analyze the design:
@smallexample
$ ghdl -a adder_tb.vhdl
@end smallexample
And build an executable for the testbench:
@smallexample
$ ghdl -e adder_tb
@end smallexample
You do not need to specify which object files are required: GHDL knows them
and automatically adds them in the executable.  Now, it is time to run the
testbench:
@smallexample
$ ghdl -r adder_tb
adder_tb.vhdl:52:7:(assertion note): end of test
@end smallexample

If your design is rather complex, you'd like to inspect signals.  Signals
value can be dumped using the VCD file format.  The resulting file can be
read with a wave viewer such as GTKWave.  First, you should simulate your
design and dump a waveform file:
@smallexample
$ ghdl -r adder_tb --vcd=adder.vcd
@end smallexample
Then, you may now view the waves:
@smallexample
$ gtkwave adder.vcd
@end smallexample

@xref{Simulation options}, for more details on the @option{--vcd} option and
other runtime options.

@node Starting with a design,  , A full adder, Starting with GHDL
@comment  node-name,  next,  previous,  up
@section Starting with a design
Unless you are only studying VHDL, you will work with bigger designs than
the ones of the previous examples.

Let's see how to analyze and run a bigger design, such as the DLX model
suite written by Peter Ashenden which is distributed under the terms of the
GNU General Public License.  A copy is kept on
@indicateurl{http://ghdl.free.fr/dlx.tar.gz}

First, untar the sources:
@smallexample
$ tar zxvf dlx.tar.gz
@end smallexample

In order not to pollute the sources with the library, it is a good idea
to create a @file{work/} subdirectory for the @samp{WORK} library.  To
any GHDL commands, we will add the @option{--workdir=work} option, so
that all files generated by the compiler (except the executable) will be
placed in this directory.
@smallexample
$ cd dlx
$ mkdir work
@end smallexample

We will run the @samp{dlx_test_behaviour} design.  We need to analyze
all the design units for the design hierarchy, in the correct order.
GHDL provides an easy way to do this, by importing the sources:
@smallexample
$ ghdl -i --workdir=work *.vhdl
@end smallexample

and making a design:
@smallexample
$ ghdl -m --workdir=work dlx_test_behaviour
@end smallexample

Before this second stage, GHDL knows all the design units of the DLX,
but no one have been analyzed.  The make command of GHDL analyzes and
elaborates a design.  This creates many files in the @file{work/}
directory, and the @file{dlx_test_behaviour} executable in the current
directory.

The simulation needs to have a DLX program contained in the file
@file{dlx.out}.  This memory image will be be loaded in the DLX memory.
Just take one sample:
@smallexample
$ cp test_loop.out dlx.out
@end smallexample

And you can run the test suite:
@smallexample
$ ghdl -r dlx_test_behaviour
@end smallexample

The test bench monitors the bus and displays each instruction executed.
It finishes with an assertion of severity level note:
@smallexample
dlx-behaviour.vhdl:395:11:(assertion note): TRAP instruction
 encountered, execution halted
@end smallexample

Since the clock is still running, you have to manually stop the program
with the @kbd{C-c} key sequence.  This behavior prevents you from running the
test bench in batch mode.  However, you may force the simulator to
stop when an assertion above or equal a certain severity level occurs:
@smallexample
$ ghdl -r dlx_test_behaviour --assert-level=note
@end smallexample

With this option, the program stops just after the previous message:
@smallexample
dlx-behaviour.vhdl:395:11:(assertion note): TRAP instruction
 encountered, execution halted
error: assertion failed
@end smallexample

If you want to make room on your hard drive, you can either:
@itemize @bullet{}
@item
clean the design library with the GHDL command:
@smallexample
$ ghdl --clean --workdir=work
@end smallexample
This removes the executable and all the object files.  If you want to
rebuild the design at this point, just do the make command as shown above.
@item
remove the design library with the GHDL command:
@smallexample
$ ghdl --remove --workdir=work
@end smallexample
This removes the executable, all the object files and the library file.
If you want to rebuild the design, you have to import the sources again,
and to make the design.
@item
remove the @file{work/} directory:
@smallexample
$ rm -rf work
@end smallexample
Only the executable is kept.  If you want to rebuild the design, create
the @file{work/} directory, import the sources, and make the design.
@end itemize

Sometimes, a design does not fully follow the VHDL standards.  For example it
uses the badly engineered @samp{std_logic_unsigned} package.  GHDL supports
this VHDL dialect through some options:
@smallexample
--ieee=synopsys -fexplicit
@end smallexample
@xref{IEEE library pitfalls}, for more details.

@node Invoking GHDL, Simulation and runtime, Starting with GHDL, Top
@comment  node-name,  next,  previous,  up
@chapter Invoking GHDL
The form of the @code{ghdl} command is

@smallexample
$ ghdl @var{command} [@var{options@dots{}}]
@end smallexample

The GHDL program has several commands.  The first argument selects
the commands.  The options are used to slightly modify the action.

No options are allowed before the command.  Except for the run commands,
no options are allowed after a filename or a unit name.

@menu
* Building commands::           
* GHDL options::                
* Passing options to other programs::  
* GHDL warnings::               
* Rebuilding commands::         
* Library commands::            
* Cross-reference command::     
* File commands::               
* Misc commands::               
* Installation Directory::      
* IEEE library pitfalls::       
* IEEE math packages::          
@end menu

@node Building commands, GHDL options, Invoking GHDL, Invoking GHDL
@comment  node-name,  next,  previous,  up
@section Building commands
The mostly used commands of GHDL are those to analyze and elaborate a design.

@menu
* Analysis command::            
* Elaboration command::         
* Run command::                 
* Elaborate and run command::   
* Bind command::                
* Link command::                
* List link command::           
* Check syntax command::        
* Analyze and elaborate command::  
@end menu

@node Analysis command, Elaboration command, Building commands, Building commands
@comment  node-name,  next,  previous,  up
@subsection Analysis command
@cindex analysis
@cindex @option{-a} command
@smallexample
$ ghdl -a [@var{options}] @var{files}
@end smallexample

The @dfn{analysis} command compiles one or more files, and creates an
object file for each source file.  The analysis command is selected with
@var{-a} switch.  Any argument starting with a dash is an option, the
others are filenames.  No options are allowed after a filename
argument. GHDL analyzes each filename in the given order, and stops the
analysis in case of error (the following files are not analyzed).
@c FIXME: check this.

@xref{GHDL options}, for details on the GHDL options.  For example,
to produce debugging information such as line numbers, use:

@smallexample
$ ghdl -a -g my_design.vhdl
@end smallexample

@node Elaboration command, Run command, Analysis command, Building commands
@comment  node-name,  next,  previous,  up
@subsection Elaboration command
@cindex elaboration
@cindex @option{-e} command
@smallexample
$ ghdl -e [@var{options}] @var{primary_unit} [@var{secondary_unit}]
@end smallexample

On GNU/Linux the @dfn{elaboration} command creates an executable
containing the code of the @code{VHDL} sources, the elaboration code
and simulation code to execute a design hierarchy. On Windows this
command elaborates the design but does not generate anything.

The elaboration command is selected with @var{-e} switch, and must be
followed by either:

@itemize @bullet
@item a name of a configuration unit
@item a name of an entity unit
@item a name of an entity unit followed by a name of an architecture unit
@end itemize

Name of the units must be a simple name, without any dot.  You can
select the name of the @samp{WORK} library with the @option{--work=NAME}
option, as described in @ref{GHDL options}.

@xref{Top entity}, for the restrictions on the root design of a
hierarchy.

On GNU/Linux the filename of the executable is the name of the
primary unit, or for the later case, the concatenation of the name of
the primary unit, a dash, and the name of the secondary unit (or
architecture).  On Windows there is no executable generated.

The @option{-o} followed by a filename can override the default
executable filename.

For the elaboration command, @code{GHDL} re-analyzes all the
configurations, entities, architectures and package declarations, and
creates the default configurations and the default binding indications
according to the LRM rules.  It also generates the list of objects files
required for the executable.  Then, it links all these files with the
runtime library.

The actual elaboration is performed at runtime.

On Windows this command can be skipped because it is also done by the
run command.

@node Run command, Elaborate and run command, Elaboration command, Building commands
@comment  node-name,  next,  previous,  up
@subsection Run command
@cindex run
@cindex @option{-r} command
Run (or simulate) a design.

@smallexample
$ ghdl -r [@var{options}] @var{primary_unit} [@var{secondary_unit}] [@var{simulation_options}]
@end smallexample

The options and arguments are the same as for the elaboration command, @pxref{Elaboration command}.

On GNU/Linux this command simply determines the filename of the executable
and executes it.  Options are ignored. You may also directly execute
the program.

This command exists for three reasons:
@itemize @bullet{}
@item
You don't have to create the executable program name.
@item
It is coherent with the @samp{-a} and @samp{-e} commands.
@item
It works with the Windows implementation, where the code is generated in
memory.
@end itemize

On Windows this command elaborates and launches the simulation.  As a consequence
you must use the same options used during analysis.

@xref{Simulation and runtime}, for details on options.

@node Elaborate and run command, Bind command, Run command, Building commands
@comment  node-name,  next,  previous,  up
@subsection Elaborate and run command
@cindex elaborate and run
@cindex @option{--elab-run} command
Elaborate and then simulate a design unit.

