[Erp5-dev] Performance on 32 vs. 64 bits machines

[Jacek Medrzycki]

Hi!

I've run some performance tests recently.

I've installed 32 bits Mdv. 2008.1 with ERP5, run tests, then on the 
same machine I've installed 64 bits Mdv 2008.1 with ERP5 and run the 
same tests. Both systems were installed on the same disks/partitions to 
exclude any hardware influence. All systems were "out of the box" 
installations, with ERP5 from their respective repositories.

The tests included some stock tests (amont others testPerformance, 
testInventory, testInventoryApi) and some customized tests (eg. multiple 
rendering a complicated page (with cacheing disabled) from one of sites 
we are currently developing), testMultiRequestPerformance on a fresh 
site, oood performance tests (based on testOoodAllFormats). All tests 
were timed with time command.

Surprisingly, I have noticed practically no performance gain for 64 bits 
over 32 bits. Oood test was even slower on 64bits. Inventory tests were 
a little bit faster on 64, but the difference is not very relevant (eg. 
555 vs 564 seconds). Other tests gave practically the same times on both 
machines.

The machine on which the tests was run: 64 bit AMD processor with 
1800MHz clock, and 1G of memory.

I have heard that there should be significant performance gain on 64 
bits, so I'm a little bit confused. Is something wrong with my test 
environment? Perhaps the machine is too weak to see the performance 
gain, or some additional config work should be done on 64, or somewhat?

Regards
Jacek
-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/x-pkcs7-signature
Size: 2218 bytes
Desc: S/MIME Cryptographic Signature
URL: <http://mail.tiolive.com/pipermail/erp5-dev/attachments/20080611/b59a29c6/attachment.bin>