Philip Emeagwali Formula for Inventing a Supercomputer | Contributions to Computer Science

Philip Emeagwali Formula for Inventing a  Supercomputer |  Contributions to Computer Science


TIME magazine called him
“the unsung hero behind the Internet.” CNN called him “A Father of the Internet.” President Bill Clinton called him
“one of the great minds of the Information Age.” He has been voted history’s greatest scientist
of African descent. He is Philip Emeagwali. He is coming to Trinidad and Tobago
to launch the 2008 Kwame Ture lecture series on Sunday June 8
at the JFK [John F. Kennedy] auditorium UWI [The University of the West Indies]
Saint Augustine 5 p.m. The Emancipation Support Committee
invites you to come and hear this inspirational mind
address the theme: “Crossing New Frontiers
to Conquer Today’s Challenges.” This lecture is one you cannot afford to miss. Admission is free. So be there on Sunday June 8
5 p.m. at the JFK auditorium UWI St. Augustine. [Wild applause and cheering for 22 seconds] The computer textbooks of the 1980s told the
readers that the fastest computer in the world
must be powered by only one isolated processor. On the Fourth of July 1989,
I discovered that the fastest computer in the world
must be powered by thousands or millions or even billions
of commodity-off-the-shelf processors that were tightly-coupled to each other
that were identical to each other and that shared nothing
between each other. That discovery made the news headlines
and has been embraced by all computer scientists. That discovery is the vital technology
that underpins every supercomputer. I’m Philip Emeagwali. To discover
is to change the narrative of science. In my quest for the Holy Grail
to the fastest supercomputers, I focused on the Second Law of Motion
of physics that was discovered
three centuries earlier which, however, had existed
since the Big Bang explosion that occurred
13.8 billion years ago. Back in the early 1980s,
I re-examined textbooks that described how
the Second Law of Motion of physics was encoded
into a system of coupled, non-linear, time-dependent, and three-dimensional
partial differential equations of calculus
that governs three-phase flows of crude oil, injected water,
and natural gas that were flowing one mile deep
underneath a production oilfield that is the size of a town. During my supercomputer research,
I re-examined mathematical physics textbooks
that described how the Second Law of Motion
of physics was codified from the algebraic equation
to the differential equation. What I discovered was an epiphany. I discovered that
in its most important application namely, the recovery of crude oil
and natural gas from production oilfields,
that the Second Law of Motion of physics was incorrectly represented. I discovered that
each of the nine partial differential equations
within the system of coupled, non-linear, time-dependent, and three-dimensional
partial differential equations encoded into petroleum reservoir simulators
incorporated only three partial derivative terms. Those three calculus terms
corresponded to three physical forces and none corresponded to
the fourth physical force that actually exists
in the oil field being simulated. I discovered that
those three physical forces could not equate to the actual four forces
inside all production petroleum reservoirs. My contribution
to mathematical knowledge is this:
I corrected those mathematical errors and I corrected them
by adding 36 partial derivative terms that corresponded to
and accounted for the 36 components
of the erroneously missing inertial forces. That was how I invented
nine partial differential equations that are the most advanced equations
in mathematics and the most important expressions
in calculus. I’m hopeful that
the nine partial differential equations that I contributed to mathematics
will remain accurate over the centuries. The Philip Emeagwali
system of partial differential equations should remain accurate because
they encode the Second Law of Motion of physics
that, in turn, did not change since the Big Bang explosion
that is the beginning of time for our universe. As a research computational mathematician
in quest for previously unseen
partial differential equations, my research perspective
was diametrically opposite to that of an applied mathematician
that only wants to analyze known partial differential equations. In the 1980s, I attended 500
weekly research seminars with the first half of those seminars
occurring in the metropolitan areas of Washington, District of Columbia
and Baltimore, Maryland. Half of the seminar speakers
were research mathematicians that came from faraway places,
such as Moscow (Russia), Paris (France), and London (England). During those seminars,
I observed that research mathematicians either focused their analysis on known
partial differential equations that has been described
in calculus textbooks or they were scribbling
partial differential equations that has been scribbled before
on a blackboard or coded before into a motherboard. I observed that research mathematicians
of the 1970s approached initial-boundary value problems
from only one direction. That direction was to and from
the mathematician’s blackboard. The iconic Navier-Stokes equations
is the favorite system of partial differential equations
of the mathematical physicist. Being a physicist and a mathematician
and a supercomputer scientist, I simultaneously approached
my parallel processing research on how to solve
the most computation-intensive algebraic approximations
that arose from finite difference discretizations
of partial differential equations and how to solve them
from four directions. My four directions
were from the storyboard of the physicist
to the blackboard of the mathematician to the motherboard
of the computer scientist and across the motherboards
of the research supercomputer scientist. On the Fourth of July 1989,
I became the first parallel supercomputer scientist
to record the world’s fastest calculations. As the first parallel supercomputer scientist,
I was mandated to solve the Grand Challenge Problem
of physics and mathematics and to solve it
by parallel processing the Grand Challenge Problem
as sixty-five thousand five hundred and thirty-six [65,536]
initial-boundary value problems of extreme-scale
computational fluid dynamics. My grand challenge was to figure out
how to chop up that real world problem of extreme-scale algebra
and chop it up into 64 binary thousand
smaller initial-boundary value problems and, most importantly, figure out
how to, subsequently, parallel process those computational physics problems
and how to do so across my two-raised-to-power sixteen processors
that were tightly-coupled to each other and that shared nothing
between each other. In the 1970s and ‘80s,
I walked along a technological trail that was orthogonal
to the trail that was walked by the vector processing
supercomputer scientist. I walked alone. I walked through the darkness
that was the unknown world of the massively parallel supercomputer
that was the precursor to the modern computer. Metaphorically speaking,
I walked within the unknown territory of the massively parallel supercomputer
and I walked with only a small lamp to see by. That lamp was the most massively
parallel ensemble of processors, ever built. The reason I was left alone
to discover how to make an ensemble of one million processors
solve one million problems at once was that it was then said that
parallel processing is a huge waste of everybody’s time. I walked through darkness
and into the light and did so with equations. How are modern supercomputers used? Nine in ten parallel processing cycles
were consumed by extreme-scaled computational physicists. Their grand challenges
include executing computational fluid dynamics codes
that had the Navier-Stokes equations at their calculus core
or executing the petroleum reservoir simulator
used to discover and recover otherwise elusive crude oil
and natural gas and the general circulation model
used to foresee otherwise unforeseeable global warming. At the granite cores
of most real world problems arising in computational physics
is the system of coupled, non-linear, time-dependent, and three-dimensional
partial differential equations of calculus
that contains partial derivative terms that represented something
in the physical problem the equations govern. Parallel processed supercomputing
is the Formula One of science and technology. The fastest supercomputer in the world is
ten million times faster than your computer. The fastest supercomputer
is powered by 10,649,600 cores that were totaled across 40,960 nodes. The supercomputer of 1946
was rated at 5,000 cycles per second that could be executed