@smallexample
$ ghdl --elab-run [@var{elab_options}] @var{primary_unit} [@var{secondary_unit}] [@var{run_options}]
@end smallexample

This command acts like the elaboration command (@pxref{Elaboration command})
followed by the run command (@pxref{Run command}).

@node Bind command, Link command, Elaborate and run command, Building commands
@subsection Bind command
@cindex binding
@cindex @option{--bind} command
Bind a design unit and prepare the link step.

@smallexample
$ ghdl --bind [@var{options}] @var{primary_unit} [@var{secondary_unit}]
@end smallexample

This command is only available on GNU/Linux.

This performs only the first stage of the elaboration command; the list
of objects files is created but the executable is not built.  This
command should be used only when the main entry point is not ghdl.

@node Link command, List link command, Bind command, Building commands
@subsection Link command
@cindex linking
@cindex @option{--link} command
Link an already bound design unit.

@smallexample
$ ghdl --link [@var{options}] @var{primary_unit} [@var{secondary_unit}]
@end smallexample

This performs only the second stage of the elaboration command: the
executable is created by linking the files of the object files list.
This command is available only for completeness.  The elaboration command is
equivalent to the bind command followed by the link command.

@node List link command, Check syntax command, Link command, Building commands
@subsection List link command
@cindex @option{--list-link} command
Display files which will be linked.

@smallexample
$ ghdl --list-link @var{primary_unit} [@var{secondary_unit}]
@end smallexample

This command is only available on GNU/Linux.

This command may be used only after a bind command.  GHDL displays all
the files which will be linked to create an executable.  This command is
intended to add object files in a link of a foreign program.

@node Check syntax command, Analyze and elaborate command, List link command, Building commands
@subsection Check syntax command
@cindex checking syntax
@cindex @option{-s} command
Analyze files but do not generate code.

@smallexample
$ ghdl -s [@var{options}] @var{files}
@end smallexample

This command may be used to check the syntax of files.  It does not update
the library.

@node Analyze and elaborate command,  , Check syntax command, Building commands
@subsection Analyze and elaborate command
@cindex Analyze and elaborate command
@cindex @option{-c} command
Analyze files and elaborate them at the same time.

On GNU/Linux:
@smallexample
$ ghdl -c [@var{options}] @var{file}@dots{} -e @var{primary_unit} [@var{secondary_unit}]
@end smallexample

On Windows:
@smallexample
$ ghdl -c [@var{options}] @var{file}@dots{} -r @var{primary_unit} [@var{secondary_unit}]
@end smallexample

This command combines analysis and elaboration: @var{file}s are analyzed and
the unit is then elaborated.  However, code is only generated during the
elaboration.  On Windows the simulation is launched.

To be more precise, the files are first parsed, and then the elaboration
drives the analysis.  Therefore, there is no analysis order, and you don't
need to care about it.

All the units of the files are put into the @samp{work} library.  But, the
work library is neither read from disk nor saved.  Therefore, you must give
all the files of the @samp{work} library your design needs.

The advantages over the traditional approach (analyze and then elaborate) are:
@itemize
@item
The compilation cycle is achieved in one command.
@item
Since the files are only parsed once, the compilation cycle may be faster.
@item
You don't need to know an analysis order
@item
This command produces smaller executable, since unused units and subprograms
do not generate code.
@end itemize
However, you should know that currently most of the time is spent in code
generation and the analyze and elaborate command generate code for all units
needed, even units of @samp{std} and @samp{ieee} libraries.  Therefore,
according to the design, the time for this command may be higher than the time
for the analyze command followed by the elaborate command.

This command is still experimental.  In case of problems, you should go back
to the traditional way.

@comment  node-name,  next,  previous,  up
@node GHDL options, Passing options to other programs, Building commands, Invoking GHDL
@comment  node-name,  next,  previous,  up
@section GHDL options
@cindex IEEE 1164
@cindex 1164
@cindex IEEE 1076.3
@cindex 1076.3
@c document gcc options
Besides the options described below, @code{GHDL} passes any debugging options
(those that begin with @option{-g}) and optimizations options (those that
begin with @option{-O} or @option{-f}) to @code{GCC}.  Refer to the @code{GCC}
manual for details.

@table @code
@item --work=@var{NAME}
@cindex @option{--work} switch
@cindex WORK library
Specify the name of the @samp{WORK} library.  Analyzed units are always
placed in the library logically named @samp{WORK}.  With this option,
you can set its name.  By default, the name is @var{work}.

@code{GHDL} checks whether @samp{WORK} is a valid identifier.  Although being
more or less supported, the @samp{WORK} identifier should not be an
extended identifier, since the filesystem may prevent it from correctly
working (due to case sensitivity or forbidden characters in filenames).

@code{VHDL} rules forbid you to add units to the @samp{std} library.
Furthermore, you should not put units in the @samp{ieee} library.

@item --workdir=@var{DIR}
@cindex @option{--workdir} switch
Specify the directory where the @samp{WORK} library is located.  When this
option is not present, the @samp{WORK} library is in the current
directory.  The object files created by the compiler are always placed
in the same directory as the @samp{WORK} library.

Use option @option{-P} to specify where libraries other than @samp{WORK}
are placed.

@item --std=@var{STD}
@cindex @option{--std} switch
Specify the standard to use.  By default, the standard is @samp{93c}, which
means VHDL-93 accepting VHDL-87 syntax.  For details on @var{STD} values see
@ref{VHDL standards}.

@item --ieee=@var{VER}
@cindex @option{--ieee} switch
@cindex ieee library
@cindex synopsys library
@cindex mentor library
Select the @code{IEEE} library to use. @var{VER} must be one of:

@table @samp
@item none
Do not supply an @code{IEEE} library.  Any library clause with the @samp{IEEE}
identifier will fail, unless you have created by your own a library with
the @code{IEEE} name.

@item standard
Supply an @code{IEEE} library containing only packages defined by
@sc{ieee} standards.  Currently, there are the multivalue logic system
packages @samp{std_logic_1164} defined by IEEE 1164, the synthesis
packages , @samp{numeric_bit} and @samp{numeric_std} defined by IEEE
1076.3, and the @sc{vital} packages @samp{vital_timing} and
@samp{vital_primitives}, defined by IEEE 1076.4.  The version of these
packages is defined by the VHDL standard used.  @xref{VITAL packages}, 
for more details.

@item synopsys
Supply the former packages and the following additional packages:
@samp{std_logic_arith}, @samp{std_logic_signed},
@samp{std_logic_unsigned}, @samp{std_logic_textio}.
@c @samp{std_logic_misc}.
These packages were created by some companies, and are popular.  However
they are not standard packages, and have been placed in the @code{IEEE}
library without the permission from the @sc{ieee}.

@item mentor
Supply the standard packages and the following additional package:
@samp{std_logic_arith}.  The package is a slight variation of a definitely
not standard but widely mis-used package.
@end table

To avoid errors, you must use the same @code{IEEE} library for all units of
your design, and during elaboration.

@item -P@var{DIRECTORY}
@cindex @option{-P} switch
Add @var{DIRECTORY} to the end of the list of directories to be searched for
library files.

The @code{WORK} library is always searched in the path specified by the
@option{--workdir=} option, or in the current directory if the latter
option is not specified.

@item -fexplicit
@cindex @option{-fexplicit} switch
When two operators are overloaded, give preference to the explicit declaration.
This may be used to avoid the most common pitfall of the @samp{std_logic_arith}
package.  @xref{IEEE library pitfalls}, for an example.

This option is not set by default.  I don't think this option is a
good feature, because it breaks the encapsulation rule.  When set, an
operator can be silently overridden in another package.  You'd better to fix
your design and use the @samp{numeric_std} package.

@item -frelaxed-rules
@cindex @option{-frelaxed-rules} switch
Within an object declaration, allow to reference the name (which
references the hidden declaration).  This ignores the error in the
following code:

@example
package pkg1 is
 type state is (state1, state2, state3);
end pkg1;

use work.pkg1.all;
package pkg2 is
 constant state1 : state := state1;
end pkg2;
@end example

Some code (such as Xilinx packages) have such constructs, which
are valid.

(The scope of the @samp{state1} constant start at the @code{constant}
word. Because the constant @samp{state1} and the enumeration literal
@samp{state1} are homograph, the enumeration literal is hidden in the
immediate scope of the constant).

@item -fpsl
@cindex @option{-fpsl} switch
Enable parsing of PSL assertions within comments.  @xref{PSL implementation},
for more details.

@item --no-vital-checks
@item --vital-checks
@cindex @option{--no-vital-checks} switch
@cindex @option{--vital-checks} switch
Disable or enable checks of restriction on VITAL units.  Checks are enabled
by default.

Checks are performed only when a design unit is decorated by a VITAL attribute.
The VITAL attributes are @samp{VITAL_Level0} and @samp{VITAL_Level1}, both
declared in the @samp{ieee.VITAL_Timing} package.

Currently, VITAL checks are only partially implemented.  @xref{VHDL
restrictions for VITAL}, for more details.

@item --syn-binding
@cindex @option{--syn-binding} switch
Use synthesizer rules for component binding.  During elaboration, if a
component is not bound to an entity using VHDL LRM rules, try to find
in any known library an entity whose name is the same as the component
name.

