mbrubeck/android-completion 212
Bash completion for "adb" from the Google Android SDK
Very simple server for the Gemini hypertext protocol
Rust wrapper type that implements hashing and comparison by pointer address
a C read-eval-print loop
Readable, memorable URLs to replace long bug numbers
A simple AI for 2048
Actor framework for Rust
Security advisory database for Rust crates published through crates.io
issue commentservo/rust-smallvec
RFC: rewrite based on min_const_generics for 2.0
I looked into some possible SmallVec optimisations yesterday. They weren't fruitful, but I thought I'd describe my findings here in case you find them useful for the 2.0 rewrite.
I was motivated by a few comments here, which described a large branch prediction penalty for spilled SmallVecs. I'd seen similar complaints before, so I wanted to investigate possible fixes.
Some data here suggests large performance improvements when using a (32-bit) cmov in place of an unpredictable branch, with minimal or zero performance loss when the branch is completely predictable.
It occurred to me that, with the SmallVec implementation you described above, the length is unchanged whether or not the array is spilled. SmallVec::deref is only branching on which value to assign to a single pointer, which is a good candidate for use of cmov.
I experimented with code which reduced all branching in SmallVec::deref to a trivial let p: *const T = if a { b } else { c }. Unfortunately, I noticed that LLVM completely refused to emit a cmov instruction when it could potentially cause an unnecessary memory access. It turns out this is a known LLVM bug.
Unless you're willing to use inline assembly behind a feature flag, this idea is probably at a dead end. You could experiment with bit-masking as a replacement for cmov, but I think that would cost at least three CPU cycles (xor, and, xor), while a mispredicted branch might cost ten or less.
comment created time in 3 hours
startedmbrubeck/agate
started time in 3 hours
startedmbrubeck/agate
started time in 4 hours
startedmbrubeck/robinson
started time in 13 hours
created repositoryaconbere/smart-relay
A kicad project that builds a programmable relay with a current and voltage sensing
created time in 2 days
created repositoryaconbere/avr-rust-attiny85-test
Smallest possible build for an attiny85 test
created time in 2 days
startedlaunchdarkly/haskell-server-sdk
started time in 2 days
startedmbrubeck/android-completion
started time in 2 days
startedrougier/mu4e-dashboard
started time in 2 days
startedmbrubeck/agate
started time in 3 days
startedesgibter/fandomstats
started time in 3 days
startedmbrubeck/robinson
started time in 3 days
startedchuntaro/emacs-keypression
started time in 3 days
fork autonome/Cumulus
☁️ A SoundCloud player that lives in your menubar.
http://gillesdemey.github.io/Cumulus
fork in 4 days
startedmbrubeck/android-completion
started time in 4 days
startedmrdoob/stats.js
started time in 5 days
startedfactbook/factbook.json
started time in 5 days
push eventunicode-rs/unicode-segmentation
commit sha a844e6ff87b2d88c38ca01d7df1447cdbe9d19ef
Update documentation
push time in 5 days
fork autonome/Web-Extension-Starter
Typescript, React, Redux, Styled-Component and Webpack based sample extension boilerplate. Runs on Chrome and Firefox. Sample chrome extension.
fork in 5 days
issue commentunicode-rs/unicode-segmentation
Ah, the publication commit wasn't pushed. I anyway did a new commit for 1.7.1 with the doc changes https://github.com/unicode-rs/unicode-segmentation/commit/3b75ee19b3c0ddacaeec03be688a7b8766833728
comment created time in 5 days
push eventunicode-rs/unicode-segmentation
commit sha 3b75ee19b3c0ddacaeec03be688a7b8766833728
Publish 1.7.1 (1.7.0 has the same non-documentation changes)
push time in 5 days
issue commentunicode-rs/unicode-segmentation
I don't believe that there has been a release of the Unicode 13 changes. I submitted the PR with those updates. If there is anything that I can do to help prep a release, please let me know.
comment created time in 5 days
issue commentunicode-rs/unicode-segmentation
The docs also need an update.
comment created time in 5 days