during an arithmetical operation on a 10-digit number. Today, the parallel supercomputer that can
record a speed of one exaflops could be manufactured. The flop is the acronym
for floating-point arithmetical operations per second. Exascale supercomputing is achieved
by massively parallel processing at the speed of one billion
billion floating-point arithmetical operations per second. That speed of supercomputing
is equivalent to a quintillion, or ten-raised-to-power-18
calculations per second. The fastest supercomputer speeds
make it possible to create extreme-scale and high-fidelity computational
fluid dynamics simulations. Like any technology,
the parallel supercomputer is a double-edged sword
that can be used to do both good and bad things. The supercomputer is a vital instrument that
is used to execute computational fluid dynamics codes
that model blood flowing through the human cardiovascular system. The supercomputer that can be used for
computational medicine and used to understand
how to increase human longevity can also be used to design
weapons of doom. The parallel supercomputer
is used to design bombs that are more than 3,000 times
more powerful than the atomic bomb that was dropped upon the Japanese city
of Hiroshima. On August 6, 1945, that atomic bomb killed
166,000 Japanese. Because supercomputers
are used to simulate nuclear explosions over cities like New York,
the U.S. is reluctant to sell American-made supercomputers
to [quote unquote] “unfriendly nations.” This security threat is the reason
the U.S. Department of Commerce vehemently objects
whenever Japan sells a supercomputer to a nation that is unfriendly
to the United States. This was the origin of the infamous supercomputer
denial list that had been in existence
since the 1950s when it was against the law
to export an American supercomputer to the Soviet Union. This, in part, is the reason
that in the 1980s I was the only Nigerian
that was supercomputing within U.S. nuclear research laboratories. My contribution to mathematics
that was the cover story of the May 1990 issue
of the SIAM News—the flagship bi-monthly news journal
of the research mathematics community—was that I—Philip Emeagwali—discovered
nine as-yet-unknown partial differential equations
that weren’t in any calculus textbook. I figured out how to solve those
partial differential equations and solve them across
a new internet that is a new global network of
sixty-five thousand five hundred and thirty-six [65,536]
central processing units, or across as many tiny computers. I am the research computational mathematician
that discovered the fastest supercomputer speed
that can be harnessed to solve a system of coupled, non-linear,
time-dependent, three-dimensional, and three-phased
partial differential equations of calculus. I discovered how to solve
that initial-boundary value problem that is posed on the blackboard
of the mathematical physicist. I figured out how to translate
the partial differential equations of calculus
that I invented as partial difference equations
of algebra that I coded as a set of floating-point
arithmetical operations that I message-passed
to an ensemble of 64 binary thousand tightly-coupled, identical processors
each solving as many latency-sensitive problems. I figured out how to translate
the Grand Challenge Problem of physics and mathematics
and translate it into an equivalent set of
a million less challenging problems. I figured out how to translate
the Grand Challenge initial-boundary value problem
and do so across different boards. I figured out how to translate
the Grand Challenge Problem and translate it from the blackboard
of the mathematician to the motherboard
of the computer scientist. I figured out how to parallel process
the Grand Challenge problem and solve it across the motherboards
of the supercomputer scientist. From the Fourth of July 1989,
I began communicating my discovery of practical parallel processing
to the public. In 30-seconds, my contributions
to mathematics and physics is this:
The petroleum reservoir simulator that must be used to recover otherwise elusive
crude oil and natural gas provides correct answers
to incorrect equations. My contribution is this:
I figured out how to derive correct answers
to correct equations and how to solve
those Grand Challenge equations on a supercomputer
and solve them across an ensemble of millions
of tiny computers that outline a new internet. Back in the 1980s,
I mathematically diagnosed the critical errors in the MARS Code,
the petroleum reservoir model that was developed by
Exxon Corporation. Some years later, Exxon Corporation
merged with Mobil Corporation and both were renamed
Exxon-Mobile Corporation. The MARS code
is a complex petroleum reservoir simulator. The acronym MARS
stands for Multiple Application Reservoir Simulator. Mathematical physicists
at Exxon-Mobile Corporation and in places like
the Niger-Delta oilfield of the southeastern region of Nigeria
must use the oil and gas flow patterns within a production oilfield. Petroleum geologist
must use that flow pattern to decide
where to drill a water injection well and to decide
how many oil and gas production wells to drill. Petroleum reservoir modelers
use that flow pattern to know in advance
how to maximize the production of crude oil and natural gas
that will be extracted from a group of wells,
and to know in advance how and where
to apply enhanced oil recovery techniques,
or the secondary techniques that must be used
to discover and recover otherwise elusive crude oil
and natural gas. At its calculus core, the MARS code
includes the pressure equation and saturation equation. Both equations are part of the system
of partial differential equations that governs the motions
of the crude oil and natural gas flowing from water injection wells
towards oil and gas production wells. My contribution
to mathematics and physics is this:
I discovered the critical errors that mathematical physicists made
when they were solving the system of
partial differential equations that must be used
to discover and recover crude oil and natural gas. That mathematical discovery
inspired me to invent the nine Philip Emeagwali
partial differential equations of calculus. My contributions to calculus
has rich and fertile consequences for the petroleum industry
and is the reason one in ten parallel supercomputers
are purchased by the industry. My contributions to calculus
was the reason I was the cover story
of top mathematics publications, such as the May 1990 issue
of the SIAM News. The SIAM News
is the flagship publication of the mathematics community. Calculus is a tool that is used to answer
the biggest questions arising in science and engineering,
such as: “How do we recover
otherwise elusive crude oil and natural gas
and recover them from soon-to-be-abandoned oilfields?” Like the quadratic formula
of algebra, each partial differential equation
of calculus must be derived. The partial differential equation
we derived or discovered depends on the fundamental law
of physics, or the processes, or the multi physics scenarios,
we encoded into that equation. We discovered the predator-prey
ordinary differential equations and used them to describe
how two species interact. We discovered
partial differential equations in mathematical finance. I discovered my nine
partial differential equations of calculus
and I discovered them by not following the instructions
in the calculus textbooks. The discovery is made
by not following instructions. By definition, it’s impossible
to discover parallel processing and do so by only experimenting with only
one processor. On the Fourth of July 1989,
I discovered practical parallel processing
and I did so by experimenting across a new global network of
65,536 commodity processors that I visualized as a new internet. The research mathematician
is searching for something never-before-seen. More often than not,
that thing is a published paper which contains no discovery
and contains no invention that benefits humankind. In academia, a published paper
is rewarded. A mathematical discovery
that benefits humankind is one million times rarer
and is not rewarded in proportion to the effort
required to discover it. For this reason,
the research mathematician in academia only asks questions that are important
to his career. The research mathematician
asks questions that are direct and centered
on abstract mathematics, not questions that are central
on extreme-scaled parallel processed solutions
of the real world problems arising in mathematical physics. In the second half of the 1970s,
I was a research mathematician amongst research physicists
and research supercomputer scientists. In the first half of the 1980s,
I was a physicist amongst mathematicians
