From: "José A. S. Alegria" <jose.alegria@...>
Date: 2003-04-23T06:08:20+09:00
Subject: Performance test: 1.8.0p2 versus 1.6.8

This is a multi-part message in MIME format.

------=_NextPart_000_0005_01C3091B.9E68EF70
Content-Type: text/plain;
	charset="iso-8859-1"
Content-Transfer-Encoding: quoted-printable

Hi all! I've just run an experiment to see how Ruby 1.8.0p2 and Ruby =
1.6.8 compared to Python 2.2.2 on Mandrake Linux 9.0. For this I used a =
1GB Dell workstation with dual Xeon CPUs and dual SCSI 18GB disks. I =
tried to test basic sequential text file I/O, string.split(";") and =
hashing. The test CSV file was reasonably large (around 400MBytes and =
3.5 million lines).

The timmings I've got are:

Ruby 1.6.8 on Linux
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D

Test 1 --> elapsed time to read the full file, line by line, and =
counting the number of lines =3D=3D> 23 seconds
Test 2 --> the same as above but now including for each line read
                          fields =3D line.split(";", -1)
                          n_fields +=3D fields.length
                =3D=3D> 57 seconds (delta=3D34 seconds for the extra =
string.split stuff)
Test 3 --> the same as Test 2 but now adding an hash table =
(table=3DHash.new(0)) to count all occurrences of different values for =
fields[5]
                          table[fields[5]] +=3D 1
                =3D=3D> 65 seconds (delta=3D8 seconds for the extra =
hashing stuff)
       =20
Ruby 1.8.0p2 also on Linux
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D

Test 1 -->   7 seconds  (3 times faster than version 1.6.8 !!!)
Test 2 --> 83 seconds (delta=3D76 secs --> quite slower than version =
1.6.8!!!)
Test 3 --> 91 seconds (delta=3D8 secs --> same speed as version 1.6.8)

First conclusion: 1.8.0p2 is much faster than version 1.6.8 for basic =
sequential text I/O but string operations like str.split(";", -1) seem =
to be 50% slower! Any reason for this?

Python 2.2.2 also on Linux
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D

Test 1 --> 8 secs (aprox. the same speed as Ruby 1.8.0p2)
Test 2 --> 62 secs (delta=3D54 secs, a bit faster than Ruby 1.6.8)
Test 3 --> 73 secs (delta=3D9 secs --> a tiny bit slower than Ruby =
1.8.0p2 or 1.6.8)

So, at least for these very very basic operations Ruby compares well =
with Python. Any idea why Ruby 1.8.0p2 is slower than version 1.6.8 for =
such a basic operation like string.split(";", -1) ?

Regards,

J. Alegria

BTW: While Python on Windows Xp (on the same machine) is almost as fast =
as on Linux the same doesnt't seem true for Ruby. The Windows version is =
slower than the linux one!

------=_NextPart_000_0005_01C3091B.9E68EF70
Content-Type: text/html;
	charset="iso-8859-1"
Content-Transfer-Encoding: quoted-printable

