LCS2016_045c < P1076

P1076 Web>VHDL2017>LCS2016_045c (2017-07-24, JimLewis) (raw view)EditAttach
<sticky><!--
   * Local SB = <sticky><b>
   * Local BS = </b></sticky>
--></sticky>

---+!! Language Change Specification for Interface Construct and Port Mode Configurations Proposal

%TOC%

| %SB%LCS Number:%BS%  | LCS-2016-045c |
| %SB%Version:%BS%     | 5 |
| %SB%Date:%BS%        | 26-Mar-17 |
| %SB%Status:%BS%      | Voting |
| %SB%Author:%BS%      | [[Main.BrentHahoe][Brent Hayhoe]] and [[Main.PatrickLehmann][Patrick Lehmann]] |
| %SB%Email:%BS%       | %USERINFO{"BrentHahoe" format="$emails"}% |
| %SB%Source Doc:%BS%  | [[InterfaceConstructandPortModeConfigurations#Analysis][Interface Construct and Port Mode Configurations]] %BR% \
                         [[UCComplexRTLSignalCPUInterface][Use Case - Complex RTL Signal CPU Interface]] |
| %SB%History Doc:%BS%  | [[LCS2016_045c_History][History]]  |
| %SB%Summary:%BS%     | This LCS details a new attribute for the mode view construct. |
| %SB%Related LCS:%BS% | [[LCS2016_045a][LCS-2016-045a]] %BR% \
                         [[TopLCS2016_045b][LCS-2016-045b]] |

---++ Voting Results: Cast your votes here

   * Yes:
      1. %USERSIG{BrentHahoe - 2017-03-21}% - Version 4
      1. %USERSIG{JimLewis - 2017-03-12}% - Version 5
      1. %USERSIG{SteveGrout - 2017-03-16}% - Version 1
      1. %USERSIG{RobGaddi - 2017-03-30}% - Version 5 - conditional of course on approval of 045a
      1. %USERSIG{PatrickLehmann - 2017-03-26}% - Version 5

   * No:
      1. %USERSIG{PeterFlake - 2017-03-16}% - Version 1

   * Abstain:

---++ Style Notes
<noautolink>
<sticky>
   * LRM text is shown between single quotes <sticky>%BLACK%'... P1076 LRM quoted text ...'%ENDCOLOR%</sticky>
   * Existing LRM text is shown in <sticky>%BLACK%<b>BLACK</b> coloured font%ENDCOLOR%</sticky>
   * Additional LRM text is shown in <sticky>%RED%<b>RED</b> coloured font%ENDCOLOR%</sticky>
   * Deleted LRM text is shown in <sticky>%RED%<strike><b>RED</b> with strike-through</strike>%ENDCOLOR%</sticky>
   * In order to try and emphasize text in italic font:
      * Italic font in existing LRM text is shown in <sticky>%GRAY%<i><b>GRAY</b> italic font</i>%BLACK%</sticky>
      * Italic font in new LRM text is shown in <sticky>%ORANGE%<i><b>ORANGE</b> italic font</i>%ENDCOLOR%</sticky>
      * Italic font in deleted LRM text is shown in <sticky>%ORANGE%<strike><i><b>ORANGE</b> with strike-through</i></strike>%ENDCOLOR%</sticky>
   * Editing instructions (and informative annotations) are shown in <sticky>%GREEN%<b>GREEN</b> coloured font%ENDCOLOR%</sticky>

---++ Reviewing Notes
*Version 5:* 26-Mar-2017
   * Improved attribute 'CONVERSE
   * Shrank example to a minimum
	 
---++ Details of Language Change:
---+++ 16.2.1 General
Predefined attributes denote values, functions, types, subtypes, %RED%mode views,%ENDCOLOR% signals, and 
ranges associated with various kinds of named entities. These attributes are described as follows. For 
each attribute, the following information is provided: %BR% \
   &emsp;&mdash;&ensp;The kind of attribute: value, type, subtype, %RED%mode view,%ENDCOLOR% range, function, or signal %BR% \
   &emsp;&mdash;&ensp;The prefixes for which the attribute is defined %BR% \
   &emsp;&mdash;&ensp;A description of the parameter or argument, if one exists %BR% \
   &emsp;&mdash;&ensp;The result of evaluating the attribute, and the result type (if applicable) %BR% \
   &emsp;&mdash;&ensp;Any further restrictions or comments that apply

---+++ 16.2.7 Predefined attributes of named mode views %GREEN%[New section]%ENDCOLOR%

