OpenEuphoria

Re: Faster lower() command

• Posted by jacques deschênes <desja at globetrotter.net> Jan 20, 2007
• 554 views

I'm using windows xp pro running on DELL inpiron 1501 with 1Go Ram
CPU AMD turion 64x2 1.60Ghz

Why euphoria interpreter should create 5 or 6 * 1000000 long sequence?
each character is represented by an integer so its a 4000000 bytes array

new test

function lower1( sequence x )
    integer c
    for i = 1 to length(x) do
        c = x[i]
        if c <= 'Z' then
            if c >= 'A' then
                x[i] = c + 32
            end if
        end if
    end for
return x
end function

function lower(object o)
  return o + (o>='A' and o<='Z')*32 -- work as well with atom or sequence
end function



atom t
constant txt=repeat('A',1000000)
sequence l

t = time()
l = lower1(txt)
? time()-t
t = time()
l = lower(txt)
? time()-t

t = time()
for i = 1 to 1000 do
  l = lower1(txt)
end for
? time() - t

t = time()
for i = 1 to 1000 do
  l = lower(txt)
end for
? time() - t

result
0.13
0.11
64.2
95.41

This time its clear that lower1() is faster than lower()
I don't know how the loop is implemented inside euuphoria interpreter but
there is place for optimization!

regards,
Jacques Deschênes


Pete Lomax wrote:
> 
> jacques deschênes wrote:
> > on my laptop
> > lower1() give .12sec
> > lower()  give .09sec
> I not only agree with other posters but am downright jaw-dropping astonished
> that you got lower() to appear faster that lower1() in ANY circumstances!
> 
> My analysis was that lower() must create 5 or 6 * 1,000,000 - long sequences
> before finally calculating the result.
> 
> > I can't figure how lower1() can be faster on your machine.
> 
> Are you by any chance on a Linux box while we're all on Windows?
> 
> Puzzled,
> Pete