This rule is known as synthesizer rule.

There are two key points: normal VHDL LRM rules are tried first and
entities are searched only in known library.  A known library is a
library which has been named in your design.

This option is only useful during elaboration.

@item --PREFIX=@var{PATH}
@cindex @option{--PREFIX} switch
Use @var{PATH} as the prefix path to find commands and pre-installed (std and
ieee) libraries.

@item --GHDL1=@var{COMMAND}
@cindex @option{--GHLD1} switch
Use @var{COMMAND} as the command name for the compiler.  If @var{COMMAND} is
not a path, then it is search in the list of program directories.

@item -v
Be verbose.  For example, for analysis, elaboration and make commands, GHDL
displays the commands executed.
@end table

@node Passing options to other programs, GHDL warnings, GHDL options, Invoking GHDL
@comment  node-name,  next,  previous,  up
@section Passing options to other programs

These options are only available on GNU/Linux.

For many commands, @code{GHDL} acts as a driver: it invokes programs to perform
the command.  You can pass arbitrary options to these programs.

Both the compiler and the linker are in fact GCC programs.  @xref{Invoking GCC,
GCC options, GCC Command Options, gcc, GCC manual}, for details on GCC
options.  

@table @code
@item -Wc,@var{OPTION}
@cindex @option{-W} switch
Pass @var{OPTION} as an option to the compiler.

@item -Wa,@var{OPTION}
@cindex @option{-Wa} switch
Pass @var{OPTION} as an option to the assembler.

@item -Wl,@var{OPTION}
@cindex @option{-Wl} switch
Pass @var{OPTION} as an option to the linker.
@end table

@node GHDL warnings, Rebuilding commands, Passing options to other programs, Invoking GHDL
@comment  node-name,  next,  previous,  up
@section GHDL warnings
Some constructions are not erroneous but dubious.  Warnings are diagnostic
messages that report such constructions.  Some warnings are reported only
during analysis, others during elaboration.

You could disable a warning by using the @option{--warn-no-XXX}
instead of @option{--warn-XXX}.

@table @code
@item --warn-reserved
@cindex @option{--warn-reserved} switch
Emit a warning if an identifier is a reserved word in a later VHDL standard.

@item --warn-default-binding
@cindex @option{--warn-default-binding} switch
During analyze, warns if a component instantiation has neither
configuration specification nor default binding.  This may be useful if you
want to detect during analyze possibly unbound component if you don't use
configuration.  @xref{VHDL standards}, for more details about default binding
rules.

@item --warn-binding
@cindex @option{--warn-binding} switch
During elaboration, warns if a component instantiation is not bound
(and not explicitly left unbound).  Also warns if a port of an entity
is not bound in a configuration specification or in a component
configuration.  This warning is enabled by default, since default
binding rules are somewhat complex and an unbound component is most
often unexpected.

However, warnings are even emitted if a component instantiation is
inside a generate statement.  As a consequence, if you use the conditional
generate statement to select a component according to the implementation,
you will certainly get warnings.

@item --warn-library
@cindex @option{--warn-library} switch
Warns if a design unit replaces another design unit with the same name.

@item --warn-vital-generic
@cindex @option{--warn-vital-generic} switch
Warns if a generic name of a vital entity is not a vital generic name.  This
is set by default.

@item --warn-delayed-checks
@cindex @option{--warn-delayed-checks} switch
Warns for checks that cannot be done during analysis time and are
postponed to elaboration time.  This is because not all procedure
bodies are available during analysis (either because a package body
has not yet been analysed or because @code{GHDL} doesn't read not required
package bodies).

These are checks for no wait statement in a procedure called in a
sensitized process and checks for pure rules of a function.

@item --warn-body
@cindex @option{--warn-body} switch
Emit a warning if a package body which is not required is analyzed.  If a
package does not declare a subprogram or a deferred constant, the package
does not require a body.

@item --warn-specs
@cindex @option{--warn-specs} switch
Emit a warning if an all or others specification does not apply.

@item --warn-unused
@cindex @option{--warn-unused} switch
Emit a warning when a subprogram is never used.

@item --warn-error
@cindex @option{--warn-error} switch
When this option is set, warnings are considered as errors.

@end table

@node Rebuilding commands, Library commands, GHDL warnings, Invoking GHDL
@comment  node-name,  next,  previous,  up
@section Rebuilding commands
Analyzing and elaborating a design consisting in several files can be tricky,
due to dependencies.  GHDL has a few commands to rebuild a design.

@menu
* Import command::              
* Make command::                
* Generate Makefile command::   
@end menu

@node Import command, Make command, Rebuilding commands, Rebuilding commands
@comment  node-name,  next,  previous,  up
@subsection Import command
@cindex importing files
@cindex @option{-i} command
Add files in the work design library.

@smallexample
$ ghdl -i [@var{options}] @var{file}@dots{}
@end smallexample

All the files specified in the command line are scanned, parsed and added in
the libraries but as not yet analyzed.  No object files are created.

The purpose of this command is to localize design units in the design files.
The make command will then be able to recursively build a hierarchy from
an entity name or a configuration name.

Since the files are parsed, there must be correct files.  However, since they
are not analyzed, many errors are tolerated by this command.

Note that all the files are added to the work library.  If you have many
libraries, you must use the command for each library.

@c Due to the LRM rules, there may be many analysis orders, producing
@c different results.  For example, if an entity has several architectures,
@c the last architecture analyzed is the default one in default binding
@c indications.

@xref{Make command}, to actually build the design.

@node Make command, Generate Makefile command, Import command, Rebuilding commands
@comment  node-name,  next,  previous,  up
@subsection Make command
@cindex make
@cindex @option{-m} command
@smallexample
$ ghdl -m [@var{options}] @var{primary} [@var{secondary}]
@end smallexample

Analyze automatically outdated files and elaborate a design.

The primary unit denoted by the @var{primary} argument must already be
known by the system, either because you have already analyzed it (even
if you have modified it) or because you have imported it.  GHDL analyzes
all outdated files.  A file may be outdated because it has been modified
(e.g. you just have edited it), or because a design unit contained in
the file depends on a unit which is outdated.  This rule is of course
recursive.

With the @option{-f} (force) option, GHDL analyzes all the units of the
work library needed to create the design hierarchy.  Not outdated units
are recompiled.  This is useful if you want to compile a design hierarchy
with new compilation flags (for example, to add the @option{-g}
debugging option).

The make command will only re-analyze design units in the work library.
GHDL fails if it has to analyze an outdated unit from another library.

The purpose of this command is to be able to compile a design without prior
knowledge of file order.  In the VHDL model, some units must be analyzed
before others (e.g. an entity before its architecture).  It might be a
nightmare to analyze a full design of several files, if you don't have
the ordered list of file.  This command computes an analysis order.

The make command fails when a unit was not previously parsed.  For
example, if you split a file containing several design units into
several files, you must either import these new files or analyze them so
that GHDL knows in which file these units are.

The make command imports files which have been modified.  Then, a design
hierarchy is internally built as if no units are outdated.  Then, all outdated
design units, using the dependencies of the design hierarchy, are analyzed.
If necessary, the design hierarchy is elaborated.

This is not perfect, since the default architecture (the most recently
analyzed one) may change while outdated design files are analyzed. In
such a case, re-run the make command of GHDL.

@c does not exists: @section GHDL robust make command

@node Generate Makefile command,  , Make command, Rebuilding commands
@comment  node-name,  next,  previous,  up
@subsection Generate Makefile command
@cindex @option{--gen-makefile} command
Generate a Makefile to build a design unit.

@smallexample
$ ghdl --gen-makefile [@var{options}] @var{primary} [@var{secondary}]
@end smallexample

This command works like the make command (@pxref{Make command}), but only a
makefile is generated on the standard output.

@node Library commands, Cross-reference command, Rebuilding commands, Invoking GHDL
@comment  node-name,  next,  previous,  up
@section Library commands
GHDL has a few commands which act on a library.

@comment  node-name,  next,  previous,  up
@menu
* Directory command::           
* Clean command::               
* Remove command::              
* Copy command::                
* Create a Library::
@end menu

@node Directory command, Clean command, Library commands, Library commands
@comment  node-name,  next,  previous,  up
@subsection Directory command
@cindex displaying library
@cindex @option{-d} command
Display the name of the units contained in a design library.
@smallexample
$ ghdl -d [@var{options}]
@end smallexample

The directory command, selected with the @var{-d} command line argument
displays the content of the work design library.  All options are
allowed, but only a few are meaningful: @option{--work=NAME},
@option{--workdir=PATH} and @option{--std=VER}.

@node Clean command, Remove command, Directory command, Library commands
@comment  node-name,  next,  previous,  up
@subsection Clean command
@cindex cleaning
@cindex @option{--clean} command
Remove object and executable files but keep the library.

@smallexample
$ ghdl --clean [@var{options}]
@end smallexample

GHDL tries to remove any object, executable or temporary file it could
have created.  Source files are not removed.

There is no short command line form for this option to prevent accidental
clean up.

@node Remove command, Copy command, Clean command, Library commands
@subsection Remove command
@cindex cleaning all
@cindex @option{--remove} command
Do like the clean command but remove the library too.

