code_genome.html

来自「有趣的模拟进化的程序 由国外一生物学家开发 十分有趣」· HTML 代码 · 共 375 行 · 第 1/2 页

  class <font color="#880000">cHardwareBase</font>;
  typedef void (<font color="#880000">cHardwareBase</font>::*<font color="#880000">tHardwareMethod</font>)();
</pre>

<p>
The first command basically states that there will be a class called
<font color="#880000">cHardwareBase</font> that will be fully defined at
some point in the future.  We need to refer to the class here, but not
directly interact with any of its methods or variables, so it is sufficient
to just tell the compiler that it is a class.  This form of statement is
called a <b>predeclaration</b>.

<p>
The second line is a <tt>typedef</tt> statement, which defines a new data \
type.  A typedef looks just like a variable declaration, but the
keyword <tt>typedef</tt> indicates that the "variable" that
would have been defined is instead a new "type".  For example, the command
"<tt>typedef unsigned char UCHAR</tt>" creates a new type called "UCHAR" that
is just an unsigned char.  In the case above, we are creating a new type
called <font color="#880000">tHardwareMethod</font>, which is a pointer to
a method in the class cHardwareBase.  Basically it will point to a location
in memory that contains a method in the yet-to-be-defined class.  This is
a tricky concept -- but it is the easiest way to point to a variable method
that we don't know about until run time (when we load the instruction set
file in).

<p>
The cInstLib class is long, so I'm going to split the code presentation into
two halves, with explanation in between.  Here is the first part:

<pre>
  class <font color="#880000">cInstLib</font> {
  private:

    // This class gives full info about a single instruction in the library.
    class <font color="#880000">cInstEntry</font> {
    public:
      <font color="#880000">cString</font> <font color="#000088">name</font>;             <font color="#886600">// Name of this instruction.</font>
      <font color="#880000">tHardwareMethod</font> <font color="#000088">function</font>; <font color="#886600">// Pointer to hardware method.</font>
  
      <font color="#886600">// Some additional details to be used in the future...</font>
      <font color="#880000">int</font> <font color="#000088">redundancy</font>;           <font color="#886600">// Weight in instruction set</font>
      <font color="#880000">int</font> <font color="#000088">cost</font>;                 <font color="#886600">// additional time spent to execute inst.</font>
      <font color="#880000">int</font> <font color="#000088">ft_cost</font>;              <font color="#886600">// additional time spent first time exec</font>
      <font color="#880000">double</font> <font color="#000088">prob_fail</font>;         <font color="#886600">// probability of failing to execute inst</font>
    };
   
    <font color="#880000">tArray&lt;cInstEntry&gt;</font> <font color="#000088">inst_array</font>;  <font color="#886600">// The instructions</font>
    <font color="#880000">tArray&lt;int&gt;</font> <font color="#000088">nop_mods</font>;           <font color="#886600">// Modification table for nops</font>
  
    <font color="#886600">// Static components...</font>
    static const <font color="#880000">cInstruction</font> <font color="#000088">inst_error</font>;
    static const <font color="#880000">cInstruction</font> <font color="#000088">inst_none</font>;
    static const <font color="#880000">cInstruction</font> <font color="#000088">inst_default</font>;
</pre>

<p>
The first thing that I do in the private section of the cInstLib definition is
define another class called <font color="#880000">cInstEntry</font>, which
contains all of the information about a single entry in an instruction
library.  Since it is defined internally and privately, cInstEntry can only
be used <i>inside</i> of cInstLib; we refer to it as a <b>helper</b> class.

<p>
cInstEntry contains only pubiic data about a single instruction.  At the
moment, the only portion of that data that we are using is the instruction
name and the hardware method that it is supposed to call.  We also have
room to record how redundant this instruction is supposed to be in the 
full instruction set, how many CPU cycles should be spent as the cost to
executing it, an additional first-time cost incurred only the first time
the instruction is used by an organism, and a probability of failure, that
the instruction will not be executed at all.  Any other useful information
about an instruction can be put here, but then it should also be made use
of elsewhere in the code.

<p>
After the definition of cInstEntry, we have the declaration of an array of
entries called <font color="#000088">inst_array</font>.  The position of
an instruction in this array is the ID number that uniquely identifies that
instruction.  Thus, whichever instruction entry is placed in the array first,
that instruction has ID of 0.  The next one has an ID of 1, then 2, and so on.

<p>
The inst_array has the instructions ordered in the same order they are
presented in the inst_lib file from which they are loaded with one
caveat: all no-operation instructions must be listed first.  This is 
because the next variable, <font color="#000088">nop_mods</font> is an
array that indicates which CPU component each nop is associated with.
The size of this array determines how many nops there are, and they 
directly correspond to the same array positions in inst_array.

<p>
In the default instruction set, the inst_array has 26 entries contained
in it, since this is the size of the instruction set, the first three of
which are <tt>nop-A</tt>, <tt>nop-B</tt>, and <tt>nop-C</tt> respectively.
The nop_mods array has three integers in it, one for each of the three
nops.

<p>
The remainder of the private section of the cInstSet definition contains
three static and constant variables that represent special instructions.
They are <font color="#000088">inst_error</font>, for an illegal instruction
(it will be given the ID number 254), <font color="#000088">inst_none</font>
to represent an empty position (ID number 255), and
<font color="#000088">inst_default</font> for the default instruction to
initialize genomes with, unless otherwise specified (usually ID number 0
-- <tt>nop-A</tt> in the default set -- but this can be changed).

