Re: Elisp native profiler

[Paul Eggert]

On 10/01/2012 10:52 AM, Eli Zaretskii wrote:
> The timer thread suspends the main thread (thus simulating a signal),
> so the main thread cannot run at that time.

Ah, OK, thanks for explaining.  Sounds good,
especially now that the CPU usage while profiling
is essentially zero.

In trunk bzr 110328 I fixed a bug that caused a crash in
some of my profiler tests -- it was a race condition
involving malloc.  I am still looking into allowing
finer-grained profiling when supported.  Simplest may
be to change the nominal resolution of profiler-sampling-interval
from milliseconds to nanoseconds (and the default value from
1 to 1000000).  Of course no machine really does nanoseconds
(yet :-) but the low-level interface is there so why not?