Re: while vs loop


#1

Hi,I saw this thread from last week: while vs loop,

I don’t think anybody mentioned the slight performance difference
between
the two,
(or maybe I got the wrong info!!)
In fact, I had always assumed the ‘loop do’ construct to be better than
‘while 1 do’, if only to be more elegant or concise for implementing
infinite loops… but it turns out it seems to be the other way:

##############
require ‘benchmark’
Benchmark.bm do |x|
x.report(‘while1’) do
n = 0
while 1 do
break if n >= 10000000
n += 1
end
end
x.report('loop ') do
n = 0
loop do
break if n >= 10000000
n += 1
end
end
end

##############
RESULTS:

               user     system      total        real

while 1 19.770000 0.230000 20.000000 ( 34.458926)
loop 25.870000 0.300000 26.170000 ( 42.804569)

###########

I’m running this benchmark on 1.8.6 patchlevel 114 (universal-darwin9.0)

So why is it behaving this way?
Are every implementations doing the same?

Thanks!

L-P

On Sat, Feb 14, 2009 at 7:16 AM, Nobuyoshi N.
<n…http://groups.google.com/groups/unlock?_done=/group/ruby-talk-google/browse_thread/thread/3ee74ef131ec90ba&msg=2f988b0deb69e61d
@ruby-lang.org> wrote:

Hi,

At Sat, 14 Feb 2009 15:03:03 +0900,
Vetrivel V. wrote in [ruby-talk:328175]:

what is the Difference between loop and while in ruby ?

loop do
end

loop is a kernel method which takes a block. A block
introduces new local variable scope.

loop do
a = 1
break
end
p a #=> causes NameError

while 1
end

while doesn’t.

while 1
a = 1
break
end
p a #=> 1


Nobu Nakada

furthermore loop do has an implicit rescue clause for a StopIteration
exception
(I believe 1.8.7 and 1.9.1 only IIRC)

therefore

loop do
some_enumerator.next
end
becomes a convenient idiom.

HTH
Robert


#2

On Feb 20, 2009, at 13:57 , Louis-Philippe wrote:

Hi,I saw this thread from last week: while vs loop,

I don’t think anybody mentioned the slight performance difference
between
the two,

If you’re going to talk performance differences between while and
loop, don’t forget for and each:

                      user     system      total        real

each 1.370000 0.010000 1.380000 ( 1.380012)
for 1.630000 0.000000 1.630000 ( 1.652562)
each-var 2.450000 0.010000 2.460000 ( 2.479478)
while 5.510000 0.010000 5.520000 ( 5.532909)
loop 7.730000 0.020000 7.750000 ( 7.786959)

require ‘benchmark’

max = (ARGV.shift || 1_000_000).to_i

puts “# of iterations = #{max}”
Benchmark::bm(20) do |x|
x.report(“each”) do
(0…max).each do
end
end

x.report(“for”) do
for i in 0…max do
# do nothing
end
end

x.report(“each-var”) do
(0…max).each do |i|
end
end

x.report(“while”) do
n = 0
while true do
break if n >= max
n += 1
end
end

x.report(“loop”) do
n = 0
loop do
break if n >= max
n += 1
end
end
end


#3

mmm, Interesting to see loop being this high in the list,
but even if they all loop from 0 to X doesn’t mean they can all
compared…
I’m not sure what was the initial question (!!!), but mine is about
infinite
loop implementations,
where list iterators of the .each flavour don’t do.

So THE fastest ruby infinite loop is ‘while 1’?

2009/2/20 Ryan D. removed_email_address@domain.invalid


#4

On Fri, Feb 20, 2009 at 6:02 PM, Louis-Philippe
removed_email_address@domain.invalidwrote:

mmm, Interesting to see loop being this high in the list,
but even if they all loop from 0 to X doesn’t mean they can all
compared…
I’m not sure what was the initial question (!!!), but mine is about
infinite
loop implementations,
where list iterators of the .each flavour don’t do.

So THE fastest ruby infinite loop is ‘while 1’?