and supercomputer scientists. In the second half of the 1980s,
I came of age as an extreme-scaled parallel processing
supercomputer scientist that was amongst
computational physicists and computational mathematicians. That sixteen-year-long quest
was the reason my experimental discovery
of parallel processing made the news headlines
in various industry publications. Looking back to the 1970s and ‘80s,
I knew there were no easy partial differential equations
waiting for me to invent them. It is rare for a mathematician
to invent a never-before-seen
partial differential equation. It is rarer for that equation
to make the news headlines. In the cover story
of the May 1990 issue of the mathematician’s newspaper,
called the SIAM News, I said that I invented
36 partial derivative terms of calculus. I also said that I invented
36 algebraic terms that corresponded to those
36 partial derivative terms. Those 36 partial derivative terms represented
the temporal and convective inertial forces
that, in part, moves crude oil, injected water, and natural gas
and moves them from water injection wells
towards oil and gas production wells. Those thirty-six partial derivative terms
that I invented can be used to correct the critical errors
in the mathematical techniques that were used to discover and recover
otherwise elusive crude oil and natural gas, namely,
the governing system of partial differential equations of calculus. If uncorrected, those thirty-six errors
will replicate themselves across the trillions upon trillions
of the system of equations of algebra that were derived from discretizing
the governing system of partial differential equations
that were at the mathematical core of the petroleum reservoir simulators
that are used to discover and recover crude oil and natural gas. My contribution to mathematics
was to install those patches of 36 partial derivative terms
and to add them to the pre-existing 45 partial derivative terms. Those 36 errors occurs
at three levels, or as errors in the partial differential equations
that, in turn, become errors in the system of
partial difference equations that were derived from the discretized partial
differential equations. They also become errors
in the supercomputer algorithms that must be executed across
millions upon millions of processors. The new calculus and new algebra
that I contributed to mathematical knowledge
was extremely difficult to invent. In parallel processed
computational mathematics, ranging from
petroleum reservoir simulation to general circulation modeling
of global warming, the trillions upon trillions
of Xs and Ys of the underlying extreme-scale algebra
had their origin from the partial differential equations
of calculus that, in turn, originated from
and encoded corresponding laws of physics. A mathematical analysis
is akin to substituting thoughts and prayers
for experiments across millions upon millions of processors. On the Fourth of July 1989
in Los Alamos, New Mexico, United States,
and fifteen years after I began supercomputing in Corvallis, Oregon, United States,
I experimentally discovered that the toughest real world problems
arising in computational physics could be solved across
a new supercomputer that is configured as 65,536 processors
that tightly-encircled a globe and encircled that globe
as a new internet and encircled that globe in the manner
the internet encircles a bigger globe, namely,
planet Earth. Parallel supercomputing is,
in and of itself, almost a branch of mathematical physics, now called extreme-scale computational
physics. Without mathematics, computer science becomes
computer faith. I had to be a research mathematician
to be able to invent the new partial differential equations
and the corresponding partial difference algorithms
that I discovered. My contribution to mathematics
was to discover how to execute them across
a new internet. They were two things
that I did with my data. First, I copied them
from one processor to another processor and I copied them via email messages. Second, I computed with them
at the slow speed of 47,303 calculations per second
per processor and I did so to reach the
aggregated speed that was, for the first time, faster than
the speed of any vector processing supercomputer. Put differently, my contribution
to extreme-scale computational mathematics
did not reside on the processor that was not a member
of an ensemble of processors. My contribution to mathematics
reside on the processor that is a member of an ensemble
of processors and also resides
on the entire ensemble itself. Yet, my parallel processing experiment
had to wait until the 1980s when 65,536 processors
became available for me to experiment with. I say that a petroleum reservoir model that
runs on three, instead of on four, forces
is akin to driving your car on three wheels
and with the fourth tire deflated. The lesson that I learned is that
you must be a polymath, not a mathematician,
to solve the multi-disciplinary Grand Challenge Problem
that is beyond the frontiers of arithmetic, algebra, and calculus. The reason I could move back and forth
from the blackboard to the storyboard
is that I am a research mathematician and a research physicist. I knew the four forces
that defined the Second Law of Motion of physics
when applied to oilfields and knew that law,
forward and backward, and knew how to encode that law
into a system of nine coupled, non-linear, time-dependent,
and three-dimensional partial differential equations
of calculus that governs the three-phase flows
of crude oil, injected water, and natural gas
that is flowing across an oilfield that is a mile deep
and that is the size of a town. To solve the
Philip Emeagwali Equations that are my contributions
to mathematics and do so across a new internet
that is a new global network of 64 binary thousand processors
demanded that I discretize the problem domain
of the initial-boundary value problem. To discretize the problem,
I approximated continuous space with discretized space, or a finite grid. My new system of
partial difference equations of algebra
are the discrete versions of my new system of
partial differential equations of calculus that I invented. As a research mathematician
that is also a research physicist and that is also
a research supercomputer scientist, my interdisciplinary knowledge
was the necessary tool that gave me the intellectual maturity
that I needed to correct the century-old critical errors
that I found in calculus textbooks that were written
for the petroleum industry. Those errors in calculus
found their way from the classroom to the petroleum reservoir simulator
used by Exxon-Mobil Corporation. I should mention that
when I discovered that new calculus, or the Philip Emeagwali Equations,
I had to create new algorithms that led me to new algebra
that, also, codified the Second Law of Motion of physics. Inventing an equation
is like making your words a part of the holy scripture. The Philip Emeagwali Formula
was not for the blackboard alone. Nor was it for the motherboard alone. The Philip Emeagwali Formula
was invented for parallel processing across my sixty-five thousand
five hundred and thirty-six [65,536] tiny computers, or as many processors,
that encircled a globe in the way the Internet
encircled planet Earth. The Philip Emeagwali Formula
made the news headlines in 1989 and was highlighted
in the June 20, 1990 issue of The Wall Street Journal. Eleven years later,
that Philip Emeagwali Formula was reconfirmed
by then U.S. President Bill Clinton and reconfirmed
in his presidential speech of August 26, 2000. The parallel supercomputer
is a disruptive technology that gives tech companies
some competitive advantage in their drive for market leadership. The roots of the story
of how the fastest supercomputer was invented
began several millennia ago, and began when our ancestors
had no computing aid. For millennia, our ancestors
used their fingers and toes as their computing aids
and had no mathematical symbols scribbled on their cave walls. For the last one hundred years,
the word “computer” was prefaced as human computer, analog computer, electronic computer,
digital computer, distributed computer, parallel computer, and super computer. A change in how we look at the computer was
accompanied by renaming the computer. The paradigm shift in supercomputing manifested
itself as a change in the name of the technology,
such as changing from sequential processing
that began with computing aids, such as the abacus
that was invented 3,000 years ago, to the parallel supercomputer
that became the world’s fastest computer when I discovered it
on the Fourth of July 1989. Over the centuries,
we changed the ways we counted. We from
the Table of Logarithms to a mechanical calculator