<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0 Transitional//EN">
<HTML><HEAD>
<META http-equiv=3DContent-Type content=3D"text/html; =
charset=3Diso-8859-1">
<META content=3D"MSHTML 6.00.2800.1141" name=3DGENERATOR>
<STYLE></STYLE>
</HEAD>
<BODY bgColor=3D#ffffff>
<DIV><FONT face=3DArial>Hi all! I've just run an experiment to see how =
Ruby=20
1.8.0p2 and Ruby 1.6.8 compared to Python 2.2.2 on Mandrake Linux 9.0. =
For this=20
I used a 1GB Dell workstation with dual Xeon CPUs and dual SCSI 18GB =
disks. I=20
tried to test basic sequential text file I/O, string.split(";") and =
hashing. The=20
test CSV file was reasonably large (around 400MBytes and 3.5 million=20
lines).</FONT></DIV>
<DIV><FONT face=3DArial></FONT>&nbsp;</DIV>
<DIV><FONT face=3DArial>The timmings I've got are:</FONT></DIV>
<DIV><FONT face=3DArial></FONT>&nbsp;</DIV>
<DIV><FONT face=3DArial>Ruby 1.6.8 on Linux</FONT></DIV>
<DIV><FONT =
face=3DArial>=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D</FONT></DIV>
<DIV><FONT face=3DArial></FONT>&nbsp;</DIV>
<DIV><FONT face=3DArial>Test 1 --&gt; elapsed time to read the full =
file, line by=20
line,&nbsp;and counting the number of lines =3D=3D&gt;&nbsp;23 =
seconds</FONT></DIV>
<DIV><FONT face=3DArial>Test 2 --&gt; the same as above but now =
including for each=20
line read</FONT></DIV>
<DIV><FONT=20
face=3DArial>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;=
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;=20
<FONT>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; fields =3D =
line.split(";",=20
-1)</FONT></FONT></DIV>
<DIV><FONT=20
face=3DArial>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;=
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&=
nbsp;&nbsp;&nbsp;=20
n_fields +=3D fields.length</FONT></DIV>
<DIV><FONT=20
face=3DArial>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;=
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;=20
=3D=3D&gt; 57 seconds (delta=3D34 seconds for the extra string.split=20
stuff)</FONT></DIV>
<DIV><FONT face=3DArial>Test 3 --&gt; the same as Test 2 but now adding =
an hash=20
table (table=3DHash.new(0)) to count all occurrences of different values =
for=20
fields[5]</FONT></DIV>
<DIV><FONT=20
face=3DArial>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;=
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&=
nbsp;&nbsp;&nbsp;&nbsp;table[fields[5]]=20
+=3D 1</FONT></DIV>
<DIV><FONT=20
face=3DArial>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;=
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;=20
=3D=3D&gt;&nbsp;65 seconds (delta=3D8 seconds for the extra hashing=20
stuff)</FONT></DIV>
<DIV><FONT=20
face=3DArial>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;</FONT></DIV=
>
<DIV>
<DIV><FONT face=3DArial>Ruby 1.8.0p2 also&nbsp;on Linux</FONT></DIV>
<DIV><FONT =
face=3DArial>=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D</FONT></DIV>
<DIV><FONT face=3DArial></FONT>&nbsp;</DIV>
<DIV><FONT face=3DArial>Test 1 --&gt;&nbsp;&nbsp; 7 seconds&nbsp; (3 =
times faster=20
than version 1.6.8 !!!)</FONT></DIV>
<DIV><FONT face=3DArial>Test 2 --&gt; 83 seconds (delta=3D76 secs --&gt; =
quite=20
slower than version 1.6.8!!!)</FONT></DIV>
<DIV><FONT face=3DArial>Test 3 --&gt; 91 seconds (delta=3D8 secs --&gt; =
same speed=20
as version 1.6.8)</FONT></DIV>
<DIV><FONT face=3DArial></FONT>&nbsp;</DIV>
<DIV><FONT face=3DArial>First conclusion: 1.8.0p2 is much faster than =
version=20
1.6.8 for basic sequential text I/O but string operations like =
str.split(";",=20
-1) seem to be 50% slower! Any reason for this?</FONT></DIV>
<DIV><FONT face=3DArial></FONT>&nbsp;</DIV>
<DIV><FONT face=3DArial>Python 2.2.2 also on Linux</FONT></DIV>
<DIV><FONT =
face=3DArial>=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
</FONT></DIV>
<DIV><FONT face=3DArial></FONT>&nbsp;</DIV>
<DIV><FONT face=3DArial>Test 1 --&gt; 8 secs (aprox. the same speed as =
Ruby=20
1.8.0p2)</FONT></DIV>
<DIV><FONT face=3DArial>Test 2 --&gt; 62 secs (delta=3D54 secs, a bit =
faster=20
than&nbsp;Ruby 1.6.8)</FONT></DIV>
<DIV><FONT face=3DArial>Test 3 --&gt; 73 secs (delta=3D9 secs --&gt; a =
tiny bit=20
slower than Ruby 1.8.0p2 or 1.6.8)</FONT></DIV>
<DIV><FONT face=3DArial></FONT>&nbsp;</DIV>
<DIV><FONT face=3DArial>So, at least for these very very basic =
operations Ruby=20
compares well with Python. Any idea why Ruby 1.8.0p2 is slower than =
version=20
1.6.8 for such a basic operation like string.split(";", -1) =
?</FONT></DIV>
<DIV><FONT face=3DArial></FONT>&nbsp;</DIV>
<DIV><FONT face=3DArial>Regards,</FONT></DIV>
<DIV><FONT face=3DArial></FONT>&nbsp;</DIV>
<DIV><FONT face=3DArial>J. Alegria</FONT></DIV>
<DIV><FONT face=3DArial></FONT>&nbsp;</DIV>
<DIV><FONT face=3DArial>BTW: While Python on Windows Xp (on the same =
machine) is=20
almost as fast as on Linux the same doesnt't seem&nbsp;true for Ruby. =
The=20
Windows version is slower than the linux one!</FONT></DIV>
<DIV><FONT face=3DArial></FONT>&nbsp;</DIV></DIV></BODY></HTML>

------=_NextPart_000_0005_01C3091B.9E68EF70--