@smallexample
$ ghdl --remove [@var{options}]
@end smallexample

There is no short command line form for this option to prevent accidental
clean up.  Note that after removing a design library, the files are not
known anymore by GHDL.

@node Copy command, Create a Library, Remove command, Library commands
@subsection Copy command
@cindex copying library
@cindex @option{--copy} command
Make a local copy of an existing library.

@smallexample
$ ghdl --copy --work=@var{name} [@var{options}]
@end smallexample

Make a local copy of an existing library.  This is very useful if you want to
add unit to the @samp{ieee} library:
@example
$ ghdl --copy --work=ieee --ieee=synopsys
$ ghdl -a --work=ieee numeric_unsigned.vhd
@end example

@node Create a Library,  , Copy command, Library commands
@subsection Create a Library
@cindex create your own library

A new library is created by compiling entities (packages etc.) into it.

@smallexample
ghdl -a --work=my_custom_lib my_file.vhd
@end smallexample

A library's source code is usually stored and compiled into its own directory,
that you specify with the --workdir option.

@smallexample
ghdl -a --work=my_custom_lib --workdir=my_custom_libdir my_custom_lib_srcdir/my_file.vhd
@end smallexample

See also the -PPATH command line option.

@node Cross-reference command, File commands, Library commands, Invoking GHDL
@comment  node-name,  next,  previous,  up
@section Cross-reference command
To easily navigate through your sources, you may generate cross-references.

@smallexample
$ ghdl --xref-html [@var{options}] @var{file}@dots{}
@end smallexample

This command generates an html file for each @var{file} given in the command
line, with syntax highlighting and full cross-reference: every identifier is
a link to its declaration.  Besides, an index of the files is created too.

The set of @var{file} are analyzed, and then, if the analysis is
successful, html files are generated in the directory specified by the
@option{-o @var{dir}} option, or @file{html/} directory by default.

If the @option{--format=html2} is specified, then the generated html
files follow the HTML 2.0 standard, and colours are specified with
@samp{<FONT>} tags.  However, colours are hard-coded.

If the @option{--format=css} is specified, then the generated html files
follow the HTML 4.0 standard, and use the CSS-1 file @file{ghdl.css} to
specify colours.  This file is generated only if it does not already exist (it
is never overwritten) and can be customized by the user to change colours or
appearance.  Refer to a generated file and its comments for more information.

@node File commands, Misc commands, Cross-reference command, Invoking GHDL
@comment  node-name,  next,  previous,  up
@section File commands
The following commands act on one or several files.  They do not analyze
files, therefore, they work even if a file has semantic errors.

@menu
* Pretty print command::        
* Find command::                
* Chop command::                
* Lines command::               
@end menu

@node Pretty print command, Find command, File commands, File commands
@comment  node-name,  next,  previous,  up
@subsection Pretty print command
@cindex @option{--pp-html} command
@cindex pretty printing
@cindex vhdl to html

Generate HTML on standard output from VHDL.

@smallexample
$ ghdl --pp-html [@var{options}] @var{file}@dots{}
@end smallexample

The files are just scanned and an html file, with syntax highlighting is
generated on standard output.

Since the files are not even parsed, erroneous files or incomplete designs
can be pretty printed.

The style of the html file can be modified with the @option{--format=} option.
By default or when the @option{--format=html2} option is specified, the output
is an HTML 2.0 file, with colours set through @samp{<FONT>} tags.  When the
@option{--format=css} option is specified, the output is an HTML 4.0 file,
with colours set through a CSS file, whose name is @samp{ghdl.css}.
@xref{Cross-reference command}, for more details about this CSS file.

@node Find command, Chop command, Pretty print command, File commands
@comment  node-name,  next,  previous,  up
@subsection Find command
@cindex @option{-f} command
Display the name of the design units in files.

@smallexample
$ ghdl -f @var{file}@dots{}
@end smallexample

The files are scanned, parsed and the names of design units are displayed.
Design units marked with two stars are candidate to be at the apex of a
design hierarchy.


@node Chop command, Lines command, Find command, File commands
@comment  node-name,  next,  previous,  up
@subsection Chop command
@cindex @option{--chop} command
Chop (or split) files at design unit.

@smallexample
$ ghdl --chop @var{files}
@end smallexample

@code{GHDL} reads files, and writes a file in the current directory for
every design unit.

The filename of a design unit is build according to the unit.  For an
entity declaration, a package declaration or a configuration the file
name is @file{NAME.vhdl}, where @var{NAME} is the name of the design
unit.  For a package body, the filename is @file{NAME-body.vhdl}.
Finally, for an architecture @var{ARCH} of an entity @var{ENTITY}, the
filename is @file{ENTITY-ARCH.vhdl}.

Since the input files are parsed, this command aborts in case of syntax
error.  The command aborts too if a file to be written already exists.

Comments between design units are stored into the most adequate files.

This command may be useful to split big files, if your computer has not
enough memory to compile such files.  The size of the executable is
reduced too.

@node Lines command,  , Chop command, File commands
@comment  node-name,  next,  previous,  up
@subsection Lines command
@cindex @option{--lines} command
Display on the standard output lines of files preceded by line number.

@smallexample
$ ghdl --lines @var{files}
@end smallexample

@node Misc commands, Installation Directory, File commands, Invoking GHDL
@comment  node-name,  next,  previous,  up
@section Misc commands
There are a few GHDL commands which are seldom useful.

@menu
* Help command::                
* Disp config command::         
* Disp standard command::       
* Version command::             
@end menu

@node Help command, Disp config command, Misc commands, Misc commands
@subsection Help command
@cindex @option{-h} command
@cindex @option{--help} command
Display (on the standard output) a short description of the all the commands
available.  If the help switch is followed by a command switch, then options
for this later command are displayed.

@smallexample
$ ghdl --help
$ ghdl -h
$ ghdl -h @var{command}
@end smallexample

@node Disp config command, Disp standard command, Help command, Misc commands
@comment  node-name,  next,  previous,  up
@subsection Disp config command
@cindex @option{--disp-config} command
@cindex display configuration
Display the program paths and options used by GHDL.

@smallexample
$ ghdl --disp-config [@var{options}]
@end smallexample

This may be useful to track installation errors.

@node Disp standard command, Version command, Disp config command, Misc commands
@comment  node-name,  next,  previous,  up
@subsection Disp standard command
@cindex @option{--disp-standard} command
@cindex display @samp{std.standard}
Display the @samp{std.standard} package:

@smallexample
$ ghdl --disp-standard [@var{options}]
@end smallexample

@node Version command,  , Disp standard command, Misc commands
@comment  node-name,  next,  previous,  up
@subsection Version command
@cindex @option{--version} command
@cindex version
Display the @code{GHDL} version and exit.

@smallexample
$ ghdl --version
@end smallexample

@node Installation Directory, IEEE library pitfalls, Misc commands, Invoking GHDL
@comment  node-name,  next,  previous,  up
@section Installation Directory
@c @code{GHDL} is installed with the @code{std} and @code{ieee} libraries.  
During analysis and elaboration @code{GHDL} may read the @code{std}
and @code{ieee} files.  The location of these files is based on the prefix,
which is (in priority order):
@enumerate
@item
the @option{--PREFIX=} command line option

@item
the @var{GHDL_PREFIX} environment variable

@item
a built-in default path.  It is a hard-coded path on GNU/Linux and the
value of the @samp{HKLM\Software\Ghdl\Install_Dir} registry entry on Windows.
@end enumerate

You should use the @option{--disp-config} command (@pxref{Disp config command} for details) to disp and debug installation problems.

@node IEEE library pitfalls, IEEE math packages, Installation Directory, Invoking GHDL
@comment  node-name,  next,  previous,  up
@section IEEE library pitfalls
When you use options @option{--ieee=synopsys} or @option{--ieee=mentor},
the @code{IEEE} library contains non standard packages such as
@samp{std_logic_arith}. @c FIXME: ref

These packages are not standard because there are not described by an IEEE
standard, even if they have been put in the @code{IEEE} library.  Furthermore,
they are not really de-facto standard, because there are slight differences
between the packages of Mentor and those of Synopsys.

Furthermore, since they are not well-thought, their use has pitfalls.  For
example, this description has error during compilation:
@example
library ieee;
use ieee.std_logic_1164.all;

--  @r{A counter from 0 to 10}.
entity counter is
   port (val : out std_logic_vector (3 downto 0);
         ck : std_logic;
         rst : std_logic);
end counter;

library ieee;
use ieee.std_logic_unsigned.all;

architecture bad of counter
is
   signal v : std_logic_vector (3 downto 0);
begin
   process (ck, rst)
   begin
     if rst = '1' then
        v <= x"0";
     elsif rising_edge (ck) then
        if v = "1010" then -- @r{Error}
           v <= x"0";
        else
           v <= v + 1;
        end if;
     end if;
   end process;

   val <= v;
end bad;
@end example

When you analyze this design, GHDL does not accept it (too long lines
have been split for readability):
@smallexample
$ ghdl -a --ieee=synopsys bad_counter.vhdl
bad_counter.vhdl:13:14: operator "=" is overloaded
bad_counter.vhdl:13:14: possible interpretations are:
../../libraries/ieee/std_logic_1164.v93:69:5: implicit function "="
    [std_logic_vector, std_logic_vector return boolean]
../../libraries/synopsys/std_logic_unsigned.vhdl:64:5: function "="
    [std_logic_vector, std_logic_vector return boolean]
../translate/ghdldrv/ghdl: compilation error
@end smallexample
Indeed, the @code{"="} operator is defined in both packages, and both
are visible at the place it is used.  The first declaration is an
implicit one, which occurs when the @code{std_logic_vector} type is
declared and is an element to element comparison, the second one is an
explicit declared function, with the semantic of an unsigned comparison.

