Flow Control Status Report 5/4/98
Participants
Francisco Hernadez
Gregg Wilder
Ernie Wahl
Don Organ
Tony Taylor
Greg Maston
Warren Davis
Phil Barch
Jim Dexter
Mike Purtell
Spass Stoiantschewsky
Sundar Lakshmipathy
Agreements
target a "trial-use" standard
we'll STIL'ize enVision as a strawman
Open Issues
Source language versus an execution environment (datalogging, test counts,
a trace of execution) .
Interface with TestMethods.
Requirements Considered
-
test sequencing/flow (multiple threads)
-
sequencing is 1 dimenisonal, flows define multiple paths
-
multiple thread processing may be necessary for mixed signal
-
data flow versus control flow
-
different flows for different test conditions (QA vs final test) - or variations
on the same flow depending on different test conditions
-
test sites (where special processing may be required), (May be 2 separate
issues - different tests for different device types, and different tests
for every Nth device) Examples:
-
testing special test-dice on a wafer.
-
environmental testing? of every Nth device.
-
characterization flow - for every Nth device.
-
separate power-up and power-down flows for different devices.
-
separate flows from Loading and Unloading a program.
-
generation of interim summary reports - after every Nth wafer
-
test flow parameters/options
-
expression processing as function of flow
-
variables will be used both for pre-conditioning as well as results representation
(which may be "tested" later - resulting in a change to the flow).
-
multi-site
-
We shouldn't do anything that will prevent an ATE vendor from implementing
a multi-site solution.
-
status message (to operator)
-
test identification
-
A unique ID for a particular test (at what level of abstraction should
the test be?)
-
binning
-
Hardware bins. Software bins (name and number). Terminal and non-terminal.
Used for collecting data for summary.
-
Rule-based (chaining) versus control flow.
-
flow statitistical reports
-
We have no responsibilities beyond our binning requirement.
-
Where is boundary between design/device information versus factory process
management? How far should STIL go?
-
Multiple threads
-
multi-site
-
concurrent execution (for mixed signal or cores) - numerical processing
is a good candidae
-
let's avoid forks and joins as a programming construct
-
easy to maintain
-
avoid redundent information
-
Others
-
a common base class for all test methods
-
flow related variables that are not associated with any specific test
object.
-
sub-flows
Next Steps
-
Formalize requirements
-
STIL'zing enVision
-
Develop a formal data model
-
development of use cases