[ExI] The Slow Winter

[Eugen Leitl]

https://www.usenix.org/system/files/1309_14-17_mickens.pdf

According to my dad, flying in airplanes used to be fun. You could smoke on
the plane, and smoking was actually good for you. Everybody was attractive,
and there were no fees for anything, and there was so much legroom that you
could orient your body parts in arbitrary and profane directions without
bothering anyone, and you could eat caviar and manatee steak as you were
showered with piles of money that were personally distributed by JFK and The
Beach Boys. Times were good, assuming that you were a white man in the
advertising business, WHICH MY FATHER WAS NOT SO PERHAPS I SHOULD ASK HIM
SOME FOLLOW-UP QUESTIONS BUT I DIGRESS. The point is that flying in airplanes
used to be fun, but now it resembles a dystopian bin-packing problem in which
humans, carry-on luggage, and five dollar peanut bags compete for real estate
while crying children materialize from the ether and make obscure demands in
unintelligible, Wookie-like languages while you fantasize about who you won’t
be helping when the oxygen masks descend.

I think that it used to be fun to be a hardware architect. Anything that you
invented would be amazing, and the laws of physics were actively trying to
help you succeed. Your friend would say, “I wish that we could predict
branches more accurately,” and you’d think, “maybe we can leverage three bits
of state per branch to implement a simple saturating counter,” and you’d
laugh and declare that such a stupid scheme would never work, but then you’d
test it and it would be 94% accurate, and the branches would wake up the next
morning and read their newspapers and the headlines would say OUR WORLD HAS
BEEN SET ON FIRE. You’d give your buddy a high-five and go celebrate at the
bar, and then you’d think, “I wonder if we can make branch predictors even
more accurate,” and the next day you’d start XOR’ing the branch’s PC address
with a shift register containing the branch’s recent branching history,
because in those days, you could XOR anything with anything and get something
useful, and you test the new branch predictor, and now you’re up to 96%
accuracy, and the branches call you on the phone and say OK, WE GET IT, YOU
DO NOT LIKE BRANCHES, but the phone call goes to your voicemail because
you’re too busy driving the speed boats and wearing the monocles that you
purchased after your promotion at work. You go to work hung-over, and you
realize that, during a drunken conference call, you told your boss that your
processor has 32 registers when it only has 8, but then you realize THAT YOU
CAN TOTALLY LIE ABOUT THE NUMBER OF PHYSICAL REGISTERS, and you invent a
crazy hardware mapping scheme from virtual registers to physical ones, and at
this point, you start seducing the spouses of the compiler team, because it’s
pretty clear that compilers are a thing of the past, and the next generation
of processors will run English-level pseudocode directly. Of course, pride
precedes the fall, and at some point, you realize that to implement
aggressive out-of-order execution, you need to fit more transistors into the
same die size, but then a material science guy pops out of a birthday cake
and says YEAH WE CAN DO THAT, and by now, you’re touring with Aerosmith and
throwing Matisse paintings from hotel room windows, because when you order
two Matisse paintings from room service and you get three, that equation is
going to be balanced. It all goes so well, and the party keeps getting
better. When you retire in 2003, your face is wrinkled from all of the
smiles, and even though you’ve been sued by several pedestrians who suddenly
acquired rare paintings as hats, you go out on top, the master of your
domain. You look at your son John, who just joined Intel, and you rest well
at night, knowing that he can look forward to a pliant universe and an easy
life.

Unfortunately for John, the branches made a pact with Satan and quantum
mechanics during a midnight screening of “Weekend at Bernie’s II.” In
exchange for their last remaining bits of entropy, the branches cast evil
spells on future generations of processors. Those evil spells had names like
“scaling induced voltage leaks” and “increasing levels of waste heat” and
“Pauly Shore, who is only loosely connected to computer architecture, but who
will continue to produce a new movie every three years until he sublimates
into an empty bag of Cheetos and a pair of those running shoes that have
individual toes and that make you look like you received a foot transplant
from a Hobbit, Sasquatch, or an infertile Hobbit/Sasquatch hybrid.” Once
again, I digress. The point is that the branches, those vanquished foes from
long ago, would have the last laugh.