%TABLE{disableallsort="on" tableframe="void" tablerules ="none" valign="baseline" dataalign="left" tablewidth ="800" columnwidths="15%,10%,70%"}%
 | %RED%M'CONVERSE%ENDCOLOR% | %RED%Kind:%ENDCOLOR%            | %RED%Mode view.%ENDCOLOR% |
 |            | %RED%Prefix:%ENDCOLOR%          | %RED%Any named mode view M of composite type T, or an alias thereof.%ENDCOLOR% |
 |            | %RED%Result:%ENDCOLOR%          |>> %RED%A mode view of composite type T with each element mode EM declared as the converse as that of mode view M.%ENDCOLOR%
   * %RED%If EM is of mode *in*, the converse mode is *out*.%ENDCOLOR%
   * %RED%If EM is of mode *out*, the converse mode is *in*.%ENDCOLOR%
   * %RED%If EM is of mode *inout*, the converse mode is *inout*.%ENDCOLOR%
   * %RED%If EM is of mode *buffer*, the converse mode is *in*.%ENDCOLOR%
   * %BLUE%If EM is of mode *null*, the converse mode is *null*.%ENDCOLOR% %GREEN%[If LCS 045b is approved]%ENDCOLOR%
   * %RED%If EM is a mode view, the converse mode is EM'CONVERSE.%ENDCOLOR% <<|

	 
%ORANGE%
_Example:_
<pre>%RED%
-- A FIFO like streaming interface:
<b>type</b> StreamingIf <b>is record</b>
  Valid  : Std_uLogic;
  Data   : Std_uLogic_Vector(7 <b>downto</b> 0);
  Ack    : Std_uLogic;
<b>end record</b> StreamingIf;

<b>view</b> MasterView <b>of</b> StreamingIf <b>is</b>
   Valid : <b>out</b>;
   Data  : <b>out</b>;
   Ack   : <b>in</b>;
<b>end view</b> MasterView;

<b>alias</b> SlaveView is MasterView'CONVERSE;
-- Equivalent mode view declaration
-- <b>view</b> SlaveView <b>of</b> MasterView <b>is</b>
--    Valid : <b>in</b>;
--    Data  : <b>in</b>;
--    Ack   : <b>out</b>;
-- <b>end view</b> SlaveView;

-- A stream processing element
<b>entity</b> SPE <b>is</b>
  <b>port (</b>
    Clock  : <b>in</b>   Std_uLogic;
    Reset  : <b>in</b>   Std_uLogic;
    Input  : <b>view</b> SlaveView;   -- input from previous SPE
    Output : <b>view</b> MasterView   -- output to next SPE
  <b>)</b>;
<b>end entity</b>;
%ENDCOLOR%</pre>

---+ Comments
%BLACK%

This LCS does not show where the attribute is used.

-- %BUBBLESIG{PeterFlake - 2017-03-16}%

!@PF: Use is shown in the new examples block.

%RED%<u><b>Version 2</b></u>

   * Attribute mode description updated for subviews.
   * Examples of use added.

-- %BUBBLESIG{BrentHahoe - 2017-03-19}%


The list of possible kinds must be edited (16.2.1).

Why are aliases excluded from the prefix?

Why is =M&#39;converse&#39;converse= not allowed?

Why uses the example 9-bit for e.g. =RdData=?

Why uses =BusSlaveView= such a heavy construct instead of an alias? =alias BusSlaveView is BusMasterView&#39;CONVERSE=.

-- %BUBBLESIG{PatrickLehmann - 2017-03-20}%


!@PL2:  Likewise.  For an example, see array attributes or LCS 018

!@PL5:  Likewise.   Example seems to be inconsistent with syntax.   If &#39;CONVERSE is a record view, then it should be stand-alone or used with an alias and not inside of a view declaration.


-- %BUBBLESIG{JimLewis - 2017-03-20}%

!@PL1: Added modifications to include mode view entity class kind.

!@PL2 & %BR% \
!@JL1: Added missing alias reference in attribute prefix.

!@PL3: Why is it disallowed?

!@PL4: Why not? This is example code. Have you not heard of parity?

!@PL5: The examples are there to indicate how one can define views and not necessarily in the most efficient manner. I have added an alias based declaration example.

!@JL2: 'CONVERSE returns a mode view. Mode view declarations may contain mode views of composite elements within that declaration.