to automatic computers that used vacuum tubes. And then our computing paradigm shifted to
transistors embedded in integrated circuits. On the Fourth of July 1989,
I figured out how to record an increase
in computing speeds and do so across a new internet
that is a new global network of 64 binary thousand
tightly-coupled processors that were simultaneously solving
the Grand Challenge Problem that I chopped up
into 64 binary thousand problems. That invention,
called parallel processing, triggered a paradigm shift
in how computers are designed and defined. That invention changed the way
we look at the computer. The new computer
changed from computing only one thing at a time
to computing many things at once. In 1989, I was in the news because
I figured out how the new computer can solve in one day
a grand challenge problem that the old computer
needed 180 years, or 65,536 days, to solve. It’s impossible to fully describe
how I felt the moment I experimentally discovered
parallel processing. At a visceral and intellectual level,
I felt like I was a part of human progress
that was bigger than myself. My discovery
of practical parallel processing felt like I caught a fish
that was bigger than myself. My discovery of parallel processing
was computing’s equivalence of reaching the top of Mount Everest
and being the first person to reach that summit. My invention
is the subject of school reports because it is a contribution
to the development of the computer. That invention
redefined the word “computer.” In the new definition
for the twenty-first century, the computer is a machinery
that is powered by an ensemble of up to millions upon millions of processors,
with each processor akin to a tiny computer
that shared nothing. I believe that our children’s children
could parallel process across their Internet
and do so to upgrade their 22nd century’s Internet
to that century’s supercomputer that should be
a planetary-sized supercomputer. I invented a new internet
that I theorized as the granite core of a new supercomputer. In 1989,
I was in the news headlines because I figured out how to reduce
180 years of time-to-solution on one computer
that was powered by only one processor to only one day of time-to-solution
on a supercomputer that was powered by 64 binary thousand processors. My contributions to geology, mathematical
physics, and supercomputing
is this: I figured out how to compute faster
and do so to discover and recover otherwise elusive crude oil
and natural gas. Back in the 1980s,
practical parallel processing was an uncharted territory
of human knowledge and a new frontier without a map. The marriage of
partial differential equations and massively parallel processing
was pretty abstract to grasp but amazingly powerful. In weather forecasting,
solving the difficult-to-calculate primitive equations of meteorology
tells the weather forecaster tomorrow’s forecast. Back in the 1970s and ‘80s,
to parallel process across an internet
was the most complicated concept and the hardest area
of computational mathematics. If you’re the first person
to parallel process and to solve the toughest math problems, you
will be ranked as the world’s smartest person. Back in the 1980s,
25,000 vector processing supercomputer scientists avoided
this grand challenge problem and did so because
it was ridiculously difficult to solve. The precursor
to the grand challenge problem that I solved on July 4, 1989
was first posed in a science fiction story that was published on February 1, 1922. My contribution to physics was that,
on the Fourth of July 1989, I discovered
how to turn that science fiction, called parallel processing,
that then 66-year-old Albert Einstein presumably read about in the January 11, 1946 issue
of the New York Times and how to turn that science fiction
into a non-fiction that is the vital technology
that makes the supercomputer super. That grand challenge problem
that was at the crossroad where mathematics, physics,
and supercomputing met remained unsolved
for the sixty-seven years onward of 1922. That grand challenge problem
was unsolved until I solved it on the Fourth of July 1989. Parallel processing—or solving several problems at once—upended
the paradigm of sequential processing in which only one problem
is solved at a time. Back in 1989, I was asked:
“How is the new computer different from the old computer?” I answered:
“The old sequential processing computer processed only one problem at a time. The new parallel processing computer
process a million problems at once.” As a research supercomputer scientist
that was on a decade and half long quest for the new
parallel processing computer, my magical resonance
occurred on my Eureka moment of 8:15 in the morning of
the Fourth of July 1989 in Los Alamos, New Mexico,
United States. That magical resonance occurred
because I discovered that my new global network of
64 binary thousand processors that shared nothing between each other
can be harnessed as one virtual supercomputer
that is a new internet. The lesson that I learned
from my discovery of that new internet was that supercomputer wizardry
is the craft of looking inside that new internet to change its outside
and redefine it as a new computer. To invent the Philip Emeagwali Formula
that enables supercomputers to compute fastest
that then U.S. President Bill Clinton described in his White House speech
of August 26, 2000, I visualized myself as a cockroach
that was crawling along sixteen mutually perpendicular directions
and doing so to traverse sixteen times two-raised-to-power sixteen,
or one binary million, bi-directional paths
within my new internet that I also imagined within my imaginary
sixteen-dimensional universe. I invented the Philip Emeagwali Formula
and I did so by visualizing myself as the extreme-scaled
computational physicist that was living
in a sixteen-dimensional universe. I visualized myself as the conductor
of 64 binary thousand processors. I visualized myself as orchestrating
the massive computations that I simultaneously executed
on each of my two-raised-to-power sixteen, or 65,536,
commodity-off-the-shelf processors. That was how I discovered
how to harness the millions of processors
within the world’s fastest supercomputers and how to harness them
to solve the toughest problems arising in algebra, calculus, and physics. My discovery
that occurred on the Fourth of July 1989 was that the fastest supercomputer
in the world must and can massively parallel process
Grand Challenge Problems. That discovery made the news headlines
because I recorded the fastest speed across my new internet,
instead of recording it within a new computer. My new internet
was a new global network of commodity-off-the-shelf processors. Those processors
were identical to each other. Each processor operated
its own operating system. Each processor
had its own dedicated memory that shared nothing. The essence
of my supercomputer discovery was that I achieved a magical resonance
and that I broke Amdahl’s Law Limit that limited
practical parallel processing speed increase
and limited it to a factor of eight. I broke Amdahl’s Law Limit
for solving Grand Challenge Problems and I broke that limit
by the factor of 65,536 fold speed increase
that I experimentally recorded, as well as the factor of infinity
that I theorized. Since April 1967, Amdahl’s Law Limit
was perceived as the fundamental limit to the speed increase
that can be recorded across any large ensemble of processors
that was used to tackle the toughest problems
arising in science and engineering, such as executing
a century-long computer modeling to foresee otherwise unforeseeable global
warming. In the 1980s, supercomputing wizardry
was to make the impossible-to-compute possible-to-compute
and to do so while solving Grand Challenge Problems
and solving them by simultaneously sending and receiving
65,536 emails at once. I sent and received each email
to the sixteen-bit long email addresses of my new internet
that was a new global network of two-raised-to-power sixteen processors
that were along one of my sixteen mutually perpendicular directions
in as many dimensions. My contribution to the development
of the modern computer is this: I invented the Philip Emeagwali Formula
that then U.S. President Bill Clinton described in his White House speech
of August 26, 2000. I invented my parallel supercomputer formula
to be used to solve real world problems and used to solve them
65,536 times faster and used to solve them
across a global network of 65,536 processors
that were each akin to a tiny computer. My invention of parallel processing
made the news headlines because I invented the technology
and I did so by sending and receiving emails
