Project Konrad: A Network Database Simulation of Mental
Associationism
Copyright Notice: This material was written and published in
|
|
First published online 12:00 BST 11th
October 2007; this version [2.2 - extended content] dated 09:00 BST 21st August
2008
1 -
Introduction
The Konrad Project represents a long-overdue application of computer software to one of the most enduring problems of mental philosophy. The software in question is the CA-IDMS network database, the latest in a long line of "Data Base Task Group" (DBTG) databases going back to the late 1960s, and the philosophical problem is that of the associative network nature of the biological mind. The author learned of the latter as a psychology undergraduate, and became deeply immersed in the former during his years as analyst-programmer in the data-processing industry in the 1980s. He then drew on both bodies of knowledge when in November 1990 he acquired a lectureship in cognitive neuropsychology at what is now University of Wales Institute, Cardiff, and his personal analysis of the hierarchical structure of cognition, complete with a long-term memory database separated out from a short-term memory executive system was published in Smith (1993c).
Little then
changed until in 1996 the author came across the theory of "second
messenger" neurotransmission, whereupon he noted a degree of resonance between descriptions of the DBTG concept of
"database currency" and the biological mechanisms of "calcium
switching" implicated in medium-term neural sensitisation. Both allow
their respective systems to maintain a particular mental theme across a
time-span larger than the span of the immediate here and now, both do this by
holding material momentarily somewhere between short term memory and long term
memory, both combine storage and retrieval functions, and - above all - both
exist to help "bind" widely scattered memory fragments into logical
wholes [for more on the biochemistry of calcium sensitisation, see
Appendix A at the end of the present document]. These similarities were duly reflected upon in Smith (1997a, 1997b,
1997e), and then presented as animated cognitive models in Smith (2000a, 2000b,
2002a; 2005c).
ASIDE: For a
fuller history of the DBTG type of database, see the companion
resource on "Data Modelling". For a gentler introduction in
PowerPoint format, see the first dozen slides of Smith (2005b).
Work on the Konrad Project - named in tribute to Konrad Zuse, a largely
unsung pioneer of digital computing - began in earnest following a in-house
research presentation on 8th December 2003, and to date [August 2008] has
generated four qualitatively distinct bodies of work, as now described.
2 - The First
Project Element - An Interdisciplinary Glossary
The
Associationist view of the network structure of the biological mind emerged
originally from classical thinkers such as Aristotle, was popularised by
Enlightenment philosophers such as Locke and Hartley, and was then invoked by
two of the 20th century's most famous new explanatory perspectives. The first
of these was the psychotherapeutic usage of the "free association"
and "association of ideas" investigatory paradigms (see, e.g., Freud,
1916, and Jung, 1918, respectively). The second was the recognition in the
early years of Artificial Intelligence research that there was a network
structure of sorts to what psychologists now describe as "semantic memory" [Glossary]
(see, e.g., Richens and Booth, 1955; Ceccato, 1961). This latter has gone on to become the research area now commonly referred
to as "semantic networks" [click here for
Wikipedia's introduction to this subject, and here for John F. Sowa's
detailed historical review].
Now the problem with semantic networks as a
general area of study is that its material belongs to no single science. If we
want to do the topic proper justice we require the full resources of the
interdisciplinary superscience known as "cognitive science", whose
member disciplines include (in alphabetical order) cognitive psychology,
computing, "Connectionism" [the science of "neural
networks"], cultural anthropology, epistemology [the philosophy of
knowledge], medicine (notably neurology and neuropsychology), mental
philosophy, palaeontology, philology [linguistics as language theory rather than language ability], physical anthropology,
robotics, and zoology (notably ethology and primatology).
Sadly, each of these member sciences is
already bursting at the academic seams with its own jargon and its own
literature, and interdisciplinary collaboration tends as a result to be rendered
inefficient by lack of mutual understanding. The first thrust of the Konrad Project was therefore to support
interdisciplinary collaboration with a back-to-basics online glossary - an
encyclopaedia of the mind - which will teach computing jargon to mental
philosophers, electronics to neuroscientists, phenomenology to physicians, and
so on. This turned out to be pretty much a never-ending task, but here is the
product as it currently stands .....
