An interesting view on parallel computing.

Discussion in 'CPUs and Overclocking' started by William Gaatjes, Dec 23, 2012.

  1. William Gaatjes

    Joined:
    May 11, 2008
    Messages:
    15,609
    Likes Received:
    236
    For those that like cpu's and parallel execution of threads (swapping threads while waiting for results, might like this small excerpt from this book from Richard Feynman : Surely you are joking.

    Funny view on how to solve a problem.
    It is all about the colored cards...

    Take note, i did not edit out some parts of the text (would have made it easier to keep the focus on the computing problem).
     
    #1 William Gaatjes, Dec 23, 2012
    Last edited: Dec 23, 2012
  2. Loading...

    Similar Threads - interesting view parallel Forum Date
    This is interesting. Kaby lake and Sandy bridge comparison. By [H]. CPUs and Overclocking Jan 14, 2017
    Anything interesting in the upcoming Atom and Celeron roadmaps (mini PCs&Chromebook)? CPUs and Overclocking Mar 11, 2016
    Kinda Interesting: Intel 47W Quad w/ 4 Ghz Turbo CPUs and Overclocking Jul 29, 2014
    There's something very interesting about the desktop Kabini's... CPUs and Overclocking Mar 25, 2014
    Commoditization of CPUs, my view of the market, and a warning to Intel... CPUs and Overclocking Apr 25, 2012

  3. nemesismk2

    nemesismk2 Diamond Member

    Joined:
    Sep 29, 2001
    Messages:
    4,791
    Likes Received:
    3
    yes i agree it's very interesting
     
  4. Borealis7

    Borealis7 Platinum Member

    Joined:
    Oct 19, 2006
    Messages:
    2,496
    Likes Received:
    30
    you didn't read the whole thing :p
    i stopped when he started talking about his wife or flat tires...but i read the last part and think i got the point.
     
    #3 Borealis7, Dec 23, 2012
    Last edited: Dec 23, 2012
  5. Ben90

    Ben90 Platinum Member

    Joined:
    Jun 14, 2009
    Messages:
    2,866
    Likes Received:
    1
    I read the whole thing and found all of it interesting. I wish I was more involved with the industry. OoO seems like one of those things that would keep you up at night thinking of creative ways to do things.
     
  6. lehtv

    lehtv Elite Member

    Joined:
    Dec 8, 2010
    Messages:
    11,870
    Likes Received:
    68
    Wall of text... please fix and maybe more people might consider reading it
     
  7. Ferzerp

    Ferzerp Diamond Member

    Joined:
    Oct 12, 1999
    Messages:
    6,316
    Likes Received:
    36
    drugs are bad mmkay
     
  8. ShintaiDK

    ShintaiDK Lifer

    Joined:
    Apr 22, 2012
    Messages:
    20,395
    Likes Received:
    128
    Gave up reading it...
     
  9. ehume

    ehume Golden Member

    Joined:
    Nov 6, 2009
    Messages:
    1,112
    Likes Received:
    4
    Thank you, William. Great piece.

    @ShintaiDK. Come back when you have time. It's a great read.
     
  10. Idontcare

    Idontcare Elite Member

    Joined:
    Oct 10, 1999
    Messages:
    21,130
    Likes Received:
    9
    Interesting that in the end when they really needed to boost single-threaded performance to get a single compute job done asap they still resorted to parallelism for managing the ECC aspects of the calculations.

    Obviously they would have been better off had they implemented an algorithm within the computations themselves, as hardware ECC does now, instead of relying on duplicity of computations to get the correct result in the end.

    Two things utterly amaze me about that era - the fact they successfully designed and built functioning atomic weapon without the aid of anything remotely as sophisticated as a modern desktop computer (or calculator for that matter), and the fact that they were able to design, build and carry-out moon landings with computer (and engineering) technology that was only slightly more advanced than what they created atomic bombs with.

    That took real brains and intelligence, to do what they did with computers that played such a minor role overall. I doubt we have the technical competence these days to accomplish the same if you took today's engineers and physicists and took away all their computer/software toys and tasked them with accomplishing the same.

