Dan Luu

@danluu

. Prev: search&cloud , TPU , x86&ARM , POWER , flash .

NYC
danluu.com
Joined December 2008
385 Photos and videos Photos and videos

Tweets

You blocked @danluu

Are you sure you want to view these Tweets? Viewing Tweets won't unblock @danluu

  1. Pinned Tweet
    8 Feb 2017

    Most of the web really sucks if you have a slow connection

    37 replies 370 retweets 552 likes
    Undo
  2. Aug 23

    One of the many delightful absurdities of this is that you can't comply with the terms of Intel's latest security patch if you continue to have public performance regression tests or any kind of nightly/CI/regular benchmark.

    6 replies 6 retweets 33 likes
    Show this thread
    Show this thread
    Undo
  3. Aug 23

    Intel taking a page from the Oracle playbook and not allowing benchmarks on (some) of their systems:

    1 reply 12 retweets 24 likes
    Show this thread
    Show this thread
    Undo
  4. Retweeted
    Aug 17

    This is cool: most bignum libraries fail hilariously at primality testing when adversaries select the candidates (ie: when an untrusted server provides parameters to client, which is then meant to check them).

    Prime and Prejudice: Primality Testing Under Adversarial Conditions: MR Albrecht, J Massimo, KG Paterson, J Somorovsky
    5 replies 50 retweets 121 likes
    Show this thread
    Show this thread
    Undo
  5. Aug 14

    People are all over how Threadripper 2990WX sucks on Windows, but for most of us, the more relevant part is how there can easily be a 20% to 25% perf difference due to minor variations in Linux setup, *with the order of results being different for different benchmarks*.

    5 replies 7 retweets 21 likes
    Show this thread
    Show this thread
    Undo
  6. Aug 14

    It's interesting how the same hardware gets such different performance on different operating systems when running the same benchmark:

    10 replies 23 retweets 54 likes
    Show this thread
    Show this thread
    Undo
  7. Retweeted
    6 Sep 2017

    This shit is why most GC people I know hate weak references with all their guts.

    8 replies 26 retweets 80 likes
    Undo
  8. Retweeted
    26 Oct 2017

    awesome, it's not every day you run across a useful, easy optimization missing from all of LLVM, GCC, and Intel CC

    8 replies 108 retweets 322 likes
    Show this thread
    Show this thread
    Undo
  9. Jul 21

    I'm not sure what my favorite part of this is. People using Google to check if Google is down? People wondering if every online service is made by Google? People asking Google for stock picking advice (not shown, lower down)?

    2 replies 3 retweets 11 likes
    Show this thread
    Show this thread
    Undo
  10. Jul 21

    What do people want to know about big tech companies (circa 2018)?

    5 replies 1 retweet 5 likes
    Show this thread
    Show this thread
    Undo
  11. Jul 21

    What do people want to know about big tech companies (circa 2016)?

    2 replies 5 retweets 15 likes
    Show this thread
    Show this thread
    Undo
  12. Jul 18

    BTW, I don't really mean legally. I don't understand why Google's reputation is fine-to-good when MS's reputation in the 90s/00s was garbage. They both seem to throw their weight around in the same way. If anything, Google seems to do it more?

    7 replies 20 retweets 48 likes
    Show this thread
    Show this thread
    Undo
  13. Jul 18

    As a mere lay person, I don't understand why the openness is the problem? Didn't MS get in trouble without being open? It seems to me that things like preventing DDG/Bing from being used are the problem; one could have an open platform that doesn't do the shady stuff Google does

    7 replies 23 retweets 52 likes
    Show this thread
    Show this thread
    Undo
  14. Jul 18

    The EU fined Google ~$5B for anti-competitive Android/search behavior. A Google employee and lawyer (not Google counsel, not speaking on behalf of Google) says Google's real crime was being too open:

    2 replies 14 retweets 20 likes
    Show this thread
    Show this thread
    Undo
  15. Jul 17

    There's a model of computation based on frictionless billiard balls. Those don't exist, so you have to use solider crabs when implementing the actual computer. Solider crabs. 🦀🦀🦀

    5 replies 42 retweets 84 likes
    Undo
  16. Retweeted
    15 Aug 2017

    "Dwarf Fortress starting during apt-get upgrade"

    7 replies 254 retweets 356 likes
    Undo
  17. Retweeted
    19 Aug 2017

    PS4 game downloads are notoriously slow. I did a bit of digging to find out the (a?) root cause.

    11 replies 66 retweets 98 likes
    Undo
  18. Retweeted
    30 Oct 2017

    a strong contender for the stupidest undefined behavior in C (thanks for reminding about this one )

    6 replies 24 retweets 79 likes
    Show this thread
    Show this thread
    Undo
  19. Jul 8

    Don't be evil, 2009 edition: "Remove the search engine setting. Hard-code the search engine to Google"

    4 replies 35 retweets 137 likes
    Show this thread
    Show this thread
    Undo
  20. Jul 6

    I don't mean to pick on this too much -- open office advocates typically have zero evidence for their claims. It's nice to see a study on this even if it's unclear what the result actually means in practice.

    10 replies 6 retweets 106 likes
    Show this thread
    Show this thread
    Undo
  21. Jul 6

    Another strange thing is the magnitude of the result: 5.8h/day -> 1.7h/day. Was that really all the result of the changing office plan and not a result of any other factors? No way to tell since there was no control and other changes that might've co-occured aren't discussed.

    4 replies 4 retweets 58 likes
    Show this thread
    Show this thread
    Undo

@danluu hasn't Tweeted yet.

Loading seems to be taking a while.

Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.

    You may also like

    ·