Re: v2.5 Opens exw files wayyyy too slow
Patrick Barnes wrote:
> On Mon, 20 Dec 2004 16:26:49 -0800, Robert Craig wrote:
>> I'm not going to make things more complicated.
>>
>> 1. If you have a decent machine (less than 3 years old),
>> you can probably parse 100,000 lines in 3 seconds.
>> 99% of programs are much smaller than that.
>> Maybe your scheme would cut that to 1 second.
>> Judith's IDE, for example, then takes 3 more seconds
>> executing code to initialize itself (nothing to do with parsing).
>
> Well, many people don't *have* decent machines, and can't afford to
> upgrade them.
... or don't want to do so. Am I supposed to buy a new PC, just in order
to reduce the parsing time of the new Euphoria interpreter?
Some people keep writing, that a new PC only costs about 500 USD.
Then the upgrade from Eu 2.4 Complete Edition to the Eu 2.5 binder costs
for me 524 USD rather than 24 USD. Any further comments necessary?
> Otherwise, yes, it's a largely valid point.
>
>> 2. 2.5 beta parses significantly faster than 2.5 alpha in all cases.
>> It's dramatically faster on old machines with small memories
>> (64 MB or less) when very large programs are parsed.
I hope 2.5 beta will run dramatically faster also on other old machines.
My PC contains a Pentiom II prozessor, 400 MHz. Although it's got
*256 MB* RAM, promgrams that e.g. include Win32Lib are parsed unbearable
slow by Eu 2.5 alpha. I'll have to wait and see, what "dramatically
faster" actually means, too.
> That is very good to hear. Any ideas as to when it will be released?
> (Heh sorry, no pressure)
>
>> 3. Old, slow machines are disappearing every day, being replaced
>> by > 2GHz machines. Already, most people have little concern about
>> parse speed.
"most people"? How much percent? Where do that statistical data come from?
>> In a couple of years no one will care about this.
Then it will be the best, to use Eu 2.5 not now, but "in a couple of
years"?
>> Why build a major new mechanism that has little use now, and
>> will be completely useless in a couple of years?
>> It will be one more thing to confuse beginners.
>
> Hey, what about my 286 in the closet!
> Yes, it may become less useful over time, but remember that the way
> CPU technology is going, we're not going to get much faster
> clock-speeds. It's more likely going to be multi-cored, which I don't
> think would help parse time much. Speaking of major new mechanisms,
> how about multi-threaded execution?
>
>> 4. If you translate, your app will start with zero parse time.
>> If you bind, your app will also now start with zero parse time (in 2.5).
>
> That's against one of the major selling points of euphoria. Namely,
> that it's Edit, Run, Edit, Run, Edit, Run. If it has to be
> bound/compiled, it adds an extra step.
Yes.
> Thanks for your response.
>
> Just one thing though... what about the large windows libraries, like
> win32lib? You can't bind them, the new programmer needs to interpret
> win32lib every time they run their little tiny windows app.
>
> It would be nice to be able to turn libraries into semi-compiled
> files, to make interpreting faster. Despite all of your explanations
> above, I still believe this is important. Why?
That's one of the the main points in this context. I wrote about it
already some weeks ago.
> Joe Newbie downloads Euphoria... realises that he needs win32lib to do
> windows programming, downloads that too.
> Now, if he has to wait several seconds for:
> }}}
<eucode>
> include win32lib.ew
> winMain( create(Window, "Hello World",0,Default,Default,200,100,0), Normal)
> </eucode>
{{{
> to run, what's he going to think?
> Without a doubt, he'll think: "This language is really slow!"
Yes, and he'll be completely right in this regard.
> I think there should at least be a way to 'bind' or 'shroud' libraries
> in such a way that a program can include them without the parser
> having to interpret it every time.
Eu 2.4 *has* this possibility. Paradoxically, it's not available in Eu
2.5 any more, while parsing time has considerably increased ...
> Do you think that's technically
> feasible? I guess a header in the file containing lookups for each
> global function and variable name would be enough. It could be only
> available to registered users, that's fine...
Regards,
Juergen
--
Have you read a good program lately?
|
Not Categorized, Please Help
|
|