    I know I couldn't. My education prepared me to rely on computers to do my thinking for me, push a button and wait for the answer type solutions. My education is basically the error-correction factor in the calculation, if the computer answer doesn't seem right then I'm expected to have some sense of this before the bridge is actually built and falls down. But give me a stack of paper and a pencil and ask me to design the bridge (or IC) by hand and I would be totally useless.

    Also the story reminds me very much of what was alive and well some 30 yrs later in what was dubbed "sneakernet" where a cluster of computers were a virtual cluster by virtue of some hapless individual who would physically walk between computers in the appropriate cycle to remove and replace floppy disks of data between each compute node.
     
  11. BrightCandle

    BrightCandle Diamond Member

    Joined:
    Mar 15, 2007
    Messages:
    4,763
    Likes Received:
    0
    Actually you could. Modern programs are far more complicated than anything these individuals with this tools could produce. It took them a long time to do quite simple things on those computers, and the process of finding and fixing defects took a lot longer, but its a similar process today. Try for example programming in Edsac assembler (you can get an emulator) and you'll find it doesn't take a lot of additional knowledge to get going. But you'll find its quite slow and error prone.

    The old hardware designs before transistors were always less than 30,000 switches or so, because beyond that the failure rate was just too high. This sort of hardware is surprisingly easy to design on paper, I did infact design a 8 bit processor in an exam using only pen and paper. Its the sort of thing a computer science graduate is capable of doing in a few hours.

    Most modern programmers don't have to deal with these low level problems anymore, which improves our productivity for the actual problem at hand. But that doesn't mean they can't.
     
  12. Idontcare

    Idontcare Elite Member

    Joined:
    Oct 10, 1999
    Messages:
    21,130
    Likes Received:
    9
    That is a good point, regarding the fact that today's engineering challenges are vastly more complex than those of 60yrs ago so of course we need computers to do today's challenges.

    Yeah I suppose the computing challenge of designing a 30k xtor IC is probably within the grasp of a modern design engineer, true that.

    But building an atomic bomb? Look at how much technology Iran has access to, despite the embargo, and yet they are still going to spend more money and more years attempting to build their first atomic bomb in comparison to the years it took the Allied forces in the 40's (or Russia for that matter).

    I like to think the difference comes down to the expertise of the engineers and mathematicians, and not the prevalence of technology itself.

    The moon landings are another example, given modern compute technology one would think getting to the moon is easily within the technological grasp of any country today that currently has the inflation-adjusted GDP of the USA in 1960. And yet it is a feat that has yet to be duplicated, even by ourselves with our access to modern engineering tools and compute.

    The difference is the people IMO. There was just a different class of people working on those projects then. As the cliche goes, they were "made of the right stuff".
     
  13. BrightCandle

    BrightCandle Diamond Member

    Joined:
    Mar 15, 2007
    Messages:
    4,763
    Likes Received:
    0
    Science is a deeply collaborative thing. It took the Allies 10's of years to build such bombs using all the science and research of many countries. The problem for Iran is that the knowledge of how to make efficient gas centrifuges is not public knowledge. There is a lot of insight that goes into engineering them and it was done with a large number of scientists and engineers doing research in that area. Iran doesn't have a lot of Uranium within its borders either, which means its going to take a lot of time to find and mine it, since no one will sell it to them. Iran will produce weapon grade uranium and likely a dual stage plutonium bomb soon if they haven't done so already, but getting the fuel made is what takes time. These were talented engineers that made the first bomb but such talent still exists today and in greater quantity.

    No body wants to get to the moon. But think about the recent entirely automated space station resupplies from a commercial company where the software is written by John Carmack, a man more famous for Quake. He launched into space with less budget than the USA, much more efficient rockets and entirely computer controlled container ships. Quite simply his solution is vastly superior to what Nasa has produced for a fraction of the cost, without the risk of human lives being lost.

    It doesn't take much computer resources to get into space, it simply requires a decent reliability record and durability. NASA has been outdone by modern advances on many fronts now, so much so you can launch a satellite for considerably less money than ever before. But put simply the cost of going to the moon isn't worth it.

