Carl Sassenrath of Amiga fame released Rebol for Gentoo and Debian last night. You can find it via this link:

http://www.rebol.net/article/0204.html

We are very interested in having it tested.

REBOL/Core includes hundreds of functions, 40 native datatypes, 10 Internet protocols, compression, built-in help, and more. /Core is the console version of REBOL. It is the fundamental system of REBOL and is used mainly as a server-side or lower-level tool (because it has no GUI, just command line).

/Core was built using Debian and it also runs fine on Gentoo. It may also run on Ubuntu/(K)ubuntu

The download is only 230K!

Note that this is not a full distro, just the program file. To give it a quick test try, run REBOL from the terminal and type:

do http://www.rebol.com/speed.r

That will download and run a simple speed test (assuming you have direct Internet access - no proxy). If you need a proxy, you can set it up. Google "REBOL proxy setup".

Please post your experiences here. We may even have some "Friday Free Stuff" for enthusiastic effort!

R&B

Hi

I have never written anyting in Rebol before. I have done some testings with "Bonnie" on the Pegasos and got some depressing resullts.

This is a little test script for testing i/o perfomance in Rebol. I have set "bytes" to 100M, so it will write a 100M file one byte at a time. Beware that this wil take some time. If you are in a hury please change "bytes" to somthing else.

Btw . I have put a slightly modiified version of "Bonnie" for morphos on www.pegasos.org.

The script, bench.r

Now more or less obsolete try version 1.1 in my second post.

**********************CUT HERE***************************
REBOL [

Title: "Bench"

Author: "Michael Grunditz"

Version: 1.0

Purpose: "Testing Rebol I/O perfomance"

]

write %junk.txt ""

x: "x"
bytes: 104857600

recycle

t1: now/precise

loop bytes[

write/binary/append %junk.txt x

]

delete %junk.txt

t2: now/precise

d: difference t2 t1

sp: to-integer (bytes) / d/3 / 1024 / 1024

print ["Writing " bytes "to disk:" d "-" sp "MB/S"]
halt
*******************CUT HERE**********************************

And another one. This does the same thing but opens the file only one time. The only problem here is that the actual writing is perfomed when the file handle is closed.

Fixed ! open/direct did solve that problem.

REBOL [

Title: "Bench"

Author: "Michael Grunditz"

Version: 1.1

Purpose: "Testing Rebol I/O perfomance"

]

testfile: open/direct %junk.txt

bytes: 104857600
recycle

t1: now/precise

loop bytes[

insert tail testfile "x"
]
close testfile

t2: now/precise ; Time before the delete , it is the writing speed we want ti know

delete %junk.txt



d: difference t2 t1

sp: to-integer (bytes) / d/3 / 1024

print ["Writing " bytes "bytes to disk:" d "-" sp "KB/S"]
halt

@micken - happy to see these efforts! Did you sort this out?

R&B

Actualy the second listing sorts things out. This shows one of the weakest points on a pegasos. I/O performance in byte to byte transfers is realy slow.

I only have standard IDE disks to test with , but it would be interesting to se if there is a differnce with fast scsi or sata.

byte to byte copying is slow by definition. I would scarcely blame the pegasos for that. Try byte per byte copying on IDE disks on a P4 and I'be surprised if you would get *much* better performance. A little bit, yes due to faster cache etc. But check the same disk in the same mode (not SATA, or faster modes) on both systems and you would find the processing is more less the same speed there.

Of course, that's not to deny that the I/O on the pegasos would not benefit from a speed bump. And I don't know how REBOL handles this, but byte/byte I/O benchmarking is hardly reallife benchmark

Konstantinos

I have done many benchmarks on very very small systems and very very large systems and every benchmark is worth doing if you can compare the results with something else.

A simple putc() test performs much better on a older pc than on a pegasos and that IS a issue.

micken

I'm not going to deny that, I agree. But they're just indicators. Personally, I tend to be very sceptical to most benchmarks unless they're actual applications with actual data to work on (eg. MySQL testsuites or scientific simulations).
Could you please back this statement? I am willing to accept results, but not this.
At worst it might just say that the glibc putc() on the powerpc is a lame implementation, compared to the one in x86. Which is quite often the truth as I've searched the glibc codebase myself many times. But it definitely doesn't say that Pegasos I/O is worse than that of an old PC.

Konstantinos

Well, if you take my rebol script as an example and change the "open mode" so that every putc is performed in ram you get about 4000K/S. That is the same that you get with "Bonnie" with the same instruction.

On a pc with linux you can get somthing like 30000K/s with putc.

If there is a bad implementation in the glibc , then it will deffinetly cause people to think that the pegasos platform is a bad idea. So this is important.

Some other day I will try to do the test with assembler using linux syscalls.

Here are my results on a PIII@700Mhz with a not-that-fast IDE disk.
And here are the results on my Pegasos II: The results speak for themselves. I still don't think the Pegasos I/O is worse, but I'm pretty sure the putc() implementation is not optimal, which is sth I intend to check more thoroughly.
Thanks for bringing that to our attention.

Konstantinos

I remember now that the real difference with pc and peg was when I played with Bonnie++. But I have not looked at that source so I don't know why.

The peg got somthing like 400K/s

I realy look forward to when I can test Solaris on pegasos.

Here is what the MOS-Bonnie gave as results:
-------Sequential Output-------- ---Sequential Input--
-Per Char- --Block--- -Rewrite-- -Per Char- --Block---
Machine MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU
1000 6208 30.6 32479 3.4 17126 9.1 8297 43.6 43717 5.1

This thread is about REBOL on pegasos
I am sorry that I started to discuss
I/O performance.

Anyway , I think that REBOL for
PEGASOS is a very cool thing. It is
a easy to read language and it is
very easy to create services. The
block methology is very fun.

@BBRV

If you feel like donate something so
is a disk very welcome esp. if it is
a scsi one. I like to continue to
investegate disk performance, both in
linux and in MorphOS. It seems like
disk performance is a little better
in mos than in linux. I think that
Solaris will performe even better.

Best regards
Michael Grunditz
michael.grunditz@telia.com