Tweetovi

Blokirali ste korisnika/cu @dEnergy_dTime

Jeste li sigurni da želite vidjeti te tweetove? Time nećete deblokirati korisnika/cu @dEnergy_dTime

  1. Every now and then, has these special moments, where the ecosystem gets something it should have had, or not have had, for years. This is one of them. Thank you, .

    Poništi
  2. proslijedio/la je Tweet
    prije 3 sata

    For those that missed the event, the recording of my talk yesterday about the progress on gr-satellites v3 is now available at

    Poništi
  3. proslijedio/la je Tweet
    prije 3 sata

    Great session yesterday at Free Radio Software devroom. Really inspiring talks, lots of people to chat with, and many others I didn't get a chance to meet in such a tightly packed event.

    Poništi
  4. prije 9 sati

    at : good talks so far, with focussing on License success (more in the context of innovative/new licenses), but drawing from many interesting case studies of established/failed(?) cases. Up next: Panel, sadly missing :(

    Poništi
  5. 2. velj

    Jean-Michel Friedt, representing , again is the perfect whirlwind closer for the devroom at : He's deterring meddling lab colleagues from taking his expensive instruments by replacing them with cheaper, better, self-designed SDR experiments

    Poništi
  6. 2. velj

    We're continue our scheduled transmissions for STOMPING the scheduling problem with a Python-based workload arbiter in a simple static DAG done-feedback meta scheduler

    Poništi
  7. 2. velj

    Good point in time to remind people at of this excellent conferencen

    Poništi
  8. 2. velj

    Really like aspiring to have RF-in-the-loop continuous integration (with a photo, showing 's RFCI infra!). I know we're not there, but definitely strives to achieve the same level of robustness

    Poništi
  9. proslijedio/la je Tweet
    2. velj

    And, as usual, () is a rockstar, this time bringing the state of to the devroom at . Just if you wonder what the osmo world thinks of this system: contributed 2G-fallback :) I _really_ like .

    Poništi
  10. 2. velj

    Just a quick shoutout to these making the devroom happen: · as whole · Esp, the fosdem A/V team · J.-M. Friedt surveying the stream · org'ing the room, with · supplying door ACL/cgroup mgmt · supplying speaker gifts

    Poništi
  11. 2. velj

    And, as usual, () is a rockstar, this time bringing the state of to the devroom at . Just if you wonder what the osmo world thinks of this system: contributed 2G-fallback :) I _really_ like .

    Poništi
  12. 2. velj

    Enter () optimizing signal processing for accelerator-aided autonomous vehicles; more love for 's gr-ieee-802-11

    Poništi
  13. 2. velj

    also, honestly, humans deal with numbers sequentially. And the first languages people actually learn that yield useful software quickly are of imperative nature. I'm certainly no exception to that rule. So, it's "natural" to want HLS-style things.

    Prikaži ovu nit
    Poništi
  14. 2. velj

    and then want that a hardware design based on that would even come close to what it's possible? The answer is simple, and very human-centric: Hardware is hard; the temporal decomposition that imperative programming languages make it way easier to reason on what you're doing;

    Prikaži ovu nit
    Poništi
  15. 2. velj

    "I tried to do a matrix inverse in High-Level Synthesis and the result was disappointig!" YES! Why would you take an inherently parallel, pipelined, distributed-data problem and translate it to an imperative language, losing all explicit data relationship knowledge on the way,

    Prikaži ovu nit
    Poništi
  16. 2. velj

    I hope the borderline irony of gifting with yet another FPGA/CPU SoC-based SDR is hopefully not lost on anybody ;) But seriously, the >Your algorithm design is dictated by the hardware it runs on wisdom should stick. The last audience question was a perfect example:

    Prikaži ovu nit
    Poništi
  17. 2. velj

    Jeffrey Vetter and Seeyong Lee (of ) talking about cellphone-to-world-leading-supercomputer scaling computing; tying SDR to Domain-specific computing. Analyzing relationships of blocks in the wild(e.g.'s gr-ieee-802-11), where time is spent, bridging scheduler gaps…

    Poništi
  18. 2. velj

    OH, btw: no SWIG, no Boost. Way easier on the dependency management.

    Prikaži ovu nit
    Poništi
  19. 2. velj

    The funky thing is that this _helps us kill PMT_. In the future (not right now, don't worry), we'll replace PMT with a standardized, higher-perf structured data xchg format. With an external library, we offer a low-complexity interface for legacy data sets, and ad-hoc translation

    Prikaži ovu nit
    Poništi
  20. 2. velj

    The power of : *whilst* sitting in a devroom at , the (very specific) dude next to me showed me that he managed to make a standalone PMT library with C++ and Python bindings, so that you can finally work with PMT objects without having .

    Prikaži ovu nit
    Poništi

Čini se da učitavanje traje već neko vrijeme.

Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.

    Možda bi vam se svidjelo i ovo:

    ·