Thurs Sep 26, 10:00 am PDT

Attendees:
Peter Wohl
Tony Taylor
Greg Maston
Jim Teisher
Doug Sprague

Documents

http://grouper.ieee.org/groups/1450/dot1/p1450.1-D15.pdf
http://grouper.ieee.org/groups/1450/dot1/p1450.1-D15-review-resolution.pdf

Agenda
ITC meeting
Scan chain grouping
SignalVariables issues
LockStep

Jim is not available, Peter arriving late, Doug will not be present. Meeting
has fallen to Tony and Greg and 8 empty chairs.
 

Scan chain groupings

Basic proposal is a mechanism of defining groups of scanchains in the
scanstructures, and a statement in patterns
to reference these groups.
 

Issues with current proposal around: common namespace across groups
and chains, and design rule checking concerns requiring checking the
entire pattern space for "scanchaingroups-on-the-fly".
 

Current conflict between:
(a) common namespace between groups and chains
(b) two separate namespaces
 

Motivation for (a) is similarity to signals&groups. Motivation for (b)
is deterministic identification of scanchain groupings, by not
allowing the scanchain-referencing statement to create new
scanchaingroups-on-the-fly in the patterns.
 

Tony proposed that the scanchain statement in the patterns appear in
the Shift block only. This would eliminate searching the pattern space
for this statement. Problem with this is that scanchains are active
outside of the Shift statement, in pre- and post- vectors, and this
solution would not identify the scope of activity properly.
 

Peter and Denis need to resolve this [AI1].
 

SignalVariables Questions

Two issues:

(1) Current spec places all new "variables" declarations into a new block
Variables. A question was raised about having the SignalVariables in
this block, because SignalVariables are referenced like SignalGroups
and having them in the Variables block is not ideal -- they ought to
be in the SignalGroups.
 

Greg identified that while SignalVariables and SignalGroups are used
"interchangeably" on the left-hand side of an assignment, they are
*not* used interchangeable on the RHS. For example (from Tony):
 

XYZ = 0101;  // yes, var XYZ looks like signalgroup
grp = 'XYZ'; // no, only signalvariables on the RHS
 

Therefore these items are *not* interchangeable with Signals or
SignalGroups - they are truly a new item that shares SOME
characteristics with groups, and some characteristics with
variables. So Greg proposes to leave them where they are; seconded by
Doug and Jim and accepted by all.
 

Tony identified the additional issue that the namespace description in
dot1 needs to be extended to identify that signalvariables are part of
the 'signal references' namespace and therefore names across current
selected signalgroups, signals, and current selected signalvariables
cannot conflict.
 

Tony to add this motivation to spec, to identify why we resolve to
keep the constructs as currently organized [AI2].
 

(2) Writeup on Variables issues from Don Organ. This writeup is
attached to the end of the current issues document. The Working Group
is asked to read this through and familiarize themselves with the
issues here so we can respond to them; [AI3] to Working Group.

LockStep

The current proposal has been reviewed down to the point of
AllowInterleave. Jim is still uncomfortable with this construct and
would prefer an alternate strategy, for example driving "shared"
signals from a single, separate pattern location. Greg identified that
this is always possible to do but requires re-writing the pattern
blocks which is what SoC test has a requirement to NOT do.
 

Tony asked how to make progress on this issue and Greg recommended
incorporating the current LockStep additions into the document. Jim
and Doug asked for a concrete example of the need for this construct
or demonstrate it's utility. Tony took [AI4] to put this in doc; Greg
and Tony took [AI5] to generate example as well.

Next meeting

next meeting @ ITC on Monday, Oct. 7
next phone meeting in 4 weeks, Oct. 24
 

Meeting adjourned at 11:00 PDT.
 

AIs
---
[AI1] Peter and Denis resolve Scanchain referencing issues
[AI2] Tony to add motivation to leave SignalVariables as-is
[AI3] Working Group review Don's variables issues
[AI4] Tony to put current interleave into draft for review
[AI5] Greg and Tony generate allowinterleave example