|
Glossary Section |
Notes |
|
Contains only general
notes; avoid unless and until specifically directed |
|
|
Readers may find the
entries for associationism, Bachman diagram, and borderline
personality disorder particularly helpful |
|
|
Contains general notes;
avoid unless and until specifically directed |
|
|
Contains introductory
notes on a number of important philosophical analyses of the highest mental
functions. Readers may find the entries for Kant and Heidegger
particularly helpful in setting the philosophical scene within which machine
consciousness research takes place. |
|
|
Readers may find the
entries for chain pointers, cognitive deficit, cognitive
modelling, <CONNECT>, and current of .....
particularly helpful |
|
|
Readers may find the
entries for Dasein, data ....., and dualisms or monisms
particularly helpful |
|
|
Readers may find the
various entries for Freud particularly helpful |
|
|
Readers may find the
entries for Gegenstandstheorie, homunculus fallacy, inner speech, and intuition particularly helpful |
|
|
Readers may find the entry
for long term working memory particularly helpful |
|
|
Readers may find the
entries for machine consciousness, massive modularity, mind-brain
debate, minute perceptions, and object ..... particularly
helpful |
|
|
Readers may find the entry
for personality, splitting of particularly helpful |
|
|
Readers may find the
entries for perception ....., phenomenology, pragmatic .....,
preconscious, protein kinase studies, and relationship .....
particularly helpful |
|
|
Readers may find the
various entries for self and soul particularly helpful |
|
|
Readers may find the
various entries for the unconscious particularly helpful |
3 - The
Second Project Element - A "Proof of Concept" DBTG Network
The second thrust of the project was then to
produce some actual DBTG software. Progress under this heading has been slow due
to the size of the computers required to execute said software [that is to say,
none of the commercial DBTG systems has ever been down-scaled to run on PCs].
Following the 2003 in-house presentation, various feelers were put out to
companies operating IDMS systems in the hope that one would be able to provide
testbed access. These requests were heard sympathetically enough, but no
concrete offer of support was forthcoming until April 2006 when International
Software Products [corporate profile]
kindly offered their services. The resulting software - the Konrad1 "proof of concept"
system - went into development on 1st November 2007, and successfully ran its
test program on 11th March 2008 [click here
to see the Press Release]. Here are some useful descriptive documents .....
Konrad1: Technical
Scope and Aims
Konrad1: Bachman Diagram [Draft] - pdf format (use for the on-screen zoom
facility)
Konrad1: Bachman Diagram [Draft] - PowerPoint poster format (with hyperlinks)
4 - The Third
Project Element - A Formal Prototype DBTG Network [WORK IN PROGRESS]
The third thrust
of the project is now to upgrade the proof of concept system into a more powerful
prototype in the light of the initial development experience. The resulting
software - the Konrad2 system - went
into development on 1st April 2008. Unlike Konrad1,
the Konrad2 schema takes account of
both the physical and the mental domain, and has been heavily abstracted from a
complex logical data model in which more than 250 discrete entity types are
identified. These logical entity types are physically implemented as eight Area
1 (physical world) and seven Area 2 (mental world) entity types. Development of
the substantive test program is in hand.
Here are some useful descriptive documents .....
Konrad2: Logical Job Pack for Conference Demonstration
PERIODIC PROGRESS REPORTS WILL APPEAR HERE
5 - The
Fourth Project Element - A "Common Workshop" Format [WORK IN PROGRESS]
To avoid repeatedly having to explain the
technicalities of network database design to scientists with no computing
background it was decided to go for a common and repeatable demonstration of
the software, complete with all the support and audit trail paperwork needed.
As a result, a typical software demonstration involves running the program
three (optionally five) times, and requires that the following files have been
carefully prepared in advance .....
LOAD FILE: This file contains initial database load
instructions, and is used (in RUN A below) to create an initial body of mental
content.
TASK FILE: This file contains a task-specific set of database
interrogation instructions, designed to put the loaded database to work in some
particular way.
RETRAINING FILE: This file contains a second set of database load
instructions, designed to extend the loaded database for the purposes discussed
below.
These various input files are then used as
follows .....
Run A: The first run of the program will load the LOAD FILE into the database,
creating the various records, placing them where they belong, and then linking
them together, all in accordance with the currently operational version of the database
schema. The database key (that is to say, the physical location) of each
record thus stored will be written out to a print file - "Print A" -
for future reference. The resulting database represents the starting point for
both Runs B and C below.
