OpenEuphoria

Re: case and foreach

• Posted by codepilot Gmail Account <codepilot at gmail.com> Mar 31, 2005
• 559 views

I think the modified "for x in s do" is just about perfect.
But what about selectively not doing the iteration on some elements,
like every even one.
for x in s do-all
end for
for i = 1 to length(seq) by 2 do--odd
x=seq[i]
end for
for i = 2 to length(seq) by 2 do--even
x=seq[i]
end for
for x in s by 2 do--odd
end for
for x=2 in s by 2 do--even
constant bitmap={1,0,1,0,1,0,1,0}
for x in s by bitmap do--any arbitrary bitmap same length as s
end for
--same as
for i = 1 to length(s) do
if not bitmap[i] then next end if
x=s[i]
end for
--but since there is not next
for i = 1 to length(s) do
if bitmap[i] then
x=s[i]
..............
.............
..........
end if
end for
Lets try this example and see the optimization
sequence a,b
a=rand(repeat(128,256))-1--256 seq all bytes
b=rand(repeat(128,256))-1--256 seq all bytes--
for i = 1 to length(a) do
a[i]+=b[i]
end for--normal method
for x in a and y in b do
x+=y
end for--new method
--new method can allow for sse2 optimization of 16bytes fit in each,
and if you could hint the code to tell the translator or interpreter
that 'a' and 'b' will all be bytes then it could do the following

mov eax, address of a
mov ebx, address of b
MOVDQA xmm1,[eax]
PADDUSB xmm1, [ebx]
MOVNTDQ [eax], xmm1

in a loop, and if it knew the length of the sequence it could even
unroll it, and make huge speed.


On Thu, 31 Mar 2005 15:03:21 -0800, Alexander Toresson
<guest at rapideuphoria.com> wrote:
> 
> posted by: Alexander Toresson <alexander.toresson at gmail.com>
> 
> Jason Gade wrote:
> >
> > I've thought about foreach and I don't really think that it needs to be
> > added to the
> > language because it can be done with a function or procedure.
> >
> > foreach(list_of_stuff, routine_id(some_action))
> >
> > or
> >
> > list_of_stuff = foreach(list_of_stuff, routine_id(some_action)) --Prefer
> > pass by reference
> > or optimization of this case
> 
> Thing is, this puts the loop and the code inside the loop in different places
> making the code harder to read. Imagine how eu code would be if the for loop
> was implemented the same way, as a function or procedure! Spaghetti-code.
> 
> > But it is something that could be helped by having pass by reference and an
> > eval()
> > function.
> >
> > One problem with a built-in foreach -- does it only iterate the top level of
> > a sequence
> > or does it dig down through sub sequences?  I think that is one reason to
> > leave it
> > as a procedure or function.
> 
> With the same logic, why should the find() function exist? It only iterates
> the top level of a sequence, yet it's very useful and your code can get much
> faster if you structure your sequences so you can use it.
> 
> Regards, Alexander Toresson
> 
> 
> 
>