From mboxrd@z Thu Jan  1 00:00:00 1970
X-Spam-Checker-Version: SpamAssassin 3.4.4 (2020-01-24) on polar.synack.me
X-Spam-Level: 
X-Spam-Status: No, score=-0.9 required=5.0 tests=BAYES_00,FORGED_GMAIL_RCVD,
	FREEMAIL_FROM,T_FILL_THIS_FORM_SHORT autolearn=no autolearn_force=no
	version=3.4.4
X-Google-Thread: 103376,b59b337045eece60,start
X-Google-Attributes: gid103376,domainid0,public,usenet
X-Google-Language: ENGLISH,ASCII-7-bit
Path: 
 g2news1.google.com!postnews.google.com!m36g2000hse.googlegroups.com!not-for-mail
From: Maciej Sobczak <see.my.homepage@gmail.com>
Newsgroups: comp.lang.ada
Subject: Structure of the multitasking server
Date: Fri, 19 Sep 2008 05:21:52 -0700 (PDT)
Organization: http://groups.google.com
Message-ID: 
 <8b4d1170-22e6-40d3-8ed1-096dc0163491@m36g2000hse.googlegroups.com>
NNTP-Posting-Host: 128.141.45.232
Mime-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit
X-Trace: posting.google.com 1221826912 19890 127.0.0.1 (19 Sep 2008 12:21:52
 GMT)
X-Complaints-To: groups-abuse@google.com
NNTP-Posting-Date: Fri, 19 Sep 2008 12:21:52 +0000 (UTC)
Complaints-To: groups-abuse@google.com
Injection-Info: m36g2000hse.googlegroups.com; posting-host=128.141.45.232;
	posting-account=bMuEOQoAAACUUr_ghL3RBIi5neBZ5w_S
User-Agent: G2/1.0
X-HTTP-UserAgent: Mozilla/5.0 (Macintosh; U; Intel Mac OS X 10.5; en-US;
	rv:1.9.0.1) Gecko/2008070206 Firefox/3.0.1,gzip(gfe),gzip(gfe)
Xref: g2news1.google.com comp.lang.ada:2044
Date: 2008-09-19T05:21:52-07:00
List-Id: <comp.lang.ada>

Hi all,

Imagine a server with fixed number of worker tasks. There is no queue
of jobs and jobs are immediately passed to one of the tasks that is
currently idle. There is a separate task (or just the main one) that
provides jobs for worker tasks.

I am concerned with the proper structure of objects - I mean in the
sense of recommended Ada practice.
Obviously there is a need for some shared resource where the
requesting task will put the job and from where the worker task will
pick it up.

This is more or less what I came up with, where the "channel" is a
single processing pipeline:

   type Worker_State is (Idle, Ready, Working);

   protected type Channel_State is
      procedure Post (J : in Job_Type);
      entry Get_Job (J : out Job_Type);
      function Busy return Boolean;
   private
      State : Worker_State := Idle;
      Job_To_Do : Job_Type;
   end Channel_State;

   protected body Channel_State is

      procedure Post (J : in Job_Type) is
      begin
         if State /= Idle then
            raise Program_Error;
         end if;

         Job_To_Do := J;
         State := Ready;
      end Post;

      entry Get_Job (J : out Job_Type) when State = Ready is
      begin
         J := Job_To_Do;
         State := Working;
      end Get_Job;

      function Busy return Boolean is
      begin
         return State /= Idle;
      end Busy;

   end Channel_State;

   type Channel;
   task type Worker_Task (Ch : access Channel);

   type Channel is record
      State : Channel_State;
      Worker : Worker_Task (Channel'Access);
   end record;

   task body Worker_Task is
      Job : Job_Type;
   begin
      loop
         Ch.all.Get_Job (Job);

         --  do the job ...

      end loop;
   end Worker_Task;

   Max_Channels : constant := 5;

   Channels : array (1 .. Max_Channels) of Channel;

My question is whether this is what a seasoned Ada programmer would
do.
Initially I tried to have two separate arrays, one for jobs and one
for worker tasks, but I found it difficult to link workers with their
respective jobs. After bundling them together in a single record that
is referenced from the task it worked and I actually find it
structured better.

The main task after constructing a job object finds some channel where
the worker task is not busy and posts the job to its shared state
component:

   loop
      Job := ...

      Found_Worker := False;
      for I in Channels'Range loop
         if not Channels (I).State.Busy then
            Channels (I).State.Post (Job);
            Found_Worker := True;
            exit;
         end if;
      end loop;

      if not Found_Worker then
         --  all pipelines are busy,
         --  but the overflow handling is not shown...
      end if;
   end loop;

All this works fine, but my question considers the choice of language
constructs and idioms.

--
Maciej Sobczak * www.msobczak.com * www.inspirel.com

Database Access Library for Ada: www.inspirel.com/soci-ada