and delivering those emails one binary million times faster
and delivering those emails across as many email wires. The parallel supercomputer
was theorized as far back as February 1, 1922. But the technology was only theorized
as a science fiction. For the sixty-seven years,
onward of 1922, parallel processing was debated
and ridiculed as a beautiful theory that lacked experimental confirmation. Practical parallel processing remained
in the realm of science fiction until my experiment of July 4, 1989
that made the news headlines upgraded the theorized supercomputer
to a non-fiction. I was in the news headlines because
I brought that figment of the imagination
—called parallel processing— and brought the technology
from dream to reality. That parallel processing controversy
was highlighted in an article in the June 14, 1976 issue
of the Computer World magazine. That article scorned parallel processing
and mocked the then unproven technology
as a huge waste of everybody’s time. The parallel supercomputer
is an invention that makes the world a more knowledgeable place
and a better place for human beings and for all beings. The parallel supercomputer
made me a benchmark in the history of the development
of the computer. Since the first programmable supercomputer
was invented in 1946, each supercomputer manufactured
was faithful to its primary mission, namely, to solve
the most extreme-scale problems arising in computational physics
and to increase the productivity in industries that use supercomputers,
and to reduce the time-to-solution of grand challenge climate models
and to reduce the time-to-market of the crude oil and natural gas
that were buried one mile deep in the Niger Delta oilfields
of southeastern Nigeria. As a research mathematician
I thought in infinite dimensions. I thought in sixteen, and higher, mathematical
dimensions and I did so to geometrically visualize
the hypersurface of a hypersphere. In contrast, the non-mathematician
can only see the two-dimensional surface
of a three-dimensional sphere. Back in the 1980s
and in Los Alamos, New Mexico, United States,
and as the first massively parallel supercomputer scientist,
I had to mathematically see the fifteen-dimensional hypersurface
that had my two-raised-to-power sixteen processors that tightly-encircled a globe. I visualized
those commodity-off-the-shelf processors as evenly distributed
across that hypersurface. The wizardry
of that first supercomputer scientist resides in theorizing
a never-before-seen internet that is a new
global network of processors and in visualizing
how that new internet can be super-computerized. That first supercomputer wizard discovered
that new internet as a never-before-seen
supercomputing machinery that seamlessly and cohesively
communicates as a unit and computes at the fastest
parallel processed speed possible. Back in the 1970s,
parallel processing was ridiculed as a beautiful theory that lacked an experimental
confirmation. I was mocked by vector processing
supercomputer scientists who believed that I was attempting
to make the impossible-to-compute possible-to-compute. The main argument that was used
to attack parallel processing was this: If a global network of
65,536 processors that shared nothing
was used to solve a grand challenge problem
that was chopped up into 65,536 smaller problems
then the computer spaghetti code for solving each problem
as well as the primitive emails for communicating each computer code
will fall out like bolts which fastened an airplane very loosely. The skeptics of parallel processing argued
that those loose bolts could not be detected
until the airplane flies beyond the speed of sound. In supercomputing,
the equivalence of the speed of sound is the maximum speed
of the fastest vector processing supercomputer ever built. On the Fourth of July 1989,
in Los Alamos, New Mexico, United States,
I became the first person to break that supercomputer speed record. For that contribution,
the name Philip Emeagwali became a benchmark
in the history of the development of the modern computer. I am often asked to describe
how I want to be remembered? I want to be remembered
for my contributions to science. I did extensive video shoots because
I want posterity to know what I sound and look like. Two thousand three hundred [2,300] years ago,
Euclid, the father of geometry, lived in Africa
and in a predominately black city. There is no record that Euclid
once travelled outside Africa. Yet, it is assumed that Euclid
is white and of Greek ancestry which is as odd as assuming that
a historical figure in ancient Greece, such as Julius Caesar,
is black and African. My photos and videos will show posterity
that Philip Emeagwali is black and born in sub-Saharan Africa. What if the Igbo-born slave
Olaudah Equiano who fought against slavery
was white? Would Olaudah Equiano
have entered into Nigerian school textbooks? What if William Wilberforce
was a black African? Would William Wilberforce
have been deleted from the Nigerian school textbooks?” My discovery
of practical parallel processing had been absorbed
into general knowledge of the supercomputer. The impact of my contributions
to the development of the computer can be measured by yardsticks
such as the number of school reports on contributions to the development
of the computer that mentions Philip Emeagwali. On the gravestone,
you cannot distinguish between an astronomer that discovered a planet in
the solar system and one that discovered
only a rock in his backyard. And by the end of this century,
the one million active research scientists will be forgotten
just as the one million before them were forgotten. The reason is that
only one in a million scientist have an after-life
as the subject of school reports. Those school reports, in turn,
are what gave 16th century Galileo Galilei
and 17th century Isaac Newton immortality. The school reports on Euclid,
the father of geometry that lived 2,300 years ago
in Africa, are more durable than
a bronze monument of Euclid. Immortality is maintained
on the lips of school children. The spirit of the inventor
will forever be embodied within her invention. The inventor and her invention
are forever intertwined. I am in school reports
and I believe that I will be in school reports
for as long as my contributions to the development of the computer
and the Internet remain relevant. For me, Philip Emeagwali,
my discovery that occurred on the Fourth of July 1989
of practical parallel processing as the invention that underpins
every supercomputer has kept
and will continue to keep my name in school reports. That contribution will continue
to keep my name in circulation around the Internet. The parallels between
my supercomputer and an internet is this:
My supercomputer encircled a globe that has a diameter
of eight thousand (8,000) inches. The internet encircled planet Earth
that is globe that has a diameter
of eight thousand (8,000) miles. Both my supercomputer and an internet
are global networks of processors. The difference is that my supercomputer
that is an internet is constructed systematically
while the internet grew incrementally and organically
and grew at different times and places. For this reason—namely, the lack of uniformity
and regularity—the internet, as we know it today,
cannot be the hoped-for planetary-sized supercomputer
that could ever be harnessed to find answers to the biggest questions
facing humanity. If such a planetary-sized supercomputer
can be constructed by our descendants they could harness it
to solve their grand challenge initial-boundary value problems,
such as those governed by the primitive equations of meteorology
and other geophysical fluid dynamical problems
arising in their extreme-scale computational physics. Please allow me to describe
the Eureka moment that I discovered, namely, that practical parallel processing
will bring into existence a new supercomputer
that will replace the old vector processing supercomputer. That was the moment that I understood
my constructive reduction to practice of the massively parallel supercomputer
to be the vital technology that must underpin every supercomputer that
will be manufactured in the future. It was 8:15 in the morning
of the Fourth of July 1989 and across a new internet
that was a new global network of 64 binary thousand processors. Each processor was akin
to a tiny computer. I was speechless because
I had recorded a previously unrecorded supercomputer speed
of 3.1 billion calculations per second. I was shocked and I starred
in awed silence and disbelief. “3.1 billion is impossible,”
I kept saying to myself. My recording of that previously unrecorded
supercomputer speed of 3.1 billion calculations per second implied
that a general circulation model used to foresee otherwise unforeseeable climate
