1. sorta kinda Re: Types but different...
jiri babor wrote:
>Yes and No, David. In fact it will run faster, and
>sometimes *much* faster (I do not know why), if you
>make everything an object, because Euphoria has
>nothing to check in case of an object.
ummm you just answered your own question :)
it's why there is no difference when you put in
'without type_check'.
EU doesn't have to check anything if
1>it's an object and type check is on,
2>if type is off.
so you get the speed in case 1 that you would get
from case 2.
anyway, here is my real reason to post, but it was
this post that reminded me about a peculiar behavior
i've noticed over and over...
please, others, if you get a chance, try this...
might be my machine, i dunno. (i have tried it
on other machines)
this regards a post i made a while back, got a sprinkling
of posts back about it, tho quite few and far between.
during benchmarks of (for ex) highly graphical-natured,
rather large EU programs, i kept finding that once bound,
the proggies got reduced frame rate readings, sometimes
1-2% and other pushing 20-25% reductions...
same exact programs, per se. (one bound, ran by typing
its name, the other ran simply as "ex name.ex")
oh, and bound using defaults to bind.ex.
this was noticed on a
486--vesaLB,
p75--stealth64,
a pair of p133's--stealth 3d2000,
p200mmx--stealth turbo2000 and voodooII
p200mmx--s3virge325PCI DX5.0, DX5.2 and DX6final drivers
all showed this peculiarity... not noticeable, even with
tickrate(1000) if the program just wasn't that large.
i know...i've been in that party bowl one too many times :)
--Hawke'
2. Re: sorta kinda Re: Types but different...
Hawke just wrote:
>jiri babor wrote:
>>Yes and No, David. In fact it will run faster, and
>>sometimes *much* faster (I do not know why), if you
>>make everything an object, because Euphoria has
>>nothing to check in case of an object.
>ummm you just answered your own question :)
No, I did not. I am still puzzled by the "sometimes *much* faster" bit.
According to Robert, if I remember correctly, you can expect about 5%
penalty, in round figures, for type checking. But in fact in many cases I
clocked much larger increases. I suspect the actual penalty per item is much
higher, and 5% is just some sort of average for a typical program. jiri
3. Re: sorta kinda Re: Types but different...
{Hawke'}
>ummm you just answered your own question :)
{jiri}
>No, I did not. I am still puzzled by the "sometimes
>*much* faster" bit.
oops! sowwy :) I misread the post, and thought you meant
something like "i am still puzzled by the difference in
speed" instead of "i am puzzled by the random and often
huge difference in speed".
in refman.doc it says:
Performance Note:
When typical user-defined types are used extensively,
type checking adds only 20 to 40 percent to execution time.
Leave it on unless you really need the extra speed.
You might also consider turning it off for just a few
heavily-executed routines.
Profiling can help with this decision.
---end
TWENTY to FORTY??? percent!?!?!
well it did say if 'types are used extensively'
let's see, that would mean 20FPS would drop to
between 12-16FPS...good for debugging perhaps,
not good otherwise.
--Hawke'