    I just can't accept that modern scientists and engineers and anything but better than their predecessors for a simple reason - we know everything they knew and more. Science continues to improve and get better, as does engineering. This isn't getting worse and people aren't getting dumber, on the contrary we have more highly educated and skilled individuals than in any other time in history, and they are making increasing more impressive devices and breakthroughs, breakthroughs our predecessors couldn't have even conceived.
     
    #12 BrightCandle, Dec 23, 2012
    Last edited: Dec 23, 2012
  14. AtenRa

    AtenRa Lifer

    Joined:
    Feb 2, 2009
    Messages:
    12,395
    Likes Received:
    1,200
    Very interesting, thx for the read :thumbsup:

    We can still do everything we make today without computers, it will only take more time and more money. I can design the Bridge without the Computer, it will take me much more time and thus more money but eventually i can design it. People making ships 2500 year ago, and my dad would design a bulk carrier without computers 30-40 years ago. It would take way less time for a naval architect to design the same ship today with the aid of computers but he could also do it without.

    As for the moon landing,
    There are two factors in my opinion preventing that, there is no reason/goal to spend the money going to the moon today and the second, we still using the same technology as we did 50 years ago, liquid fuel and rockets.
    50 years have past and we as Humanity haven't invested in new propulsion technologies like Nuclear or Plasma etc.
     
  15. ehume

    ehume Golden Member

    Joined:
    Nov 6, 2009
    Messages:
    1,112
    Likes Received:
    4
    I recall playing in my university's computer center some evenings in the spring of 1968. To do our computing jobs we replicated what the guys did on the day shift: a stack of Fortran punch cards was placed into the punch card reader. We would dismount a 2-foot tape reel, walk it over to the compiler -- a separate machine -- where we mounted the reel on the left spindle and the machine compiled the Fortran into Autocoder. We would then take off the right reel and walk it over to the main computer (I can't remember what it was; it was now the second string machine: the IBM 360 was brand new and we didn't touch it) where it would do our calculations. We would then dismount the output reel and carry it over to the line printer, mount it and watch to see if our programs ran . . . or not.

    We thought it was amazing. I used formulas from The Universe and Dr. Einstein and some basic algebra to calculate various examples of relativistic time dilation, the effect of acceleration on arrival times, etc. My friends used the line printer to print out pages of EAT EAT EAT, which they thought was hilarious (well, we were late teenagers).

    Fun.
     
  16. HeXen

    HeXen Diamond Member

    Joined:
    Dec 13, 2009
    Messages:
    7,663
    Likes Received:
    15
    yeah, there are all kinds of new propulsion systems. like the one that uses cylinders made for land and sea use. or electromagnetic propulsions, or magnetized-beam plasma propulsion which is still being worked on i believe... Plenty of investments are made, likely the main problem is the money to use them to get to the moon or they are not ready yet or capable to do that specific type of travel. However point is it appears there is plenty of investments being made all over the world, but many governments can't afford to invest too much these days leaving it to others to do so.
     
  17. WhoBeDaPlaya

    WhoBeDaPlaya Diamond Member

    Joined:
    Sep 15, 2000
    Messages:
    7,209
    Likes Received:
    319
    :whiste:

    [​IMG]
     
  18. zephyrprime

    zephyrprime Diamond Member

    Joined:
    Feb 18, 2001
    Messages:
    7,496
    Likes Received:
    1
    The funny thing about the situation with Iran and other modern states and nuclear weapons is that the primary problem they have is completely different than the primary problem the guys in the Manhattan project had. In the book, Feynman was talking about having to worry about detonating the atmosphere of the earth ala nuclear fusion via the nuclear fission of the atom bomb and other stuff like that. But that isn't even an issue that Iran has to deal with. They don't have to do the calculations at all for stuff like that because they already know the answer. The answer is that the atmosphere doesn't detonate.

    You worry that you rely on computers to do the thinking for you. Maybe so. But Iran also has the same advantage you have in that they don't need to do a lot of thinking. The general knowledge of how to build a nuke is actually pretty widely known. You just have to get the specifics on the chain reaction and etc to figure the whole puzzle out.