Run B: The second run of the program will conduct a pre-planned navigation of
the standing database produced during Run A, to simulate a discrete act of
cognition (perceiving and responding to a simple sentence, say). The input file
on this occasion will be the TASK FILE. Every record interrogated or updated
during this process will be written out to a print file - "Print B" -
for subsequent inspection. If updates have been applied, they will then be "rolled back"
[that is to say, reversed out] in readiness for Run C [for more on the technicalities of database
rollback, see Appendix B at the end of the present document].
Run C: The third run of the program executes precisely the same input file as Run B, but against a database accessed now via a subtly incomplete database subschema, selected from a pool of such subschemas maintained for research purposes [details in Sections 6 and 7]. This deliberately restricted view of the available data will impair the efficiency, specific content, execution time, or other key features of the database navigation. Each selected subschema will simulate a particular cognitive or personality deficit by imposing a corresponding "network deficit". Again, every record interrogated or updated during this process will be written out to a print file - "Print C" - for subsequent inspection. If Run D [see next] is required, then any updates which have been applied during Run C will be "rolled back" before proceeding. The difference between Print B and Print C represents the network deficit brought about by the known restrictions in the chosen subschema. These prints may then be analysed at leisure. [For more on the technicalities of the subschema device, see Appendix C at the end of the present document.]
ASIDE: The next two runs of the program were suggested by co-worker Minnette
Celliers in October 2007, and address the fact that many network deficits bring
with them a need for rehabilitation of some sort.
Run
D (Optional): This run of the
program uses the same restrictive subschema as Run C, but applies the
RETRAINING FILE of input commands. This file contains additional database load
and structuring commands designed to circumvent [by which we mean
"treat" or "attempt to remediate" or "intervene
therapeutically"] the problems introduced by the subschema. Every record
interrogated or updated during this process will be written out to a print file
- "Print D" - for subsequent inspection.
Run E (Optional): This final run of the program uses the same restrictive subschema as Runs C and D, but represents the TASK FILE in the hope that some improvement will have been brought about by the "remediation" carried out in Run D. The difference between Print C and Print E represents the reduced [good] or increased [bad] network deficit attributable to said remediation. The nature of any differences may then be analysed at leisure.
6 - The
Proposed Research Subschemas [WORK IN PROGRESS]
As the Konrad2
prototype takes shape, demonstration workshops will be constructed around
specific research subschemas. As outlined in the preceding section, each demonstration
will involve five passes through the main update program, and the controlling
Job Control Language [see outline job pack]
will be co-developed with the first of the subschemas to go live. The likely
candidate for this honour is .....
AKRASIC: The word "akrasia" [Greek
ακρασια] means to lack balance in the
consciously willed control over one's behaviour [see detailed etymology]. To be
"akrasic" [the word may also be softened to "acrasic"] is thus
to lack willpower when faced with a temptation or craving of some sort. It is
to allow your heart to rule your mind, rather than to obey the rules of
rationality, duty, and conscience. As such, the term is the formal descriptor
not just for everyday weaknesses of will, but also for the substance-dependency
and impulse-control side of psychiatry. The AKRASIC subschema will accordingly
mask out data elements [entity types and/or their attributes, and/or their
relationship parameters] involved in keeping the system's behaviour within
normal limits of consumption. Such a restrictive subschema might (a) help
explain the low success rate of lifestyle initiatives across the dependency spectrum
[food, drug, alcohol, tobacco, exercise, etc.], and (b) provide a testbed
resource for evaluating specific rehabilitative initiatives. The software will
allow continuous variable settings for the intensities of both the addictive
craving and the simulated willpower, and will thus support either correlational
or sample-contrastive research designs.
ASIDE: For more on the relationship between desire, emotional experience, and
conscious understanding, begin under soul, tripartite in the companion
Glossary.
7 -
Prospective Developments
Later versions of the software will gradually
add further subschemas to the repertoire. The following are currently in the
queue .....
AGRAMMATIC: This subschema will mask out data elements
involved in converting Konrad's deep
praxemic code into grammatically appropriate and fluently delivered surface
language, and is designed to assist psycholinguistics researchers interested in
the deep mechanisms of speech production in Broca's aphasia.
Such a restrictive subschema might (a) help explain the organisation of the
brain's language functions, and (b) provide a testbed resource for evaluating
specific speech-rehabilitative initiatives.