With some analyser, the explicit declaration has priority over the implicit
declaration, and this design can be analyzed without error.  However, this
is not the rule given by the VHDL LRM, and since GHDL follows these rules,
it emits an error.

You can force GHDL to use this rule with the @option{-fexplicit} option.
@xref{GHDL options}, for more details.

However it is easy to fix this error, by using a selected name:
@example
library ieee;
use ieee.std_logic_unsigned.all;

architecture fixed_bad of counter
is
   signal v : std_logic_vector (3 downto 0);
begin
   process (ck, rst)
   begin
     if rst = '1' then
        v <= x"0";
     elsif rising_edge (ck) then
        if ieee.std_logic_unsigned."=" (v, "1010") then
           v <= x"0";
        else
           v <= v + 1;
        end if;
     end if;
   end process;

   val <= v;
end fixed_bad;
@end example

It is better to only use the standard packages defined by IEEE, which
provides the same functionalities:
@example
library ieee;
use ieee.numeric_std.all;

architecture good of counter
is
   signal v : unsigned (3 downto 0);
begin
   process (ck, rst)
   begin
     if rst = '1' then
        v <= x"0";
     elsif rising_edge (ck) then
        if v = "1010" then
           v <= x"0";
        else
           v <= v + 1;
        end if;
     end if;
   end process;

   val <= std_logic_vector (v);
end good;
@end example

@node IEEE math packages,  , IEEE library pitfalls, Invoking GHDL
@comment  node-name,  next,  previous,  up
@section IEEE math packages
@cindex Math_Real
@cindex Math_Complex

The @samp{ieee} math packages (@samp{math_real} and
@samp{math_complex}) provided with @code{GHDL} are fully compliant with
the @code{IEEE} standard.

@node Simulation and runtime, GHDL implementation of VHDL, Invoking GHDL, Top
@comment  node-name,  next,  previous,  up
@chapter Simulation and runtime

@menu
* Simulation options::          
* Debugging VHDL programs::     
@end menu

@node Simulation options, Debugging VHDL programs, Simulation and runtime, Simulation and runtime
@comment  node-name,  next,  previous,  up
@section Simulation options
In most system environments, it is possible to pass options while
invoking a program.  Contrary to most programming languages, there is no
standard method in VHDL to obtain the arguments or to set the exit
status.

In GHDL, it is impossible to pass parameters to your design.  A later version
could do it through the generics interfaces of the top entity.

However, the GHDL runtime behaviour can be modified with some options; for
example, it is possible to stop simulation after a certain time.

The exit status of the simulation is @samp{EXIT_SUCCESS} (0) if the
simulation completes, or @samp{EXIT_FAILURE} (1) in case of error
(assertion failure, overflow or any constraint error).

Here is the list of the most useful options.  Some debugging options are
also available, but not described here.  The @samp{--help} options lists
all options available, including the debugging one.

@table @code
@item --assert-level=@var{LEVEL}
@cindex @option{--assert-level} option
Select the assertion level at which an assertion violation stops the
simulation.  @var{LEVEL} is the name from the @code{severity_level}
enumerated type defined in the @code{standard} package or the
@samp{none} name.

By default, only assertion violation of severity level @samp{failure}
stops the simulation.

For example, if @var{LEVEL} was @samp{warning}, any assertion violation
with severity level @samp{warning}, @samp{error} or @samp{failure} would
stop simulation, but the assertion violation at the @samp{note} severity
level would only display a message.

@samp{--assert-level=none} prevents any assertion violation to stop
simulation.

@item --ieee-asserts=@var{POLICY}
@cindex @option{--ieee-asserts} option
Select how the assertions from @samp{ieee} units are
handled. @var{POLICY} can be @samp{enable} (the default),
@samp{disable} which disables all assertion from @samp{ieee} packages
and @samp{disable-at-0} which disables only at start of simulation.

This option can be useful to avoid assertion message from
@samp{ieee.numeric_std} (and other @samp{ieee} packages).

@item --stop-time=@var{TIME}
@cindex @option{--stop-time} option
Stop the simulation after @var{TIME}.  @var{TIME} is expressed as a time
value, @emph{without} any space.  The time is the simulation time, not
the real clock time.

For examples:

@smallexample
$ ./my_design --stop-time=10ns
$ ./my_design --stop-time=ps
@end smallexample

@item --stop-delta=@var{N}
@cindex @option{--stop-delta} option
Stop the simulation after @var{N} delta cycles in the same current time.
@c Delta cycles is a simulation technic used by VHDL to

@item --disp-time
@cindex @option{--disp-time} option
@cindex display time
Display the time and delta cycle number as simulation advances.

@item --disp-tree[@var{=KIND}]
@cindex @option{--disp-tree} option
@cindex display design hierarchy
Display the design hierarchy as a tree of instantiated design entities.
This may be useful to understand the structure of a complex
design. @var{KIND} is optional, but if set must be one of:
@table @samp
@item none
Do not display hierarchy.  Same as if the option was not present.
@item inst
Display entities, architectures, instances, blocks and generates statements.
@item proc
Like @samp{inst} but also display processes.
@item port
Like @samp{proc} but display ports and signals too.
@end table
If @var{KIND} is not specified, the hierarchy is displayed with the
@samp{port} mode.

@item --no-run
@cindex @option{--no-run} option
Do not simulate, only elaborate.  This may be used with
@option{--disp-tree} to display the tree without simulating the whole
design.

@item --vcd=@var{FILENAME}
@item --vcdgz=@var{FILENAME}
@cindex @option{--vcd} option
@cindex @option{--vcdgz} option
@cindex vcd
@cindex value change dump
@cindex dump of signals
@option{--vcd} dumps into the VCD file @var{FILENAME} the signal
values before each non-delta cycle.  If @var{FILENAME} is @samp{-},
then the standard output is used, otherwise a file is created or
overwritten.

The @option{--vcdgz} option is the same as the @option{--vcd} option,
but the output is compressed using the @code{zlib} (@code{gzip}
compression).  However, you can't use the @samp{-} filename.
Furthermore, only one VCD file can be written.

@dfn{VCD} (value change dump) is a file format defined
by the @code{verilog} standard and used by virtually any wave viewer.

Since it comes from @code{verilog}, only a few VHDL types can be dumped.  GHDL
dumps only signals whose base type is of the following:
@itemize @bullet
@item
types defined in the @samp{std.standard} package:
@itemize @bullet
@item
@samp{bit}
@item
@samp{bit_vector}
@end itemize
@item
types defined in the @samp{ieee.std_logic_1164} package:
@itemize @bullet
@item
@samp{std_ulogic}
@item
@samp{std_logic} (because it is a subtype of @samp{std_ulogic})
@item
@samp{std_ulogic_vector}
@item
@samp{std_logic_vector}
@end itemize
@item
any integer type
@end itemize

I have successfully used @code{gtkwave} to view VCD files.

Currently, there is no way to select signals to be dumped: all signals are
dumped, which can generate big files.

It is very unfortunate there is no standard or well-known wave file
format supporting VHDL types.  If you are aware of such a free format,
please mail me (@pxref{Reporting bugs}).

@item --fst=@var{FILENAME}
@cindex @option{--fst} option
Write the waveforms into a @code{fst}, that can be displayed by
@code{gtkwave}. The @code{fst} files are much smaller than VCD or
@code{GHW} files, but it handles only the same signals as the VCD format.

@item --wave=@var{FILENAME}
@cindex @option{--wave} option
Write the waveforms into a @code{ghw} (GHdl Waveform) file.  Currently, all
the signals are dumped into the waveform file, you cannot select a hierarchy
of signals to be dumped.

The format of this file was defined by myself and is not yet completely fixed.
It may change slightly.  The @samp{gtkwave} tool can read the GHW files.

Contrary to VCD files, any VHDL type can be dumped into a GHW file.

@item --sdf=@var{PATH}=@var{FILENAME}
@item --sdf=min=@var{PATH}=@var{FILENAME}
@item --sdf=typ=@var{PATH}=@var{FILENAME}
@item --sdf=max=@var{PATH}=@var{FILENAME}
@cindex @option{--sdf} option
Do VITAL annotation on @var{PATH} with SDF file @var{FILENAME}.

@var{PATH} is a path of instances, separated with @samp{.} or @samp{/}.
Any separator can be used.  Instances are component instantiation labels,
generate labels or block labels.  Currently, you cannot use an indexed name.

If the option contains a type of delay, that is @option{min=},
@option{typ=} or @option{max=}, the annotator use respectively minimum,
typical or maximum values.  If the option does not contain a type of delay,
the annotator use the typical delay.

@xref{Backannotation}, for more details.

@item --help
Display a short description of the options accepted by the runtime library.
@end table

@node Debugging VHDL programs,  , Simulation options, Simulation and runtime
@comment  node-name,  next,  previous,  up
@section Debugging VHDL programs
@cindex debugging
@cindex @code{__ghdl_fatal}
Debugging VHDL programs using @code{GDB} is possible only on GNU/Linux systems.