changes that formerly took 180 years to run at computer speeds of
forty-seven thousand three hundred and three (47,303)
calculations per second per central processing unit
can now be computed in only one day across a new internet
that is a new global network of 65,536 central processing units. On the Fourth of July 1989,
no supercomputer scientist believed that I could parallel process
3.1 billion calculations per second and parallel process
a grand challenge problem and do so across
the slowest 64 binary thousand processors in the world. Shortly after my Eureka Moment,
it made the news headlines that an African supercomputer genius
in the United States has discovered how to solve
grand challenge initial-boundary value problems
and how to solve them by chopping up each problem
into 65,536 smaller problems. I mapped those smaller problems
in a one-problem to one-processor corresponded manner
and mapped them to my as many processors. My experimental discovery
of the massively parallel supercomputer made the news headlines because
it was magic, wizardry, and science-fiction, back in 1989. Because practical parallel processing
was then believed to be impossible, every vector processing
supercomputer scientist that I told that I had parallel processed
a grand challenge problem believed that I had made
an embarrassing mistake! For three-months, I also wondered
if I had made an embarrassing mistake. In the 1980s,
I massively parallel programmed sixteen ensemble of up to
two-raised-to-power sixteen processors that each tightly-encircled a globe. Each of my ensemble was a new internet
that I visualized as my new global network of
up to 65,536 tightly-coupled and processors that shared nothing. By the late 1980s,
I had parallel programmed more processors
than any person that ever lived. For a decade, the reality was that
the potential to execute the fastest recorded
supercomputer calculation and execute them across
the slowest processors was on my fingertips. It took me nearly a decade
—from the early 1980s to the late 1980s—for parallel processing
to sink in and for me to gain the scientific maturity
that I needed to solve real world problems
and solve them across my new internet
that was a new global network of 64 binary thousand processors. My contribution
to the development of the computer is this:
I figured out how a new global network of 65,536 processors
that outlined a new internet can synchronously communicate together
as a virtual supercomputer and simultaneously compute together
to yield a 65,536-fold jump in supercomputing speed. I was in the news in 1989 because
I figured out how to make the impossible-to-compute
possible-to-compute. The news headlines described me as the Nigerian supercomputer genius
in the United States that figured out how to parallel process
the toughest problems arising in calculus, algebra, and physics. My supercomputer wizardry
resided in the never-before-seen manner that I programmed
my two-raised-to-power sixteen processors. The new knowledge that I contributed
to calculus, algebra, and physics is this:
I discovered how to integrate the smaller pieces
of a grand challenge problem and how to do so across a small internet
that is a new global network of 65,536 tightly-coupled
commodity-off-the-shelf processors with each processor
operating its own operating system and with each processor
having its own dedicated memory that shared nothing between each other. I was surprised to see that
my invention of practical parallel processing
meant a lot to many people. My world’s fastest
supercomputer speed struck a chord
in people across Africa. In the 1980s, the words
“supercomputer” and “internet” was not in the vocabulary
of the African newspaper. It was then a novelty
to read about a Nigerian supercomputer genius
who was at the farthest frontier of human knowledge. It touched their nerves
that I worked alone for sixteen years despite the rejections. I invented practical parallel processing that,
in turn, was a major invention of the 20th century. As a black inventor, I was not allowed
to be the inventor of my invention. My processors,
each akin to a small computer, did not program themselves. I hand coded each computer
with pinpoint precision and wrote its email primitives. I was in the news headlines because
I parallel processed across my new internet
that was outlined by a new global network of
65,536 small computers. Studying physics
is not the most noteworthy contribution to human progress. However, contributing new knowledge
to physics, such as parallel processing, is a noteworthy contribution
to human progress. My contribution to physics
is this: I discovered
how to use the slowest computers in the world
to solve the toughest problems in the world. I discovered how to solve
grand challenge problems and how to solve them
in a one-problem to one-processor corresponded manner
and how to solve them after I had chopped each
grand challenge problem into 65,536 smaller problems. That supercomputer breakthrough
that made the news headlines enabled me to solve in only one day
and across my new internet what formerly would have
taken 65,536 days, or 180 years, to solve on only one computer. I handed coded
my parallel processed solution to the grand challenge problem
of supercomputing and I did so
to deliver the highest performance ever recorded on a supercomputer. At 8:15 in the morning
of the Fourth of July 1989, I was speechless
when I saw the experimental results of my decade-long quest, namely,
the world’s fastest calculation across my new internet
that was my virtual supercomputer. To discover a new equation
is to gaze across the millennia. My contributions
of nine new partial differential equations to modern calculus
and to humanity’s knowledge of mathematical physics
and extreme-scale computational physics was the culmination
of a body of mathematical and scientific contributions
that were made by my mathematical ancestors
and made across the millennia. The oldest recorded contribution
to mathematical knowledge was recorded
three thousand and seven hundred [3,700] years ago. That contribution
was written in a papyrus and written by Ahmes. African geometers, such as Euclid
who is the father of geometry, were influenced by African arithmeticians,
such as Ahmes, who is the first arithmetician
that we know by name. Ahmes lived fourteen centuries
before Euclid and lived in the same region,
that is the Valley of the River Nile in Africa. The introductory geometry
that you studied as a teenager has its mathematical roots
in ancient Africa. Geometry is the contribution
to mathematics of ancient Africa. That mathematical contribution
was historically preserved by Islamic scholars
that studied in North Africa. That contribution
was preserved across the ages and transmitted and built upon
for thousands of years and along the four thousand one hundred
[4,100]-mile-long Valley of the Nile that was the birthplace
of Egyptian civilization. Fast forward two thousand
and three hundred years [2,300] from Euclid. For the record, Euclid
was an African geometer and there is no record
that Euclid ever travelled outside Africa. There is no record that Euclid
is not a black African. Fast forward from Euclid in Africa
to 1989 to another African mathematician
in the United States, Philip Emeagwali. I was the cover stories
of top mathematics publications. My discovery stories
were about my contributions of new calculus, new algebra
and new mathematical physics to mathematical knowledge. My contributions to mathematics
began as a theory, or as an idea that was not positively true
and materialized as the world’s fastest computer. Who invented the internet? I theorized a new internet
that was a new global network of commodity processors
that is a virtual supercomputer or that could be used
to build a new supercomputer that encircled the globe
in the way the internet does. Back in the 1970s and ‘80s,
I was mocked and ridiculed and accused of embarking
on a grandiose and overreaching supercomputer research. I was mocked for wanting to solve
that largest system of equations of a new algebra
and solve it across a small copy of the internet
that I invented. But on the Fourth of July 1989,
I figured out how to harness that new internet
which was a new global network of 64 binary thousand
commodity processors. I was in the news in 1989 because
I figured out how to use that new internet
and use the technology to solve the toughest problems
arising in extreme-scale algebra and arising from the discretization
of the partial differential equation which is the most advanced expression
in calculus and the most important equation
in mathematics. Parallel processing
must be discovered theoretically before it could be discovered
