Re: [gentoo-dev] python-2.3.2 testing required

[Alastair Tse <liquidx@gentoo.org>]

[-- Attachment #1: Type: text/plain, Size: 4749 bytes --]

On Wed, 2003-11-12 at 18:46, Alastair Tse wrote:
> The reason why I'm not making this default is because UCS4 python uses
> more memory. An example is supybot (Python IRC bot) that uses 8M for
> UCS2 and 13M for UCS4. But note that this example is not scientific
> because the machines were different in kernel version, compiler and
> compiler optimisations.

I've found a little spare time this weekend to do a little bit of memory
benchmarking to prove/disprove my point about UCS4 using more memory
than UCS2.

I wrote and conducted 2 simple tests that I thought were relevant to
Python on Gentoo. The two tests I conducted were:

1. Generating a large number of Python Unicode Strings and recording the
memory usage.
2. Running "emerge" on various different options and recording the
memory usage.

The results demonstrate that UCS4 is more memory hungry _only_ if a
script/module/application uses unicode strings. This means any bindings
that use PyUnicode_* objects (for example, pygtk) or any script that
uses unicode strings. If a script/module/application does not use
unicode objects, it suffers from no noticable memory impact.

The numbers reported are averages from 3 or more runs. In nearly all
cases, the memory usage was constant.

Results:
========
1 : Generating Unicode Multi-Byte Strings (1 to 10000) strings
(String Size of 256 mbchars stored in a regular python list)
-------------------------------------------------------------------
Strings: (UCS2) Mem   RSS  Shared (UCS4) Mem  RSS  Shared    %+
1               1839  710  1535          1839 711 1535       0
10              1871  712  1535          1871 717 1535       0
100             1904  765  1535          1971 830 1535       3.5
1000            2465  1336 1535          3102 1960 1535      25.84
10000           8213  7052 1535          14445 13309 1535    75.80

2 : Generating Unicode ASCII Strings (1 to 10000) strings
(String Size of 256 chars stored in a regular python list)
-------------------------------------------------------------------
Strings: (UCS2) Mem   RSS  Shared (UCS4) Mem  RSS  Shared    %+
1               1839  710  1535          1839 711  1535      0
10              1871  712  1535          1871 717  1535      0 
100             1904  765  1535          1971 830  1535      3.5
1000            2465  1336 1535          3102 1960 1535      25.84
10000           8213  7053 1535          14445 13309 1535    75.80

3: Max Memory Usage under "emerge -p kde"
-------------------------------------------------------------------
      Mem  RSS  Shared
UCS2: 3222 1893 1955
UCS4: 3123 1769 1955

4: Max Memory Usage under "emerge search kde"
-------------------------------------------------------------------
      Mem  RSS  Shared
UCS2: 3221 1898 1955
UCS4: 3160 1803 1955

Discussion
==========

There are two immediate observations. One is that UCS4 does use more
memory compared to UCS2 when unicode strings are involved. From Test 1
and 2, the VM has an overhead of 1.8M and as more strings are created,
their memory usage difference steadily increase to 75% difference.

The other observation is that if there are is no unicode usage in
application, like "emerge", there is virtually no impact. Actually, in
this case, you'll find that UCS4 uses about 60K ot 100K less memory than
UCS2. I don't have an explanation for that behaviour.

Other observations that can be made which do not relate to the UCS2/UCS4
benchmark is that it doesn't matter if you are primarily dealing with
ASCII or Multi-Byte (eg, CJK characters) strings. As soon as they are
cast as unicode objects, they use more memory. Note that the two runs
have identical memory usage, that is not a mistake.

Another one is that 'emerge' uses the same amount of memory regardless
of what is being run. I had an informal test running just "emerge info"
and it still used approximately the same memory as running more
complicated things like merging packages or searching the package
database.

Other Details
=============
The above results were run with dev-lang/python-2.3.2-r1 with:
Kernel 2.6.0-test9-mm1
Glibc-2.3.2-r8 (w/ nptl)
GCC-3.3.2
Portage 2.0.49-r16

The raw logs for the tests and the scripts used can be found at:
http://dev.gentoo.org/~liquidx/python-test/

Remarks
=======

After running these tests, I still divided about whether UCS4 should be
enabled by default. I'm not seeing the added benefits of UCS4 in
contrast with the memory usage increase it brings. Yet, it also seems
like the "right" thing to do for m17n support.

Cheers,
-- 
Alastair 'liquidx' Tse
 >> Gentoo Developer
 >> http://www.liquidx.net/ | http://dev.gentoo.org/~liquidx/


[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 189 bytes --]