<p>
Here is the public portion of the cInstLib definition:

<pre>
  public:
    <font color="#008800">cInstLib()</font> { ; }
    <font color="#008800">cInstLib</font>(const <font color="#880000">cInstLib</font> & <font color="#000088">in_inst_lib</font>);
    <font color="#008800">~cInstLib()</font> { ; }
   
    <font color="#880000">cInstLib</font> & <font color="#008800">operator=</font>(const <font color="#880000">cInstLib</font> & <font color="#000088">_in</font>);
   
    <font color="#886600">// Accessors</font>
    const <font color="#880000">cString</font> & <font color="#008800">GetName</font>(const <font color="#880000">cInstruction</font> & <font color="#000088">inst</font>) const;
    <font color="#880000">tHardwareMethod</font> <font color="#008800">GetFunction</font>(const <font color="#880000">cInstruction</font> & <font color="#000088">inst</font>) const;
    <font color="#880000">int</font> <font color="#008800">GetNopMod</font>(const <font color="#880000">cInstruction</font> & <font color="#000088">inst</font>) const
      { return <font color="#000088">nop_mods</font>[<font color="#000088">inst</font>.<font color="#008800">GetOp()</font>]; }

    <font color="#880000">cInstruction</font> <font color="#008800">GetInst</font>(const <font color="#880000">cString</font> & <font color="#000088">in_name</font>) const;
   
    <font color="#880000">int</font> <font color="#008800">GetSize()</font> const { return <font color="#000088">inst_array</font>.<font color="#008800">GetSize()</font>; }
    <font color="#880000">int</font> <font color="#008800">GetNumNops()</font> const { return <font color="#000088">nop_mods</font>.<font color="#008800">GetSize()</font>; }
   
    <font color="#886600">// Instruction Analysis.</font>
    <font color="#880000">int</font> <font color="#008800">IsNop</font>(const <font color="#880000">cInstruction</font> & <font color="#000088">inst</font>) const
      { return (<font color="#000088">inst</font>.<font color="#008800">GetOp()</font> &lt; <font color="#000088">nop_mods</font>.<font color="#008800">GetSize()</font>); }
    <font color="#880000">cString</font> <font color="#008800">FindBestMatch</font>(const <font color="#880000">cString</font> & <font color="#000088">in_name</font>) const;
  
    <font color="#886600">// Insertion of new instructions...</font>
    <font color="#880000">int</font> <font color="#008800">Add</font>(const <font color="#880000">cString</font> & <font color="#000088">_name</font>, <font color="#880000">tHardwareMethod</font> <font color="#000088">_fun</font>,
            <font color="#880000">int</font> <font color="#000088">redundancy</font>=1, <font color="#880000">int</font> <font color="#000088">ft_cost</font>=0, <font color="#880000">int</font> <font color="#000088">cost</font>=0, <font color="#880000">double</font> <font color="#000088">prob_fail</font>=0.0);
    <font color="#880000">int</font> <font color="#008800">AddNop</font>(const <font color="#880000">cString</font> & <font color="#000088">_name</font>, <font color="#880000">tHardwareMethod</font> <font color="#000088">_fun</font>, <font color="#880000">int</font> <font color="#000088">reg</font>,
            <font color="#880000">int</font> <font color="#000088">redundancy</font>=1, <font color="#880000">int</font> <font color="#000088">ft_cost</font>=0, <font color="#880000">int</font> <font color="#000088">cost</font>=0, <font color="#880000">double</font> <font color="#000088">prob_fail</font>=0.0);
   
    <font color="#886600">// Static methods...</font>
    static const <font color="#880000">cInstruction</font> & <font color="#008800">GetInstDefault()</font> { return <font color="#000088">inst_default</font>; }
    static const <font color="#880000">cInstruction</font> & <font color="#008800">GetInstError()</font>   { return <font color="#000088">inst_error</font>; }
    static const <font color="#880000">cInstruction</font> & <font color="#008800">GetInstNone()</font>    { return <font color="#000088">inst_none</font>; }
  };
</pre>

<p>
The constructors, destructor, and assignment operator are all rather standard
for this class -- an instruction library can either be created from scratch
or else copied over from a pre-existing one.

<p>
The accessors in this class allow specific information to be obtained about
instructions passed in.  The <font color="#008800">GetName()</font> method
returns the human-readable name for a specific instruction,
<font color="#008800">GetFunction()</font> returns a pointer to the method
associated with an instruction, and <font color="#008800">GetNopMod()</font>
returns the id of the component modified by a nop-instruction passed in.
Conversely, <font color="#008800">GetInst()</font> will return an instruction
object when an instruction name (as a string) is passed in.
Additionally, there are the accessors <font color="#008800">GetSize()</font>
to obtain the total number of instructions in the set, and
<font color="#008800">GetNumNops()</font> will return the number of nop
instructions in the set.

<p>
Two instruction analysis methods also exist; one to determine if an
instruction is a nop called <font color="#008800">IsNop()</font>, and
another method called <font color="#008800">FindBestMatch()</font> takes
in an unknown instruction name, and returns the closest name match that it
can find in the set.  This allows Avida to make suggestions to the user
in the case of a mis-spelled instruction name.

<p>
An instruction set is initially loaded by successive calls to the
<font color="#008800">Add()</font> and <font color="#008800">AddNop()</font>,
filling in the appropriate arguments.  These methods are called from a
utility class (<font color="#880000">cHardwareUtil</font>) that manages the
loading of the inst_set file.  Next class we will be integrating the classes
discussed here with the classes that directly implment the hardware.

<p>
Finally, three static methods are included to provide access to the
three pre-defined instructions.