OpenEuphoria

Re: Web Access? ++ speed freaks

• Posted by Lmailles <Lmailles at AOL.COM> May 25, 1998
• 862 views

[web access]

> Is there anyone on this list who has no, or expensive, web access?
>  Specifically, is there any need for e-mail access to Euphoria archives?
>  (program archives - not the listserver messages)

In England we have to pay for local phone calls, and my machine is a 386, so
web access is very slow, even with images turned off, it takes a number of
seconds to render a page and scrolling sends you to sleep.

>  If there is significant need, maybe I could build a mail-bot to
>  return lists and files.

If you could manage to figure out a simple interface (very tough) then this
would be fantastic. Keep us posted on the proceedings


[Speed freaks and length()]

I ran a benchmark, to prove that use of length() caused significant overhead.
To make it fair, I did a typical usage, with a slice. The results shocked me.
With the defaults tick_rate there was no measurable difference between
length() and leng up to 10000 repetitions. I had to do 100000 before I got a
.39 second difference on my 386-25. The program is below

"Temptation will always be there, the opportunity won't" - Les Mailles

Daniel


begin test
file--------------------------------------------------------------------------
----

--efficiency of length() in typical usage tester

constant repetitions=100000 --number of times to repeat the lookup
sequence junk,garbage,lookup
atom now
integer leng
junk=repeat("0z;",300)

--initialise test values
lookup={}
for a=1 to repetitions do
    lookup = lookup & rand(300)
end for

--test 1
--set clock
now=time()
leng=length(junk) --pre-calculate length
for a=1 to repetitions do
    --do a slice, and don't let it make a pointer
    garbage=junk[lookup[a]..leng]&1
end for
? time()-now

--test 2
--set clock
now=time()
--do not pre-calculate length
for a=1 to repetitions do
    --do a slice, and don't let it make a pointer
    garbage=junk[lookup[a]..length(junk)]&1
end for
? time()-now