tiscript vs spidermonkey (jsdb/gluescript)

[FoxZ]

wow... tiscript seen 3 times more SLOW than spidermonkey engine (not
JIT enabled).

kopit ?

[Andrew]

Yes, there is a room of optimizations in TIScript. Some functions need
to be inlined, bytecode optimized for at least for jump-to-jumps and
peephole.

But speed of embeddable scripting engines is not so critical. They are
just glue at the end.
Native methods (DOM access) are things that making the difference.

Here is something close to real:
http://terrainformatica.com/tests/std-table-fixed-10000-rows.htm
This table contains 10`000 rows. First button named Update runs loop
through all rows and does update of cell's text in first column.

Timings on my machine:

Sciter (TIScript + H-SMILE core) - 781ms
Firefox (3.5, Gecko + Spidermonkey?) - from 1450ms to 2200ms
Google Chrome(2.0, WebKit + V8) - 1210ms
Opera (9.64) - 2141ms

So as you see on typical UI tasks Sciter with its tiscript VM works
around two times better than conventional browsers. Even Chrome with
its optimized VM.

And yet memory consumption...
With that sample table loaded FF takes 76mb of RAM while Sciter shows
33mb.
Opera: 44mb and Chrome:64mb

Cheers.

[FoxZ]

wow... very good job !!!
sciter seen MORE simply than xulrunner !

do u think u can speedup tiscript engine ?

webkit javascriptcore is best fast javascript engine for now
(x2 times more speed than tracemonkey, x6 times more speed tiscript).

I dont take care DOM access timing because I dont need huge GUI.

I follow ur project with care.

syl
foxz at free point fr

[Andrew]

On Sep 23, 3:28 am, FoxZ <f...@free.fr> wrote:
> wow... very good job !!!
> sciter seen MORE simply than xulrunner !

Sciter was designed to be an embeddable engine, that's why.
Trident, Gecko and WebKit was designed and good for general and safe
web browsing - to be part of browser (a.k.a. User Agent).
But desktop UI tasks require different architecture/feature set, IMO.

>
> do u think u can speedup tiscript engine ?
>
> webkit javascriptcore is best fast javascript engine for now
> (x2 times more speed than tracemonkey, x6 times more speed tiscript).

To be able to say anything I'd like to know first what tasks you want
to do in TIScript.

E.g. if you want to do, say, ray tracing then it is better to make
native primitives for that.
That will be Optimization the Ultimate.

Even in its current state TIScript appears as work faster on typical
tasks than e.g. Python or Ruby (languages that have similar feature
set with the tiscript).

I am also who think that premature optimization is not that makes
sense to start with.