AMNESIC: This subschema will mask out data elements
involved in the consolidation [definition] of
short-term memory into new long-term memory, and is designed to demonstrate how
a failure of the mechanisms by which this process normally takes place can lead
to the sufferer becoming subjectively "frozen in time" as a result.
Prior familiarity with the neurological condition known as "anterograde
amnesia" [see
syndrome outline] would be an advantage to researchers planning to use this
subschema.
APE_MAN: This subschema will mask out data elements
involved in converting Konrad's deep
praxemic code into grammatically appropriate and fluently delivered surface
language, and is designed to assist cognitive anthropologists and evolutionary
psychologists model the philogeny of human language skills.
ASIDE: Konrad2's logical data model
incorporates and is broadly compatible with a number of state-of-the-art models
of cognition[1].
It goes to great length to identify the substages of perceptual aesthesis, the
contents of phenomenal awareness, and the complex of unconscious and
preconscious vectors involved in behavioral initiation. Any of these
subprocesses (or permutations of several) can be wholly or partly put out of
commission, thus allowing phylogenetically inferior grades of mind to be
simulated, even to the extent that one (virtual) Neandertal (say) could be
observed communicating to another, long after the extinction of the natural
specimens!
ASLEEP: In the biological nervous system, sleep is
initiated at brainstem level [see
process summary], and involves a more or less total interruption of waking-state
sensory and motor pathways, periodically accompanied by dreaming processes
which, upon awakening, may be only partly reportable. This subschema will mask
out data elements critical to waking state cognition, thus allowing the
software to simulate both the sleeping and (more interestingly) the dreaming.
ASIDE: The fact that all database accesses - regardless of whether they are
subsequently subjectively reportable or not - are recorded in the system's
Log-File as they occur, means that Konrad2
offers a window on mental activity during dreaming which is not available with
biological systems. Konrad2 can tell
you what it is "thinking", in other words, without being woken up and
interrogated, and it can do this moreover without loss of accessible content
due to qualitative or quantitative restrictions on conscious access (or
retrieval therefrom) at the time of waking!
ASPERGER and AUTISTIC: These subschemas will mask out data
elements involved in the process of meta-representation [Glossary],
in the hope that this will deprive ongoing cognition of access to a personally
discrete and functionally intact self, and thus help account for autistic
spectrum dysfunctionality (high and low end respectively). Prior
familiarity with autistic spectrum conditions [see syndrome outline] would be
an advantage to researchers planning to use this subschema.
BORER: In
CARELESS:
This subschema will mask out data
elements involved either in the prevention of skill-based errors [details]
(or in the mitigation of their severity if they cannot be prevented). It is
intended to assist ergonomics researchers in the field of system, site, or
product safety.
ASIDE (1): The fact that Konrad2
is an inanimate software simulation allows it to be put in harm's way without
the ethical restrictions which constrain accident research with human subjects.
For example, this subschema can be used to simulate the effects of triggering
variables [e.g., fatigue, distraction, strength of grip, etc.] on the frequency
of (virtual) knife slips in the (virtual) elderly, using a simple repeated
measures experimental design.
ASIDE (2):
The fact that the core database can be "rolled back" between trials
[as explained in Section 5 above] also means that each new trial can presume
equal levels of prior skill. This gives researchers total control over the practice
effects which plague repeated measures studies in human subjects [see theoretical
discussion].
DYSEXECUTIVE: This subschema will mask out data elements
involved in the execution of effective mental planning and strategic control,
in the hope that this will simulate key aspects of the neurological condition
known as dysexecutive syndrome [more on
this]. Such a restrictive subschema might (a) help explain the organisation
of the brain's highest higher mental functions [Glossary] and (b)
provide a testbed resource for evaluating specific neuro-rehabilitative
initiatives, both adult and paediatric [more on
this].
EXTREMIST: This subschema will mask out data elements
involved in the structures of personality and self concept (especially those
for prospective autobiographical memory), in the hope that this will upset the
normal balance between the respective mental vectors for self-preservation and
conscientious self-sacrifice. Such a restrictive subschema might (a) help
account for the sort of suicidality seen in terrorist bombers, and (b) provide
a testbed resource for evaluating potential counter-terrorist proposals.