When John went to work in 2003, he had an indomitable spirit and a love for
danger, reminding people of a less attractive Ernest Hemingway or an
equivalently attractive Winston Churchill. As a child in 1977, John had met
Gordon Moore; Gordon had pulled a quarter from behind John’s ear and then
proclaimed that he would pull twice as many quarters from John’s ear every 18
months. Moore, of course, was an incorrigible liar and tormentor of youths,
and he never pulled another quarter from John’s ear again, having immediately
fled the scene while yelling that Hong Kong will always be a British
territory, and nobody will ever pay $8 for a Mocha Frappuccino, and a variety
of other things that seemed like universal laws to people at the time, but
were actually just arbitrary nouns and adjectives that Moore had scrawled on
a napkin earlier that morning. Regardless, John was changed forever, and when
he grew up and became a hardware architect, he poured all of his genius into
making transistors smaller and more efficient. For a while, John’s efforts
were rewarded with ever-faster CPUs, but at a certain point, the transistors
became so small that they started to misbehave. They randomly switched
states; they leaked voltage; they fell prey to the seductive whims of cosmic
rays that, unlike the cosmic rays in comic books, did not turn you into a
superhero, but instead made your transistors unreliable and shiftless, like a
surly teenager who is told to clean his room and who will occasionally just
spray his bed with Lysol and declare victory.

As the transistors became increasingly unpredictable, the foundations of
John’s world began to crumble. So, John did what any reasonable person would
do: he cloaked himself in a wall of denial and acted like nothing had
happened. “Making processors faster is increasingly difficult,” John thought,
“but maybe people won’t notice if I give them more processors.” This, of
course, was a variant of the notorious Zubotov Gambit, named after the
Soviet-era car manufacturer who abandoned its attempts to make its cars not
explode, and instead offered customers two Zubotovs for the price of one,
under the assumption that having two occasionally combustible items will
distract you from the fact that both items are still occasionally
combustible. John quietly began to harness a similar strategy, telling his
marketing team to deemphasize their processors’ speed, and emphasize their
level of parallelism.

At first, John’s processors flew off the shelves. Indeed, who wouldn’t want
an octavo-core machine with 73 virtual hyper-threads per physical processor?
Alan Greenspan’s loose core policy and weak parallelism regulation were
declared a resounding success, and John sipped on champagne as he watched the
money roll in. However, a bubble is born so that a bubble can pop, and this
one was no different. John’s massive parallelism strategy assumed that lay
people use their computers to simulate hurricanes, decode monkey genomes, and
otherwise multiply vast, unfathomably dimensioned matrices in a desperate
attempt to unlock eigenvectors whose desolate grandeur could only be imagined
by Edgar Allen Poe. 

Of course, lay people do not actually spend their time trying to invert
massive hash values while rendering nine copies of the Avatar planet in
1080p. Lay people use their computers for precisely ten things, none of which
involve massive computational parallelism, and seven of which involve
procuring a vast menagerie of pornographic data and then curating that data
using a variety of fairly obvious management techniques, like the creation of
a folder called “Work Stuff,” which contains an inner folder called “More
Work Stuff,” where “More Work Stuff” contains a series of ostensible
documentaries that describe the economic interactions between people who
don’t have enough money to pay for pizza and people who aren’t too bothered
by that fact. Thus, when John said “imagine a world in which you’re
constantly executing millions of parallel tasks,” it was equivalent to saying
“imagine a world that you do not and will never live in.” Indeed, a world in
which you’re constantly simulating nuclear explosions while rendering massive
3-D environments is a world that’s been taken over by members of a high
school A.V. club. The members of a high school A.V. club lack the chops to
establish a global dictatorship, if only because doing such a thing would
require them to reduce their visits to Renaissance festivals, and those
turkey legs need help to be consumed in the style of a 15th century Italian
aristocrat.

John was terrified by the collapse of the parallelism bubble, and he quickly
discarded his plans for a 743-core processor that was dubbed The Hydra of
Destiny and whose abstract Platonic ideal was briefly the third-best chess
player in Gary, Indiana. Clutching a bottle of whiskey in one hand and a
shotgun in the other, John scoured the research literature for ideas that
might save his dreams of infinite scaling. He discovered several papers that
described software-assisted hardware recovery. The basic idea was simple: if
hardware suffers more transient failures as it gets smaller, why not allow
software to detect erroneous computations and re-execute them? This idea
seemed promising until John realized THAT IT WAS THE WORST IDEA EVER. Modern
software barely works when the hardware is correct, so relying on software to
correct hardware errors is like asking Godzilla to prevent Mega-Godzilla from
terrorizing Japan. THIS DOES NOT LEAD TO RISING PROPERTY VALUES IN TOKYO.
It’s better to stop scaling your transistors and avoid playing with monsters
in the first place, instead of devising an elaborate series of monster
checksand-balances and then hoping that the monsters don’t do what monsters
are always going to do because if they didn’t do those things, they’d be
called dandelions or puppy hugs.

