I am glad that my report helped to improve native-comp branch.

Thanks,
Ivan

On Sun, Apr 26, 2020 at 3:00 PM Andrea Corallo <akrl@sdf.org> wrote:

> yyoncho <yyoncho@gmail.com> writes:
>
> > Hi,
> >
> > I did benchmarks of json parsing here it is what I get:
> >
> >   | test         | non-gc avg (s) | gc avg (s) | gcs avg | tot avg
> > (s) | tot avg err (s) |
> >   |
> >
> --------------+----------------+------------+---------+-------------+-----------------
> > |
> >   | json-parsing |           2.57 |       3.32 |     318 |
> >  5.89 |            0.10 |
> >   |
> >
> --------------+----------------+------------+---------+-------------+-----------------
> > |
> >   | total        |           2.57 |       3.32 |     318 |
> >  5.89 |            0.10 |
> >
> >
> > Native:
> >
> >   | test         | non-gc avg (s) | gc avg (s) | gcs avg | tot avg
> > (s) | tot avg err (s) |
> >   |
> >
> --------------+----------------+------------+---------+-------------+-----------------
> > |
> >   | json-parsing |           1.71 |       4.31 |     343 |
> >  6.02 |            0.01 |
> >   |
> >
> --------------+----------------+------------+---------+-------------+-----------------
> > |
> >   | total        |           1.71 |       4.31 |     343 |
> >  6.02 |            0.01 |
> >
> > It looks like this is bellow the average speedup you might wanna take
> > a look:
> >
> > Here it is the benchmark code:
> >
> > (require 'cl-lib)
> > (require 'json)
> >
> > (defvar json-parsing-input (with-temp-buffer
> >     (insert-file-contents-literally  "benchmarks/sample.json")
> >     (buffer-string)))
> >
> >
> > (defun elb-json-parsing-entry ()
> >   (cl-loop repeat 1000
> >   do (json-read-from-string json-parsing-input)))
> >
> > Thanks,
> > Ivan
>
> Hi Ivan,
>
> I've got time to re-look into this.
>
> The perf difference was due to GC time (as Eli pointed out all json
> computation is in the C world).
>
> I believe this delta was due to the greater number of objects allocated
> in the native comp variant with the consequent more time needed to
> traverse them during GC.
>
> Since these measures I've made several changes to reduce the number of
> unnecessary live objects.  Repeating the test on the current native-comp
> against master I see on my machine a performance delta of 2% in favor of
> master.  Considering I expect this effect to fade out on a real session
> with more allocated objects I consider this being acceptable.
>
> Andrea
>
> --
> akrl@sdf.org
>