On the design of the SYSENTER Intel syscall instruction, and how the elegant idea fell apart https://groups.google.com/forum/m/#!topic/comp.arch/CjDs4MJCBow%5B1-25%5D …pic.twitter.com/VIZlsh87cE
You can add location information to your Tweets, such as your city or precise location, from the web and via third-party applications. You always have the option to delete your Tweet location history. Learn more
@RichFelker because a lot of common syscalls, like say fetching time, require no security checks and can be done in a few cycles
@RichFelker presumably someone had a dream for making syscalls less horrible and the dream slowly shattered
@FioraAeterna My view is just that software arch., not hw/insn timings, is the overwhelmingly dominating factor limiting syscall perf.
@FioraAeterna One big problem is that the Linux syscall ABI makes all registers call-saved, so they have to save/restore any kernel may use.
@RichFelker reminds me of one of the many strategies Synthesis OS (http://valerieaurora.org/synthesis/SynthesisOS/ …) used to avoid syscall overhead
@RichFelker it JIT-compiled syscall functions to avoid unnecessary paths, including minimal register saving/restoring
@RichFelker (of course this was 1992, so the entire lockless JIT kernel was written in 68K assembly)
@RichFelker I highly recommend reading at least some of it, it has all kinds of fun nonsense, all in the name of 44100hz interrupts on a 68K
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.