Phone Conference P1450.3 Working Group
Thursday, Aug 4, 2005 - 10:00-10:30 Pacific Time
Attendees:
Tony Taylor (chair)
Greg Maston (scribe)
Daniel Fan
Jose Santiago
John Cosley
Docs:
- review new D10 with the addition of TRC memory checks
Discussion:
IEEE meeting clearances
Nothing under discussion or presentation for this meeting was identified as being proprietary or restricted.
Memory Checks
Started discussing the SignalCharacteristics, with the first question: should
MaxVectorMemory be removed from this block and put only in
PatternCharacteristics? If so, should other *memory statements be removed from
here as well? Greg says yes, these statements are out of context with where the
counting occurs (unless counting statements are added here as well).
Identified addition to syntax to reference multiple PatternCharacteristic blocks
in one SignalCharacters block.
Added MaxVectorsCount statement to the PatternCharacteristics block, form
slightly different than original doc. Greg was fine with syntax here.
Brief discussion about the issue of context contributing to these counts.
Sometimes the amount of memory required for a construct depends on what came
before. Greg identified that this is the "sub-behaviors" issue, identified in
Annex B, "limitations to this approach". Sub-behaviors can become arbitrarily
complex, depending not just on the prior statement, but the constructs present
within some
statements as well (for instance, single-vector loops vs. multi-vector loops).
Tony proposed an additional statement to support counting when Vectors consume
less than an integral count, some sort of statement to indicate when 2 or 3 or
more Vectors add 1 to the count.
Tony proposed changing MaxVectors to MaxVectorMemory, to reduce the implication
that this was simply a count of STIL Vectors.
Much discussion ensued over the issue of shared resource memory effects. How do
we count when, for instance, 2 instruments are applied across a different set of
(overlapping) signals, and both access a common sea of memory?
Daniel identified a desire to use the TRC to assist in resource allocation /
configuration identification. When multiple configurations are possible, how
does the TRC identify which is desirable?
Some discussion about the mechanism to identify signals into
SignalCharacteristics blocks - the attributes Scan, Data, etc being defined in
Table 2. Tony to move Table 2 into the syntax definition.
Concern that Table 2 is too restrictive and inappropriately defined
(particularly for multi-bit contexts) to provide sufficient "bucketing" of
signal types for checking signal-type-specific rules.
Next meeting
Date: Next meeting Thursday, Aug
18
Time: 10:00 to 11:30 Pacific
Action Items:
AI-1: Bruce - create TRC for the Teseda tester using the D10
AI-2: Tony - update D10 per above decisions re: mem checking
AI-3: Dan - create TRC for
Credence testers using D10