At this point, John was living under a bridge and wearing a bird’s nest as a
hat. Despite his tragic sartorial collaborations with the avian world, John
still believed that somehow, some way, he could continue to make his
transistors smaller. Perhaps the processor could run multiple copies of each
program, comparing the results to detect errors? Perhaps a new video codec
could tolerate persistently hateful levels of hardware error?  All of these
techniques could be implemented. However, John slowly realized that these
solutions were just things that he could do, and inventing “a thing that you
could do” is a low bar for human achievement. If I were walking past your
house and I saw that it was on fire, I could try to put out the fire by
finding a dingo and then teaching it how to speak Spanish. That’s certainly a
thing that I could do. However, when you arrived at your erstwhile house and
found a pile of heirloom ashes, me, and a dingo with a chewed-up Rosetta
Stone box, you would be less than pleased, despite my protestations that
negative scientific results are useful and I had just proven that
Spanishilliterate dingoes cannot extinguish fires using mind power.

It was at this moment, when John had hit the bottom, that he discovered
religion.

John began to attend The Church of the Impending Power Catastrophe. He sat in
the pew and he heard the cautionary tales, and he was afraid. John learned
about the new hyperthreaded processor from AMD that ran so hot that it burned
a hole to the center of the earth, yelled “I’ve come to rejoin my people!”,
discovered that magma people are extremely bigoted against processor people,
and then created the Processor Liberation Front to wage a decades-long,
hilariously futile War to Burn the Intrinsically OK-With-Being-Burnt Magma
People. John learned about the rumored Intel Septium chip, a chip whose
prototype had been turned on exactly once, and which had leaked so much
voltage that it had transformed into a young Linda Blair and demanded an
exorcism before it embarked on a series of poor career moves that culminated
in an inevitable spokesperson role for PETA. The future was bleak, and John
knew that he had to fight it. So, John repented his addiction to scaling, and
he rededicated his life to reducing the power consumption of CPUs. It was a
hard path, and a lonely path, but John could find no other way. Formerly the
life of the party, John now resembled the scraggly, one-eyed wizard in a
fantasy novel who constantly warns the protagonist about the variety of
things that can lead to monocular bescragglement. At team meetings, whenever
someone proposed a new hardware feature, John would yell “THE MAGMA PEOPLE
ARE WAITING FOR OUR MISTAKES.” He would then throw a coffee cup at the
speaker and say that adding new hardware features would require each
processor to be connected to a dedicated coal plant in West Virginia. John’s
coworkers eventually understood his wisdom, and their need to wear
coffeeresistant indoor ponchos lessened with time. Every evening, after John
left work, he went to the bus stop and distributed power literature to
strangers, telling them to abandon transistor scaling and save their souls.
Standing next to John, another man wore a sandwich board that said that the
Federal Reserve was using fluorinated water to hide the fact that we never
landed on the moon. The sandwich board required no transistors at all. It
made John smile.

When John comes home for the holidays, you’re glad that he’s back, but you
miss the old twinkle in his eye. Your thoughts wander to your own glory days
thirty years ago, when Aerosmith mistook young John for a large Xanax tablet
and tried to trade him for a surface-to-air missile that could be used
against anti-classic rock regimes. Oh, how you laughed! The subsequent visit
by Child Protection Services was less amusing, but that was the way that
hardware architects lived: working hard, partying hard, and occasionally
waking up in Tijuana to discover that your left kidney is missing and your
toddler has been shipped to a Columbian arms smuggler. It was crazy, but you
wouldn’t change a thing. Your generation had lived so many dreams, and slain
so many foes. 

Today, if a person uses a desktop or laptop, she is justifiably angry if she
discovers that her machine is doing a non-trivial amount of work. If her hard
disk is active for more than a second per hour, or if her CPU utilization
goes above 4%, she either has a computer virus, or she made the disastrous
decision to run a Java program. Either way, it’s not your fault: you brought
the fire down from Olympus, and the mortals do with it what they will. But
now, all the easy giants were dead, and John was left to fight the ghosts
that Schrödinger had left behind. “John,” you say as you pour some eggnog,
“did I ever tell you how I implemented an out-of-order pipeline with David
Hasselhoff and Hulk Hogan’s moustache colorist?” You are suddenly aware that
you left your poncho in the other room.

James Mickens is a researcher in the Distributed Systems group at Microsoft’s
Redmond lab. His current research focuses on web applications, with an
emphasis on the design of JavaScript frameworks that allow developers to
diagnose and fix bugs in widely deployed web applications. James also works
on fast, scalable storage systems for datacenters.  James received his PhD in
computer science from the university of Michigan, and a bachelor’s degree in
computer science from georgia Tech. mickens at microsoft.com