@code{GDB} is a general purpose debugger for programs compiled by @code{GCC}.
Currently, there is no VHDL support for @code{GDB}.  It may be difficult
to inspect variables or signals in @code{GDB}, however, @code{GDB} is
still able to display the stack frame in case of error or to set a breakpoint
at a specified line.

@code{GDB} can be useful to precisely catch a runtime error, such as indexing
an array beyond its bounds.  All error check subprograms call the
@code{__ghdl_fatal} procedure.  Therefore, to catch runtime error, set
a breakpoint like this:
@smallexample
(gdb) break __ghdl_fatal
@end smallexample
When the breakpoint is hit, use the @code{where} or @code{bt} command to
display the stack frames.

@node GHDL implementation of VHDL, GHDL implementation of VITAL, Simulation and runtime, Top
@comment  node-name,  next,  previous,  up
@chapter GHDL implementation of VHDL

This chapter describes several implementation defined aspect of VHDL in GHDL.

@menu
* VHDL standards::              
* PSL implementation::          
* Source representation::       
* Library database::            
* VHDL files format::           
* Top entity::                  
* Using vendor libraries::      
* Interfacing to other languages::  
@end menu

@node VHDL standards, PSL implementation, GHDL implementation of VHDL, GHDL implementation of VHDL
@comment  node-name,  next,  previous,  up
@section VHDL standards
@cindex VHDL standards
@cindex IEEE 1076
@cindex IEEE 1076a
@cindex 1076
@cindex 1076a
@cindex v87
@cindex v93
@cindex v93c
@cindex v00
@cindex v02
This is very unfortunate, but there are many versions of the VHDL language.

The VHDL language was first standardized in 1987 by IEEE as IEEE 1076-1987, and
is commonly referred as VHDL-87. This is certainly the most important version,
since most of the VHDL tools are still based on this standard.

Various problems of this first standard have been analyzed by experts groups
to give reasonable ways of interpreting the unclear portions of the standard.

VHDL was revised in 1993 by IEEE as IEEE 1076-1993.  This revision is still
well-known.

Unfortunately, VHDL-93 is not fully compatible with VHDL-87, i.e. some perfectly
valid VHDL-87 programs are invalid VHDL-93 programs.  Here are some of the
reasons:

@itemize @bullet
@item
the syntax of file declaration has changed (this is the most visible source
of incompatibility),
@item
new keywords were introduced (group, impure, inertial, literal,
postponed, pure, reject, rol, ror, shared, sla, sll, sra, srl,
unaffected, xnor),
@item
some dynamic behaviours have changed (the concatenation is one of them),
@item
rules have been added.
@end itemize

Shared variables were replaced by protected types in the 2000 revision of
the VHDL standard.  This modification is also known as 1076a.  Note that this
standard is not fully backward compatible with VHDL-93, since the type of a
shared variable must now be a protected type (there was no such restriction
before).

Minors corrections were added by the 2002 revision of the VHDL standard.  This
revision is not fully backward compatible with VHDL-00 since, for example,
the value of the @code{'instance_name} attribute has slightly changed.

You can select the VHDL standard expected by GHDL with the
@samp{--std=VER} option, where @var{VER} is one of the left column of the
table below:

@table @samp
@item 87
Select VHDL-87 standard as defined by IEEE 1076-1987.  LRM bugs corrected by
later revisions are taken into account.
@item 93
Select VHDL-93; VHDL-87 file declarations are not accepted.
@item 93c
Select VHDL-93 standard with relaxed rules:
@itemize @bullet
@item
VHDL-87 file declarations are accepted;
@item
default binding indication rules of VHDL-02 are used.  Default binding rules
are often used, but they are particularly obscure before VHDL-02.
@end itemize
@item 00
Select VHDL-2000 standard, which adds protected types.
@item 02
Select VHDL-2002 standard (partially implemented).
@end table

You cannot mix VHDL-87 and VHDL-93 units.  A design hierarchy must have been
completely analyzed using either the 87 or the 93 version of the VHDL standard.

@node PSL implementation, Source representation, VHDL standards, GHDL implementation of VHDL
@comment  node-name,  next,  previous,  up
@section PSL implementation
GHDL understands embedded PSL annotations in VHDL files, but not in
separate files.

As PSL annotations are embedded within comments, you must analyze and elaborate
your design with option @option{-fpsl} to enable PSL annotations.

A PSL assertion statement must appear within a comment that starts
with the @code{psl} keyword.  The keyword must be followed (on the
same line) by a PSL keyword such as @code{assert} or @code{default}.
To continue a PSL statement on the next line, just start a new comment.

A PSL statement is considered as a process.  So it is not allowed within
a process.

All PSL assertions must be clocked (GHDL doesn't support unclocked assertion).
Furthermore only one clock per assertion is allowed.

You can either use a default clock like this:
@example
  -- psl default clock is rising_edge (CLK);
  -- psl assert always
  --   a -> eventually! b;
@end example
or use a clocked expression (note the use of parenthesis):
@example
  -- psl assert (always a -> next[3](b)) @@rising_edge (clk);
@end example

Of course only the simple subset of PSL is allowed.

Currently the built-in functions are not implemented.

@node Source representation, Library database, PSL implementation, GHDL implementation of VHDL
@comment  node-name,  next,  previous,  up
@section Source representation
According to the VHDL standard, design units (i.e. entities,
architectures, packages, package bodies and configurations) may be
independently analyzed.

Several design units may be grouped into a design file.

In GHDL, a system file represents a design file.  That is, a file compiled by
GHDL may contain one or more design units.

It is common to have several design units in a design file.

GHDL does not impose any restriction on the name of a design file
(except that the filename may not contain any control character or
spaces).

GHDL do not keep a binary representation of the design units analyzed like
other VHDL analyzers.  The sources of the design units are re-read when
needed (for example, an entity is re-read when one of its architecture is
analyzed).  Therefore, if you delete or modify a source file of a unit
analyzed, GHDL will refuse to use it.

@node Library database, VHDL files format, Source representation, GHDL implementation of VHDL
@section Library database
Each design unit analyzed is placed into a design library.  By default,
the name of this design library is @samp{work}; however, this can be
changed with the @option{--work=NAME} option of GHDL.

To keep the list of design units in a design library, GHDL creates
library files.  The name of these files is @samp{NAME-objVER.cf}, where
@var{NAME} is the name of the library, and @var{VER} the VHDL version (87
or 93) used to analyze the design units.

You don't have to know how to read a library file.  You can display it
using the @option{-d} of @code{ghdl}.  The file contains the name of the
design units, as well as the location and the dependencies.

The format may change with the next version of GHDL.

@node VHDL files format, Top entity, Library database, GHDL implementation of VHDL
@comment  node-name,  next,  previous,  up
@section VHDL files format
@cindex file format
@cindex logical name
VHDL has features to handle files.

GHDL associates a file logical name (the VHDL filename) to an operating
system filename.  The logical name @samp{STD_INPUT} is associated to
the standard input as defined by @samp{stdin} stream of the C library,
while the logical name @samp{STD_OUTPUT} is associated to the standard
output, as defined by the @samp{stdout} stream of the C library.  Other
logical name are directly mapped to a filename as defined by the first
(@samp{path}) argument of the @samp{fopen} function of the C library.
For a binary file, the @samp{b} character is appended to the mode argument
(binary mode).

If multiple file objects are associated with the same external file, a stream
is created for each object, except for the standard input or output.

GHDL has no internal restrictions on the number of file objects that are
associated at one time with a given external file, but the operating system
may restrict the maximum number of file open at the same time.

For more details about these point, please refer to your operation system
documentation.

@c tell more about possible errors.

There are two kinds of files: binary or text files.

Text files are files of type @samp{std.textio.text}.  The format is the
same as the format of any ascii file.  In VHDL-87, only the first 128
characters (7 bits) are allowed, since the character type has only 128
literals.  The end of line is system dependent.  Note that the stdio
functions with the text mode are used to handle text files: the fgets
function is used to read lines.  Please, refer to the manual of your C
library for more information.

There are two kind of binary files, according to the type mark of the
file.  According to the VHDL standard, binary files must be read using
the same type they are written.

If the type mark is a non-composite type (integer, floating type
enumeration, physical), the file is a raw stream:
elements are read or written using the same format as is used to represent
the data in memory.  This is highly non-portable, but you should be able
to read file written by a non-@code{GHDL} program.

If the type mark is a composite type (record or array), the file is composed
of a 2 lines signature, followed by a raw stream.

@node Top entity, Using vendor libraries, VHDL files format, GHDL implementation of VHDL
@comment  node-name,  next,  previous,  up
@section Top entity
There are some restrictions on the entity being at the apex of a design
hierarchy:

@itemize @bullet
@item
The generic must have a default value, and the value of a generic is its
default value;
@item
The ports type must be constrained.
@end itemize

@node Using vendor libraries, Interfacing to other languages, Top entity, GHDL implementation of VHDL
@comment  node-name,  next,  previous,  up
@section Using vendor libraries
Many vendors libraries have been analyzed with GHDL.  There are
usually no problems.  Be sure to use the @option{--work=} option.
However, some problems have been encountered.