ASIDE: Again, the fact that Konrad2 is an inanimate software simulation relieves researchers of
the ethical restrictions which constrain research into human belief and
motivation. For example, this subschema could be used to simulate the effects
of an attack on a person's religious beliefs without causing the slightest
human distress. In addition, the computer's processing speed also allows for
the compression of elapsed time, being limited primarily by the research team's
ability to prepare the necessary input files in advance.
HEARING_VOICES: This subschema will mask out data elements
involved in the phenomenon of inner speech [Glossary],
in the hope that this will upset the multiple feedback loops involved therein.
It is designed to assist research into the cognitive deficit explanation of
schizophrenia. Prior familiarity with Frith, Rees,
and Friston's (1998) "Forward Model" of mental organisation, and
with the role played by efference copy [Glossary]
therein, would be an advantage to researchers planning to use this subschema.
JARGONAPHASIC: As AGRAMMATIC, but for researchers
interested in Wernicke's
aphasia.
LOW_ACHIEVER: This subschema will mask out data elements
involved in the structures of personality and self concept (especially those
for prospective autobiographical memory), in the hope that this will upset the
normal balance between the respective mental vectors for educational ambition
and educational effort. It is designed to assist educational, psychological,
and sociological research into the nature of student disaffection with the
educational system.
ASIDE: The fact that Konrad2
is an inanimate software simulation allows it to be "abused" (in the
broadest sense of the word) without the ethical restrictions which constrain
research with human subjects. For example, this subschema can be used to
simulate the effects of triggering variables [e.g., social deprivation, peer
pressure, TV violence, computer games, or what you will] on the frequency of
(virtual) disaffection in the (virtual) schoolchild, using a simple repeated
measures experimental design.
MARYS_ROOM: This subschema will mask out data elements
involved in perceptual aesthesis [Glossary],
and is designed to make tangible and testable the philosophical issues of
perceptual qualia [Glossary]
raised by Jackson's (1982) "Mary's
Room" thought experiment.
OEDIPAL: This subschema will mask out data elements
involved in successfully resolving the famous "Oedipus Conflict"
stage of early childhood as seen by Freudian Theory [Glossary],
in the hope that this might throw light on the validity of that quintessentially
Freudian analytic. Prior familiarity with the role played by identification
[Glossary]
during development, would be an advantage to researchers planning to use this
subschema.
PAEDOPHILE: This variant of the AKRASIC subschema will
mask out data elements involved in keeping the system's sexual behaviour within normal limits of propriety, in the hope
that this will upset the normal balance between libidinous desire and its overt
expression. Such a restrictive subschema might (a) help better account for
predatory sexual behaviour towards minors, and (b) provide a testbed resource
for evaluating potential remediation programmes.
ASIDE (1): Again, the fact that Konrad2 is an inanimate software simulation relieves researchers of
the ethical restrictions which constrain criminological research into the
seamier side of human motivation. For example, this subschema could be used to
simulate the effects of forced medication without transgressing anybody's human
rights.
ASIDE (2):
The fact that all database accesses - regardless of whether they are
subsequently subjectively reportable or not - are recorded in the system's
Log-File as they occur, means that Konrad2
offers a window on behavioural motivation which is not available with biological
systems. All cognitions become explicitly visible, including (a) those which
are known to the subject but deliberately kept secret from the researcher, and
(b) those which are more or less unconscious even in the subject. The machine's
audit trail, in other words, becomes a truth trail.
RAPIST: As for PAEDOPHILE, but without the
age-of-victim restriction.
ZOMBIE: This variant of the DYSEXECUTIVE subschema
will mask out data elements involved in providing the processes of conscious mental planning and strategic
control, and is designed to make tangible and testable the philosophical issues
behind mental philosophy's famous "Zombie Test" [Glossary].
Appendix A -
Calcium-Sensitised Medium-Term Memory
The main thrust of modern memory research suggests that there are three types of biological memory, namely electrical, structural, and "calcium-sensitised". The electrical type supports the span of immediate consciousness (two or three seconds), the structural type supports permanent memory, and the calcium-sensitised type provides a means of the former "tagging" the latter with things which have just been accessed and might be needed again within the next hour or so; it is also the physiological mechanism which underlies the phenomenon of memory consolidation. It is the calcium-sensitised memory variant which concerns us here, because it allows direct access to items within long term memory, provided only that they are in the necessary state of heightened excitation.