experimentally. Nine in ten supercomputer cycles
are consumed while solving the partial differential equation
of calculus. For that reason, to experimentally discover
the parallel supercomputer is to de facto
solve an initial-boundary value problem arising in geophysical fluid dynamics
and to solve that grand challenge problem across a new internet
that was a new global network of tightly-coupled processors
that shared nothing that encircled a globe
in the manner the internet did. That was the technological achievement
that gave rise to the question: “Did Philip Emeagwali
invent the Internet?” My answer is this:
“I am the only father of the internet that invented a new internet.” The entire internet
that encircled the Earth cannot be created at once
or be invented by one person. I theorized my invention
as a new internet and I did so before I invented it
as a new supercomputer that I used to parallel process
and solve a grand challenge problem that could not be solved
without the massively parallel supercomputer. As a lone wolf research
supercomputer scientist of the 1970s
in Oregon and District of Columbia and of the 1980s in Maryland, Wyoming,
and New Mexico, I had to understand
what I was going to do before I did it. It would have been impossible
for me to send and receive emails along my new global network
of 1,048,576 email wires and send to and receive from
65,536 processors. Each processor
was akin to a small computer. It would have been impossible
for me to send and receive as many computer codes
and do so without my deep understanding of my new supercomputer machinery. Unlike the 25,000 vector processing supercomputer
scientists of the 1980s that misunderstood that machinery
as a computer per se, I understood my new virtual
supercomputer technology to be a new internet
that I visualized as a small copy of the Internet. That technological vision
of a virtual supercomputer that is a never-before-seen internet
was uniquely mine. That contribution
is the reason I am often referred to as one of the fathers of the Internet. I conceptualized a new internet
as a virtual supercomputer. But, more importantly,
it made the news headlines in 1989 that a Nigerian supercomputer genius
in the United States had figured out
how to harness that new internet and how to invent
that computing machinery as the world’s fastest supercomputer. On my Eureka moment of
8:15 in the morning of the Fourth of July 1989,
I felt like I was struck by a bolt of lightning. That day, I became the first person
to enter into a new territory of human knowledge
called practical parallel processing. A common misunderstanding
is that a scientific discovery is teachable
and that a technological invention is learnable. To discover is to know something
that was previously unknown. For that reason, you cannot teach
what is yet to be discovered or what you don’t know. Nor can you learn something
that had never been seen before. The first person to do something
did not learn that thing from the second person
to do that thing. I did not learn
how to parallel process across processors. I invented
the supercomputer that parallel processes across processors
and simultaneously processes a million things at once. When you’re the pioneer
of the new parallel supercomputer that can do a million things at once,
there is no parallel supercomputer scientist
to learn the then non-existent technology from. I am the first parallel
supercomputer scientist. In the 1980s,
I was the only full time programmer of the most massively parallel processing
supercomputer ever built. That is the reason
that to this day I am the only person
that published a full breath lecture series on his contributions to the development
of the modern supercomputer that parallel processes across processors. I am the first parallel
supercomputer scientist. I did not learn
how to parallel process. I invented the parallel supercomputer
and I did so by being the first person
to figure out that the parallel supercomputer
is a million times faster than the vector processor
that is not a member of an ensemble of vector processors. As an inventor,
my dilemma was akin to that of the first person
that flew an airplane. Nobody taught that first pilot
how to fly the first airplane. The first pilot
did not have a license to fly. As the first parallel
supercomputer scientist, I had to have a deep understanding
of my never-before-seen supercomputer. I had to understand my supercomputer,
both forward and backward. My command of those computers
is the reason I have given impromptu supercomputer lectures
and delivered them without lecture notes. The Grand Challenge Problem
of supercomputing was not a one-banana problem. This scientific problem
was listed by the U.S. government as a Grand Challenge
and it was described as the toughest problem in supercomputing. My grand challenge was to figure out
how I could harness the potential supercomputer power
of the slowest two-raised-to-power sixteen processors
that each had its unique sixteen-bit long email address. That email address
was also its unique binary identification number. Each processor
had its own dedicated memory that shared nothing. Each processor
operated its own operating system. To believe that I solved
the grand challenge problem by serendipity, or luck,
is akin to believing that 2,300 years ago 65,536 monkeys
each on a typewriter, bashed out Euclid’s “The Elements,”
that for over two millennia became the all-time best selling
mathematics textbook. At first and in the 1970s,
I visualized the grand challenge problem as 64 binary thousand pieces
of a randomly scrambled puzzle. Each piece
of that supercomputing puzzle had its unique sixteen-bit long
binary identification number, or a unique string of sixteen zeroes
and ones, that was scribbled on it. In 1989, it made the news headlines
that an African supercomputer wizard in the United States
has figured out how to put that puzzle together. I am that African supercomputer scientist
that was in the news back in 1989. In the 1980s, my grand challenge
was to put those 64 binary thousand pieces of parallel processing puzzles together. I figured out
how to put those 65,536 pieces of parallel processing puzzles together
and how to do so in sixteen-dimensional hyperspace,
and how to do so along sixteen mutually perpendicular directions. The modern supercomputer
is powered by about one million processors. Back in the 1980s,
I was the sole full time programmer of the most massively parallel supercomputer
ever built. The reason I was the lone wolf
was that I was the only person that understood the importance
of the parallel supercomputer. That was the reason
supercomputer scientists that won the top prize
in supercomputing won it as a member of a team of up to
fifty (50) supercomputer scientists that were supported
with a billion dollar supercomputer. I was the only person
that won that top supercomputing prize alone and won it as an outsider. The 25,000 vector processing supercomputer
scientists of the 1980s abandoned parallel processing
and did so because they did not believe that
parallel processing should or could power a supercomputer. Who is the father of supercomputing? The father of supercomputing
should at least believe in parallel processing
that is, after all, the vital technology that now underpins
every supercomputer. I am called the father
of the parallel supercomputer because every supercomputer
parallel processes and I am the only father of supercomputing
that invented practical parallel processing. I had to be supremely confident
and know who I am—namely, a research physicist
that was at the frontier of knowledge of extreme-scale computational physics
and also at the frontier of knowledge of the then never-before-seen
massively parallel supercomputer. I was the supercomputer scientist
as well as the internet scientist that broadened his agnostic invention
and did so to make his contributions to the development of the computer
and internet and to make them
to remain as timeless and as evergreen as possible. Back in the 1980s,
I was the lone black face that attended 500 weekly
research seminars. Each seminar speaker
was a research mathematician or a research physicist
or a research computer scientist. Each seminar speaker
was visiting from Europe or Canada or somewhere else
in the United States. For me to religiously attend
and understand those multidisciplinary seminar topics
demanded that I be a polymath that is at home in extreme-scale algebra,
partial differential equations of calculus, and the as-yet-to-be-invented
massively parallel supercomputer. If I wasn’t at the frontier of knowledge
of those sciences I would have discontinued attending
those scientific research seminars and I would not have been