GHDL follows the VHDL LRM (the manual which defines VHDL) more
strictly than other VHDL tools.  You could try to relax the
restrictions by using the @option{--std=93c}, @option{-fexplicit},
@option{-frelaxed-rules} and @option{--warn-no-vital-generic}.

@node Interfacing to other languages,  , Using vendor libraries, GHDL implementation of VHDL
@comment  node-name,  next,  previous,  up
@section Interfacing to other languages
@cindex interfacing
@cindex other languages
@cindex foreign
@cindex VHPI
@cindex VHPIDIRECT
Interfacing with foreign languages is possible only on GNU/Linux systems.

You can define a subprogram in a foreign language (such as @code{C} or
@code{Ada}) and import it in a VHDL design.

@subsection Foreign declarations
Only subprograms (functions or procedures) can be imported, using the foreign
attribute.  In this example, the @code{sin} function is imported:

@example
package math is
  function sin (v : real) return real;
  attribute foreign of sin : function is "VHPIDIRECT sin";
end math;

package body math is
  function sin (v : real) return real is
  begin
    assert false severity failure;
  end sin;
end math;
@end example

A subprogram is made foreign if the @var{foreign} attribute decorates
it.  This attribute is declared in the 1993 revision of the
@samp{std.standard} package.  Therefore, you cannot use this feature in
VHDL 1987.

The decoration is achieved through an attribute specification.  The
attribute specification must be in the same declarative part as the
subprogram and must be after it.  This is a general rule for specifications.
The value of the specification must be a locally static string.

Even when a subprogram is foreign, its body must be present.  However, since
it won't be called, you can made it empty or simply but an assertion.

The value of the attribute must start with @samp{VHPIDIRECT } (an
upper-case keyword followed by one or more blanks).  The linkage name of the
subprogram follows.


@menu
* Restrictions on foreign declarations::  
* Linking with foreign object files::  
* Starting a simulation from a foreign program::  
* Linking with Ada::            
* Using GRT from Ada::          
@end menu

@node Restrictions on foreign declarations, Linking with foreign object files, Interfacing to other languages, Interfacing to other languages
@subsection Restrictions on foreign declarations

Any subprogram can be imported.  GHDL puts no restrictions on foreign
subprograms.  However, the representation of a type or of an interface in a
foreign language may be obscure.  Most of non-composite types are easily imported:
@table @samp
@item integer types
They are represented on a 32 bits word.  This generally corresponds to
@code{int} for @code{C} or @code{Integer} for @code{Ada}.
@item physical types
They are represented on a 64 bits word.  This generally corresponds to the
@code{long long} for @code{C} or @code{Long_Long_Integer} for @code{Ada}.
@item floating point types
They are represented on a 64 bits floating point word.  This generally
corresponds to @code{double} for @code{C} or @code{Long_Float} for @code{Ada}.
@item enumeration types
They are represented on 8 bits or 32 bits word, if the number of literals is
greater than 256.  There is no corresponding C types, since arguments are
not promoted.
@end table

Non-composite types are passed by value.  For the @code{in} mode, this
corresponds to the @code{C} or @code{Ada} mechanism.  The @code{out} and
@code{inout} interfaces of non-composite types are gathered in a record
and this record is passed by reference as the first argument to the
subprogram.  As a consequence, you shouldn't use @code{in} and
@code{inout} modes in foreign subprograms, since they are not portable.

Records are represented like a @code{C} structure and are passed by reference
to subprograms.

Arrays with static bounds are represented like a @code{C} array, whose
length is the number of elements, and are passed by reference to subprograms.

Unconstrained array are represented by a fat pointer.  Do not use unconstrained
arrays in foreign subprograms.

Accesses to an unconstrained array is a fat pointer.  Other accesses correspond to an address and are passed to a subprogram like other non-composite types.

Files are represented by a 32 bits word, which corresponds to an index
in a table.

@node Linking with foreign object files, Starting a simulation from a foreign program, Restrictions on foreign declarations, Interfacing to other languages
@subsection Linking with foreign object files
You may add additional files or options during the link using the
@option{-Wl,} of @code{GHDL}, as described in @ref{Elaboration command}.
For example:

@example
$ ghdl -e -Wl,-lm math_tb
@end example
will create the @file{math_tb} executable with the @file{lm} (mathematical)
library.

Note the @file{c} library is always linked with an executable.

@node Starting a simulation from a foreign program, Linking with Ada, Linking with foreign object files, Interfacing to other languages
@subsection Starting a simulation from a foreign program
You may run your design from an external program.  You just have to call
the @samp{ghdl_main} function which can be defined:

in C:
@smallexample
extern int ghdl_main (int argc, char **argv);
@end smallexample

in Ada:
@smallexample
with System;
@dots{}
function Ghdl_Main (Argc : Integer; Argv : System.Address)
   return Integer;
pragma import (C, Ghdl_Main, "ghdl_main");
@end smallexample

This function must be called once, and returns 0 at the end of the simulation.
In case of failure, this function does not return.  This has to be fixed.

@node Linking with Ada, Using GRT from Ada, Starting a simulation from a foreign program, Interfacing to other languages
@subsection Linking with Ada
As explained previously in @ref{Starting a simulation from a foreign program},
you can start a simulation from an @code{Ada} program.  However the build
process is not trivial: you have to elaborate your @code{Ada} program and your
@code{VHDL} design.

First, you have to analyze all your design files.  In this example, we
suppose there is only one design file, @file{design.vhdl}.
@smallexample
$ ghdl -a design.vhdl
@end smallexample
Then, bind your design.  In this example, we suppose the entity at the
design apex is @samp{design}.
@smallexample
$ ghdl --bind design
@end smallexample
Finally, compile, bind your @code{Ada} program at link it with your @code{VHDL}
design:
@smallexample
$ gnatmake my_prog -largs `ghdl --list-link design`
@end smallexample

@node Using GRT from Ada,  , Linking with Ada, Interfacing to other languages
@comment  node-name,  next,  previous,  up
@subsection Using GRT from Ada
@quotation Warning
This topic is only for advanced users knowing how to use @code{Ada}
and @code{GNAT}.  This is provided only for reference, I have tested
this once before releasing @code{GHDL} 0.19 but this is not checked at
each release.
@end quotation

The simulator kernel of @code{GHDL} named @dfn{GRT} is written in
@code{Ada95} and contains a very light and slightly adapted version
of @code{VHPI}.  Since it is an @code{Ada} implementation it is
called @dfn{AVHPI}. Although being tough, you may interface to @code{AVHPI}.

For using @code{AVHPI}, you need the sources of @code{GHDL} and to recompile
them (at least the @code{GRT} library).  This library is usually compiled with
a @code{No_Run_Time} pragma, so that the user does not need to install the
@code{GNAT} runtime library.  However, you certainly want to use the usual
runtime library and want to avoid this pragma.  For this, reset the
@var{GRT_PRAGMA_FLAG} variable.
@smallexample
$ make GRT_PRAGMA_FLAG= grt-all
@end smallexample

Since @code{GRT} is a self-contained library, you don't want
@code{gnatlink} to fetch individual object files (furthermore this
doesn't always work due to tricks used in @code{GRT}).  For this,
remove all the object files and make the @file{.ali} files read-only.
@smallexample
$ rm *.o
$ chmod -w *.ali
@end smallexample

You may then install the sources files and the @file{.ali} files.  I have never
tested this step.

You are now ready to use it.

For example, here is an example, @file{test_grt.adb} which displays the top
level design name.
@example
with System; use System;
with Grt.Avhpi; use Grt.Avhpi;
with Ada.Text_IO; use Ada.Text_IO;
with Ghdl_Main;

procedure Test_Grt is
   --  VHPI handle.
   H : VhpiHandleT;
   Status : Integer;
   
   --  Name.
   Name : String (1 .. 64);
   Name_Len : Integer;
begin
   --  Elaborate and run the design.
   Status := Ghdl_Main (0, Null_Address);
   
   --  Display the status of the simulation.
   Put_Line ("Status is " & Integer'Image (Status));
   
   --  Get the root instance.
   Get_Root_Inst(H);
   
   --  Disp its name using vhpi API.
   Vhpi_Get_Str (VhpiNameP, H, Name, Name_Len);
   Put_Line ("Root instance name: " & Name (1 .. Name_Len));
end Test_Grt;
@end example

First, analyze and bind your design:
@smallexample
$ ghdl -a counter.vhdl
$ ghdl --bind counter
@end smallexample

Then build the whole:
@smallexample
$ gnatmake test_grt -aL@var{grt_ali_path} -aI@var{grt_src_path} -largs
 `ghdl --list-link counter`
@end smallexample

Finally, run your design:
@smallexample
$ ./test_grt
Status is  0
Root instance name: counter
@end smallexample

@node  GHDL implementation of VITAL, Flaws and bugs report, GHDL implementation of VHDL, Top
@comment  node-name,  next,  previous,  up
@chapter GHDL implementation of VITAL
@cindex VITAL
@cindex IEEE 1076.4
@cindex 1076.4
This chapter describes how VITAL is implemented in GHDL.  Support of VITAL is
really in a preliminary stage.  Do not expect too much of it as now.

