Async-signal-safe access to __thread variables from dlopen()ed libraries?

[Andrew Hunter]

On Tue, Sep 24, 2013 at 3:24 PM, Rich Felker <dalias@aerifal.cx> wrote:
> I would be surprised if it ever took more than 5 microseconds. For
> what it's worth, millisecond latency is serious and could be
> considered a problem for many realtime uses. Microsecond, not so much.
>

To be more precise, I would also be surprised if it passed 5
microseconds or so (barring long-tail issues like kernel lock
contention, etc, that really aren't much worse for this than the
original version)  I was just being very, very conservative in my
initial statement.