@RichFelker presumably someone had a dream for making syscalls less horrible and the dream slowly shattered
-
-
Replying to @FioraAeterna
@FioraAeterna My view is just that software arch., not hw/insn timings, is the overwhelmingly dominating factor limiting syscall perf.1 reply 0 retweets 2 likes -
Replying to @RichFelker
@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.2 replies 0 retweets 2 likes -
Replying to @RichFelker
@FioraAeterna If everything were call-clobbered that would help. But sheer over-long code paths are also a big part of the problem...1 reply 0 retweets 0 likes -
Replying to @RichFelker
@RichFelker@FioraAeterna Caller or callee, doesn't matter; used registers have to be saved.1 reply 0 retweets 0 likes -
Replying to @SamuelAFalvoII
@SamuelAFalvoII@FioraAeterna Caller doesn't have to save any ABI-call-clobbered registers it's not currently using. That's almost all.1 reply 0 retweets 0 likes -
Replying to @RichFelker
@RichFelker@FioraAeterna But in POSIX, anytime you return from a system call, you might have to dispatch to a signal handler, no?2 replies 0 retweets 0 likes -
Replying to @SamuelAFalvoII
@SamuelAFalvoII@FioraAeterna If the syscall was already entitled to clobber those regs, signal handler is entitled to just save junk.2 replies 0 retweets 0 likes -
Replying to @RichFelker
@SamuelAFalvoII@FioraAeterna Signal handlers only need to faithfully save registers when interrupting asynchronously. This is impl detail.1 reply 0 retweets 0 likes -
Replying to @RichFelker
@RichFelker@FioraAeterna I believe POSIX signal handlers are specified to be asynchronous.1 reply 0 retweets 0 likes
@SamuelAFalvoII @FioraAeterna Yes and no; too subtle for 140 chars. But _when_ they trigger at a syscall, there's nothing to save.
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.