@menu
* VITAL packages::              
* VHDL restrictions for VITAL::  
* Backannotation::              
* Negative constraint calculation::  
@end menu

@node VITAL packages, VHDL restrictions for VITAL, GHDL implementation of VITAL, GHDL implementation of VITAL
@comment  node-name,  next,  previous,  up
@section VITAL packages
The VITAL standard or IEEE 1076.4 was first published in 1995, and revised in
2000.

The version of the VITAL packages depends on the VHDL standard.  VITAL
1995 packages are used with the VHDL 1987 standard, while VITAL 2000
packages are used with other standards.  This choice is based on the
requirements of VITAL: VITAL 1995 requires the models follow the VHDL
1987 standard, while VITAL 2000 requires the models follow VHDL 1993.

The VITAL 2000 packages were slightly modified so that they conform to
the VHDL 1993 standard (a few functions are made pure and a few one
impure).

@node  VHDL restrictions for VITAL, Backannotation, VITAL packages, GHDL implementation of VITAL
@comment  node-name,  next,  previous,  up
@section VHDL restrictions for VITAL
The VITAL standard (partially) implemented is the IEEE 1076.4 standard
published in 1995.

This standard defines restriction of the VHDL language usage on VITAL
model.  A @dfn{VITAL model} is a design unit (entity or architecture)
decorated by the @code{VITAL_Level0} or @code{VITAL_Level1} attribute.
These attributes are defined in the @code{ieee.VITAL_Timing} package.

Currently, only VITAL level 0 checks are implemented.  VITAL level 1 models
can be analyzed, but GHDL doesn't check they comply with the VITAL standard.

Moreover, GHDL doesn't check (yet) that timing generics are not read inside
a VITAL level 0 model prior the VITAL annotation.

The analysis of a non-conformant VITAL model fails.  You can disable the
checks of VITAL restrictions with the @option{--no-vital-checks}.  Even when
restrictions are not checked, SDF annotation can be performed.

@node  Backannotation, Negative constraint calculation, VHDL restrictions for VITAL, GHDL implementation of VITAL
@comment  node-name,  next,  previous,  up
@section Backannotation
@cindex SDF
@dfn{Backannotation} is the process of setting VITAL generics with timing
information provided by an external files.

The external files must be SDF (Standard Delay Format) files.  GHDL
supports a tiny subset of SDF version 2.1, other version number can be
used, provided no features added by the next version are used.

Hierarchical instance names are not supported. However you can use a list of
instances.  If there is no instance, the top entity will be annotated and
the celltype must be the name of the top entity.  If there is at least one
instance, the last instance name must be a component instantiation label, and
the celltype must be the name of the component declaration instantiated.

Instances being annotated are not required to be VITAL compliant.  However
generics being annotated must follow rules of VITAL (e.g., type must be a
suitable vital delay type).

Currently, only timing constraints applying on a timing generic of type
@code{VitalDelayType01} has been implemented.  This SDF annotator is
just a proof of concept.  Features will be added with the following GHDL
release.

@node  Negative constraint calculation,  , Backannotation, GHDL implementation of VITAL
@comment  node-name,  next,  previous,  up
@section Negative constraint calculation
Negative constraint delay adjustment are necessary to handle negative
constraint such as a negative setup time.  This step is defined in the VITAL
standard and should occur after backannotation.

GHDL does not do negative constraint calculation.  It fails to handle models
with negative constraint.  I hope to be able to add this phase soon.

@node Flaws and bugs report, Copyrights, GHDL implementation of VITAL, Top
@comment  node-name,  next,  previous,  up
@chapter Flaws and bugs report

The current version of GHDL is really a beta version.  Some features of
VHDL have not been implemented or are only partially implemented.  Besides,
GHDL has not been extensively tested yet.

@menu
* Deficiencies::                
* Reporting bugs::              
* Future improvements::         
@end menu

@node Deficiencies, Reporting bugs, Flaws and bugs report, Flaws and bugs report
@comment  node-name,  next,  previous,  up
@section Deficiencies
Here is the non-exhaustive list of flaws:

@itemize @bullet
@item
So far, @code{GHDL} has been compiled and tested only on @samp{i386-linux} systems.
@item
Overflow detection is not yet implemented.
@item
Some constraint checks are missing.
@item
VHDL-93 is not completely implemented.
@item
There are no checks for elaboration order.
@item
This list is not exhaustive.
@item
@dots{}
@end itemize

@node Reporting bugs, Future improvements, Deficiencies, Flaws and bugs report
@comment  node-name,  next,  previous,  up
@section Reporting bugs
In order to improve GHDL, we welcome bugs report and suggestions for
any aspect of GHDL.  Please use the bug tracker on
@indicateurl{http://gna.org/projects/ghdl}.  You may also send an
email to @email{ghdl@@free.fr}.

If the compiler crashes, this is a bug.  Reliable tools never crash.
  
If your compiled VHDL executable crashes, this may be a bug at
runtime or the code produced may be wrong.  However, since VHDL
has a notion of pointers, an erroneous VHDL program (using invalid
pointers for example) may crash.

If the compiler emits an error message for a perfectly valid input or
does not emit an error message for an invalid input, this may be a bug.
Please send the input file and what you expected.  If you know the LRM
well enough, please specify the paragraph which has not been well
implemented.  If you don't know the LRM, maybe your bug report will be
rejected simply because there is no bug.  In the latter case, it may be
difficult to discuss the issue; and comparisons with other VHDL tools
is not a very strong argument.

If a compiler message is not clear enough for you, please tell me.  The
error messages can be improved, but I have not enough experience with
them.

If you have found a mistake in the manual, please send a comment.  If
you have not understood some parts of this manual, please tell me.
English is not my mother tongue, so this manual may not be well-written.
Again, rewriting part of it is a good way to improve it.

If you send a @code{VHDL} file producing a bug, it is a good idea to try
to make it as short as possible.  It is also a good idea to make it
looking like a test: write a comment which explains whether the file
should compile, and if yes, whether or not it should run successfully.
In the latter case, an assert statement should finish the test; the
severity level note indicates success, while a severity level failure
indicates failure.

For bug reports, please include enough information for the maintainers to
reproduce the problem. This includes:

@itemize @bullet
@item
the version of @code{GHDL} (you can get it with @samp{ghdl --version}).
@item
the operating system
@item
whether you have built @code{GHDL} from sources or used the binary
distribution.
@item
the content of the input files
@item
a description of the problem and samples of any erroneous input
@item
anything else that you think would be helpful.
@end itemize

@node Future improvements,  , Reporting bugs, Flaws and bugs report
@comment  node-name,  next,  previous,  up
@section Future improvements
I have several axes for @code{GHDL} improvements:
@itemize @bullet
@item
Documentation.
@item
Better diagnostics messages (warning and error).
@item
Full support of VHDL-87 and VHDL-93.
@item
Support of VHDL-02.
@item
Optimization (simulation speed).
@item
Graphical tools (to see waves and to debug)
@item
Style checks
@item
VITAL acceleration
@end itemize

@c And without any order:
@c VHPI
@c FOREIGN
@c AMS
@c verilog

@node Copyrights, Index, Flaws and bugs report, Top
@comment  node-name,  next,  previous,  up
@chapter Copyrights

The GHDL front-end, the @samp{std.textio} package and the runtime
library (grt) are copyrighted Tristan Gingold, come with @emph{absolutely
no warranty}, and are distributed under the conditions of the General
Public License.

The @samp{ieee.numeric_bit} and @samp{ieee.numeric_std} packages are
copyrighted by the IEEE.  The source files may be distributed without
change, except as permitted by the standard.
@comment FIXME: this sounds strange
This source file may not be
sold or distributed for profit.  See the source file and the IEEE 1076.3
standard for more information.

The @samp{ieee.std_logic_1164} package is copyrighted by the IEEE.  See
source file and the IEEE 1164 standard for more information.

The @samp{ieee.VITAL_Primitives}, @samp{ieee.VITAL_Timing} and
@samp{ieee.VITAL_Memory} packages are copyrighted by IEEE.  See source
file and the IEEE 1076.4 standards for more information.

The @samp{ieee.Math_Real} and @samp{ieee.Math_Complex} packages are
copyrighted by IEEE.  These are draft versions which may used and distributed
without restriction.  These packages cannot be sold or distributed for profit.
See source files for more information.

The packages @samp{std_logic_arith}, @c @samp{std_logic_misc},
@samp{std_logic_signed}, @samp{std_logic_unsigned} and
@samp{std_logic_textio} contained in the @samp{synopsys} directory are
copyrighted by Synopsys, Inc.  The source files may be used and
distributed without restriction provided that the copyright statements
are not removed from the files and that any derivative work contains the
copyright notice.  See the source files for more information.

The package @samp{std_logic_arith} contained in the @samp{mentor}
directory is copyrighted by Mentor Graphics.  The source files may be
distributed in whole without restriction provided that the copyright
statement is not removed from the file and that any derivative work
contains this copyright notice.  See the source files for more information.

As a consequence of the runtime copyright, you may not be allowed to
distribute an executable produced by @code{GHDL} without the VHDL
sources.  To my mind, this is not a real restriction, since there is no
points in distributing VHDL executable.  Please, send a comment
(@pxref{Reporting bugs}) if you don't like this policy.

@node Index,  , Copyrights, Top
@unnumbered Index
@printindex cp

@bye