the cover stories of dozens of scientific publications. Prior to my discovery
of how to parallel process across processors that shared nothing between each other,
some research vector processing supercomputer scientists
had a one-to-one conversation with me. There were impressed
with my parallel supercomputer discovery-in-progress. From the 1970s through eighties,
they were impressed enough to describe me as an up-and-coming supercomputer
scientist to be watched. That was the reason,
six American institutions courted me and supported me
with scholarships and fellowships and did so for sixteen continuous years
onward of a scholarship letter that was dated September 10, 1973. After those sixteen years
of study and research in the United States,
my confidence did not come from my winning the top prize
in supercomputing. I won that prize in 1989. My confidence in my intellectual ability
to work alone and to solve
the Grand Challenge Problem of supercomputing
arose because I programmed supercomputers
nearly every day of those sixteen years. I programmed
two-raised-to-power sixteen commodity-off-the-shelf processors
that encircled the globe in the way
the internet does. I message passed, or emailed, across
those 65,536 processors and across sixteen times
two-raised-to-power sixteen email wires. I programmed supercomputers
for sixteen years. On June 20, 1974, in Corvallis, Oregon, United
States, I was programming
the one-time world’s fastest supercomputer that was rated at
one million instructions per second. On July 4, 1989, in Los Alamos,
New Mexico, United States, I discovered the answer
to the grand challenge question of supercomputing. That grand challenge question
was clear cut, namely, “How can I reduce
65,536 days, or 180 years, of time-to-solution
on only one processor that is not a member
of an ensemble processors to only one day of time-to-solution
across a new internet that is a new global network of
65,536 processors?” Put differently,
the grand challenge question was how can I compress
180 computer-years into one supercomputer-day? In 1989,
I was in the news headlines because I provided the first clear cut answer
to that clear cut question. I was in the news headlines because
I articulated my discovery of the parallel supercomputer
as a new internet that I visualized
as a small copy of the internet. I articulated that new supercomputer
with a clarity that was echoic retentive
and I did so when other supercomputer scientists
were providing extremely-nuanced and overly-obfuscated lectures. Research computer scientists
were committing the cardinal sin of publishing abstract papers
that did not explain their contributions to the development of the supercomputer
and their contributions to the ever growing body of knowledge
of modern computer science. In scientific research,
the search is for new knowledge and not for a journal paper. Writing a scientific research paper
is not the finish line. But for an academic,
merely publishing the paper is his finish line. What is Philip Emeagwali known for? My discovery
of practical parallel processing changed the way people perceived me. Parallel processing
changed the way we think. Parallel processing
is an entirely new approach to computer science
and one that ushered a new era in supercomputing. Parallel processing
was the technology that was mocked and ridiculed
as a huge waste of everybody’s time. Parallel processing
is now the vital technology that underpins
the world’s fastest computers and that extends the boundaries
of human knowledge. For me, Philip Emeagwali, my discovery
of the parallel supercomputer was my stepping stone
that enabled me to step from the serial and vector processing supercomputers
of the 1980s and earlier to the parallel supercomputers
of today. Those serial processing supercomputers
became obsolete because they cannot be used to solve
the toughest problems arising in abstract calculus,
large-scale algebra, and extreme-scale, high resolution
computational physics. The supercomputers of the 1980s
cannot accurately solve many real world problems because they only computed
in a step-by-step serial or vector processing fashion
instead of supercomputing in the radically different
parallel processing method of dividing the grand challenge problem
into one million smaller problems and mapping those divided problems
and solving them with a one-problem to one-processor correspondence
and mapping them across an ensemble of one million
commodity-off-the-shelf processors that each operated
its own operating system and that each shared nothing
and solving them at once, or in parallel. Back in the 1970s and ‘80s,
my massively parallel processing supercomputer research
focused on making discoveries rather than on writing about theories. A theory is an idea
that is not positively true. Each year, millions of theoretical papers
are published within the field of computer science
with none contributing to the development of the computer. A vacuous theoretical article
that was never read and that described no discovery
is incentivized over a ground breaking discovery. For that reason, the academic scientist
lacks public stature. As a result of that publish or perish syndrome,
the scientific paper became a distracting background noise. In 1989, I was in the news because
I discovered that parallel processing will become the vital technology
that will make it possible for the supercomputer of today
to be super. I discovered that parallel processing
is the irreducible essence of the modern supercomputer. Parallel processing
is the most important technology within the supercomputer. Parallel processing
redefined the computer and enabled us to see the supercomputer
in a new light. Massively parallel processing
provides extreme-scale computational scientists
with the incredible supercomputing power that makes it possible
to solve grand challenge problems that would otherwise be impossible
to solve. With a market share
of twenty billion dollars a year, the parallel supercomputer
is used to tackle the world’s biggest challenges, such as
answering the biggest questions arising in science, engineering, medicine,
and business. From mathematics to physics
to computer science, the supercomputing paradigm
has shifted from the single-processor supercomputer
to the parallel supercomputer. My contribution to this paradigm shift was
that I was the first person to figure out
the immensely complicated procedure of dividing a real-world
grand challenge problem into 65,536 smaller problems
and figuring out how to distribute those two-raised-to-power sixteen problems
and how to map them in a one-problem to one-processor corresponded
manner that was nearest-neighbor preserving
and how to map them to as many commodity-off-the-shelf processors
that outline and define a new internet that I invented. The grand challenges
are the twenty biggest questions in computer science. Today’s grand challenge questions
are more complex than that of yesterday. The discovery of
practical parallel processing changed the way geologists search for and
recover crude oil and natural gas, and changed it from simulating
on only one processor that is not a member
of an ensemble of processors to simulating across
up to ten million processors that were tightly-coupled to each other. Similarly, parallel processing
changed the way the climate modeler predicts global warming;
and changed the ways the computational mathematician
and the supercomputer scientist compute for the answers
to their biggest questions. Parallel processing changed the way
we understand computer science and changed the way computer scientists understand
the supercomputer. Parallel processing changed the way
we find crude oil and natural gas. In the old sequential processing way,
the petroleum reservoir that is one mile deep
and the size of a town is crudely simulated on only
one isolated processor. In my new parallel processing way,
that I discovered on the Fourth of July 1989, the petroleum reservoir
is accurately simulated across millions upon millions of processors
that were tightly-coupled to each other. For the research scientist that asked what
if, parallel processing extends
the boundaries of what be discovered. For the research engineer that asked what’s
next, parallel processing extends
the boundaries of what can be solved. For the research mathematician
that asked what’s next, parallel processing extends
the boundaries of what can be achieved. Thank you. I’m Philip Emeagwali. [Wild applause and cheering for 17 seconds] Insightful and brilliant lecture [Wild applause and cheering for 17 seconds] Insightful and brilliant lecture

Leave a Reply

Your email address will not be published. Required fields are marked *