Anything purporting to be an infinite loop which runs faster than
something
which is an infinite loop is, by definition NOT an infinite loop!

Rick DeNatale

Blog: http://talklikeaduck.denhaven2.com/
Twitter: http://twitter.com/RickDeNatale
WWR: http://www.workingwithrails.com/person/9021-rick-denatale
LinkedIn: http://www.linkedin.com/in/rickdenatale


#5

On Feb 20, 2009, at 15:02 , Louis-Philippe wrote:

mmm, Interesting to see loop being this high in the list,
but even if they all loop from 0 to X doesn’t mean they can all
compared…
I’m not sure what was the initial question (!!!), but mine is about
infinite
loop implementations,
where list iterators of the .each flavour don’t do.

So THE fastest ruby infinite loop is ‘while 1’?

wrong


#6

Rick DeNatale wrote:

Anything purporting to be an infinite loop which runs faster than something
which is an infinite loop is, by definition NOT an infinite loop!

Not sure Georg Cantor would agree with you. One infinite loop might
only go through a countable (but infinite) number of iterations, i.e.
aleph null, whereas another might go through an uncountable number of
iterations.

Suppose one loop iterates over just the rationals, while another
iterates over the reals. The former iterates fewer times than the
latter, even though both are infinite.

Still, your point is well taken. :slight_smile:


#7

On Sat, Feb 21, 2009 at 3:42 PM, Rick DeNatale removed_email_address@domain.invalid
wrote:

On Fri, Feb 20, 2009 at 6:02 PM, Louis-Philippe removed_email_address@domain.invalidwrote:

Anything purporting to be an infinite loop which runs faster than something
which is an infinite loop is, by definition NOT an infinite loop!

This was one of the most famous stories in our department . We were
developing a Debugger and my colleague was debugging an infinite loop,
which happened to be inside another infinite loop. Of course he did
not discover the problem of the outer infinite loop before having
fixed the inner one.
When our project leader asked him about the progress he had made, he
genuinely replied: “It still loops, but much faster”. We happened to
pull his leg for quite some time of course :)?
Robert


#8

This is indeed very fun and mind boggling!Still, pragmatically, not all
infinite loops are equal, in the performance of their implementations
that
is.
So taking Ryan’s idea to twist iterators to perform infinite loops, here
is
the benchmark again.

For the comparison to be as close as possible between the control
structures, I use the same inner routine for all, ignoring the fact that
I
could derive incrementation from the infinite range iteration, because
this
is not the subject of the benchmark.

Ruby Infinite loops Inplementations:
user system total real
while1 31.960000 0.150000 32.110000 ( 36.749597)
loop 42.240000 0.190000 42.430000 ( 45.533708)
Infinite.each 108.400000 0.970000 109.370000 (117.421059)
for Infinite 112.070000 1.010000 113.080000 (126.712828)

#############

Benchmark.bm do |x|
x.report(‘while1’) do
n = 0
while 1 do
break if n >= 1000000
n += 1
end
end
x.report(‘loop’) do
n = 0
loop do
break if n >= 1000000
n += 1
end
end
x.report(‘Infinite.each’) do
n = 0
(0…(1/0.0)).each do
break if n >= 1000000
n += 1
end
end
x.report(‘for Infinite’) do
n = 0
for i in (0…(1/0.0)) do
break if n >= 1000000
n += 1
end
end
end

#############

2009/2/21 Robert D. removed_email_address@domain.invalid


#9

On Sat, Feb 21, 2009 at 4:49 PM, Jeff S. removed_email_address@domain.invalid
wrote:

Rick DeNatale wrote:

Anything purporting to be an infinite loop which runs faster than
something
which is an infinite loop is, by definition NOT an infinite loop!

Not sure Georg Cantor would agree with you. One infinite loop might only go
through a countable (but infinite) number of iterations, i.e. aleph null,
whereas another might go through an uncountable number of iterations.
Might it? I do not think so! I believe that a Turing complete language
can only loop countable times, which of course should be fixed ;).
R.