"Full-Stack Correctness in Wasm – Eliminating Bugs Inside and Outside the Sandbox"
https://www.youtube.com/live/tiksgrSC3Ig?si=VqEsDdBEuEeZiHWL&t=2495
Great talk by @cfallin
=> More informations about this toot | View the thread
Til that the equivalent of the GIL in the default C Ruby VM is called GVL, and the Ruby community are having isomorphic discussions around it: https://byroot.github.io/ruby/performance/2025/01/29/so-you-want-to-remove-the-gvl.html
=> More informations about this toot | View the thread
'Meta-tracing interpreters in WebAssembly' https://m.youtube.com/watch?v=tiksgrSC3Ig&t=1h46m54s
Looks very interesting. Unfortunately the technical middle of the talk is missing from the stream 😕
=> More informations about this toot | View the thread
it's simply the fact that gdb turns address space randomization off by default. if I run it with setarch -R, the reproducer also crashes reliably.
=> More informations about this toot | View the thread
Now i turned it into an anti-Heisenbug! I minimized the crash into something that happens only when I run it in gdb, and not when run standalone. Better than the reverse, I suppose?!
=> More informations about this toot | View the thread
Now I'm learning about the gnu hash dynamic section in elf shared libraries (it's in the parser of that where libunwind segfaults. Looking at the code I have no idea how it could work)
=> More informations about this toot | View the thread
I'm currently debugging a random segfault in libunwind. Where have I gone wrong in my life?
=> More informations about this toot | View the thread
The existence of 'Die drei ???' (German version of the 'Three Investigators') and 'Die drei !!!' (spinoff with girls) clearly implies the existence of 'Die drei ‽‽‽'
=> More informations about this toot | View the thread
Oh nice, there's a plugin to jupyter lab to get language server protocol support in notebooks: https://github.com/jupyter-lsp/jupyterlab-lsp?tab=readme-ov-file
=> More informations about this toot | View the thread
Just finished "The Naturalist Society" by Carrie Vaughn. It's so wonderful! I loved it a lot 😊 . Chill, magic with a cool magic system, amazing characters. Great start into my 2025 book year.
=> More informations about this toot | View the thread
Til about MemorySSA in LLVM:
https://llvm.org/docs/MemorySSA.html
The first half of this talk is a nice intro:
https://youtu.be/1e5y6WDbXCQ?si=JJLVG0tR2u-OrG8F
My mental model so far is 'passing the state of the heap around'.
=> More informations about this toot | View the thread
@shriramk Shriram! Look what the vandals have done to our beautiful boy! Culture is dead.
=> View attached media | View attached media
=> More informations about this toot | View the thread
Bit hard to see, but I managed to get the moon through the window of the Duisburg main station 😊🚉🌕
=> More informations about this toot | View the thread
Another month, another compiler book with a cute reinterpretation of 'dealing with dragons' on the cover.
=> More informations about this toot | View the thread
"Musings on Tracing in PyPy" https://pypy.org/posts/2025/01/musings-tracing.html
blog post version of a Twitter thread/discussion last summer
=> More informations about this toot | View the thread
'Unlocking High Performance in Mojo through User-Defined Dialects'
Describes how Mojo uses Memory SSA to express rewrites such as removing l.append(...); l.pop()
.
The talk stops before it gets to the details of the interesting part, unfortunately.
https://youtube.com/watch?v=yuSBEXkjfEA&si=qBdxhwF5DJaBxDrL
=> More informations about this toot | View the thread
Most German of new year's wishes:
'happy 2025
Party - yes!
Cleaning too!!!'
=> More informations about this toot | View the thread
"Unwinding support for the JIT compiler", extremely carefully researched CPython issue by Pablo Galindo Salgado and Brandt Bucher about how to allow various stack unwinding tools (GDB, libunwind, libdw, the native unwinder) to unwind through JIT generated functions: https://github.com/python/cpython/issues/126910
=> More informations about this toot | View the thread
Got side-tracked a little bit yesterday and started making the machine code backends of PyPy's JIT a bit faster a generating code. I don't know that code super well (I suck at machine code so I try to stay away from them). It's 40% faster so far, the technique is always the usual thing of doing fewer dictionary lookups. And every time I get rid of a dict somewhere I find another one in another file :-(((
=> More informations about this toot | View the thread
Jits are hard
=> More informations about this toot | View the thread
=> This profile with reblog | Go to cfbolz@mastodon.social account This content has been proxied by September (3851b).Proxy Information
text/gemini