Twitter | Pretraživanje | |
Thomas Lively
47
Tweetovi
62
Pratim
81
Osobe koje vas prate
Tweetovi
Thomas Lively 30. sij
It’s exciting to see SIMD get closer and closer to standardization and shipping. It’s certainly been a long road to get here. Huge thanks to for all the work she’s been doing to drive the proposal forward.
Reply Retweet Označi sa "sviđa mi se"
Thomas Lively 18. sij
Odgovor korisniku/ci @tlively52
At least the upside of prioritizing PRs by number is that I'll have a much shorter wait the next time around. I wonder if anyone has ever squatted a PR number to get ahead in the queue, though 🤔
Reply Retweet Označi sa "sviđa mi se"
Thomas Lively 18. sij
I just waited 10 days for my two-line bugfix to make its way to the head of the Rust merge queue and I was so excited to see it finally get there. And then the tests failed because of a network error. Guess I'll just wait some more 🙃🙃🙃
Reply Retweet Označi sa "sviđa mi se"
Thomas Lively 28. pro
Odgovor korisniku/ci @brionv
How are you using Binaryen? C API? Binaryen.js? CMake subdirectory?
Reply Retweet Označi sa "sviđa mi se"
Thomas Lively 24. stu
Odgovor korisniku/ci @RReverser @barrelshifter
True, locals would be the extra fun part to handle. But on the other hand, it is possible to have runs of code with no locals at all that can be trivially matched. That’s the stack machine advantage.
Reply Retweet Označi sa "sviđa mi se"
Thomas Lively 24. stu
Odgovor korisniku/ci @RReverser @barrelshifter
This would be really cool to implement in Binaryen, though! WebAssembly being a stack machine with structured control flow should make outlining easier and more powerful. Multivalue will help, too!
Reply Retweet Označi sa "sviđa mi se"
Thomas Lively 17. stu
Odgovor korisniku/ci @MaxGraey
Yes, it would be possible for an engine to do that optimization based on runtime profiling, but it’s currently difficult for the toolchain to provide optimization hints. I know there’s interest in using custom sections for optimation hints, though.
Reply Retweet Označi sa "sviđa mi se"
Thomas Lively 17. stu
Odgovor korisniku/ci @MaxGraey
It’s not clear what the benefit would be, given that there’s currently no way of encoding that information explicitly in a wasm module and there are no guarantees about what engines will do with the branch arms.
Reply Retweet Označi sa "sviđa mi se"
Thomas Lively 17. stu
Odgovor korisniku/ci @B_Schelling @llvmorg
No, thank you for finding our bugs for us 😅
Reply Retweet Označi sa "sviđa mi se"
Thomas Lively 11. stu
So exciting! I’m looking forward to it 😄
Reply Retweet Označi sa "sviđa mi se"
Thomas Lively 2. stu
Odgovor korisniku/ci @zeuxcg
Very cool! Are there any instructions you feel are missing from the proposal that could help you go faster?
Reply Retweet Označi sa "sviđa mi se"
Thomas Lively 28. lis
Odgovor korisniku/ci @xexd
At least it’s not a thesis resented to the department of computer science!
Reply Retweet Označi sa "sviđa mi se"
Thomas Lively 19. lis
It’s happening! 😍🚀
Reply Retweet Označi sa "sviđa mi se"
Thomas Lively proslijedio/la je tweet
Alon Zakai 5. lis
Emscripten reached 20,000 commits earlier today. Thanks to everyone that contributed, code and otherwise!
Reply Retweet Označi sa "sviđa mi se"
Thomas Lively 15. ruj
Odgovor korisniku/ci @brionv
We also just added narrowing operations! They don’t have a built in store, but if you need to narrow and store multiple vectors they would still be useful.
Reply Retweet Označi sa "sviđa mi se"
Thomas Lively 11. ruj
Feature detection libraries like this will dovetail nicely with the upcoming conditional compilation proposal to allow users to trade off number of separate builds and binary size while still targeting multiple feature sets.
Reply Retweet Označi sa "sviđa mi se"
Thomas Lively 13. kol
Odgovor korisniku/ci @wasmsf @teropa @WasmWeekly
That would be super cool! You could even try it out now using emscripten’s SIMD support I know Rust and AssemblyScript support SIMD as well. Would love to see what you create!
Reply Retweet Označi sa "sviđa mi se"
Thomas Lively 24. srp
Next stop, multivalue! 🚂
Reply Retweet Označi sa "sviđa mi se"
Thomas Lively 7. srp
Odgovor korisniku/ci @brionv
This is a problem we would like to solve... somehow. We just don’t know how general a dynamic feature detection feature should be. I guess we should probably come up with some options and do science on them.
Reply Retweet Označi sa "sviđa mi se"
Thomas Lively 4. srp
Curious about where WebAssembly is heading? SIMD is a performance-critical part of that picture. In this talk I explained what SIMD is and how you can try it out today. Let us know how it works for you!
Reply Retweet Označi sa "sviđa mi se"