Twitter | Pretraživanje | |
Pekka Enberg
Software Engineer and PhD Student . Interested in low-latency networked systems, operating systems, and cloud and edge computing.
4.731
Tweetovi
662
Pratim
1.288
Osobe koje vas prate
Tweetovi
Pekka Enberg 2. velj
Interesting article on Google’s efforts on quantum software! Many development environments today are libraries for designing quantum circuits, which feels similar to developing applications in Verilog on classical computers. Curious on developments in higher level abstractions!
Reply Retweet Označi sa "sviđa mi se"
Pekka Enberg proslijedio/la je tweet
David Greco 31. sij
Odgovor korisniku/ci @penberg @rnella01 i 2 ostali
Good post! I would give a look at the chapel programming language, a modern multi-paradigm parallel programming language
Reply Retweet Označi sa "sviđa mi se"
Pekka Enberg 31. sij
New blog post 👇 On Parallel Programming () (Thanks to and for the feedback while I was writing this.)
Reply Retweet Označi sa "sviđa mi se"
Pekka Enberg proslijedio/la je tweet
Felix Geisendörfer 26. sij
1) Are you using via for mac? Have you ever noticed `EXPLAIN ANALYZE` slowing down your queries by like 60x? Let's dig into some and internals to figure out what's going on!
Reply Retweet Označi sa "sviđa mi se"
Pekka Enberg 25. sij
Tutkimuksen merkitys isossa kuvassa ei määräydy sen mukaan mitä juuri nyt pidetään tärkeänä. Eikä yksityistä rahaa välttämättä löydy, vaikka tiedolle olisi kysyntää. Tämä siis tilanne esimerkiksi lääketieteessä antibioottien kohdalla ().
Reply Retweet Označi sa "sviđa mi se"
Pekka Enberg 21. sij
Interesting paper from 2018 on consumer device workloads! Is there something recently published for data center workloads that would include AI/ML?
Reply Retweet Označi sa "sviđa mi se"
Pekka Enberg 17. sij
Odgovor korisniku/ci @trav_downs @lemire @shipilev
What could the OS do less here? The Linux unmap() does the following: (1) find the VMA covering the unmapped region, (2) give back pages to the page allocator, (3) zap the page table mappings. (Btw, the pages are put to per-CPU page list for recycling in the page allocator).
Reply Retweet Označi sa "sviđa mi se"
Pekka Enberg 17. sij
Odgovor korisniku/ci @trav_downs @lemire @shipilev
Let me weaken my claim then: it is very difficult for the OS to optimize this. The issue is that munmap(), semantically, is telling the OS that application wants to throw away the _mapping_. The OS could speculate that new mmap() is coming, but that could hurt other workloads.
Reply Retweet Označi sa "sviđa mi se"
Pekka Enberg proslijedio/la je tweet
Daniel Lemire 14. sij
How fast can you allocate a large block of memory in C++?
Reply Retweet Označi sa "sviđa mi se"
Pekka Enberg 16. sij
Odgovor korisniku/ci @trav_downs @lemire @shipilev
Perhaps this was already discussed, but it seems like glibc calloc() does a mmap()/munmap() for every (large) allocation/deallocation. There's no way for the OS to optimize this if user space explicitly throws away the memory mappings.
Reply Retweet Označi sa "sviđa mi se"
Pekka Enberg proslijedio/la je tweet
Sarah Allen 16. sij
been learning about writing a library in that can be called from C or , wrote a blog post about it
Reply Retweet Označi sa "sviđa mi se"
Pekka Enberg 15. sij
Odgovor korisniku/ci @varjag @trav_downs i 2 ostali
Can you elaborate on what you mean with extensive documentation in the context of Linux? This was one of the areas I found most lacking compared to BSDs, at least in the source tree. Curious to know if that has changed.
Reply Retweet Označi sa "sviđa mi se"
Pekka Enberg 15. sij
Odgovor korisniku/ci @varjag @trav_downs i 2 ostali
Linux also has a "kunit" framework for kernel unit tests now:
Reply Retweet Označi sa "sviđa mi se"
Pekka Enberg 12. sij
Good read, even if I don’t fully agree with the conclusion. I think aiming for ”clean code” is important, but you must resist the urge to over-generalize and abstract too much.
Reply Retweet Označi sa "sviđa mi se"
Pekka Enberg 9. sij
Odgovor korisniku/ci @argenetix
I don’t really track traffic to my home page and only switched to CloudFlare (that has some stats) recently. I hope they are saying good things about the paper 😄
Reply Retweet Označi sa "sviđa mi se"
Pekka Enberg proslijedio/la je tweet
The Independent 8. sij
Quantum computing adopted by Delta and Daimler in industry first
Reply Retweet Označi sa "sviđa mi se"
Pekka Enberg 7. sij
Nice overview of Snabb packet network toolkit by (from 2017)!
Reply Retweet Označi sa "sviđa mi se"
Pekka Enberg proslijedio/la je tweet
electrified filth 6. sij
Just cut rio 0.5! It's a pure-rust misuse-resistant io_uring library. io_uring will change everything about IO. The goldrush has started but folks don't realize it yet. lifetimes+RAII tricks ensure safety. go measure the o_direct example vs what you currently use :]
Reply Retweet Označi sa "sviđa mi se"
Pekka Enberg 6. sij
Odgovor korisniku/ci @lukego @jonmasters
Fine-grained control over your machine with a monolithic, shared-memory kernel that implements POSIX semantics is very hard if not impossible. I think what Linus is saying makes lots of sense for Linux, but I am unsure if that’s how we should build our systems.
Reply Retweet Označi sa "sviđa mi se"
Pekka Enberg proslijedio/la je tweet
Jonathan Aldrich 5. sij
Join me! ACM members, sign a petition supporting open access, asking to withdraw signature from anti-OA letter and make OA available at cost. OA is good for science, and good for ACM!
Reply Retweet Označi sa "sviđa mi se"