Report on WinCE Rebol --- Speed Comparison
[1/7] from: dness::home::com at: 27-Jun-2001 22:23
Since questions have been raised here about REBOL on a Pocket PC class machine,
I thought it might be useful to run and simple experiment and report on it here.
The Experiment
I adapted the distributed `factorial.r' program to calculate a fibonacci
number. I did this in the most straightforward way (involving double recursion)
in a simple attempt to produce a program that would consume a detectable amount
of execution time so that comparisons could be made.
The Contexts
I ran the program in five different contexts:
Context 1: Rebol\View 1.2.1.3.1 on a 333mhz iDOT computer with 384mb ram
Context 2: Rebol Core 2.5.0.3.1 on the same machine
Context 3: Rebol\View 1.2.1.3.1 on an 800mhz Sony computer with 512mb ram
Context 4: Rebol Core 2.5.0.3.1 on the same machine
Context 5: Rebol\core 2.5.0.15.5 on a 206mhz Compaq iPAQ with 32mb
In all cases f(25) was calculated and returned (no surprise here) the correct
result.
The Results
Context Timing
1 333View 13 seconds
2 333Core 10 seconds
3 800View 3 seconds
4 800Core 2 seconds
5 iPAQ 1min 16seconds
For comparison with `other worlds' I used the same algorithm coded in both
the `J' and `K' languages (mathematical array processing interpretive languages
both related to APL). The timings were:
J 5.6 seconds and K 0.6 seconds.
I'd be happy to answer any questions or perform other experiments if it would
be meaningful to anyone.
[2/7] from: larry:ecotope at: 27-Jun-2001 20:39
Hi David,
I am curious about your results for Core and View on Windows. I defined the
following tree-recursive fibonacci function at the console using a 450MHz
PIII:
>> fib: func [n][either n < 2 [n] [(fib n - 2) + (fib n - 1)]]
For Core 2.5.0.3.1
>> t: now/time/precise fib 25 now/time/precise - t
== 0:00:01.32
>> t: now/time/precise fib 35 now/time/precise - t
== 0:02:53.78
For View 1.2.1.3.1
>> t: now/time/precise fib 25 now/time/precise - t
== 0:00:01.32
>> t: now/time/precise fib 35 now/time/precise - t
== 0:02:47.03
So the times are pretty much the same for View and Core. Your timings show
View to be noticeably slower and even the 800MHz times are slower than mine.
Wondering how you did your timing? Did it include starting the exe's?
-Larry
[3/7] from: chrismorency:videotron:ca at: 28-Jun-2001 0:19
Hi,
Where can we find this "adaptation" of the factorial.r ? So we could all
reports results ;)
Best,
Chris
[4/7] from: dness:home at: 28-Jun-2001 0:24
Larry Palmiter wrote:
> Hi David,
> I am curious about your results for Core and View on Windows. I defined the
<<quoted lines omitted: 15>>
> Wondering how you did your timing? Did it include starting the exe's?
> -Larry
For the view cases, I clearly brought up the console and then did a
`do %fib.r' that contained the test code. Similarly on the iPAQ.
For the `core' cases, I don't remember if I did a `rebol fib.r' or
rather a `rebol' followed by a `do %fib.r'. In my timings I didn't
use `precise' as I didn't know about it (until I read your note, that is).
For comparison, bringing up a `view' console and then
running the code as you present it directly on my 800mhz machine produces
0.49 sec (compared to your 1.32) and 57.84 sec compared to your 2:47.83,
which suggests a bit more than the 450/800 comparison one might expect, but
nevertheless is in the same `ball park'.
[5/7] from: dness:home at: 28-Jun-2001 0:56
Christian Morency wrote:
> Hi,
>
> Where can we find this "adaptation" of the factorial.r ? So we could all
> reports results ;)
>
> Best,
> Chris
>
It's essentially the code presented in Larry Palmiter's note above
which I'll copy here in case you missed it:
Larry Palmiter wrote:
[6/7] from: chrismorency:videotron:ca at: 28-Jun-2001 2:21
He he,
First thanks to David for providing me the code ;) and here's the results on
my E125 (after soft reset) and Clio :
E-125 : 46.386 seconds... which is strange considering it's a 150mhz (MIPS)
and the iPAQ is 206mhz (StrongArm)
Clio C1000 : 1 min 2 seconds.
Btw, when I mentioned the clio was slow, it was mainly because of the screen
refresh of the console window, and not really it's operation... but yet, it
stills is a ~80 mhz !
Best,
Chris
[7/7] from: mat:plothatching at: 28-Jun-2001 10:08
Heya Christian,
CM> E-125 : 46.386 seconds... which is strange considering it's a 150mhz (MIPS)
CM> and the iPAQ is 206mhz (StrongArm)
CM> Clio C1000 : 1 min 2 seconds.
Compaq Aero 8000: 23 seconds (no precise on this Rebol version?!) This
is a full-sized 800 x 600 sub notebook style WinCE machine. 133MHz
SH4. Excellent unit but discontinued.
0:00:00.4 <- Athlon 1400MHz :)
--
Mat Bettinson - EuroGamer's Gaming Evangelist with a Goatee
http://www.eurogamer.net | http://www.eurogamer-network.com
Notes
- Quoted lines have been omitted from some messages.
View the message alone to see the lines that have been omitted