Phone Conference P1450.1 Working Doc Subgroup
Thurs August 1, 10:00 am PDT

Attendees:
Doug Sprague
Greg Maston
Tony Taylor
Jim Teisher
Dave Kellerman
Jason Doege

Documents
p1450.1 D14 draft July 28.
D14 review-resolution document July 28.

Agenda
GR-7: I have done major surgery on section 10 (Variables). Please try to
review it before the meeting.

TT-2: Removal of BistStructs - Note: email notice has been sent out.

GM-6: LockStep - Please review prior to the meeting.

DM-1: \m \j - Please review prior to the meeting.

GM-9: Structure of doc and placement of Variables block.

ITC meeting (new item added by Tony in separate email)

Discussion

ITC Meeting: Tony proposes an all-day Monday meeting before ITC. This conflicts
with P1500 meeting. Jim, Greg and Tony are willing (Greg appreciates
having an excuse to not attend the P1500 meeting); Dave and Jason may be
pulled away; Doug will not be there. [AI1] Tony will reserve a room at
ITC for us.

GR-7, section 10: brief presentation by Tony on the new Variables
block based on changes from last meeting.

Questions previously emailed: why are reals not here? Response is that
reals are defined in the Spec block. But still there is a general
concern that user will see the "Variables" block here and be
frustrated that reals aren't here as well. And if we reference Real
support to the Spec block, is there an issue that Integers aren't
over in the Spec block as well?

Proposal to add a qualifier on the Constants to identify them as
IntegerConstants, to facilitate future expansion of other types of
constants, such as RealConstants. [AI2] for Tony to implement.

Discussion of the perspective of generalization of the constructs to
support a more robust, general-purpose programming language
context. Identification that the primary purpose of this effort is
still as an interchange format, and that supporting a limted
definition for interchange is often in conflict with generalization.

[AI3] Tony to add paragraph in the intro section of section 10, to
identify the primary intent of Variables (as control-flow and option
specification, and provide 'aliasing' of values to meaningful names
with Constants and Enumerations), to identify that Reals are defined in
the Spec block, and to differentiate the purpose of this block from
the Spec.

Discussion about the change of the Enum definitions statement to
"Values" to match the Values statements of other blocks; no issues
were raised about this change.

Discussion about the Usage statement under SignalVariable. Currently
this statement defines a limited number of Usages that may be
ambiguous as to intent; concern about how to change this list as
additional uses get identified. Also the Usage statement under the
Integer declaration currently identifies a STIL block as the "usage"
mechanism. The two statements need to be uniform. Doug identified a
perference of identifying the intent of the usage at the point of
application of that variable, rather than at the definition. This was
tabled for additional discussion with the additional AI below.

[AI4] Jim will generate a proposal on how to identify usage.

[AI5] to Tony to correct example in 10.2.

TT-2, BistStructures REmoval: Two email responses affirming removal;Doug still checking @
IBM. Additional consensus on removal from all separate discussions on
this topic. Section will remain until the next meeting, after which it
will be removed unless objections are identified.

GM-6: LockStep and pattern merging - Brief presentation of the intent
of LockStep, to identify pattern compatibility and specific
requirements on synchronization. Identification that perhaps this
might be better termed "protocol synchronization". Tony indicated that
this construct identifies a limited subset of behavioral requirements
to support parallel patterns, and that this construct is
sufficiently restrictive on the parallel pattern environment such that
it is supportable by any tester that can run these patterns as a
single broadside set of vectors (whereas the general ParallelPatList
may not be supportable).

Concern was raised about modification of the pattern execution, in
particular insertion of addition test cycles in one of the patterns in
order to pad/align the Shift operation with a parallel pattern that
has longer shifts. This type of pattern modification, not being
identified until the patterns are assembled, may not be comfortable
at the manufacturing level.

Discussion about the general constructs presented by Jason on pattern
merging. Jason identified two additional behaviors to support pattern
merging: 'dead-cycle insertion' to wait through capture operations,
and pattern reordering to maximize potential signal overlap (minimize
non-overlapping patterns that need to be executed separately). Jason
is concerned that the tiling/sequencing and lockstep operations ought
to be merged.

Tony identified the potential of using the BreakPoint construct to
indicate explicit locations for "waiting", and the BreakPoint block as
the "dead-cycle" specification if waiting is necessary to maintain
synchronism between parallel LockStepped patterns.

Discussion about the opportunities to reorder patterns as part of this
process, to maximize parallel execution of overlapping signals by
ordering common values on overlapping signals to execute in
parallel. Problem/issue with this strategy is it requires specific
pattern constructs to allow overlapping -- previous-unload and
next-load must be specified "unmerged" and then merged after the
pattern-reordering process, in order to support pattern reordering.

Greg identified that a standard shall identify a robust and
deterministic definition such that all users of the standard
interprete exactly the same behavior from a definition, and that
pattern reordering - while a potentially viable behavior - may be more
of a usage capability unless the reordering properties can be
defined deterministically.

[AI6] Greg and Jason to talk separately on this matter.

Request: Before closing the meeting, Tony made a specific request for various
individuals, Doug and Greg specifically, to please address their open
issues in the current issues doc that have their names on it,
specifically any paragraph rewrite references. [AI7] to Doug and Greg.

Meeting adjourned at 11:30 am PDT.

AIs
 

[AI1]: Tony reserve meeting room for Monday at ITC.
[AI2]: Tony to identify Constants as integer-types to facilitate expansion.
[AI3]: Tony add paragraph to intro of sect. 10 to identify intent
[AI4]: Jim will generate a proposal on how to identify usage.
[AI5]: Tony to correct example in 10.2.
[AI6]: Greg and Jason to talk separately on this matter.
[AI7]: Doug and Greg to close on their paragraph rewrites.