This is how we have explained our central technical thesis elsewhere .....
"[Our] general conclusion is simply that the
database concepts CURRENT OF SET and CURRENT OF RUN can both be implemented
biologically. The former could be
implemented by using calcium-switched tagging of the engrams involved, and the
latter by whatever full-blown electrical activity constitutes consciousness. When searched for via a particular set owner,
a "current-of-set" set member would always automatically be the
"preferred" set member, and would be reaccessable directly, regardless of how much effort
it had taken to locate it the hard way in the first place. Not only would this provide biological memory
with a safe method of indexing on several dimensions simultaneously, but it
would also, in turn, significantly enhance memory's value as a general purpose
cognitive resource, capable of responding adaptively and quickly to the
unpredictabilities of the real world" (Smith, 1997e,
p11).
For more detail, check out the entry for protein
kinase studies in the Project Konrad
glossary.
Appendix B -
Database Rollback
Databases
offer their owners a raft of significant benefits, but at the same time the
data they contain is rendered highly vulnerable by its sheer size and
importance. To reduce this vulnerability, databases routinely come complete
with a package of protective measures administered by a dedicated database
administration (DBA) team. Two of the most important DBA tasks are (a) to know
with absolute certainty what updates have taken place and what updates are
still awaiting processing, and (b) to have techniques of recovery available
should a particular set of updates be mistakenly processed twice (this is why
we so rarely get paid twice at month-end). The technical solution to this need
is for every update to be "journalised" as it takes place, that is to
say, for every pre-updated record to be copied out into a separate
offline file, and time-stamped accordingly. If an error is subsequently detected, the
journal file can be re-declared as an input file and used to reverse out each
incorrect update, thus returning the database to
where it had been originally. This process is
known as "rolling back" the database in question.
Because it has been implemented as a
database, Konrad's network
mind can also be rolled back in this way, giving it the power to "unlearn" things it has
previously been told about. This is a fundamentally important
capability because it is something biological systems cannot do. By then
deploying the subschema and rollback facilities together, the basic processing
sequence described in Section 5 can be turned into a powerful virtual world where a lifetime's "prior knowledge"
can be uploaded in a few minutes, where neurogenic or psychogenic trauma can be
switched on and off at will, and where time can go backwards as well as
forwards!
Appendix C -
Schemas, Subschemas, and Storage Schemas
Between 1969 and 1971, the US Department of Defense's CODASYL committee compiled two major statements of database principles (CODASYL, 1969, 1971; subsequently incorporated into ANSI/SPARC, 1976). These principles were inspired by the single central axiom that the internal complexities of a database should at all times remain totally "transparent" to the end-user. A DBMS, in other words, should allow users to concentrate upon their data rather than upon the tool they happened to be using to view it. This transparency was eventually obtained by implementing the data model in three time-separated sub-stages, each separately programmed, and each passing critical output to the one following. These three stages were as follows .....
(1)
Set Up a "Database Schema":
The first step is to convert the data model into a physically equivalent set of
declarations and descriptions known collectively as a "database
schema". Unlike the data model, however, the database schema is now in
a form which can be stored within, and manipulated by, the DBMS. This is a more
technical view of the data than hitherto, and constitutes the first major step
in bridging the gap between the data as the user knows it and the hardware on
which it is eventually to be stored.
(2)
Set Up Database "Subschemas":
The second step is to create a "departmental" view of the data. This
is another technical view, and reflects the fact that no single application
program will ever need access to all the available data. Each individual
end-user - and that includes even the most senior executives - only needs
access to a fraction of the total available data, and for him/her to be shown
too much is at best inefficient, and at worst a breach of system security. This
"need to know" facility is provided by subsets of the schema known as
"subschemas", each one allowing an individual
application program to access only the data it is legitimately concerned with. Konrad2 simply borrows this commercially
inspired facility, and uses it to create artificially impaired variants of its
core database. [Illustrative Argument:
If you prevent access to the record types for visual form, then you have gone a
long way towards simulating psychic blindness.]
(3)
Set Up Database "Storage Schemas": The third and final step is to create a "machine level" view
of the data. This is achieved by declaring what is known as a "storage
schema" to the DBMS, which the DBMS then uses to translate every
user-initiated store and retrieve instruction into a set of equivalent physical
store and retrieve instructions.