%RED%<u><b>Version 3</b></u>

   * Added modifications to section 16.2.1
   * Added alias to attribute prefix section 16.2.6
   * Added extra example illustrating use of *alias* to declare a mode *view*

-- %BUBBLESIG{BrentHahoe - 2017-03-20}%


!@BH WRT &#39;CONVERSE the example I object to is:
<pre><b>view</b> BusSlaveView <b>of</b> BusRecordType <b>is</b>
   BusMasterView'CONVERSE
<b>end</b> <b>view</b> BusSlaveView;</pre>
Mode view declarations contain mode_view_element_definition.  They may contain a choice whose mode is specified with a mode view as you demonstrated with SyncBusSlaveView.  However, there are no syntax provisions for BusMasterView'CONVERSE by itself.   

I think this example should be done with an alias as follows
<pre><b>alias</b> BusSlaveView <b>is</b>  BusMasterView'CONVERSE ;</pre>

-- %BUBBLESIG{JimLewis - 2017-03-20}%


In 16.2.1, why did you change &#34;named entities&#34; to &#34;named entity classes&#34;.   The LRM uses &#34;named entity&#34; and &#34;named entities&#34; to mean any named object in numerous places.   OTOH, &#34;named entity classes&#34; is probably not defined anywhere.


-- %BUBBLESIG{JimLewis - 2017-03-20}%


I&#39;m with Jim there, I think that you can have 

<pre>

<b>type</b> HoldTheBus <b>is record</b>
  busstop : BusRecordType
<b>end record</b> HoldTheBus;

<b>view</b> BusSlaveView <b>of</b> HoldTheBus <b>is</b>
   busstop : view BusMasterView'CONVERSE
<b>end</b> <b>view</b> BusSlaveView;

</pre>

but don&#39;t see how you can legally have

<pre><b>view</b> BusSlaveView <b>of</b> BusRecordType <b>is</b>
   BusMasterView'CONVERSE
<b>end</b> <b>view</b> BusSlaveView;</pre>


-- %BUBBLESIG{RobGaddi - 2017-03-23}%


!@RG & !@JL WRT BusMasterView'CONVERSE

Point taken. I've been thinking about it this way for so long that it seemed natural.

I think *alias* is the correct way to make this converse declaration.

We now do not need the second example.

!@JL WRT 16.2.1 "entity classes"

The BNF in the attribute defines what an entity class is. We seem to have a lot of confusion over 
entities and the entity class *entity* in the text. i thought that this may clarify the text.

OTOH, we also talk about objects a lot when we mean object class. I'm not sure how we should go with this 
one and am happy to leave it as entities. There is a technical grammar here to be followed. I'm not the 
one to specify that.


---+++ Version 4

   * Removed "entity classes" reference.
   * Changed 'CONVERSE mode view declaration to be *alias* based.
   * Removed second example.

-- %BUBBLESIG{BrentHahoe - 2017-03-21}%


Do we really need such a big example in the LRM? I suggest a smaller FIFO/Streaming bus example (=Data=, =Valid=, =Ack=).

The *Result* row in the attribute description should mention that the modes are in the same order. I somehow dislike the &#60;...&#62; notation.

The wording element mode does not apply to slices.

| M'CONVERSE | Kind:        | Mode view. |
|            | Prefix:      | Any named mode view M of composite type T, or an alias thereof, as defined in 6.5.2.1. |
|            | Result:      | %RED%A mode view of composite type T with each element mode EM declared as the converse as that of mode view M.%ENDCOLOR% |
|            |              | * %RED%If EM is of mode *in*, the converse mode is *out*.%ENDCOLOR%       |
|            |              | * %RED%If EM is of mode *out*, the converse mode is *in*.%ENDCOLOR%       |
|            |              | * %RED%If EM is of mode *inout*, the converse mode is *inout*.%ENDCOLOR%        |
|            |              | * %RED%If EM is of mode *buffer*, the converse mode is *in*.%ENDCOLOR%       |
|            |              | * %RED%If EM is of mode *null*, the converse mode is *null*.%ENDCOLOR%        |
|            |              | * %RED%If EM is a mode view, the converse mode is EM'CONVERSE.%ENDCOLOR%       |
|            | %RED%Restrictions:| %RED%A mode of *linkage* is disallowed within a mode view.%ENDCOLOR% |


-- %BUBBLESIG{PatrickLehmann - 2017-03-25}%

---+++ Version 5

%COMMENT%
</sticky>
</noautolink>
Topic revision: r23 - 2017-07-24 - 19:46:30 - JimLewis
P1076