    The only reason that Iran and other countries has such big challenge developing nukes isn't because it's hard for them to figure out the theoretical physics. It's because their industry is so crappy that they can't acquire the raw materials (enriched uranium). On the other hand, back during the manhattan project, the focus was always on the physics because that was what was new back then and that was the real challenge. Yes, it took more money to fund the industrial side than to fund the theoretical side but handling the industrial side of production for a nuke back then wasn't even a big deal because the US's industrial plant was more than adequate for the task. Feynman talks about going to oversee the industrial production of the atom bomb in the book but it wasn't a big deal. The big deal was the theoretical work back at white sands missile base.
     
  19. AluminumStudios

    AluminumStudios Senior member

    Joined:
    Sep 7, 2001
    Messages:
    628
    Likes Received:
    0
    Very interesting, thanks for sharing.

    Feynman was really brilliant and is often listed as a major figure that physicists respect and aspire to the level of. I want to read that book now ...

    It's also interesting because it kind of demonstrates that weather you are using modern computers, old punchcard computers, or calculators, the process by which problems are broken down and solved is a key in getting it done as quickly as possible.
     
  20. cytg111

    cytg111 Diamond Member

    Joined:
    Mar 17, 2008
    Messages:
    4,043
    Likes Received:
    544
    Theres a reason i excelled in math and not language, this is too much information/noise, brain bails out.

    edit. but reading the comments i gather that there was no big paradigm shift in "the wall of text" with regards to concurrent computation. Right?
     
    #19 cytg111, Dec 24, 2012
    Last edited: Dec 24, 2012
  21. Idontcare

    Idontcare Elite Member

    Joined:
    Oct 10, 1999
    Messages:
    21,130
    Likes Received:
    9
    No shift, that was the point. The reality of today is a reality that transcends time for fundamental reasons and as such it was the reality "back in the day" as well - humans are still humans and we go about solving our problems today the same as we attempted to do 60 yrs ago.
     
  22. cytg111

    cytg111 Diamond Member

    Joined:
    Mar 17, 2008
    Messages:
    4,043
    Likes Received:
    544
    Thx, and

    Damn.
     
  23. William Gaatjes

    Joined:
    May 11, 2008
    Messages:
    15,609
    Likes Received:
    236
    Indeed. When we at work get interns, and they get stuck for whatever reason, the first hint i give them is to break down the problem into smaller problems. These smaller problems are then easier to fix. The challenge is of course to set the rules for the segmenting of the problem into smaller problems. Therein is hidden ones imagination and intelligence. Also, when one does not have a clue, we always tell them to start somewhere and just analyze to gather data as a means to really understand the problem. Switching back and forth from a broad view about a situation to a more detailed specific view. To find the cause and effect relationship.

    At the moment, we have extremely smart interns. I can come along because of my experience but they are half my age and are faster problem solvers then i ever could do at that age.
    They enjoy the technique and the problem solving. They do not nag or blame others. They just see a problem and want to solve it.:thumbsup::thumbsup::thumbsup:
     
    #22 William Gaatjes, Dec 26, 2012
    Last edited: Dec 26, 2012
  24. Railgun

    Railgun Golden Member

    Joined:
    Mar 27, 2010
    Messages:
    1,279
    Likes Received:
    1
    He can't write worth a damn. Incoherent mess of words.
     
  25. William Gaatjes

    Joined:
    May 11, 2008
    Messages:
    15,609
    Likes Received:
    236
    I do not agree. Just dissect the text into sections and it is an easy read.
    You just need to keep track of the subject he (Feynman) is referring to.
    Of course he was not a writer such as NuclearNed. Feynman wrote it down as if he is speaking towards the reader directly. At least that is how i experienced the book.
     
  26. cytg111

    cytg111 Diamond Member

    Joined:
    Mar 17, 2008
    Messages:
    4,043
    Likes Received:
    544
    - Thats a rare quality, something i suspect that many looses with age. That "blaming finger" is a curse in our industry.